The Schlafmutzes Mini MT’s design and progress

[JonW]

Hi Fellas,

Thanks for all the thoughts. One goal of this project is to get some nice speakers. But another is to learn more about how speakers are designed. So discussions like this are great for me. Although I won’t touch the active versus passive debate for now.

Sadly this isn't true, you're getting results that are way off spec, in some cases by more than 100%. You can't use a 4 ohm driver with a crossover optimised for the 8 ohm equivalent and this is how your measurements have really turned out.

Usually with this kind of thing the value of the reference resistor is critical, it will determine the absolute values of your impedance measurements. Ie the trend will be perfect, but the minima/maxima will all be off. Usually this is linear, it will be inaccurate, but it should still be very precise. In other words repeat measurements will always give the same, but wrong, answer. As you're getting wildly different results from one driver to the next, this only leaves a couple of things. The first is that one of the connectors in the measurement chain has a bad contact, this could be as simple as a phono connector needing to be reseated/cleaned, or it could be that a pot somewhere needs a clean.

The impedance measurement process works by measuring the absolute level of the signal directly after the amplifier and compares it with the signal level measured across the loudspeaker once it's been divided by the reference. If there is a faulty connection somewhere and the signal level (somewhere) is being affected because of this, then you will get wrong results.

The second problem could be if there's something groovy going on in the software somewhere, that could be causing this problem.

The best way to make a reference resistor is with several parallel resistors. Multimeters can get inaccurate when measuring low resistances and can be off by a few ohms in some cases. If you can get a bunch of 100 ohm resistors or similar and measure 4 or 8 of them, record the values of the measured resistors and solder them in parallel. The combined resistances will give you close to 25 or 12.5 ohms. Now you don't measure the paralleled resistors, what you do is calculate what the resistance should be from the individual measurements performed on the individual resistors. This should help reduce the effect of multimeter inaccuracies and will thus help to remedy inaccuracies due to the reference resistor.

You mention that you are using justMLS for the impedance measurements. I found this to be a bit of a bother and went straight to LIMP instead. LIMP is a part of ARTA and is designed solely for making impedance measurements. Providing you're not setting the drive level too high by accident, then you only need 1 component, the reference, resistor, to make great measurements. All the extra components do is protect the input to the soundcard in case of an overload, but if you're using an integrated amplifier with this then this isn't much of a worry. The LIMP help file details how to set things up. JustMLS might not be the problem, but it's easy enough to try something else (just to be sure) and LIMP's hardware setup is as simple and easy as it gets, so that might help in finding if you've got problems elsewhere.

From what I understand though JustMLS has a measurement mode called cyclic measurements? This basically repeats one measurement after another, for ever, until you tell it to stop. JustMLS also shows the signal level for each channel in the top left hand corner. If you pop on cyclic measurements then the signal level should remain identical after each cycle. If it does, great, but now you can leave the cycles going and pay attention to the measured signal levels as you agitate any connectors, like twisting phono plugs in their sockets etc. If you do this and the signal level changes anywhere then you've isolated a bad connection.

Another thing to be wary of is noise. If there's any random signal noise then this could modulate the absolute level of the impedance measurements if the signal levels are too low. From what I recall JustMLS doesn't come with any method to take 5 measurements and then average them out. This can be very useful in reducing the effects of noise, LIMP will do this too so there's another reason to perhaps give it a try.

I too hate making impedance measurements. As they are low ohm measurements they are very prone to small problems in the measurement chain, such as a bad switch contact or something. I once had a similar problem to what you've got here, to find that the problem was being caused by a dodgy switch contact. This wouldn't have even been noticed in lots of other applications, but here, where the (non linear in my case) small 1-2 ohm resistance of a slightly tarnished contact wrecked everything. This is one of the reasons I moved over to LIMP, no jig, just 1 resistor and a pair of phono sockets.

Thanks for taking the time to provide such a detailed response. Everything you say makes perfect sense and is good for me to learn about. Let me ask you a seemingly simple question. Admitedly coming from a palce of ignorance. If everything is fine except my plots are running consistently high (in this case) or low, why not just subtract (in this case) or add a few ohms to all data points?

Like you say, it’s precise and inaccurate. But, as I understand it, the inaccuracy is consistent and could be corrected with a consistent correction. Then you’ve brought the plot to exactly what you are shooting for. If the error is systematic and consistent, this idea sounds OK to me. But if the variance in the baseline changed with frequency, this idea would be bad.

