The Schlafmutzes Mini MT’s design and progress

Collapse
X
 
  • Time
  • Show
Clear All
new posts
  • JonW
    Super Senior Member
    • Jan 2006
    • 1582

    #91
    Originally posted by ---k---
    If you strip the screw hole, take a wood match stick. Put some wood glue on it and stick it in the hole. Let it set up a little, then drill you screw in with a little more glue. Good as new. I've done it many times. I learned that trick here. I'm probably not doing it correctly, but it works for me.
    Right. I’ve not had to try that out, but I have read about it here at htguide. First, I’ll try to not need it.

    Comment

    • JonW
      Super Senior Member
      • Jan 2006
      • 1582

      #92
      No news yet on the bad tweeter front. But I did get to spend time on crossovers this weekend. All only on the computer so far. I took the measurement data for the in room, deluxe W16 speaker. Tried to get versions of potential crossovers resembling both LR2 and LR4. Hitting LR4 was easier. LR2 took more hammering at the computer but I got a model that looks OK. I need to massage the models a little more. Might have some changes. Then I’ll make up pictures and post both options in the coming days. It’s gratifying to see the cabinets, drivers, and measurements coming together into something that could resemble a speaker one day, when it grows up.

      Comment

      • 5th element
        Supreme Being Moderator
        • Sep 2009
        • 1671

        #93
        Considering the peak in the 3rd harmonic, in the excel driver, you'd probably be better off with 4th order at below 2kHz and even lower if 2nd order. What are you aiming for currently?
        What you screamin' for, every five minutes there's a bomb or something. I'm leavin' Bzzzzzzz!
        5th Element, otherwise known as Matt.
        Now with website. www.5een.co.uk Still under construction.

        Comment

        • Face
          Senior Member
          • Mar 2007
          • 995

          #94
          4th may work, but even with 4th, a tank may be needed to suppress the breakup 50db below reference.

          Edit: now that I'm looking at it's raw response, a tank or notch will be needed, regardless. http://www.madisoundspeakerstore.com...-woofer-4-ohm/
          SEOS 12/AE TD10M Front Stage in Progress

          Comment

          • JonW
            Super Senior Member
            • Jan 2006
            • 1582

            #95
            Hi Guys,

            Yeah, my guess is that the LR4 will win out over the LR2. But I wanted to see how an LR2 might look. I’ve never done one. Plus crossover design is a place for me to learn a lot by mucking around on the computer.

            Yes, the W16 gets noisy above 1,000 Hz. But the tweeter gets noisy below 2,000 Hz. Where to cross? I’ll guess that the woofer distortion is more objectionable here. And there is the cone floppiness that comes in higher. Need to kill that. So the steeper LR4 may be the way to go.

            Rather than picking a specific crossover frequency, I let the models move around to get the overall best matches of FR, phase, impedance, etc. There is an annoying peak at ~900 Hz that made life a little difficult. In the end, both the LR2 and LR4 cross around 2,000 Hz. After some polishing, I’ll post both versions for comparison and see what you think. This discussion will be easier with pictures.

            Comment

            • 5th element
              Supreme Being Moderator
              • Sep 2009
              • 1671

              #96
              Originally posted by JonW
              My guess is that the LR4 will win out over the LR2. But I wanted to see how an LR2 might look. I’ve never done one. Plus crossover design is a place for me to learn a lot by mucking around on the computer.
              It's only worth toying with LR2 filters with drivers that will support them. Most do not, especially metal cones. The W15CY, with it's 8.2kHz primary break up, would be just about alright if paired with something like the DXT and wave-guide that could support an LR2 down at 1500Hz or so.

              Stick with a LR4 below 2k with the W16. In fact I would want to optimise this design based off of how low the tweeter could play without distress. Zaph's measurements imply that it could cope with a LR4 at 1500Hz, but might get a bit iffy if run at very high levels. In your application, nearfield, I highly doubt that this would ever be an issue.

              Having a lower xover point will also help keep the integration tight, for the relatively close listening position, based off of the C2C spacing etc.

              Originally posted by JonW
              Yes, the W16 gets noisy above 1,000 Hz. But the tweeter gets noisy below 2,000 Hz. Where to cross? I’ll guess that the woofer distortion is more objectionable here. And there is the cone floppiness that comes in higher. Need to kill that. So the steeper LR4 may be the way to go.
              Neither the tweeter or the woofer should be getting or showing noise anywhere during the measurements. If it looks like things are getting noisy (fatter fuzzy trace etc) then the system is at fault. I do recommend you read the guide I wrote and have a bash with ARTA, just to make sure that you're doing everything right. This may seem like a headache, but considering the effort you're putting into this, you've already managed to deal successfully with the impedance headache, it cannot hurt making sure that the FR stuff is as good as it can get too. The ARTA guide, in principle, is directly relate-able to JustMLS too, so you might be able to make improvements to your measurements without even needing to use ARTA.

              The tweeter shows low distortion down pretty low and can be used down that low too. The only issue will be how gracefully the tweeter behaves be as you approach and exceed xmax under high drive levels.

              The woofer has a bell like resonance and this propagates into the distortion at integer multiples below the main resonance. In other words you play a 2kHz tone through the woofer, the motor generates a third harmonic at 6kHz and as the cone rings at about 6k, it resonates in unison with the 3rd harmonic, amplifying it and degrading the performance. This is what I believe Jon coined cone resonance distortion amplification or similar.

              You want to make sure that you cross over low enough so that this peak is off the audible radar, or possibly of greater importance, so that the peak is attenuated and the distortion profile is smooth without any bits that stick out like a sore thumb.

              Now the peak in this case pushes the third harmonic up around 10dB higher than it would otherwise be and is at about 1.9k. A standard LR4 at 1.9k would be 6dB down already at this frequency, so pushing the xover down to around 1.5k would be optimal from that point of view, that's if the tweeter can handle it. If you're after a small faceplate, neo tweeter, that can handle low Xovers, then look no further than the tweeter you chose, this is probably one of the few applications that will actually take advantage of everything that tweeter has to offer.

              Originally posted by JonW
              Rather than picking a specific crossover frequency, I let the models move around to get the overall best matches of FR, phase, impedance, etc. There is an annoying peak at ~900 Hz that made life a little difficult. In the end, both the LR2 and LR4 cross around 2,000 Hz. After some polishing, I’ll post both versions for comparison and see what you think. This discussion will be easier with pictures.
              This is what I tend to do also, usually LR4 designs at around 2kHz fall into place pretty easily with drivers mounted on straight baffles. The slightly slanted baffle geometry you've chosen should assist with the phase alignment of an LR2 though. If you want to experiment with an LR2 design, the ER15 version is the one to do it with, the well behaved cone + the robust tweeter, should give you a reasonable amount of flexibility without you needing to make significant compromises to arrive at it.

              If the 900Hz peak is showing up in both designs then it is probably diffraction related and depending on how bad it is, might require a notch of its own. Speaking of notches though, as Face says, you probably will require a notch on the W16 @ ~6k to make sure the resonance is totally out of the picture.
              What you screamin' for, every five minutes there's a bomb or something. I'm leavin' Bzzzzzzz!
              5th Element, otherwise known as Matt.
              Now with website. www.5een.co.uk Still under construction.

              Comment

              • JonW
                Super Senior Member
                • Jan 2006
                • 1582

                #97
                Originally posted by 5th element
                It's only worth toying with LR2 filters with drivers that will support them. Most do not, especially metal cones. The W15CY, with it's 8.2kHz primary break up, would be just about alright if paired with something like the DXT and wave-guide that could support an LR2 down at 1500Hz or so.

                Stick with a LR4 below 2k with the W16. In fact I would want to optimise this design based off of how low the tweeter could play without distress. Zaph's measurements imply that it could cope with a LR4 at 1500Hz, but might get a bit iffy if run at very high levels. In your application, nearfield, I highly doubt that this would ever be an issue.

                Having a lower xover point will also help keep the integration tight, for the relatively close listening position, based off of the C2C spacing etc.
                Yes, I agree with everything that you are saying here. It all makes good sense. The curves shook out to blend well at an LR4 of right about 2,000 Hz.



                Originally posted by 5th element
                Neither the tweeter or the woofer should be getting or showing noise anywhere during the measurements. If it looks like things are getting noisy (fatter fuzzy trace etc) then the system is at fault. I do recommend you read the guide I wrote and have a bash with ARTA, just to make sure that you're doing everything right. This may seem like a headache, but considering the effort you're putting into this, you've already managed to deal successfully with the impedance headache, it cannot hurt making sure that the FR stuff is as good as it can get too. The ARTA guide, in principle, is directly relate-able to JustMLS too, so you might be able to make improvements to your measurements without even needing to use ARTA.
                By saying that the tweeter gets “noisy,” what I meant was that is where the distortion goes up a lot. The measurements are fine. Sorry for the misunderstanding.

                Originally posted by 5th element
                The tweeter shows low distortion down pretty low and can be used down that low too. The only issue will be how gracefully the tweeter behaves be as you approach and exceed xmax under high drive levels.

                The woofer has a bell like resonance and this propagates into the distortion at integer multiples below the main resonance. In other words you play a 2kHz tone through the woofer, the motor generates a third harmonic at 6kHz and as the cone rings at about 6k, it resonates in unison with the 3rd harmonic, amplifying it and degrading the performance. This is what I believe Jon coined cone resonance distortion amplification or similar.

                You want to make sure that you cross over low enough so that this peak is off the audible radar, or possibly of greater importance, so that the peak is attenuated and the distortion profile is smooth without any bits that stick out like a sore thumb.

                Now the peak in this case pushes the third harmonic up around 10dB higher than it would otherwise be and is at about 1.9k. A standard LR4 at 1.9k would be 6dB down already at this frequency, so pushing the xover down to around 1.5k would be optimal from that point of view, that's if the tweeter can handle it. If you're after a small faceplate, neo tweeter, that can handle low Xovers, then look no further than the tweeter you chose, this is probably one of the few applications that will actually take advantage of everything that tweeter has to offer.
                Right. If you look at Zaph’s measurements on the tweeter (on my actual pair of tweeters), you’ll see that the distortion really goes up as you drop below 2,000 Hz. So a lower crossover point gets you less woofer distortion but more tweeter distortion. I’ll guess that the woofer distortion is more objectionable, though.

                I’ll post my first LR4 attempt in a moment. See what you think. Satisfyingly smooth plots worked out with a 2,000 Hz crossover point. But I see what you are saying about going lower. Makes sense.



                Originally posted by 5th element
                This is what I tend to do also, usually LR4 designs at around 2kHz fall into place pretty easily with drivers mounted on straight baffles. The slightly slanted baffle geometry you've chosen should assist with the phase alignment of an LR2 though. If you want to experiment with an LR2 design, the ER15 version is the one to do it with, the well behaved cone + the robust tweeter, should give you a reasonable amount of flexibility without you needing to make significant compromises to arrive at it.
                Exactly what I was thinking- the ER15. But this project is a learning experience for me. So I wanted to start with the more difficult challenge of trying LR2 on the metal cone. After I get a W16 crossover that I like, adapting things to the ER15 should be relatively easy. Although we’ll see.

                Originally posted by 5th element
                If the 900Hz peak is showing up in both designs then it is probably diffraction related and depending on how bad it is, might require a notch of its own. Speaking of notches though, as Face says, you probably will require a notch on the W16 @ ~6k to make sure the resonance is totally out of the picture.
                Yup, it’s in all the measurements. Diffraction is what I was thinking. You’ll get to see it in a minute.

                Comment

                • JonW
                  Super Senior Member
                  • Jan 2006
                  • 1582

                  #98
                  Here are my first attempts at crossovers. For the in room, deluxe flavor of the speakers. It’s just in the computer. I have not yet listened to them yet. First will be the LR4 then the LR2.

                  First up is the LR4:
                  The frequency response is kind of flat, the phase tracks a bit in the crossover region, the impedance is flat, and there is a null when the tweeter polarity is reversed. It’s close to an acoustic LR4.

                  Frequency response:
                  Click image for larger version

Name:	Slide01_zpsbbbfe52b.webp
Views:	88
Size:	32.9 KB
ID:	931393

                  Phase:
                  Click image for larger version

Name:	Slide02_zpsa7e13f14.webp
Views:	90
Size:	32.8 KB
ID:	931394

                  Impedance:
                  Click image for larger version

Name:	Slide03_zpsf1c6feda.webp
Views:	90
Size:	29.5 KB
ID:	931395

                  Reversed tweeter null:
                  Click image for larger version

Name:	Slide04_zpsaa830b44.webp
Views:	88
Size:	33.4 KB
ID:	931396

                  Circuit:
                  Click image for larger version

Name:	Slide05_zps04e44edf.webp
Views:	91
Size:	16.2 KB
ID:	931397

                  If you are playing along at home, here is how to understand the circuit:

                  -Resistors R2, R4, R6, R8, and R9 are not actually parts. They are there to account for the DC resistance of the inductors. I plan to use 20 and 18 gauge inductors here due to the form factor of the coils being taller and narrower versus, say, 14 gauge coils. There will not be much cabinet space to mount the crossovers, so the narrower coils make sense. Bonus: They are cheaper.

                  -R1 is a tweeter padding resistor (9 Ohms).

                  -R5 is a woofer padding resistor (3 Ohms). I’ve never seen one of these in a crossover before. I realize that there will be an efficiency hit. And it might make sense to skip this part and then drop the tweeter padding resistor from 9 to 6 Ohms. But the FR is a little flatter going this way. I can voice it both ways and see. Maybe the efficiency hit is not a problem for the near field version of the speaker. But maybe it will be a problem for using this as a normal speaker, on stands and away from walls? I don’t know. Watts are cheap these days.

                  -C2, C3, and L2 are a third order electric slope for the tweeter.

                  -L4, L5, and C6 are a third order electric slope for the woofer.

                  -C5 is there to kill the metal cone break up mode.

                  -L1, C1, and R3 make an impedance flattening circuit for the tweeter. It kills a big peak.

                  -L3, C4, and R7 make an impedance flattening circuit for the woofer. It kills a big peak. I wonder if this LCR may be too heavy handed of a thing to do. I’ve never seen anyone do this for a woofer. There must be a reason and I’m curious to learn about it. And the values are kind of crazy high (15 mH inductor and 250 uF cap). Using a Bennic cap, this LCR adds about $108 to the crossover cost. Will it be worth it? My last project had an LCR to kill a tweeter impedance peak and it sounded a little better with versus without the LCR. I will voice it here with and without this LCR and see.

                  I wonder if I have dealt with the metal cone breakup node well enough. Maybe I should consider an LCR filter there instead of the C5 series cap used here.

                  Also, with a crossover like this, there are 5 inductors to be crammed together into a teeny cabinet. Maybe not good.

                  Anyways, I’m not too experienced at this. So any thoughts are appreciated.
                  Last edited by theSven; 02 April 2023, 12:10 Sunday. Reason: Update image location

                  Comment

                  • JonW
                    Super Senior Member
                    • Jan 2006
                    • 1582

                    #99
                    And here is the LR2 version for comparison. I don’t like the looks of it as much but I wanted to try it out and see if it could work, at least on the computer. The tweeter polarity is reversed from normal in this circuit. It doesn't look to me as though the woofer breakup is dealt with well. (Maybe an LCR filter would help.) I do like the phase and the reduced parts count, though.

                    Click image for larger version

Name:	Slide06_zpsbe923050.webp
Views:	80
Size:	30.9 KB
ID:	931398

                    Click image for larger version

Name:	Slide07_zps459edd49.webp
Views:	91
Size:	32.8 KB
ID:	931399

                    Click image for larger version

Name:	Slide08_zpse4ae6ccd.webp
Views:	91
Size:	26.8 KB
ID:	931400

                    Click image for larger version

Name:	Slide09_zpsa6fb8119.webp
Views:	93
Size:	32.1 KB
ID:	931401

                    Click image for larger version

Name:	Slide10-1_zps0acdaded.webp
Views:	90
Size:	16.0 KB
ID:	931402
                    Last edited by theSven; 02 April 2023, 12:12 Sunday. Reason: Update image location

                    Comment

                    • JonMarsh
                      Mad Max Moderator
                      • Aug 2000
                      • 15259

                      Originally posted by JonW
                      And here is the LR2 version for comparison. I don’t like the looks of it as much but I wanted to try it out and see if it could work, at least on the computer. The tweeter polarity is reversed from normal in this circuit. It doesn't look to me as though the woofer breakup is dealt with well. (Maybe an LCR filter would help.) I do like the phase and the reduced parts count, though.









                      If you email me the data files I'll do one of my 3rd order all pass networks for fun! quasi series/parallel or full parallel. These drivers are just itchin for that, I think... :W
                      Last edited by theSven; 02 April 2023, 12:13 Sunday. Reason: Update image location
                      the AudioWorx
                      Natalie P
                      M8ta
                      Modula Neo DCC
                      Modula MT XE
                      Modula Xtreme
                      Isiris
                      Wavecor Ardent

                      SMJ
                      Minerva Monitor
                      Calliope
                      Ardent D

                      In Development...
                      Isiris Mk II updates- in final test stage!
                      Obi-Wan
                      Saint-Saëns Symphonique/AKA SMJ-40
                      Modula PWB
                      Calliope CC Supreme
                      Natalie P Ultra
                      Natalie P Supreme
                      Janus BP1 Sub


                      Resistance is not futile, it is Volts divided by Amperes...
                      Just ask Mr. Ohm....

                      Comment

                      • 5th element
                        Supreme Being Moderator
                        • Sep 2009
                        • 1671

                        It would be instructive for you to post the drivers responses unfiltered and also the transfer functions of the xovers. Something doesn't look quite right with that 800Hz bump, it's probably just diffraction, but something about the overall shape looks a little odd. Even if you were to flatten the bump you'd still have a cliff in the response, which doesn't normally occur with diffraction, at least in my experience.
                        What you screamin' for, every five minutes there's a bomb or something. I'm leavin' Bzzzzzzz!
                        5th Element, otherwise known as Matt.
                        Now with website. www.5een.co.uk Still under construction.

                        Comment

                        • JonMarsh
                          Mad Max Moderator
                          • Aug 2000
                          • 15259

                          Originally posted by 5th element
                          It would be instructive for you to post the drivers responses unfiltered and also the transfer functions of the xovers. Something doesn't look quite right with that 800Hz bump, it's probably just diffraction, but something about the overall shape looks a little odd. Even if you were to flatten the bump you'd still have a cliff in the response, which doesn't normally occur with diffraction, at least in my experience.
                          That bump is pretty typical when there's choke point in the rear wave causing a reflection back through the cone- though it could be something else completely different here. The little NHT cabinets do something similar. Small mid woofers and front panels are always a design problem- used to be I only used aluminum front panel skins with things like Seas 5-1/4" woofers on the front panel just to solve those problems- and that was in the early 80's!
                          the AudioWorx
                          Natalie P
                          M8ta
                          Modula Neo DCC
                          Modula MT XE
                          Modula Xtreme
                          Isiris
                          Wavecor Ardent

                          SMJ
                          Minerva Monitor
                          Calliope
                          Ardent D

                          In Development...
                          Isiris Mk II updates- in final test stage!
                          Obi-Wan
                          Saint-Saëns Symphonique/AKA SMJ-40
                          Modula PWB
                          Calliope CC Supreme
                          Natalie P Ultra
                          Natalie P Supreme
                          Janus BP1 Sub


                          Resistance is not futile, it is Volts divided by Amperes...
                          Just ask Mr. Ohm....

                          Comment

                          • JonW
                            Super Senior Member
                            • Jan 2006
                            • 1582

                            Originally posted by JonMarsh
                            If you email me the data files I'll do one of my 3rd order all pass networks for fun! quasi series/parallel or full parallel. These drivers are just itchin for that, I think... :W
                            Hey Jon,

                            Thanks! I just shot you an email with 4 data files- tweeter and woofer each for FR and impedance. I can send any other files that you want, too. If you get to them, great. If you’re too busy, I totally understand.

                            But there is a catch: You now have to tell me what you have in mind. What’s all this mumbo jumbo about quasi series/parallel and such? And why does it apply well here?

                            Comment

                            • JonW
                              Super Senior Member
                              • Jan 2006
                              • 1582

                              Originally posted by 5th element
                              It would be instructive for you to post the drivers responses unfiltered and also the transfer functions of the xovers. Something doesn't look quite right with that 800Hz bump, it's probably just diffraction, but something about the overall shape looks a little odd. Even if you were to flatten the bump you'd still have a cliff in the response, which doesn't normally occur with diffraction, at least in my experience.
                              The raw drivers are shown below. I’ll have to get back to you on the transfer function, maybe later tonight. I don’t have that with me at the moment.

                              Originally posted by JonMarsh
                              That bump is pretty typical when there's choke point in the rear wave causing a reflection back through the cone- though it could be something else completely different here. The little NHT cabinets do something similar. Small mid woofers and front panels are always a design problem- used to be I only used aluminum front panel skins with things like Seas 5-1/4" woofers on the front panel just to solve those problems- and that was in the early 80's!
                              You mean they had discovered aluminum back in the 1980’s?

                              The ~900 Hz bump shows up in both tweeters and the W16 woofer, but less so (or not at all) in the ER15 woofer. The perfect bump overlap between the tweeter and W16 woofer had me thinking that it’s diffraction. But maybe Jon has a point with reflections.


                              Here is the deluxe W16 and tweeter raw measurements:
                              Click image for larger version

Name:	Deluxe_L_RawDrivers_zps82cd20b5.webp
Views:	87
Size:	14.7 KB
ID:	931403

                              Here is the basic ER15 and tweeter raw measurements:
                              Click image for larger version

Name:	Basic_L_RawDrivers_zpse255478f.webp
Views:	87
Size:	14.9 KB
ID:	931404
                              Last edited by theSven; 02 April 2023, 12:14 Sunday. Reason: Update image location

                              Comment

                              • 5th element
                                Supreme Being Moderator
                                • Sep 2009
                                • 1671

                                Originally posted by JonMarsh
                                That bump is pretty typical when there's choke point in the rear wave causing a reflection back through the cone- though it could be something else completely different here. The little NHT cabinets do something similar. Small mid woofers and front panels are always a design problem- used to be I only used aluminum front panel skins with things like Seas 5-1/4" woofers on the front panel just to solve those problems- and that was in the early 80's!
                                Indeed, I have seen the little NHTs but Jon's cabinets looks about as open as they could be, all things considered and certainly more open than some of my designs that don't show this type of thing (with the W15CY too), but I guess that can be quite design dependent.

                                Jon, your raw measurements are clearly showing signs of system noise and mild gating issues. Whether or not eliminating these would cause some of the irregularities to disappear I do not know, but I would want to take better FR measurements than you have before committing to a design. Now some noise and mild gating issues can be solved with a small amount (1/12th octave at the most) of smoothing, but if done correctly you shouldn't need to use any smoothing whatsoever.

                                I'm not trying to pick holes or have a go at you when I keep mentioning this, I am only saying it because of what I'm seeing and knowing at the same time how easy it is to overcome these issues (usually) with a few small changes in your measurement process. Maybe the end point won't change by much at all and the measurements could be far worse, but they could also be better.

                                LspCADs representation of the files does look better, but this is probably because of the smoothing it's applied. If you take a screen shot from LspCAD without any filter components and set the smoothing to 1/24th, how do the individual responses look then? This should effectively help with the noise, but wont smooth anything else out.
                                What you screamin' for, every five minutes there's a bomb or something. I'm leavin' Bzzzzzzz!
                                5th Element, otherwise known as Matt.
                                Now with website. www.5een.co.uk Still under construction.

                                Comment

                                • JonW
                                  Super Senior Member
                                  • Jan 2006
                                  • 1582

                                  Originally posted by 5th element
                                  Indeed, I have seen the little NHTs but Jon's cabinets looks about as open as they could be, all things considered and certainly more open than some of my designs that don't show this type of thing (with the W15CY too), but I guess that can be quite design dependent.

                                  Jon, your raw measurements are clearly showing signs of system noise and mild gating issues. Whether or not eliminating these would cause some of the irregularities to disappear I do not know, but I would want to take better FR measurements than you have before committing to a design. Now some noise and mild gating issues can be solved with a small amount (1/12th octave at the most) of smoothing, but if done correctly you shouldn't need to use any smoothing whatsoever.
                                  So how does one get less noisy data? My last project was a few years back but those data were much less noisy. Using all the exact same gear. So it’s not an issue with my software or hardware. Maybe I had the sound card or prepro levels too high? I made sure that the sound card was not clipping and that the level meters in JustMLS were in the yellow, as mentioned in the manual.