As far as I can tell, when comparing my plots to those of Seas and Scanspeak, all looks as expected except for the shift up in the baseline. I’ve got data reflecting these drivers in these cabinets. It seems like a simple way to proceed to me. But I suspect that you will tell me that there is a reason why it’s a bad idea. Thanks again.

[cjd]

Getting it slightly wrong still throws everything off. The crossover network acts directly on the impedance - the frequency response changes based on the resulting transfer function alone.

The suggestion to simply SPL Trace the manufacturer data is likely to be the most straight-forward and accurate alternative, short of determining where you've got data coming in funky. However, I'd take the time to isolate and resolve the problem. That way you KNOW your data is all good and representative of exactly what you have to work with.

If it makes you feel better, I finished up cutting through the two inner layers of baffle today (Nebbiolo) only to realize that I was going to have a hole once I cut the corner chamfers! So now I have to fit and glue in a fix before I do the final glue of the baffles to the box.

C

[JonW]

Getting it slightly wrong still throws everything off. The crossover network acts directly on the impedance - the frequency response changes based on the resulting transfer function alone.

The suggestion to simply SPL Trace the manufacturer data is likely to be the most straight-forward and accurate alternative, short of determining where you've got data coming in funky. However, I'd take the time to isolate and resolve the problem. That way you KNOW your data is all good and representative of exactly what you have to work with.

If it makes you feel better, I finished up cutting through the two inner layers of baffle today (Nebbiolo) only to realize that I was going to have a hole once I cut the corner chamfers! So now I have to fit and glue in a fix before I do the final glue of the baffles to the box.

C

Thanks again, Chris. I’ll be more inclined to retake the data (a third time) if I think that there is a chance that I’ll be able to find the problem. And also if I can understand why offsetting the plots I have now is bad. I’d view making the offsets akin to going the SPL trace route, but better because the plots were taking with my drivers in my cabinets.

Taking measurements is frustrating because there are so many places where things can go wrong. For example, something simple like a bad contact, as mentioned by 5th Element. Probably like most people, I’m now in a domestic situation where each time I want to collect data, I have to set up the measurement gear, take the measurements, and then break it all down to put it back in the closest. I can’t keep it set up. That increases the chances of troubles being intermittent and hard to isolate.

Sorry about your woodworking mishap. I’d like to say that I have no idea what that is like. But, well…

[JonW]

Let me also ask a broader question here: What am I looking to do with the impedance data when designing the crossover?

I know that I want to make sure the impedance never dips too low throughout the frequency range. I don’t want to see any wild swings from high to low or vice versa. On my last project, I had an impedance peak. I smoothed it out with an LCR circuit. Voicing with versus without the LCR filter and I liked it better with, so my impression is that a flattish impedance plot is a good goal. Anything else that I am looking for? Because those things seem to be simple. And if I have this correct (which I may well not), it’s not like you need the most accurate or the most precise data to get at where you want. Just know where your baseline is, where the peaks are, and how much rise there is in the plots. (Like I already have if a consistent baseline offset of a few ohms is applied.)

It’s not like the SPL plots where you have to dig into the details to figure out exactly where one driver carries the load versus another. Plus with SPL you also want to get everything flat. So I’d think that SPL can’t work at all without super accurate and precise date. With phase, the concept is really simple (keep the drivers in phase) and you also need really accurate and precise data or you’ll never get it right. But maybe impedance is more of a rougher “get in the ballpark” kind of thing?

At least that’s how I understand it. Thanks for the help.

[cjd]

Adjusting your current plots individually to manufacturer spec would be akin to tracing, yes.

The only thing you might miss is if you in fact have one driver that really IS out of spec.

As far as what you are looking for - net impedance indicates some level of how difficult it will be to drive - phase angle of impedance also plays a role, as + or - equates to inductive or capacitive (and I have to look up which is which) - and capacitive is what you really need to watch out for since that's the worst for an amp.

If there is an impedance variation through the transfer function region (inductive rise or Fs peak) it will change how the crossover itself works on the response. And post-implementation flattening of impedance (as I offer on the Ansonica) is more to make things nice for the amp, which in turn might lead to an audible difference (tubes especially, as I understand it.)

I'm sure I'm missing stuff, but this is most of what you'd likely need to concern yourself with.