                                  Originally posted by 5th element
                                  I'm not trying to pick holes or have a go at you when I keep mentioning this, I am only saying it because of what I'm seeing and knowing at the same time how easy it is to overcome these issues (usually) with a few small changes in your measurement process. Maybe the end point won't change by much at all and the measurements could be far worse, but they could also be better.
                                  I see what you are saying. I can retake the measurements if there is good reason to think that the final crossover will change as a result. But if it’s just to make plots look prettier and does not influence how the speaker will sound…

                                  Originally posted by 5th element
                                  LspCADs representation of the files does look better, but this is probably because of the smoothing it's applied. If you take a screen shot from LspCAD without any filter components and set the smoothing to 1/24th, how do the individual responses look then? This should effectively help with the noise, but wont smooth anything else out.
                                  Coming right up, in a moment…

                                  Comment

                                  • JonW
                                    Super Senior Member
                                    • Jan 2006
                                    • 1582

                                    Here are the raw drivers in LspCAD…

                                    No smoothing:
                                    Click image for larger version

Name:	Slide2-6_zps21823b2e.webp
Views:	84
Size:	34.7 KB
ID:	931405

                                    1/24 smoothing:
                                    Click image for larger version

Name:	Slide3-5_zpse3d39ac9.webp
Views:	86
Size:	38.7 KB
ID:	931406

                                    1/12 smoothing:
                                    Click image for larger version

Name:	Slide4-2_zps871c8bd1.webp
Views:	89
Size:	38.2 KB
ID:	931407

                                    And here is that LR4 transfer function:
                                    Click image for larger version

Name:	Slide1-6_zps2f9a6db7.webp
Views:	84
Size:	34.4 KB
ID:	931408
                                    Last edited by theSven; 02 April 2023, 12:16 Sunday. Reason: Update image location

                                    Comment

                                    • 5th element
                                      Supreme Being Moderator
                                      • Sep 2009
                                      • 1671

                                      Originally posted by JonW
                                      So how does one get less noisy data? My last project was a few years back but those data were much less noisy. Using all the exact same gear. So it’s not an issue with my software or hardware. Maybe I had the sound card or prepro levels too high? I made sure that the sound card was not clipping and that the level meters in JustMLS were in the yellow, as mentioned in the manual.
                                      The best way around this is to keep the drive levels high. You might have high signal levels at the inputs to the software you're using, but if this has come about by cranking the analogue gain of the mic pre, then you're going to be raising the noise floor up along with the gain. Turn the gain down and crank the loudspeakers up and you're increasing the signal to noise ratio.

                                      The mic pre is also the part that is the most likely to introduce any erroneous signals that you don't want to be there. This is primarily because it is working at high gain and is therefore more sensitive to noise, another reason to keep the drive levels a few dBs higher and the mic pre, gain down.

                                      Something else to bear in mind is that if you've got the mic at 10cm for a measurement, but then want to take the mic into the far-farfield, say 1 meter away, sometimes the thing you'd do is turn the mic pre up to compensate, the better thing to do is turn the loudspeakers up instead.

                                      The final thing that software can help you with is with averaging. ARTA can do this, but I don't think JustMLS can. What you do is tell it to take say 5 measurements one after the other and take the average . As the signal coming out of the loudspeaker should remain the same each time and because the noise is theoretically entirely random, ARTA latches on to this and can reduce the effects of the noise rather significantly at times.

                                      Originally posted by JonW
                                      I see what you are saying. I can retake the measurements if there is good reason to think that the final crossover will change as a result. But if it’s just to make plots look prettier and does not influence how the speaker will sound…
                                      That's the thing, you just don't know until you take measurements that show you exactly what the driver is doing, rather than them being influenced by some room reflections and the noise.


                                      Looking at the unsmoothed LspCAD plot helps shed a little more light onto the situation, as the other graphs were a little low in resolution. From LspCAD they don't look too noisy, but you can easily see, by the up-down zigzag nature of the plots that you've got some reflections cluttering up the results. As you apply smoothing a lot of these just disappear, but the trouble is, say that peak at 900Hz, the one that's causing some of the issues, it is present in all three drivers. This implies that it isn't driver related but comes from somewhere else. This could be diffraction related, but usually you'd expect this to be perhaps present in the woofer, but not in the tweeter, as the location from the centre of the drivers to the baffles boundaries, is different enough from one driver to the next as to usually make these issues quite different.

                                      The two other hints that this might not be diffraction is that diffraction issues are usually 'rounded' in their appearance, rather than being so spiky like this is. And, that diffraction humps for cabinets of this size, are usually much higher in frequency. I've got a similarly wide cabinet for the W15 and a slightly wider cabinet, of the same overall geometry, for a 6" driver, and the main hump shows up at 1500Hz in one and 1800Hz in the other, the narrower the cabinet the higher it goes.

                                      You do ask a good question though when you ask whether or not it will influence the way the loudspeaker sounds. The short answer is yes, the long answer would be, yes, but does it affect things to a degree which will be audible?

                                      The trouble with the 900Hz spike is that it could very well be influencing and offsetting the amount of baffle step compensation you're dialing in and it could also be influencing how you tailor the low pass on the woofers. This isn't likely to be by any more than 0.5-1dB, clearly audible, but perhaps not strictly that critical. The other issue is with the smoothing correcting for the up-down-up-down nature of what are measurement aberrations, this is an averaging process and the zigzags are around 2dB. You could end up with the averaged SPL being off, only by a small amount, say 0.5-1dB again and probably closer to the 0.5 figure, but this will influence how the filters are designed. Once again this isn't a figure to get your panties in a bunch about and as your transfer functions show, the overall filters are doing what you'd expect them too, without them doing anything unnecessary, for a design such as this. The thing this will potentially affect is the transition band of the xovers and the level of baffle step, not major things, but things that do affect the overall tonal balance of the loudspeaker.

                                      If, like JonMarsh, you end up wanting to design very subtle EQ curves into your loudspeakers, say a 2dB downward tilt from 100-20kHz, to ensure the loudspeakers sound a specific way, if your measurements are out by even a small amount, it can throw things off a little. Again, you can tune these things by ear and Jon ultimately does, but as you can appreciate, having the measurements not contain these issues, even if they are only small, is only going to give you better end results.

                                      You will of course end up with good results with the measurements you've got currently, as the most important aspect, in my opinion, is making sure that you've made the measurements correctly so as to preserve the phase relationship between the two drivers. If you do have the time to perfect and retake your measurements though, I can only recommend that you do, as once you've got the process completely figured out, it will not only benefit this project, but others that are probably yet to come.
                                      What you screamin' for, every five minutes there's a bomb or something. I'm leavin' Bzzzzzzz!
                                      5th Element, otherwise known as Matt.
                                      Now with website. www.5een.co.uk Still under construction.

                                      Comment

                                      • JonW
                                        Super Senior Member
                                        • Jan 2006
                                        • 1582

                                        Thanks again for all the help. I’m learning more on this project than I expected to.

                                        Got it. Higher amp/speaker output and lower mic pre/sound card level. Makes perfect sense.

                                        It’s difficult to compare subjective sound levels of MLS pulses 4 years apart. But I seem to recall that, years ago, the MLS pulses were loud enough that I had to wear ear muffs. This time I did not. So you could be on to something here with signal-to-noise of my measurements.

                                        Thinking about 900 Hz, that corresponds to something like 38 cm, way bigger than any dimension of my cabinets. Plus you make a very good point about it being in both the tweeter and woofer, which should not be the case if it’s diffraction. So maybe it’s not baffle diffraction. Systematic noise or reflections? Would systematic noise like this indicate clipping rather than signal-to-noise that is too low?

                                        I am also wondering about gating time. I think that I used 7 milliseconds which, as I understand it, is typical. And that’s what I used last time to good effect. If it’s not just a signal-to-noise issue as mentioned above, perhaps I should try some shorter gating times. With my on wall measurements, to my surprise, I did see the nulls from the wall reflections (see post #30 on page 1 of this thread). That might indicate that I’m measuring for too long.

                                        Maybe I’ll retake the measurements if I can get a good list of things to do differently ahead of time. I really don’t want to set things up, take measurements, not have anything improve, then have to break everything down again only to need to set it up again, etc. With a little kid on the loose, I just can’t leave any electronics gear out unless I’m right there using it.

                                        Let me ask about the crossover, if I can bother you for some opinions.

                                        -Does the LR2 look generally poor and the LR4 generally OK? (FR, phase, impedance, etc.)

                                        -For the LR4, is the woofer impedance flattening LCR a silly thing to do?

                                        -Likewise, is the 3 Ohm “padding resistor” on the woofer silly? It did help flatten things out a little.

                                        -Does the metal cone resonance at ~5,000 Hz look like it was dealt with sufficiently? There is the steep LR4 slope and the series cap. Or maybe consider an LCR to kill it more?

                                        -Anything else that stands out with regard to designing a crossover?

                                        Comment

                                        • 5th element
                                          Supreme Being Moderator
                                          • Sep 2009
                                          • 1671

                                          Originally posted by JonW
                                          Got it. Higher amp/speaker output and lower mic pre/sound card level. Makes perfect sense.
                                          Not necessarily the sound card, just anywhere there is analogue gain. When you adjust the slider in windows all you're doing 99% of the time is applying digital attenuation. For example if you had the slider at max with a signal level of -1dB and a noise floor of -100dB, you'd have 99dB SnR. If you were to now turn down the slider by 10dB you'd be at -11dB for the signal and -110 for the noise, still at 99dB SnR. Of course you could then say, but I'll just turn up the drive level to get the signal back to -1dB and we'll be much better off, the trouble is that it doesn't work that way. In the first instance with the slider turned up, to reach -1dB you are going to be very close to clipping the input of the ADC with the analogue signal. If you turn down the windows slider you haven't done anything to alter the analogue signal, so if you try and turn it up, all that will happen is the input to the ADC will clip and you wont gain anything at all. All you're doing, when you turn the slider down, is throwing away 10dBs worth of digital signal to noise and squashing the analogue signal into what is effectively a smaller number of bits, you're losing resolution. Now this doesn't matter a jot when you've still got enough bits left to accurately quantise the analogue signal, but really you want to keep the windows slider turned to the max.

                                          Some sound cards, or interfaces, do have some analogue gain, the vast majority however do not.

                                          Originally posted by JonW
                                          Thinking about 900 Hz, that corresponds to something like 38 cm, way bigger than any dimension of my cabinets. Plus you make a very good point about it being in both the tweeter and woofer, which should not be the case if it’s diffraction. So maybe it’s not baffle diffraction. Systematic noise or reflections? Would systematic noise like this indicate clipping rather than signal-to-noise that is too low?
                                          I doubt that it's being caused by noise, that is unless you've got a machine close by that whines at 900Hz. This is far more likely to be a reflection of some sort.

                                          Originally posted by JonW
                                          I am also wondering about gating time. I think that I used 7 milliseconds which, as I understand it, is typical. And that’s what I used last time to good effect. If it’s not just a signal-to-noise issue as mentioned above, perhaps I should try some shorter gating times. With my on wall measurements, to my surprise, I did see the nulls from the wall reflections (see post #30 on page 1 of this thread). That might indicate that I’m measuring for too long.
                                          7ms is a rather large gate for your usual domestic environment. I think I usually use around the 3ms mark when inside.

                                          When you're setting up the room to make measurements what you need to do is place both the mic and loudspeaker as far away from any side walls, the floor andthe ceiling as you can, whilst also keeping about 1 meter between the mic and the speaker. If you don't you'll get reflections arriving too quickly and you'll muck up the measurements.

                                          As an example, imagine that you're doing your measurements in space, obviously there are no nearby reflective surfaces and it's like you're in a well designed anechoic chamber. Now imagine that you're going to place a wall 10cm behind the microphone. What happens here is that the direct sound from the loudspeaker travels to the mic, where it is recorded as the start of the impulse, but what also happens is the sound travels a further 10cm to the wall, reflects off of the wall, travels another 10cm and is recorded by the mic again. This means you've got a path difference of 20cm between the direct, wanted sound, and the reflected, unwanted sound. This is bad as you will need a very short gate length to keep out of the reflection. This will severely limit the low frequency accuracy of the measurements and they will be as good as 'useless'.

                                          Now imagine we've done the opposite and instead of placing a wall behind the mic, we've placed it behind the loudspeaker. What happens now is pretty much exactly the same thing. The direct sound travels directly to the mic, but in this case the sound also diffracts around the loudspeaker cabinet, travels 10cm to the wall behind it, reflects, travels another 10cm back to the front of the loudspeaker and goes its merry way to the microphone. Once again we've got a path difference of 20cm between the direct sound and the reflection, necessitating the use of a very short gate.

                                          There are a couple of things to mention here. The first is with the second example. This is basically what you were doing with your measurements earlier on, placing the loudspeaker against the wall, by using a long gate length and letting in the reflections. In this case this is what you wanted to happen so it could be considered a roaring success. There are times where you actually want to see how the room is influencing the sound, where increasing the gate length is necessary to get what you want.

                                          The second is with the size of the wall. You will have noticed that I mentioned that the sound has to diffract around the loudspeaker enclosure before it can be reflected by the wall. In this case the cabinet forms a physical low pass that actively stops the high frequencies from actually getting to the wall, so they cannot be reflected. The same is true of the wall size, that is, it plays an important role in the frequencies that it will reflect. If the wall happened to be 1 meter square it would be very good at reflecting high frequencies, but it would have very little affect on low frequencies. This is because the low frequency waves have very large wavelengths compared to the walls dimensions and they would see the wall as being insignificant and pass on by. You can therefore see how the size of the reflective surface comes into play, if the object doing the reflecting or diffracting etc is small, it will only have any effect at high frequencies. If you're getting reflections mucking up the results, but they only appear to be affecting a small band at high frequencies, then noticing this can make it easier to trouble shoot what could be causing the problem. Some people have issues with their mic and mic stand and putting layers of felt around them can help to sort these out.


                                          Originally posted by JonW
                                          Let me ask about the crossover, if I can bother you for some opinions.
                                          What you're after here is symmetry in the reverse null. The closer you are to achieving true 2nd or 4th order LW slopes, the more symmetrical the null will be for the given listening axis. This will also help to ensure that the bahaviour off axis vertically will also have smooth characteristics. The C2C spacing and xover frequency used will ultimately determine the size and shape of the primary listening lobe, but the way in which the drivers interact around the crossover point will also play a decent role.

                                          Usually one needs to deviate away from perfect filters because asymmetry is required to give you decent enough phase tracking throughout the cross over region. This is a decent enough approach, but do be wary of deviating too far away from what the slopes should be like. You can sometimes end up creating a situation where you've manipulated the driver to the extent that you will see a very sharp and narrow null with the polarity reversed, but because the acoustic response falls away from ideal very quickly, the phase tracking either side of the null collapses, thus creating the very narrow dip where it should really be much broader.

                                          Your 4th order slopes look absolutely fine and from the point of phase integration so do the second order slopes, the only trouble here is that the peaky nature of the W16 upsets the symmetry of the null with a peak at 900Hz and a dip at 5000Hz. You're getting less than ideal interaction between the drivers here indicating that the 4th order is probably the way to go.

                                          Impedance flattening circuits are not silly at all imo, whether or not they will actually improve the sound is another thing entirely. There are of course two types of flattening circuit. One that adjusts the impedance before the filter and ones that adjust it after. The first are absolutely necessary in some cases to make your filters perform correctly, the second simply make the loudspeaker easier to drive and don't affect the FR. You seem to be using a combination of the two, where there are components before the circuit in some and not in others.

                                          Any series resistance in line with the woofer is a bad idea and should not be needed.

                                          -Does the metal cone resonance at ~5,000 Hz look like it was dealt with sufficiently? There is the steep LR4 slope and the series cap. Or maybe consider an LCR to kill it more?
                                          With the 4th order slopes everything looks dealt with to a decent enough degree.

                                          -Anything else that stands out with regard to designing a crossover?
                                          Only that the series resistance on the woofer is not a good idea and that the impedance flattening circuits are actually adding a lot of components to the design and are probably not going to do much, if anything at all. This is significant when you've got very limited box volume to start with. A circuit to flatten the impedance throughout the crossover region might be worth considering, especially if you've got valves in mind, but the circuit to remove the peak in the bass is a bad idea. 250u caps and 15mH inductors? Those are going to be pretty large and probably expensive.

                                          You could try an L-pad on the tweeter, place the Lpad right at the tweeters terminals, before the rest of the filter components, as this could give you extra shaping options to tame the slightly rising response.
                                          What you screamin' for, every five minutes there's a bomb or something. I'm leavin' Bzzzzzzz!
                                          5th Element, otherwise known as Matt.
                                          Now with website. www.5een.co.uk Still under construction.

                                          Comment

                                          • JonW
                                            Super Senior Member
                                            • Jan 2006
                                            • 1582

                                            Thanks for all the help. Where are you located? I travel a fair amount for work. If I’m ever in the area, I can buy you a beer some time.

                                            I’ll retry the FR measurements. The two things that I’ll focus on are louder speaker output to the microphone and shorter gate times. Hopefully those will help.

                                            I am taking my measurements in a different and smaller room than last time. Life in the house has changed since then. This time, I was no less than 2 meters from a wall and almost that much from the floor and ceiling. The shorter gate time may help here. I’ve always used 1 meter from speaker to microphone. Everything you say about reflections makes perfect sense. It’s a pretty easy concept to grasp.

                                            For my last project, the measurements had much less noise. Below is the final crossover model. That’s without any smoothing. Pretty nice compared to what I’ve got with this current project. So let’s see if I can get something a little easier on the eyes here.


                                            Originally posted by 5th element
                                            Impedance flattening circuits are not silly at all imo, whether or not they will actually improve the sound is another thing entirely. There are of course two types of flattening circuit. One that adjusts the impedance before the filter and ones that adjust it after. The first are absolutely necessary in some cases to make your filters perform correctly, the second simply make the loudspeaker easier to drive and don't affect the FR. You seem to be using a combination of the two, where there are components before the circuit in some and not in others.

                                            Any series resistance in line with the woofer is a bad idea and should not be needed.

                                            (snip)

                                            Only that the series resistance on the woofer is not a good idea and that the impedance flattening circuits are actually adding a lot of components to the design and are probably not going to do much, if anything at all. This is significant when you've got very limited box volume to start with. A circuit to flatten the impedance throughout the crossover region might be worth considering, especially if you've got valves in mind, but the circuit to remove the peak in the bass is a bad idea. 250u caps and 15mH inductors? Those are going to be pretty large and probably expensive.
                                            Hmmm. OK. For my last project, shown below, the L5, C6, R7 filter was before the rest of the circuit and it only flattened a tweeter impedance peak. I listened with versus without this LCR and, to a small degree, it sounded better with it there. So I kept it in the final version. Maybe making the speaker easier to drive can, in some circumstances, have it sound better. I used solid state amps, not tubes.

                                            An easy thing to try here is voicing the speaker with versus without that woofer LCR filter. Just clip it in and out. I must admit that I am hoping not to hear a difference because, like you say, it’s a lot of large, expensive, extra parts to go into a small box.

                                            OK, I did not think that the woofer series resistor would be received well. But, somehow, 9 Ohms on the tweeter and 3 Ohms on the woofer looked a bit flatter than 6 Ohms on the tweeter and nothing on the woofer. I guess two thoughts come to mind. One is voice it both ways and see if there is any audible difference. The other is that maybe fresh measurement data will show things to be flatter, less noisy, and no need for the woofer series resistor.


                                            Originally posted by 5th element
                                            You could try an L-pad on the tweeter, place the Lpad right at the tweeters terminals, before the rest of the filter components, as this could give you extra shaping options to tame the slightly rising response.
                                            When voicing the crossover model, the plan is to change the resistor values and see what sounds best. Rather than use an L-pad, I will just clip in/out resistors of different values. I did that last time and it worked out well.

                                            Originally posted by 5th element
                                            I doubt that it's being caused by noise, that is unless you've got a machine close by that whines at 900Hz.
                                            That might be my little kid.



                                            Click image for larger version

Name:	Slide4-1.webp
Views:	85
Size:	28.6 KB
ID:	931409

                                            Click image for larger version

Name:	Slide1-5.webp
Views:	84
Size:	14.9 KB
ID:	931410
                                            Last edited by theSven; 02 April 2023, 12:17 Sunday. Reason: Update image location

                                            Comment

                                            • ---k---
                                              Ultra Senior Member
                                              • Nov 2005
                                              • 5202

                                              Jon,
                                              You know Herrick Labs has an anechoic room. According to the website 12'x12. Maybe you should load the trunk of your car up with some of that nice beer you keep promising and drive over there and see if you can make some new friends. If you're really lucky, they'll probably have all the gear (and better) hooked up ready to go. You wouldn't need to bring your own.

                                              Purdue's School of Mechanical Engineering is one of the largest in the country, conducting world-class research in manufacturing, propulsion, sustainable energy, nanotechnology, acoustics, materials, biomedicine, combustion, computer simulation, HVAC and smart buildings, human-machine interaction, semiconductors, transportation, thermodynamics, fluid dynamics, solid mechanics, vibration, heat transfer, controls, design, and more.
                                              - Ryan

                                              CJD Ochocinco! ND140/BC25SC06 MTM & TM
                                              CJD Khanspires - A Dayton RS28/RS150/RS225 WMTMW
                                              CJD Khancenter - A Dayton RS28/RS150/RS180 WTMW Center

                                              Comment

                                              • JonW
                                                Super Senior Member
                                                • Jan 2006
                                                • 1582

                                                Originally posted by ---k---
                                                Jon,
                                                You know Herrick Labs has an anechoic room. According to the website 12'x12. Maybe you should load the trunk of your car up with some of that nice beer you keep promising and drive over there and see if you can make some new friends. If you're really lucky, they'll probably have all the gear (and better) hooked up ready to go. You wouldn't need to bring your own.

                                                https://engineering.purdue.edu/Engr/...Anechoic/index
                                                Hey Ryan,

                                                I had no idea! Wow. That's walking distance from here. Looking at the people in that department, I don't recognize a single name. Never even been in that building. I seem to recall hearing that the building is slated to be torn down. Not sure if that is correct or not. I'll ask around.

                                                Comment

                                                • ---k---
                                                  Ultra Senior Member
                                                  • Nov 2005
                                                  • 5202

                                                  I had no idea either. It was just a crazy thought that Google confirmed for me. Engineers are fun people. Shot them an email and see what happens.

                                                  And yes they are building a monsterous new lab. I've no clue what the dates are on everything. I haven't read the Alumni magazine in a while.



                                                  Sent from my DROID2 GLOBAL using Tapatalk 2
                                                  - Ryan

                                                  CJD Ochocinco! ND140/BC25SC06 MTM & TM
                                                  CJD Khanspires - A Dayton RS28/RS150/RS225 WMTMW
                                                  CJD Khancenter - A Dayton RS28/RS150/RS180 WTMW Center

                                                  Comment

                                                  • 5th element
                                                    Supreme Being Moderator
                                                    • Sep 2009
                                                    • 1671

                                                    Originally posted by JonW
                                                    Thanks for all the help. Where are you located? I travel a fair amount for work. If I’m ever in the area, I can buy you a beer some time.
                                                    On the other side of the pond in the UK

                                                    Originally posted by JonW
                                                    I am taking my measurements in a different and smaller room than last time. Life in the house has changed since then. This time, I was no less than 2 meters from a wall and almost that much from the floor and ceiling. The shorter gate time may help here.
                                                    Usually the floor is the closest surface, unless you get the speakers right up off the floor so that the foor and ceiling are an equal distance away. I live in a house with high ceilings too, so that can help if I really need to lift things off the ground.

                                                    Originally posted by JonW
                                                    For my last project, the measurements had much less noise. Below is the final crossover model.
                                                    Those are far better, we're talking wiggles of around +-1dB at worst there. They are still showing some signs of contamination, but it's small enough not to cause much of an issue. The start of the measurement guide I wrote covers how to set the gate and what goes wrong if you don't do it properly, it is applicable as much to JustMLS as it is to ARTA so maybe spend 5-10 mins there just to make sure you know exactly what it is you're looking for.