C

[BobEllis]

You might try measuring some crossover components to check your rig. A resistor should be the same regardless of frequency, caps decrease with frequency and inductors increase with frequency. The reactive parts ought to be straight line. If not, keep tweaking your rig

[JonW]

You might try measuring some crossover components to check your rig. A resistor should be the same regardless of frequency, caps decrease with frequency and inductors increase with frequency. The reactive parts ought to be straight line. If not, keep tweaking your rig

That's a clever idea. More clever than I am. I'm at my most humble when subjected to the vagaries of measurements.

[JonW]

Adjusting your current plots individually to manufacturer spec would be akin to tracing, yes.

The only thing you might miss is if you in fact have one driver that really IS out of spec.

I was wondering about that. I put my meter across the binding posts. The two W16 woofers, for example, both read 6.0 Ohms. So I think that the drivers are fine. The software is also likely to be fine, given that I got good data years back, on my last project. It's probably bad wiring or a bad operator. I know where my money is.

As far as what you are looking for - net impedance indicates some level of how difficult it will be to drive - phase angle of impedance also plays a role, as + or - equates to inductive or capacitive (and I have to look up which is which) - and capacitive is what you really need to watch out for since that's the worst for an amp.

If there is an impedance variation through the transfer function region (inductive rise or Fs peak) it will change how the crossover itself works on the response. And post-implementation flattening of impedance (as I offer on the Ansonica) is more to make things nice for the amp, which in turn might lead to an audible difference (tubes especially, as I understand it.)

I'm sure I'm missing stuff, but this is most of what you'd likely need to concern yourself with.

C

Thanks. That all makes sense. It seems that I'm thinking about impedance somewhat properly.

For what it's worth, I found a slightly audible difference when flattening that impedance peak- using a solid state amp.

I owe you a beer for all this help, the next time I'm in Chicago.

[JonW]

You mention that you are using justMLS for the impedance measurements. I found this to be a bit of a bother and went straight to LIMP instead. LIMP is a part of ARTA and is designed solely for making impedance measurements. Providing you're not setting the drive level too high by accident, then you only need 1 component, the reference, resistor, to make great measurements. All the extra components do is protect the input to the soundcard in case of an overload, but if you're using an integrated amplifier with this then this isn't much of a worry. The LIMP help file details how to set things up. JustMLS might not be the problem, but it's easy enough to try something else (just to be sure) and LIMP's hardware setup is as simple and easy as it gets, so that might help in finding if you've got problems elsewhere.

I was just reading about LIMP. Looks interesting and probably worth a try. All the jigs, etc. of JustMLS are quite possibly causing me troubles. The simplicity of just a resistor and no voltage probes is appealing. I think that my sound card (M Audio Mobilepre USB) can output as described in the LIMP manual without the need for an amp. (Figure 3.1 of page 8 in the LIMP manual). If I don't need an amp or the voltage probes, there is a higher chance of getting things to work. Looks like the program my also run on my ancient computer. If it can really be that easy, I may give it a shot.

[cjd]

you can also try ARTA, which will run (but not save data) for free - just to see if it helps you figure out what's what.

[5th element]

Still disagree. If you need it that simple, woofer-tester.

Getting frequency response is as easy as mentioned, but may be unusable for crossover work. Depends on the details, I suppose - in the end, a dialed measurement rig makes either easy. But good, usable impedance data has fewer dependencies on other elements. Frequency data gets time-domain munged way too easily.

C

I guess it depends on the person doing the measurements. I am always set up for doing FR measurements as the mic + pre is always plugged into the sound card. For me, taking useful FR measurements is a piece of cake, it's basically clear the room, point the mic and shoot. Took me a little bit of time to figure out exactly what it was I needed to do, but since that its always been piece of cake.

Impedance measurements on the other hand I don't really need most of the time as I'm active and have been for a while. Sure, it's not hard to whip out the current resistor fixed to the phono sockets, but before that, when I had the jig, things were often complicated for one reason or another, usually some small hardware issue, so maybe this is the fault of the hardware, but that's sort of the point.

The hardware and calibration for making impedance measurements is more complicated than for FR measurements. So if you're going at this for the first time, it's far more likely that you will have the hardware performing properly for the FR measurements and not for the impedance, even if you're gating the FR measurements wrong or taking subsequent measurements in the wrong way to preserve correct phase data, the hardware is doing exactly what it should. Making an impedance measurement is a lot easier in some regards, but trouble shooting a hardware fault can sometimes be a nightmare.

[5th element]

If everything is fine except my plots are running consistently high (in this case) or low, why not just subtract (in this case) or add a few ohms to all data points?