                                                    Originally posted by JonW
                                                    Hmmm. OK. For my last project, shown below, the L5, C6, R7 filter was before the rest of the circuit and it only flattened a tweeter impedance peak. I listened with versus without this LCR and, to a small degree, it sounded better with it there. So I kept it in the final version. Maybe making the speaker easier to drive can, in some circumstances, have it sound better. I used solid state amps, not tubes.
                                                    Without a doubt, it will make things easier to drive and more consistent. Solid state devices are affected too by the varying impedance and for the same reason as valves, it's just that the issues are a lot smaller with SS. The output impedance forms a potential divider with the loudspeakers impedance, valves typically have highish output impedances, so as the loudspeakers impedance varies you get a change in the voltage across the loudspeakers terminals as frequency changes. This obviously affects the frequency response. Solid state amplifiers have much lower output impedances so this effect is significantly reduced, but it still occurs. Every amplifier appreciates an easier load to drive too.

                                                    Originally posted by JonW
                                                    An easy thing to try here is voicing the speaker with versus without that woofer LCR filter. Just clip it in and out. I must admit that I am hoping not to hear a difference because, like you say, it’s a lot of large, expensive, extra parts to go into a small box.
                                                    As the R of the filter is relatively high, this does allow you to buy a smaller inductor with a relatively high DCR, this should keep the size down, but a 15mH inductor is madness imo With the size constraint in mind I would simply choose to leave it out as you do have other priorities, box volume! If you want to experiment with the LCR, by all means do so. If you prefer things with it, you could always wire it up to the terminals of the amplifier instead, or have it hanging down the back of the loudspeaker cabinet. Unless there's a pressing need to keep everything completely inside the loudspeaker.

                                                    Originally posted by JonW
                                                    OK, I did not think that the woofer series resistor would be received well. But, somehow, 9 Ohms on the tweeter and 3 Ohms on the woofer looked a bit flatter than 6 Ohms on the tweeter and nothing on the woofer. I guess two thoughts come to mind. One is voice it both ways and see if there is any audible difference. The other is that maybe fresh measurement data will show things to be flatter, less noisy, and no need for the woofer series resistor.
                                                    You should be able to get where you want to go without the series resistor and by shaping the response in another way. Also, it might pay to simulate the box and to see how 3 ohms of series resistance affects the alignment. Adding in series resistances directly modifies the electrical damping of the system and affects the overall Qts of the driver and therefore the box alignment. This typically ends up with the requirement for an even bigger box, something that's already probably erring on the small side. This is why using a series resistance is always shunned, not only are you throwing away precious efficiency, but you're going to end up needing a bigger box (usually a bad thing) too.

                                                    Originally posted by JonW
                                                    When voicing the crossover model, the plan is to change the resistor values and see what sounds best. Rather than use an L-pad, I will just clip in/out resistors of different values. I did that last time and it worked out well.
                                                    An L-pad is actually perfectly suited to doing this and more so than a simple series resistance. All you need to typically do is vary the shunt resistance to alter the level. I too have used single series resistances, but the Lpads are a little bit more flexible when it comes to response shaping. That's if it's needed, can't hurt to give it a go.

                                                    Originally posted by JonW
                                                    That might be my little kid.
                                                    Lol, try it yourself, take a measurement when whistling a tone and see how it affects things.
                                                    What you screamin' for, every five minutes there's a bomb or something. I'm leavin' Bzzzzzzz!
                                                    5th Element, otherwise known as Matt.
                                                    Now with website. www.5een.co.uk Still under construction.

                                                    Comment

                                                    • JonW
                                                      Super Senior Member
                                                      • Jan 2006
                                                      • 1582

                                                      Originally posted by ---k---
                                                      I had no idea either. It was just a crazy thought that Google confirmed for me. Engineers are fun people. Shot them an email and see what happens.

                                                      And yes they are building a monsterous new lab. I've no clue what the dates are on everything. I haven't read the Alumni magazine in a while.
                                                      Interesting. Maybe I'll find an excuse to walk over toward that building and poke around. I might get farther by chatting with people I meet. I haven't been there yet, but maybe I can think up an excuse.

                                                      Comment

                                                      • JonW
                                                        Super Senior Member
                                                        • Jan 2006
                                                        • 1582

                                                        I’ll post an image in a few moments. But I gave the measurements another shot, to no avail. This is very frustrating.

                                                        Originally posted by 5th element
                                                        You should be able to get where you want to go without the series resistor and by shaping the response in another way. Also, it might pay to simulate the box and to see how 3 ohms of series resistance affects the alignment. Adding in series resistances directly modifies the electrical damping of the system and affects the overall Qts of the driver and therefore the box alignment. This typically ends up with the requirement for an even bigger box, something that's already probably erring on the small side. This is why using a series resistance is always shunned, not only are you throwing away precious efficiency, but you're going to end up needing a bigger box (usually a bad thing) too.
                                                        Ahh, I didn’t realize that the series resistor will change the Qts, etc. of the driver. But once you say that, it makes perfect sense. Thanks.

                                                        I just removed the resistor from the crossover. A lot of the overall response changes. So I will have to alter various component values. It’s not a problem at all but it will take some mucking around. I’ll make and post the revised circuit and associated plots in the coming days.

                                                        Originally posted by 5th element
                                                        On the other side of the pond in the UK
                                                        Wait. I think that I’ve heard of that place. I was in Newcastle a couple years ago for work. And I did a year of my undergraduate studies in Norwich. Very fond memories.

                                                        Comment

                                                        • JonW
                                                          Super Senior Member
                                                          • Jan 2006
                                                          • 1582

                                                          Another attempt at measurements and nothing but frustration… I tried what was discussed above- raising the db level coming out of the speakers to get higher signal-to-noise. And also a shorter gate time of 3 ms. The results are below- much worse than what I had before.

                                                          It’s a poor mechanic who blames his tools. And I don’t at all claim to be good at this. But I am starting to wonder if some of my measurement woes are stemming from hardware issues. When I had the measurement system set up, I got a few near perfect looking plots on a woofer. As I was tapping “measure” a few times to make sure that everything was looking good before I started saving files, changing drivers, etc. After a few taps, the measurements got ugly, like those shown below. And I did not touch a thing. I did not even step away from the computer. So things were OK then something went bad, never to recover.

                                                          Among other physical things like changing all wiring, etc, I tried restarting the computer. Then the computer did not even recognize the USB dongle for LspCAD for a few tries, but eventually did.

                                                          It’s a 12 year old Dell laptop. Plenty powerful enough to run LspCAD/JustMLS. We don’t use it for anything else, so it’s nice to have a dedicated computer for speakers. I’m wondering if it’s days are over. But I don’t necessarily want to blame the tools for my troubles. Because a more experienced person could probably have these measurements working in a flash.

                                                          Oh well. Up next I guess I will stick with my prior, somewhat noisy measurement files. I’ll take that woofer series resistor out of the LR4 circuit. Then rejigger all the other component values. And I’ll post the results when I get to all of that. That kind of work is much more fun than measurements.

                                                          Image not available
                                                          Last edited by theSven; 25 June 2023, 17:39 Sunday. Reason: Remove broken image link

                                                          Comment

                                                          • 5th element
                                                            Supreme Being Moderator
                                                            • Sep 2009
                                                            • 1671

                                                            I am assuming that you are going with a two channel measurement for the measurements? The type where the mic is input on one channel and the other channel measures the signal before it reaches the driver.

                                                            If you are having issues with the hardware, which seems plausible, considering you had issues with the impedance measurements too, then it might be worth taking things back to their simplest.

                                                            Two channel measurements are technically superior to single channel ones because they remove any issues in the signal generator and the amplifier, but they do introduce another area where things can go wrong. They are also largely unnecessary as most amplifiers and sound cards have ruler flat frequency responses over the bandwidth that we're interested in, especially if the sound card is run at 96 or 192kHz.

                                                            Lets simplify things and perform single channel, near-field, measurements. All this requires is that you plug the mic into one of the input channels of the sound card/sound interface and have the output going through the amplifier and then the loudspeakers. Nothing else, no jig, no extra wires or connectors, just the mic going into the mic pre.

                                                            We are also going to be performing near-field measurements because these will remove, once and for all, any room effects, with or without the gate.

                                                            So, place the mic 1cm from the cone or dome of the driver under test, keep the mic gain down and turn the volume up so that the signal level in windows is decent enough. The only thing that you have to watch out for with very close mic, near-field, measurements like this is that you're not overloading the microphone itself, in this case you don't need the loudspeaker turned up that much for good results.

                                                            If you take a near-field measurement, you still get reflections, but because of the mics close proximity to the driver, the direct sound is so much louder than any reflected sound such that the reflections have very little effect. This is very useful as it means, regardless of how you set the gate, that you're going to be taking reflection free measurements.

                                                            Now the W16 has a very rigid cone and, just like the W15s that I've measured many times, should, ignoring the cone resonance, give you a ruler flat response without any smoothing needing to be applied. The only issue you can sometimes get with near-field measurements is if the driver is giving off any hiss due to noise. In this case, because the microphone is placed literally on top of the cone the noise seems a lot louder and it can make the trace a little 'fuzzy' note that this should not be zigzags. If this happens it isn't the end of the world as we know where it's coming from.

                                                            Now during this time I have been assuming that the mic has only been routed through to either the left or the right channel. If things aren't working out, or even if they are, swap the microphone over to the other channel and see how things look there, it should be identical, or at least very, very similar.
                                                            What you screamin' for, every five minutes there's a bomb or something. I'm leavin' Bzzzzzzz!
                                                            5th Element, otherwise known as Matt.
                                                            Now with website. www.5een.co.uk Still under construction.

                                                            Comment

                                                            • JonW
                                                              Super Senior Member
                                                              • Jan 2006
                                                              • 1582

                                                              Yeah, between the troubles with FR and impedance and my getting OK plots for a few minutes then nothing good, I do wonder if it’s a hardware issue. But let’s not eliminate the potential of operator error.

                                                              Yup, I was doing 2 channel measurements. But I also tried it in single channel mode, too. No difference.

                                                              The simpler setup of only mic in and out to amp is a good idea. No jig. For what it’s worth, I have 2 jigs and I tried both, with no difference.

                                                              My wife has a more recent laptop and I’ll try that out and see if anything improves.

                                                              I also like your idea of switching channels. I should have thought of that one. Trying near field also makes sense.

                                                              OK, I have some more things to fight with over the coming days. I’ll report back what happens.

                                                              Comment

                                                              • JonW
                                                                Super Senior Member
                                                                • Jan 2006
                                                                • 1582

                                                                Well, I burned more time on the measurements and didn’t get too far. Used a new laptop and changed every single cable for which I had a duplicate.

                                                                In JustMLS, with the normal cabling, I could not even get an MLS pulse sound all the time. Sometimes it would make an abbreviated “phst” but most of the time nothing. I switched all the cabling from channel 1 to 2 and it would play the MLS pulses.

                                                                When I tried to calibrate the sound card, sometimes only 1 channel would be seen (blue) and other times two channels would appear. But the blue channel was flattish and the red channel had a huge drop of more than 40 db, with waves in the curve.

                                                                I could not get any measurements that looked real, using channel 1 or 2, with or without the jig. Without the jig, I could not get any sound to play using either channel.

                                                                Next, I tried the program Fuzzmeasure. It doesn’t use jigs so it’s a simpler setup. I could not get my microphone calibration file to be read by the program (I think that I need to reformat it) but that did not change the story.

                                                                Using channel 1 out to the amp, speaker, etc. and channel 1 in with the microphone, I could not get any sound to play from the speaker. I think that channel 1 of my sound card is dead.

                                                                Switching to channel 2 in and channel 2 out, I could get sound, get the Fuzzmeasure level meter to hear something from the microphone, and a small light flash for signal into the sound card. Although never as strong of a light as from Just MLS, even with the sound card knob turned all the way up and the speaker volume pretty loud.

                                                                I got measurements but nothing looked too great. Here is what I got for having the microphone 1 cm from an ER15 woofer:
                                                                Click image for larger version  Name:	WnearPic_zps659d4fc7.webp Views:	54 Size:	30.9 KB ID:	931411



                                                                And here is an overlay of the tweeter and woofer (ER15) far field measurements, with the microphone 1 meter from the speaker:

                                                                Image not available



                                                                Very noisy. Nothing good. So… I don’t know. My guess is that my problems are from a dead channel 1 of the sound card and a channel 2 that is not feeling well. Hmmm… What to do next? At least I have noisy measurements to work with for this project.
                                                                Last edited by theSven; 25 June 2023, 17:43 Sunday. Reason: Remove broken image link

                                                                Comment

                                                                • BobEllis
                                                                  Super Senior Member
                                                                  • Dec 2005
                                                                  • 1609

                                                                  Sound card issues. Miniplugs are failure prone, stresses on the plugs can cause the solder joints to break.

                                                                  You can get a decent USB sound card inexpensively, or the sky is the limit. That solved many of my laptop sound card measurement issues. I'd shoot for at least 24/96 capability. Just avoid the ones with "sonic enhancers" if that "feature" cannot be disabled. (SB comes to mind as a potential problem). I love my M-Audio FW410 but don't know anything about their new USB interfaces.

                                                                  Comment

                                                                  • JonMarsh
                                                                    Mad Max Moderator
                                                                    • Aug 2000
                                                                    • 15259

                                                                    I have two TC Konnekt Firewire interfaces I use with my macbook pro's with Fuzzmeasure; a TC Konnkt 8 (now discontinued), and a TC Konnekt Impact twin (the current model of the 8). The twin will go up to 24/192, but I still use the 8 most of the time, as it's in my kit box for measurements.

                                                                    Click image for larger version

Name:	612Q7qagJNL._SL1074_.jpg
Views:	87
Size:	94.1 KB
ID:	931412

                                                                    A lower cost interface is the TC Desktop Konnekt 6, which is still 24/192.

                                                                    Click image for larger version

Name:	81vFoRdcAJL._SL1500_.jpg
Views:	85
Size:	229.9 KB
ID:	931413

                                                                    What's the length of your measurement gate with Fuzzmeasure- looks like you've got a lot more low frequency data than you need, which means you'll also have more HF data and more reflections that you want. Remember the info I sent you by email?

                                                                    You don't need a 1,000 ms sweep. 200ms is usually fine; 400 if you want lot of LF data. Total window length after you drag the beginning marker and end marker in place should be about 100 msec for a woofer measurement, 10-20 msec for a tweeter measurement. To get the best S/N ratio and less noise, don't use a rectangular window as you have here, but Half Hamming, as I proposed in my email.

                                                                    It's all about the details when it comes to measurement. Normally you should use 1/12 octave smoothing; 1/6 only if you have a really poor situation or you're measuring at 2meters or more. In your screen capture here, it looks like you're not using any smoothing at all, which only compounds the issues from less than optimum window length and gate shape.

                                                                    Note your first measurement looks pretty good, but 1 cm is a bit close for a woofer, to get a realistic near field measurement, because you're quite close to the center of the cone but much farther from the edges. A good distance is between the cone R and D (radius and diameter)- so for that ER15, 3-5" is fairly ideal, won't show room issues, and shouldn't show BSC yet either; that curve should wind up looking a lot like a factor IB measurement. Note that Seas doesn't do IB measurements, though; specifying specific baffle sizes instead, unless Scanspeak, for example.

                                                                    Thinking about why you've got so much noise in your SPL plot, look at your impulse response- you've got tons of reflections going out to 25 msec in big discrete clumps- that creates comb filtering which is generating all the "noise" in your SPL plots. You have to think a bit about the physics of this, and then what's happening becomes much clearer. All the clues are there. The speed of sound is about 1126 feet per second at sea level; you can round that to 1,000 ft/sec or 1 ft per msec to have an idea of where the reflections are being generated (at least in time...)

                                                                    If you have big surfaces nearby, drape them in comforter blankets, for example- this will knock down a lot of the HF stuff.

                                                                    Here are some examples, showing the complete program window.

                                                                    This is a full range measurement at 1-1/2 meters for the Isiris, in room.

                                                                    Click image for larger version

Name:	Full1m5Isiris_zpsf518d8d2.png
Views:	87
Size:	235.5 KB
ID:	931414

                                                                    In this one, I want some reasonable level of low frequency data, but knowing that I've got this bass trap like effect from the hall way at around 38 Hz, there's no point in going for a lot of detail down low, so with a 100 msec overall gate I can get about 10 Hz resolution for data steps, which obviously is PLENTY higher up in the spectrum, and adequate from an LF measurement perspective. Note that inherently these are both related to the window length, in an inverse proportion. Shorter the window, the higher the cut off frequency and data step spacing. On top of that I used a Half Hamming window (see the gray curve over the impulse response, which models the normalized attenuation with time). This lowers the signal level for late reflections, but doesn't help a lot with early stuff from primary ceiling and wall bounce, so I do use 1/6 octave smoothing so I can see the system trends instead of the room. THAT DOESN"T MEAN THE ROOM ISN'T A PROBLEM IN THE REAL WORLD!! The better the room treatments and the less early reflections, the better the system will sound, but I'm more interested in information that correlates well with my near field measurements and determines if the crossover strategy is working as expected.

                                                                    The left window marker must be moved up right to the edge of the impulse so that the measurement result will be essentially minimum phase. Fuzzmeasure has a function for taking a measurement and converting it to a minimum phase copy, but why not get it pretty much right to begin with?



                                                                    This is a more specific measurement; looking at the midrange driver performance with crossover on the baffle, so I want to see the combined effect of baffle step from the cabinet AND the total midrange transfer function including BSC compensation.

                                                                    Click image for larger version

Name:	Mid1mFM_zpsd5031632.png
Views:	85
Size:	252.2 KB
ID:	931415

                                                                    Here, I'm only using a 20 msec window, which gives me about 50 Hz resolution, which is fine for a driver who's range of interest I'm concerned with is 100 Hz and up. Here, the Half Hamming window does provide some meaning full attenuation of the first reflection in the 6-7 msec range from the primary arrival. This helps with the intrinsic "noise" performance of the measurement; doesn't make it anechoic, but definitely cleaner than a rectangular window. And this measurement can be directly compared with the LspCAD calculated output from the midrange driver plus crossover- if everything has been done right, they should agree quite closely.
                                                                    Last edited by theSven; 02 April 2023, 12:18 Sunday. Reason: Update image location
                                                                    the AudioWorx
                                                                    Natalie P
                                                                    M8ta
                                                                    Modula Neo DCC
                                                                    Modula MT XE
                                                                    Modula Xtreme
                                                                    Isiris
                                                                    Wavecor Ardent

                                                                    SMJ
                                                                    Minerva Monitor
                                                                    Calliope
                                                                    Ardent D

                                                                    In Development...
                                                                    Isiris Mk II updates- in final test stage!
                                                                    Obi-Wan
                                                                    Saint-Saëns Symphonique/AKA SMJ-40
                                                                    Modula PWB
                                                                    Calliope CC Supreme
                                                                    Natalie P Ultra
                                                                    Natalie P Supreme
                                                                    Janus BP1 Sub


                                                                    Resistance is not futile, it is Volts divided by Amperes...
                                                                    Just ask Mr. Ohm....

                                                                    Comment

                                                                    • JonW
                                                                      Super Senior Member
                                                                      • Jan 2006
                                                                      • 1582

                                                                      Hi Bob and Jon,

                                                                      Thanks for the help! Wonderful info. I’m not there yet, but maybe getting closer.

                                                                      OK, the first thing that I need to do now is try a new sound card. Thanks for the confirmation there. I didn’t realize the soldering issues but it makes sense.

                                                                      Out of a nice coincidence, it just so happens that I have one of those TC Electronics Konnekt 6’s in the house. I got it to dip my toes into the world of computer audio and external DAC’s. Turns out that, when used for the DAC only, it sounds great. Hi rez music sounded better through the Konnekt 6 than using the DAC in my Oppo blu ray player. I did a quick review here:


                                                                      Then I need to get the measurement parameters worked out. Last night was the first time I ever tried using Fuzzmeasure. I knew that I was measuring for too long- just as Jon is saying here. But I did not get a chance to play with it much. I tried Fuzzmeasure after setting up all the cabling, installing LspCAD and the MAudio drivers on the other PC, fighting with JustMLS for a while, downloading Fuzzmeasure, installing the M Audio drivers on the Mac... By that point, my little kid really needed some attention. So I didn’t get much time to figure out the program well.

                                                                      Yes, I need to measure for shorter times. It makes sense that I need to avoid all the reflections and such. For a 1 meter measurement of a tweeter and then a woofer in Fuzzmeasure, should I just put 10 ms or 20 ms in the “duration” box of the “sweep settings’ icon? Or are there other things that I should change? In JustMLS I think that the equivalent would be to use a 3 ms or 7 ms gate time.

                                                                      Some other questions about Fuzzmeasure:
                                                                      -Will LspCAD read the measurement files from Fuzzmeasure?
                                                                      -Where are the phase data?
                                                                      -Fuzzmeasure could not read my microphone calibration file. My guess is that I just need to edit the file to make it into different columns or something. Anything that I should know here?
                                                                      -Smoothing is really OK to use here? I have not done that with JustMLS and I thought that it was a bad thing.
                                                                      -Any reason to use Fuzzmeasure versus JustMLS? (Other than the slick interface.) It’s prettier but an extra expense and I already have JustMLS. I’m trying Fuzzmeasure in demo mode now.

                                                                      Edit:
                                                                      Looks like I cannot got from the PC USB to the Konnket Firewire. So JustMLS with the Konnekt is out. Next I will have to try Fuzzmeasure on the Mac with the Konnekt. We’ll see. It’s nice to have some hope here. Thanks again for the help.

                                                                      Comment

                                                                      • JonMarsh
                                                                        Mad Max Moderator
                                                                        • Aug 2000
                                                                        • 15259

                                                                        Originally posted by JonW
                                                                        Hi Bob and Jon,

                                                                        Thanks for the help! Wonderful info. I’m not there yet, but maybe getting closer.

                                                                        OK, the first thing that I need to do now is try a new sound card. Thanks for the confirmation there. I didn’t realize the soldering issues but it makes sense.

                                                                        Out of a nice coincidence, it just so happens that I have one of those TC Electronics Konnekt 6’s in the house. I got it to dip my toes into the world of computer audio and external DAC’s. Turns out that, when used for the DAC only, it sounds great. Hi rez music sounded better through the Konnekt 6 than using the DAC in my Oppo blu ray player. I did a quick review here:


                                                                        Then I need to get the measurement parameters worked out. Last night was the first time I ever tried using Fuzzmeasure. I knew that I was measuring for too long- just as Jon is saying here. But I did not get a chance to play with it much. I tried Fuzzmeasure after setting up all the cabling, installing LspCAD and the MAudio drivers on the other PC, fighting with JustMLS for a while, downloading Fuzzmeasure, installing the M Audio drivers on the Mac... By that point, my little kid really needed some attention. So I didn’t get much time to figure out the program well.

                                                                        Yes, I need to measure for shorter times. It makes sense that I need to avoid all the reflections and such. For a 1 meter measurement of a tweeter and then a woofer in Fuzzmeasure, should I just put 10 ms or 20 ms in the “duration” box of the “sweep settings’ icon? Or are there other things that I should change? In JustMLS I think that the equivalent would be to use a 3 ms or 7 ms gate time.

                                                                        Some other questions about Fuzzmeasure:
                                                                        -Will LspCAD read the measurement files from Fuzzmeasure?
                                                                        -Fuzzmeasure could not read my microphone calibration file. My guess is that I just need to edit the file to make it into different columns or something. Anything that I should know here?
                                                                        -Smoothing is really OK to use here? I have not done that with JustMLS and I thought that it was a bad thing.
                                                                        -Any reason to use Fuzzmeasure versus JustMLS? (Other than the slick interface.) It’s prettier but an extra expense and I already have JustMLS. I’m trying Fuzzmeasure in demo mode now.

                                                                        Next I will try JustMLS with the Konnekt 6 sound card. (Although I may not have a USB to Firewire adapter around. The Konnekt 6 is Firewire only and my old PC laptop is USB only. Hmmm. Maybe I can find an adapter locally.) If there is no luck there, I will move on to Fuzzmeasure with the Konnekt 6. Or I may decide to stay wrapped in the warm Mac blanket and just use Fuzzmeasure anyways. We’ll see. It’s nice to have some hope here. Thanks again for the help.
                                                                        I was in the middle of creating my response to you when I decided I needed my Starbucks fix, so the post has been updated since it was first posted. What kind of Mac are you running? Most have Firewire; some, like my Retina MBP don't, in which case you'll need the USB to Firewire adapter- BTW, that's an active adapter with a chip and circuitry inside powered by the USB bus. It's not just a cable.