Like you say, it’s precise and inaccurate. But, as I understand it, the inaccuracy is consistent and could be corrected with a consistent correction. Then you’ve brought the plot to exactly what you are shooting for. If the error is systematic and consistent, this idea sounds OK to me. But if the variance in the baseline changed with frequency, this idea would be bad.

As far as I can tell, when comparing my plots to those of Seas and Scanspeak, all looks as expected except for the shift up in the baseline. I’ve got data reflecting these drivers in these cabinets. It seems like a simple way to proceed to me. But I suspect that you will tell me that there is a reason why it’s a bad idea. Thanks again.

There isn't anything wrong with doing that as per say, I have from time to time when I know exactly where the error is coming from, it'd just be nicer to find out exactly where the problem is coming from. If the problem is nonlinear then it could very well affect high frequencies in a slightly different way, or affect areas of greater impedance magnitude than others. Tarnished contacts are not linear with frequency and when you also add in the fact that the resistor + driver is acting as a potential divider, adding in the tarnish to the resistor brings in a resistive element that will very well vary in its properties depending on the voltage present across it, ie as the impedance varies, the voltage dropped across the tarnish changes and so does its resistance. Sure I am probably talking worse case upon worse case scenario, but I think it'd be worth figuring this out now as you will probably want to do this all again in another year or so when you get the itch again.

But maybe impedance is more of a rougher “get in the ballpark” kind of thing?

At least that’s how I understand it. Thanks for the help.

As has been said, no! A crossover network is basically a collection of resistor dividers with the inductors and capacitors varying their 'resistance' with frequency.

Go into LspCAD and do some simulating. You should be able to simulate a passive network into a purely resistive load and you can see what the transfer function is like. So set up a say a 2nd order low pass into the resistor, maybe add in a notch filter too aimed at the frequency you expect the SEAS resonance to be at. Now, alter the value of the resistor acting as the load, keep 'live update' tagged and watch how the frequency response/transfer function changes as the resistor value does.

The impedance data is just as critical to the end performance of your loudspeaker as the frequency response measurements are.

As far as what you are looking for - net impedance indicates some level of how difficult it will be to drive - phase angle of impedance also plays a role, as + or - equates to inductive or capacitive (and I have to look up which is which) - and capacitive is what you really need to watch out for since that's the worst for an amp.

Capacitive loads are the worst only because they have the ability to induce a conditionally stable amplifier into oscillations which can cause it to explode. If the amplifier is properly designed and is unconditionally stable then both inductive and capacitive loads are as bad as each other. They both push the current draw out of phase with the voltage peaks and place additional stresses on the output stage. This causes the distortion the amplifier produces to go up, causes parts of it to get hotter and if the amplifier has inadequate protection, cause it to explode. If the amplifier has adequate protection then it can cause the protection to kick in, as the amp struggles to drive the difficult load and reduces the overall sound quality.

Flattening impedance peaks will directly flatten the phase angles too. If you've got a high impedance, say 16 ohms that happens to combine with a -45degree phase angle then you're doing okay. If you've got part of the spectrum where the start of the peak causes say a 3 ohm dip to match up with a -45 degree angle, then you've got a problem. Flattening the peak may push the impedance down to 2.8 ohms, but if the phase angle is +-5 degrees by doing so then your amplifier will thank for you for it, by performing better and possibly sounding better as a result.

I was just reading about LIMP. Looks interesting and probably worth a try. All the jigs, etc. of JustMLS are quite possibly causing me troubles. The simplicity of just a resistor and no voltage probes is appealing. I think that my sound card (M Audio Mobilepre USB) can output as described in the LIMP manual without the need for an amp. (Figure 3.1 of page 8 in the LIMP manual). If I don't need an amp or the voltage probes, there is a higher chance of getting things to work. Looks like the program my also run on my ancient computer. If it can really be that easy, I may give it a shot.

Whether or not you need an amplifier depends on the sound card, can it drive an 8 ohm loudspeaker? Some can, some can't. It doesn't really matter,if it can't just plug the sound card into an amplifier and drive it with that. You do use voltage probes to a certain extent, it's just given a different name, the left input acts as one probe and the right input acts as the other.

[JonW]

you can also try ARTA, which will run (but not save data) for free - just to see if it helps you figure out what's what.

Right. I may gather up the measurement mess and give things another shot.