                                                                        If you noodle around the Fuzzmeasure site, you can find some blogs and info about little details like impedance measurement fixture, probably info about mic cal file format- have you been through the Fuzzmeasure Help or "Getting Started with Fuzzmeasure Pro", or other topics accessed through the help menu? I use an ACCO Pacific instrumentation mic that is dead flat from nearly DC to 40 kHz so I have the luxury of not messing with calibration files. It is most likely just a formatting issue; check the help and load it into Text Edit or TextWrangler (free from BBEdit)

                                                                        Smoothing is OK to use considering what you're primarily doing is removing comb filtering effects from early reflections. Measure the drivers at 6" to establish what the driver itself does, you'll usually see a very smooth curve EXCEPT when you get into areas where the driver has linear distortion- energy storage. Even then, it won't usually generate the effects you're seeing due to comb filtering. At 1 meter, you see the effects of cabinet BSC and diffraction plus the room. See how far you can filter the near field response before you start loosing real detail about the driver- if you have 1/12 octave resolution, that means you can resolve data that varies by the pitch of each step in the western musical scale (12 notes total per octave). That should be more than sufficient for crossover and loudspeaker development.

                                                                        for example, this measurement compares a near field mid measurement and 1 meter measurement; this shows how the baffle step of the cabinet alters the response, and how the midrange crossover BSC compensation fixes that to hit the desired transfer function.

                                                                        Click image for larger version

Name:	MidCombinedNFEvalSS_zpse371d81e.png
Views:	85
Size:	115.0 KB
ID:	931416

                                                                        As I mentioned in my email, sine chirp measurement inherently has a better signal to noise ratio and less stress on the drivers than an MLS test signal. Don't take my word for the benefits of swept sine deconvolution; or even the Fuzzmeasure guy's word; it's why Praxis and Leap both use this technique. But Praxis and Leap are far more expensive... I flirt with the idea of buying Leap now and then, but then I remember that I'd rather buy diamond tweeters than $3K development packages, as I'm not doing this for a living. If I were, I'd have Leap. As it is, I spent about $10K of my own money last year for software that I use at work which my company is too retarded to support or buy. And that's part of how you become a Senior Principal Engineer, I guess... It's your call on the value proposition- Just MLS is free, and get's someone started if you've bought LspCAD, but for free, I'd suggest REW- Room EQ Wizard- more of a learning curve, but looks quite powerful.

                                                                        BTW, I believe a purchased version of Fuzzmeasure unlocks more features and functionality. Have you looked at the loadable modules, or read about them in the documentation? There's a major, major upgrade to Fuzzmeasure in development, but it likely won't be done until near the end of this year. He's hired another guy to work with him, because he's become a victim of his own success (good software and too much business!)

                                                                        Fuzzmeasure generates standard FRD files. The only way you can make problems with an FRD text file is that some analysis programs like LspCAD cannot handle really huge FRD files- i.e., if you setup your measurement with a long window for lots of LF resolution, and a top end cutoff of 40 kHz, you'll develop data input problems in LspCAD when it's loading the file. This is another reason to structure your measurement efforts- there's stuff to do which is just investigating the driver capabilities and driver cabinet interaction, to make decisions about how to approach crossover development, and THEN make an appropriate measurement for the data range and resolution needed for the crossover development.

                                                                        this must sound awful structured and formal; consider that I'm not an edumacated engineer, but just a blues/rock musician from East Texas who worked his way through a psych degree playing in bands; all of this structured approach is just a result of 35+ years of experience doing this kind of stuff.

                                                                        Last, remember this a difference between information and insight. You're doing the measurements to get insights about the drivers capabilities and how you can work with it- refining information and using reasoning creates insight. Part of that refinement process is getting rid of the noise- and that means measurement artifacts as well as other distractions. And considering what you get due to reflections in your test room, this should be food for thought about room treatments and system positioning- both just as important as the basic speaker design.

                                                                        Unfortunately we rarely see DIY projects about room treatments and acoustic development; I'm tempted to setup an area for that and copy some relevant threads in there.
                                                                        Last edited by theSven; 02 April 2023, 12:19 Sunday. Reason: Update image location
                                                                        the AudioWorx
                                                                        Natalie P
                                                                        M8ta
                                                                        Modula Neo DCC
                                                                        Modula MT XE
                                                                        Modula Xtreme
                                                                        Isiris
                                                                        Wavecor Ardent

                                                                        SMJ
                                                                        Minerva Monitor
                                                                        Calliope
                                                                        Ardent D

                                                                        In Development...
                                                                        Isiris Mk II updates- in final test stage!
                                                                        Obi-Wan
                                                                        Saint-Saëns Symphonique/AKA SMJ-40
                                                                        Modula PWB
                                                                        Calliope CC Supreme
                                                                        Natalie P Ultra
                                                                        Natalie P Supreme
                                                                        Janus BP1 Sub


                                                                        Resistance is not futile, it is Volts divided by Amperes...
                                                                        Just ask Mr. Ohm....

                                                                        Comment

                                                                        • JonW
                                                                          Super Senior Member
                                                                          • Jan 2006
                                                                          • 1582

                                                                          Hi Jon,

                                                                          Yes, I just saw your added info, about- thanks! It sheds further light on my murky workings. I’m in the middle of trying get some real work done here, today. My coffee is beside me right now.

                                                                          I’m using a Macbook Pro from 2010. It has a Firewire port so I’m good for getting to the Konnekt 6. Yes, I read the “getting started” and watched the 3 intro videos for Fuzzmeasure. All are helpful but I’ll need more in depth info for what we are doing here. OK, I’ll reformat the microphone calibration file if I get that far. It’s probably not too hard. And good to know that Fuzzmeasure will make FRD files readable by LspCAD. Yeah, I’d think that a sine signal will be better than an MLS “phst” because the sine doesn’t have the driver trying to do everything at once.

                                                                          You spent $10k on software for work? E gad. I don’t want to be a Senior Principle Engineer. But I might spring $150 for Fuzzmeasure if it will solve my problems in the measurement world. That would be a small price to pay to regain my sanity. And no, this does not sound too formal at all. I really, really appreciate the help. I do deal with problems like this at work. But speakers are supposed to be fun, a break from work, ya know?

                                                                          Right, I want to get data at 1 meter distance for the crossover development, with BSC effects, etc. At this point, I’m less concerned with near field info.

                                                                          Where are the phase data that I will need?

                                                                          Comment

                                                                          • JonMarsh
                                                                            Mad Max Moderator
                                                                            • Aug 2000
                                                                            • 15259

                                                                            You change your output graph type by using the Frequency|Display type menu command selecting minimum phase response; then select "Unwrap Phase" if you want it a bit clearer. The Frequency menu item selects parameters regarding output, including switching to distortion measurement.

                                                                            Example here. This is the same measurement as the previous midrange SPL response.

                                                                            Click image for larger version

Name:	FMPhaseRepsonse_zpsa32f745b.png
Views:	85
Size:	217.6 KB
ID:	931417

                                                                            Phase is automatically included in FRD output file.

                                                                            This is exactly what you'd expect to see for a filtered driver output (includes crossover)

                                                                            Also, remember, EVERY SPL measurement in Fuzzmeasure can be converted to a distortion plot of that measurement. Slick, if you ask me...
                                                                            Last edited by theSven; 02 April 2023, 12:19 Sunday. Reason: Update image location
                                                                            the AudioWorx
                                                                            Natalie P
                                                                            M8ta
                                                                            Modula Neo DCC
                                                                            Modula MT XE
                                                                            Modula Xtreme
                                                                            Isiris
                                                                            Wavecor Ardent

                                                                            SMJ
                                                                            Minerva Monitor
                                                                            Calliope
                                                                            Ardent D

                                                                            In Development...
                                                                            Isiris Mk II updates- in final test stage!
                                                                            Obi-Wan
                                                                            Saint-Saëns Symphonique/AKA SMJ-40
                                                                            Modula PWB
                                                                            Calliope CC Supreme
                                                                            Natalie P Ultra
                                                                            Natalie P Supreme
                                                                            Janus BP1 Sub


                                                                            Resistance is not futile, it is Volts divided by Amperes...
                                                                            Just ask Mr. Ohm....

                                                                            Comment

                                                                            • JonW
                                                                              Super Senior Member
                                                                              • Jan 2006
                                                                              • 1582

                                                                              Ahhh, excellent. It's all there. Thanks!

                                                                              I'd rather not burn the time to learn a new software program. It will take some mucking around before I am semi-competent. And I'm comfortable with JustMLS. But the Konnekt is Firewire only and it does not look easy to get Firewire installed on the old Dell laptop. So my hand may be forced here. Fuzzmeasure, in all its slickness, here we come.

                                                                              Comment

                                                                              • ---k---
                                                                                Ultra Senior Member
                                                                                • Nov 2005
                                                                                • 5202

                                                                                This is why I let CJD design my crossovers.

                                                                                Sent from my DROID2 GLOBAL using Tapatalk 2
                                                                                - Ryan

                                                                                CJD Ochocinco! ND140/BC25SC06 MTM & TM
                                                                                CJD Khanspires - A Dayton RS28/RS150/RS225 WMTMW
                                                                                CJD Khancenter - A Dayton RS28/RS150/RS180 WTMW Center

                                                                                Comment

                                                                                • JonMarsh
                                                                                  Mad Max Moderator
                                                                                  • Aug 2000
                                                                                  • 15259

                                                                                  Originally posted by ---k---
                                                                                  This is why I let CJD design my crossovers.

                                                                                  Sent from my DROID2 GLOBAL using Tapatalk 2
                                                                                  We each have our talents... Now if only I could find someone to do my woodworking! :W
                                                                                  the AudioWorx
                                                                                  Natalie P
                                                                                  M8ta
                                                                                  Modula Neo DCC
                                                                                  Modula MT XE
                                                                                  Modula Xtreme
                                                                                  Isiris
                                                                                  Wavecor Ardent

                                                                                  SMJ
                                                                                  Minerva Monitor
                                                                                  Calliope
                                                                                  Ardent D

                                                                                  In Development...
                                                                                  Isiris Mk II updates- in final test stage!
                                                                                  Obi-Wan
                                                                                  Saint-Saëns Symphonique/AKA SMJ-40
                                                                                  Modula PWB
                                                                                  Calliope CC Supreme
                                                                                  Natalie P Ultra
                                                                                  Natalie P Supreme
                                                                                  Janus BP1 Sub


                                                                                  Resistance is not futile, it is Volts divided by Amperes...
                                                                                  Just ask Mr. Ohm....

                                                                                  Comment

                                                                                  • 5th element
                                                                                    Supreme Being Moderator
                                                                                    • Sep 2009
                                                                                    • 1671

                                                                                    It is worth bearing in mind here that Jon's method for acquiring low frequency data is a little off the beaten path vs the traditional way in that this is done, in a standard, non anechoic, environment.

                                                                                    The traditional way to do this is to take a far-field measurement at 1 meter in a room clear from reflective surfaces, that will give you reflection free data down to around 300-500Hz. Usually a gate of around 3ms will do, any more and you start introducing the room, which in my opinion you do not want to do, regardless of what Jon says. If you read the start of my ARTA article it talks about the impulse response and shows you how to gate it properly to ensure that you're keeping the reflections out, but also using the most of what you have.

                                                                                    With loudspeakers of an average width, average diameter drivers and average C2C spacing, which these are, the 1 meter distance + short gate will give you enough low frequency resolution to capture pretty much all the effects from cabinet diffraction and the primary peak/dip combo that occurs before bafflestep rolls in, or off. Usually the peak/dip and step starts to occur at around 1kHz, give or take, so if you've got useful data down to around 500Hz, you're going to have captured accurately a few dBs worth of the battle step transition.

                                                                                    After having done this you then need to take a near-field response and you do this typically at about 1-2cm from the cone. As Jon mentions this is too close to account for any geometrically defined frequency response issues, such as beaming, cone break-up, or the effects of phase plugs, but in most situations this does not matter. For practical purposes you only need accurate data, this time round, up to around 500Hz. As the W16, or pretty much any other driver up to around 10", will be completely pistonic and free from beaming up to this frequency, taking the measurement at 1cm is perfectly adequate.