The hardware and calibration for making impedance measurements is more complicated than for FR measurements. So if you're going at this for the first time, it's far more likely that you will have the hardware performing properly for the FR measurements and not for the impedance, even if you're gating the FR measurements wrong or taking subsequent measurements in the wrong way to preserve correct phase data, the hardware is doing exactly what it should. Making an impedance measurement is a lot easier in some regards, but trouble shooting a hardware fault can sometimes be a nightmare.

Yes, you are spot on here.

Whether or not you need an amplifier depends on the sound card, can it drive an 8 ohm loudspeaker? Some can, some can't. It doesn't really matter,if it can't just plug the sound card into an amplifier and drive it with that. You do use voltage probes to a certain extent, it's just given a different name, the left input acts as one probe and the right input acts as the other.

Reading about my sound card more, life will not be as simple as it could be. The instructions say that the speaker out can go to powered speakers. Or headphones. No mention of passive speakers, so I think that I’ll need the amp. And the probes. Which adds in more hardware places for things to go awry.

I was hoping to get more time with this project this past weekend but, alas... Maybe soon. I think that I will put the measurement system together one more time. See if I can get JustMLS to behave. If not, I’ll see if I have any better luck with LIMP. If none of that works, you can just look here for my ad, selling all my measurement gear.

[cjd]

I use a simple loopback cable with divider networks and some protection diodes wired in - and all that fits into the housing of the 1/4" or XLR jacks! Basically plug-n-play. I use ARTA. Just have to remember to turn off phantom power when I'm doing impedance measurements.

I think I was getting inverted impedance data when I finally figured out the phantom power thing, but it was very frustrating. Extra because impedance I use the 1/4" on the same input, and XLR for the mic.

C

[---k---]

Jon,
You're more ambitious than I am when it comes to measuring and designing crossovers. If more projects are in your future, I wonder if a PE Omnimic might work well for you.

[Silversmoky]

Hi Jon. Good to see you are back at it. I was just getting into speaker designing and building when you started your Spassvogel project and I gleaned alot of information from your thread as you asked alot of the same questions that I had at the time. It's amazing and humbling indeed at how much all these guys know about this stuff! Thanks guys, from a guy that doesn't post much!

Anyway, I started with similar gear as you and I would definitely suggest using/trying the ARTA programs. I initially started with JustMLS and I know that some guys have good success with it, but compared to ARTA there isn't much comparison IMO. ARTA/LIMP/STEPS are all very nice and much easier to use and they aren't as fussy as JustMLS. I think you will get much more consistent measurements, both frequency response and impedence. I also used to just use alligator clips and such to make my jigs, and always got pretty decent results but finally I took the time to build a dedicated jig box and I wouldn't go back. Similar to 5th Element I can literally be measuring in minutes with very consistent results. Might be worth it to take the time to build one of those little guys too. For me it takes some of the guesswork out of it, wondering if you have a bad connection or something as everything is pretty much set. Hope that helps a little.

[JonW]

I use a simple loopback cable with divider networks and some protection diodes wired in - and all that fits into the housing of the 1/4" or XLR jacks! Basically plug-n-play. I use ARTA. Just have to remember to turn off phantom power when I'm doing impedance measurements.

I think I was getting inverted impedance data when I finally figured out the phantom power thing, but it was very frustrating. Extra because impedance I use the 1/4" on the same input, and XLR for the mic.

C

Yep, I know frustrating. ARTA, LspCAD, other options. I think that there is a steep learning curve with any of these packages. You pick your software, learn it, and stay with it unless there is good reason to switch. I gave LIMP a try and, although the software is a little easier than JustMLS, the hardware part worked out to be very similar. (See my post below, in a few moments.)

If more projects are in your future, I wonder if a PE Omnimic might work well for you.

Hey Ryan,

As far as I can tell, the PE Omnimic is for SPL measurements, not impedance. I think that I have a good handle on SPL measurements (see post 30, on page 1). Impedance, on the other hand... Well, the data I get are what I need to know, if I just shift the baseline a bit.

Jon,
You're more ambitious than I am...

Not really. I will describe my failures in a moment, below.

[JonW]

Hi Jon. Good to see you are back at it. I was just getting into speaker designing and building when you started your Spassvogel project and I gleaned alot of information from your thread as you asked alot of the same questions that I had at the time. It's amazing and humbling indeed at how much all these guys know about this stuff! Thanks guys, from a guy that doesn't post much!