                                                                                    To put this into perspective if you made a near-field 1cm measurement of a 15" driver, with the mic placed close to the surround, then compared it with the mic placed at the dust cap, beyond a certain frequency you would get quite different results. At higher frequencies, not only does the cone start to beam and narrow its radiation pattern, but the cone itself stops operating as a piston and starts to break up. What this means is that the cone itself starts to flex and you get waves propagating up and down the cone that cause interference patterns within the cone such that the cone itself will emit sound louder from some regions than others. In other words the cone isn't the perfect radiator it was at low frequencies. At low frequencies the cone acted as a single radiating surface, but not any more, break-up has occurred and now it effectively behaves as several individual sources of sound. As Jon mentions this is otherwise known as energy storage because the cone itself is starting to resonate and store the energy that is being delivered to it.

                                                                                    As you'd expect, to accurately capture how these individual sources sum, you need to move the microphone far enough away from the cone that the distance is several times the cones diameter. If the mic is 1cm away from the cone, then if one area of the cone over there is behaving in quite a different way to the area beneath the mic, then it is not going to do a very good job of measuring this.

                                                                                    The whole 1cm near-field measurement is only accurate providing that the cone isn't beaming, or undergoing break-up.

                                                                                    In your case a 1cm measurement would be perfect for a near-field measurement. What the near-field measurement does is accurately capture the way in which the loudspeaker behaves at low frequencies. This does not include the cabinet in any way, so baffle step losses are lost. Do not worry though, this isn't typically a problem.

                                                                                    Remember before I mentioned that the far-field measurement will have captured the start of bafflestep and the diffraction associated with it, so we've already measured most, if not all, of the important diffraction related issues because these typically occur at high frequencies. What's left, is what's left of the bafflestep transition, that is the lower predictable part of it. As we know, this is 6dB total in a absolute sense and depending on the width of the loudspeaker, you may have already captured 2-3dB of this.

                                                                                    When you perform the near-field measurement, you get the program you're using to modify the near-field response to include a simulated, ideal, version of the bafflestep transition that your cabinet should have. This is the predictable part, so you're not actually losing anything by simulating this.

                                                                                    What you then do is merge the far-field and bafflestep modified near-field response, using the far-field for the high frequency data and the near-field for the low frequency data. In other words you're making the best use of both measurement techniques for a completely reflection free overall measurement.

                                                                                    When you are doing this merging you will have to pick the frequency you want to merge at and alter the relative levels of the responses so that they match up well. This is a lot easier than it sounds and JustMLS actually has the most intuitive and easy way of doing this. One thing you absolutely have to remember is that the far-field response is your reference. As you know, the far-field response, in both phase and SPL magnitude, contains all the relevant data pertaining to how the multiple drivers of the system are interacting, if you alter the far-field during the merging process all of this will be lost and all bets are off. So, when you're going about the merging process, you leave the far-field alone and manipulate the near-field response, in terms of SPL and phase, to match that of the far-field and not the other way around.

                                                                                    Once you've got the near-field and far-field responses overlayed, because the far-field response contains the start of bafflestep and the near-field has a simulated version, the two should overlay fairly easily with a reasonable degree of overlap. What this means is that the measured far-field start of bafflestep and the simulated near-field region of bafflestep are agreeing with one another, this is exactly what you want to see. It means you've measured the far-field stuff accurately and it means the simulation on the near-field is working exactly as intended and is a good way of confirming that your measurement process has been a success. Once the two are overlayed and matched up, you finalise the merge and export the data as an frd file.

                                                                                    I will point out here that Jon's method of increasing the gate size to capture the low end and then apply some smoothing, to reduce the room reflections, is a valid way of doing it, but, as he says, is entirely down to how your room behaves. I simply cannot do this in any of the rooms in my house as the response ends up a bloody mess. Now Jon has done enough measurements to know that he can do this in his room and end up with accurate enough data when doing so. Ie we've got the near-field overlayed with the far-field for a start and the two correlate really well, Jon will have also not doubt done measurements like I've suggested in the past and compared them with a large gated far-field. Again we all do this anyway at some point and for me, the room dominates far too much to make doing this viable in my environment and I am always quietly amazed at how room free Jon's far-field, full bandwidth in room measurements are.

                                                                                    All I am trying to say here is watch how you go and learn to walk before trying to run.

                                                                                    I haven't used fuzzmeasure as I don't have a MAC, but from what I've seen it appears to be very capable. Learning how to use the gating functions properly will be critical however

                                                                                    And remember, if you have the gate optimally set then there will be no influence from the room present in the measurements. If you increase the gate slowly beyond this point you will be able to see just how much effect the reflections you're now letting in are having. Increasing the gate will give you more low frequency information, but the trade off is room reflections. If the effect of these reflections is minimal (such as in Jon's case) then you can keep increasing the gate, if however (like me) they destroy the accuracy of the high frequencies, you'll have to keep the gate short.
                                                                                    What you screamin' for, every five minutes there's a bomb or something. I'm leavin' Bzzzzzzz!
                                                                                    5th Element, otherwise known as Matt.
                                                                                    Now with website. www.5een.co.uk Still under construction.

                                                                                    Comment

                                                                                    • JonMarsh
                                                                                      Mad Max Moderator
                                                                                      • Aug 2000
                                                                                      • 15259

                                                                                      Window type and shape is key- note that I pointed out for HF measurements like tweeter and mid I usually use only 10-20 msec. This will still pick up reflections from adjacent boundaries up to 12 feet away, depending on mic position. A key to dealing with that is the special window shapes like Half Hamming, which reduces the level of reflections influence on the upper range while keeping some LF extension in the measurement- Praxis uses this also, as does LEAP. And don't forget that nasty dip I have at 38 Hz in my current living room- but it's no big deal for crossover measurement.

                                                                                      I do see a lot of guys simply use a very short rectangular window (10 msec) with no data below 200 Hz and not splice in the near field LF response. I suppose you can sort of get away with that for a small mini-monitor like Jon is doing here, but then for larger systems baffle step may not be completely resolved until 150 Hz or lower, so that's dangerous...

                                                                                      BTW, I used to use Praxis, which costs several times what Fuzzmeasure does (has a hardware USB pod required for it to work), but I find Fuzzmeasure generally just as powerful and more straightforward to use- someday I'm going to spend sometime with REW, but I never seem to have the time!

                                                                                      And sitting in a corner mostly unused except for emergency impedance measurements is my trusty old CLIO system in a dedicated PC, which also uses MLS as a stimulus signal.

                                                                                      One thing I think we can both agree on, is that good measurements are critical, and to be labored over when necessary, because other wise the design process is GIGO! :W
                                                                                      the AudioWorx
                                                                                      Natalie P
                                                                                      M8ta
                                                                                      Modula Neo DCC
                                                                                      Modula MT XE
                                                                                      Modula Xtreme
                                                                                      Isiris
                                                                                      Wavecor Ardent

                                                                                      SMJ
                                                                                      Minerva Monitor
                                                                                      Calliope
                                                                                      Ardent D

                                                                                      In Development...
                                                                                      Isiris Mk II updates- in final test stage!
                                                                                      Obi-Wan
                                                                                      Saint-Saëns Symphonique/AKA SMJ-40
                                                                                      Modula PWB
                                                                                      Calliope CC Supreme
                                                                                      Natalie P Ultra
                                                                                      Natalie P Supreme
                                                                                      Janus BP1 Sub


                                                                                      Resistance is not futile, it is Volts divided by Amperes...
                                                                                      Just ask Mr. Ohm....

                                                                                      Comment

                                                                                      • 5th element
                                                                                        Supreme Being Moderator
                                                                                        • Sep 2009
                                                                                        • 1671

                                                                                        ARTA uses the 50% Hanning by default, which I'm guessing is what you mean by Half hamming? I must admit to not knowing the mathematics behind Fourier Transform and what the different windows mean, but at some point I might go down that route.

                                                                                        I do find it useful to open up the gate when I want to see trends etc in the response, but when I can get completely reflection free data, accurate down to around 300Hz, I find that plenty in every day stuff when combined with a near-field splice. When I need a little extra and want to actually try and see the full bafflestep transition, I hop into the garden and hoist things off the ground, but this is quite a lot of effort to go too and I only do it when and if I think it will help in any way.

                                                                                        You're mentioning of people designing two ways with only the far-field response down to 200-300Hz. I have done this a number of times and it works very well, but I was hesitant to recommend it as it usually requires a bit of experience, just to make sure that you get the bafflestep levels correct because you have to 'guess' at some of it. I usually work with active filters, both digital and anaogue and with those it's easy to adjust the balance should you get it wrong. Things get a little harder when you're dealing with passive as buying components is a little more permanent than turning a pot or changing a filter coefficient.

                                                                                        Still, for this near-field application, depending on what surface the loudspeakers are placed on, this is rather like floor loading a woofer, which does away with baffle step completely. In this case, the computer desk, or whatever, will have a finite size, so the speakers should still need some sort of compensation to sound correct, it just depends where and by how much they need it. And then there's the question of do you try and correct for it at all? As if you do, it kinda makes the speakers only suitable for sitting on a desk of that size.

                                                                                        Jon has mentioned measuring the loudspeakers against a wall, which implies that they may be used as such, which again does away with the speakers needing much, if any compensation and would perhaps simplify things a tad.
                                                                                        What you screamin' for, every five minutes there's a bomb or something. I'm leavin' Bzzzzzzz!
                                                                                        5th Element, otherwise known as Matt.
                                                                                        Now with website. www.5een.co.uk Still under construction.

                                                                                        Comment

                                                                                        • JonW
                                                                                          Super Senior Member
                                                                                          • Jan 2006
                                                                                          • 1582

                                                                                          Originally posted by ---k---
                                                                                          This is why I let CJD design my crossovers.
                                                                                          No kidding! I must be a control freak or a masochist or something to deal with all of this for my fun time away from work! 8O

                                                                                          Originally posted by JonMarsh
                                                                                          We each have our talents... Now if only I could find someone to do my woodworking! :W
                                                                                          If only I could find a talent... :B

                                                                                          Comment

                                                                                          • JonW
                                                                                            Super Senior Member
                                                                                            • Jan 2006
                                                                                            • 1582

                                                                                            Hi Fellas,

                                                                                            Wow. Great discussion. Even if no one else is reading this, I’m learning a lot. I have not yet had a chance to try any new measurements. Soon, hopefully.

                                                                                            Just to clarify: I’m trying to take measurements and design crossovers for both situations of having the speaker up against a wall and also on a stand, in the middle of a room. (Dealing with the nulls from canceling of output due to bouncing off that wall is a topic for another day.)

                                                                                            I have only had a chance to play around with Fuzzmeasure a little bit. But so far, it seems to be *way* more powerful than JustMLS. In Fuzzmeasure, if I understand this correctly, you take a measurement and then, after the fact, decide which time window from that measurement you want to use. (Not sure if all time lengths are saved to the file or just the window that you select.) Very powerful for all the issues (BSC, room bounce, reflections, etc.) that you have been discussing. With JustMLS, you have one gate time and that’s it.

                                                                                            I was planning on skipping the near field woofer measurements, figuring that I only need the 1 meter data to make the crossover. But the discussion here has convinced me to try and also take the near field data and merge it in. Those data may or may not be helpful for this project. However, it will be good to know how to do all of that. My next project may (or may not) be a 3 way, where the woofer data will be more important to have.

                                                                                            So my understanding here is that I use Fuzzmeasure to take 1 meter measurements of the tweeter and woofer. Then, after the fact, use half Hamming, narrow the time window to, say, 20 milliseconds (or 3 or 7 or 10), then save it, to get the far field data that I need. Take analogous off-axis data. After that, take the woofer near field data and then merge them. Correct?

                                                                                            How does one go about actually merging the near and far field data from Fuzzmeasure? The JustMLS manual explains it there.

                                                                                            Comment

                                                                                            Working...
                                                                                            😀
                                                                                            😂
                                                                                            🥰
                                                                                            😘
                                                                                            🤢
                                                                                            😎
                                                                                            😞
                                                                                            😡
                                                                                            👍
                                                                                            👎
                                                                                            Searching...Please wait.
                                                                                            An unexpected error was returned: 'Your submission could not be processed because you have logged in since the previous page was loaded.

                                                                                            Please push the back button and reload the previous window.'
                                                                                            An unexpected error was returned: 'Your submission could not be processed because the token has expired.

                                                                                            Please push the back button and reload the previous window.'
                                                                                            An internal error has occurred and the module cannot be displayed.
                                                                                            There are no results that meet this criteria.
                                                                                            Search Result for "|||"