Hi Silversmoky,

Nice to 'see' you again. Yes, the crowd here is super helpful. I’ve learned so much. And getting back to speakers after a long break is fun. “Humbling” doesn’t even begin to describe it. This is all supposed to be fun, a hobby, and a break from work. I could just stay in the office if I wanted this much aggravation.

Anyway, I started with similar gear as you and I would definitely suggest using/trying the ARTA programs. I initially started with JustMLS and I know that some guys have good success with it, but compared to ARTA there isn't much comparison IMO. ARTA/LIMP/STEPS are all very nice and much easier to use and they aren't as fussy as JustMLS. I think you will get much more consistent measurements, both frequency response and impedence. I also used to just use alligator clips and such to make my jigs, and always got pretty decent results but finally I took the time to build a dedicated jig box and I wouldn't go back. Similar to 5th Element I can literally be measuring in minutes with very consistent results. Might be worth it to take the time to build one of those little guys too. For me it takes some of the guesswork out of it, wondering if you have a bad connection or something as everything is pretty much set. Hope that helps a little.

Thanks. Good perspective. I monkeyed around with LIMP last night. I got it working on my first try, kind of. But not really any better than JustMLS. In the end, the cabling was just as complex. And I’m guessing that it’s something in my cabling or connections somewhere. The data look very good, just offset. I think that all of these programs have steep learning curves (for me, at least). So, with the time invested, there has to be a good reason to switch. Maybe you have provided enough of a reason for me to give ARTA a try. But, at this point, the only thing that I don’t like about LspCAD/JustMLS is this impedance baseline offset issue. Everything else is great- taking SPL measurements goes smoothly for me now that I have a system worked out. And I really quite like doing the crossover design work in LspCAD. I can’t imagine any other program being easier. I think that it’s so cool that you can just move around crossover parts or change the value of a given part. Then, instantly, you see how the SPL, phase, impedance, etc. all change. LspCAD is certainly not Apple/Mac slick, but it does everything I want quite well. Idunno. Let me get to the end of this speaker project. Then I’ll have some more experience upon which I can reflect. And then I’ll think about where I want to go next.

[JonW]

Well, I tried JustMLS again and got the same results- good plots that look just like from the manufacturers, but the baselines were too high. I wiggled every connection several times, tried different reference resistors, checked every part of the wiring, changed the levels of the prepro and sound card, used a different amp, reread the manual, and probably some other things, too. I even gave LIMP a try without much luck (also curves that were shifted higher, but much higher than JustMLS). For LIMP, I needed an amp and the voltage probes so, in the end, the wiring is not really any simpler than with JustMLS. The problem here is obviously me. I think that I need a hug.

I’ve got to pick and choose my battles. (That’s true for life, in general, I guess.) I now have good SPL data. The impedance data are of my drivers in my cabinets. With a quick baseline offset, I will have the data that I need. I’ve learned a lot from you folks here, which I do appreciate. At this point, there is no sense making myself crazy (I’m gettin’ close, I tell ya!) to find a problem that is not really stopping the design of the speakers. For future projects, the worst case is that I’m back in this same situation again, with offset but otherwise usable impedance data. And I will be a bit wiser. OK, skip the wiser part.

[lhwidget]

Jon,
Just one last test you may want to try... Pick a resistor and install it in your test set-up just like you've been installing your drivers when you test impedance. If you have a 10 Ω resistor (1W is barely big enough, 5W or 10W would be better), that would be great.

You should get a dead flat impedance plot at 10 Ω. If you don't, and can't find your problem, try one of these two things;

1) Adjust your reference resistor value until the plot moves to 10 Ω, then repeat your impedance measurements using that value for your reference resistor, or
2) Use the measured impedance to determine the value of the offset in your measurement rig, then you will have a decent estimate of the value to subtract from the measured driver curves.

This won't fix any non-linear problems, but it looks like your set-up is working, and the reference value is off for some reason.

Good Luck!

[5th element]

I find ARTA/LIMP/STEPS to be a fantastic package all round. ARTA is far easier, from my point of view, for doing FR measurements, especially when you buy the full version and can save the impulse files. That way you can come back to them at any time and sort out all the gating/post processing later, saving you from having to make sure you get this right when doing the measuring, which can be a real time saver when you don't have much time for taking the measurements themselves.

The fact that you're getting an offset in and of itself isn't that much of a surprise, I mean granted, one does expect you to be able to figure this out and get it sorted, the bit that bothers me though is that you get such different results between drivers of the same type. I don't think I've ever encountered an impedance measuring problem where the results were skewed so much that I got significantly different results when going from say one W16 to another W16. I would expect the measurement system to give me close to identical, but incorrect, results.

My setup though, if I remember correctly, basically takes the output from the sound card and feeds it into the amplifier. The output of the amplifier then connects directly to the right input of the sound card. The output also connects to one side of my reference resistor. The other side of the reference resistor connects to the loudspeaker and to the left intput of the sound card. Finally the other connector of the loudspeaker connects to the signal return (black post) of the amplifier. I don't use anything more or anything less when I'm doing this and I get great results.

If you're getting odd results with a setup like that, then perhaps the left and right inputs are round the wrong way? Another possibility is with the amplifier you are using, is it bridged per chance, what type of amplifier is it? Also, as power amplifiers sometimes come with 'ground lift' switches to help combat ground loops, does yours perhaps have one of these? As standard Laptops and any associated equipment, tend to work entirely floating, that is they don't have any direct connection to the safety earth pin in plugs, if you've got a ground switch lifted in the power amp then you could be getting some interesting currents being induced in the shields and this could be mucking up the ability of the software to determine what the real measured voltages are on its inputs. It's a long shot, but it's difficult to figure out what it could possibly be.

With regards to the Maudio interface being capable of driving loudspeakers directly, this isn't the kind of thing that is usually listed in the specifications. Computer based sound cards tend to be capable of driving small 8 ohm drivers to quiet levels. I mean lots of people have tiny crappy unpowered loudspeakers plugged into their computers and these do fine. Most integrated circuits contain protection so that the outputs don't fry when short circuited or over driven. I would expect the maudio to be protected too. Another thing to bare in mind is that the sound card doesn't actually end up driving the loudspeaker directly, there is the reference resistor in series with it. If the output says that it can drive headphones then it will be fine with the loudspeaker + whatever series resistance you choose to use.

[lhwidget]

Oops, I forgot that the same drivers were yielding different results.

5th element is right, you need to work on your impedance measurement until you are getting consistent results. With either a driver or a test resistor you should be able to take 3-4 measurements that are close to identical.

Now that I think about it, it's a good practice to always start a measurement session with a test on a resistor with a known value to make sure the system is working properly.

[JonMarsh]

A lot of discussion here I've missed out on. Good measurements are not an option, and the variation you see is very diagnostic of issues with accuracy and precision.

I don't know if you can afford to spend a little more on gear, but there's nothing like a dedicated tool designed for the job.

My old CLIO system does woofer measurement accurately and consistently, and so does this:

http://woofertester.com/wt2product.htm

NS12-513-4a



Runs fine in WIN7 and VISTA.


I've measured bazillions of SS tweeters, and their impedance has always been very close to factory spec.

without good measurement data, it's GIGO. Doesn't have to be a $2K instrumentation mic, but must be a solid preamp and basic mic, PREFERABLY with some way to establish either acoustic reference levels at the mic, or measure and verify the drive levels right at the speaker terminals. (I use an HP digital scope to confirm drive levels with sine chirps that Praxis or Fuzzmeasure uses). Good ole 2.83 VRMS. Works out to 4V Peak; really easy to monitor at 2V/div.

Best practices mean making all measurements on all drivers without touching any level controls; then you're guaranteed having data with the correct relative level calibration.

[JonW]

I find ARTA/LIMP/STEPS to be a fantastic package all round. ARTA is far easier, from my point of view, for doing FR measurements, especially when you buy the full version and can save the impulse files. That way you can come back to them at any time and sort out all the gating/post processing later, saving you from having to make sure you get this right when doing the measuring, which can be a real time saver when you don't have much time for taking the measurements themselves.

Thanks again for all the help. Interesting. That sounds like a strength of ARTA. I had trouble getting JustMLS to work well on my first project. But I got it, eventually. And now taking FR measurements seems to go smoothly. In fact, for this project I was surprised at how well it went, given that it had been a few years since I had done it last.

The fact that you're getting an offset in and of itself isn't that much of a surprise, I mean granted, one does expect you to be able to figure this out and get it sorted, the bit that bothers me though is that you get such different results between drivers of the same type. I don't think I've ever encountered an impedance measuring problem where the results were skewed so much that I got significantly different results when going from say one W16 to another W16. I would expect the measurement system to give me close to identical, but incorrect, results.

I’m now guessing that it’s some reference/offset issue compounded by a bad connection somewhere. When I took out the measurement rig for the third time, the measurements looked more consistent, although still with a shifted baseline. I can’t say that with too much confidence, however, because I stopped saving files and trying different drivers when I could not get the baseline correct.

My setup though, if I remember correctly, basically takes the output from the sound card and feeds it into the amplifier. The output of the amplifier then connects directly to the right input of the sound card. The output also connects to one side of my reference resistor. The other side of the reference resistor connects to the loudspeaker and to the left intput of the sound card. Finally the other connector of the loudspeaker connects to the signal return (black post) of the amplifier. I don't use anything more or anything less when I'm doing this and I get great results.

So do you not use the voltage probes? I thought that you then risk damaging the sound card. Whebn using LIMP, by the time that I hooked up the amp, probes, etc. it was about as complex of a system as with JustMLS.

If you're getting odd results with a setup like that, then perhaps the left and right inputs are round the wrong way? Another possibility is with the amplifier you are using, is it bridged per chance, what type of amplifier is it? Also, as power amplifiers sometimes come with 'ground lift' switches to help combat ground loops, does yours perhaps have one of these? As standard Laptops and any associated equipment, tend to work entirely floating, that is they don't have any direct connection to the safety earth pin in plugs, if you've got a ground switch lifted in the power amp then you could be getting some interesting currents being induced in the shields and this could be mucking up the ability of the software to determine what the real measured voltages are on its inputs. It's a long shot, but it's difficult to figure out what it could possibly be.

Hmmm... Interesting. I’m using a Yamaha reciver (RX-V659) as a prepro/volume knob and an Adcom GFA-555 amp. The amp can be bridged but I did not have it in bridged mode. I also treid skipping the Adcom and just using the amp inside the receiver. Same results. Neither the Yamaha nor the Adcom have a third grounding plug on their electric cords. No ground lift switches that I know about.

With regards to the Maudio interface being capable of driving loudspeakers directly, this isn't the kind of thing that is usually listed in the specifications. Computer based sound cards tend to be capable of driving small 8 ohm drivers to quiet levels. I mean lots of people have tiny crappy unpowered loudspeakers plugged into their computers and these do fine. Most integrated circuits contain protection so that the outputs don't fry when short circuited or over driven. I would expect the maudio to be protected too. Another thing to bare in mind is that the sound card doesn't actually end up driving the loudspeaker directly, there is the reference resistor in series with it. If the output says that it can drive headphones then it will be fine with the loudspeaker + whatever series resistance you choose to use.

Good to know. I tried LIMP with the sound card only, no amp, and it didn’t work. Then I realized that I had not connected all the negative wires as needed. By that point, I already had the receiver, amp, and probes hooked up. Maybe it would work with only the sound card and a resistor. Hmmm...

[JonW]

Jon,
Just one last test you may want to try... Pick a resistor and install it in your test set-up just like you've been installing your drivers when you test impedance. If you have a 10 Ω resistor (1W is barely big enough, 5W or 10W would be better), that would be great.

You should get a dead flat impedance plot at 10 Ω. If you don't, and can't find your problem, try one of these two things;

1) Adjust your reference resistor value until the plot moves to 10 Ω, then repeat your impedance measurements using that value for your reference resistor, or
2) Use the measured impedance to determine the value of the offset in your measurement rig, then you will have a decent estimate of the value to subtract from the measured driver curves.

This won't fix any non-linear problems, but it looks like your set-up is working, and the reference value is off for some reason.

Good Luck!

Thanks- those are good ideas! Simple, too. I’ll swap the driver for a resistor the next time I have JustMLS running. Right, that will only work for linear problems, which is what I think that I have. If I compare my plots to those from the manufacturers, everything looks the same with regard to peak magnitude and such. Just the baseline is shifted.

For what its worth, in JustMLS, if I type in a different reference resistor value than what I actually have wired up, the baseline will shift. Not surprisingly. So I can shift things to where they should be that way. But it doesn’t locate the problem or get me any closer than just shifting my baseline after the fact.

Oops, I forgot that the same drivers were yielding different results.

5th element is right, you need to work on your impedance measurement until you are getting consistent results. With either a driver or a test resistor you should be able to take 3-4 measurements that are close to identical.

Now that I think about it, it's a good practice to always start a measurement session with a test on a resistor with a known value to make sure the system is working properly.

I’m pretty sure that the different drivers yielding different results is a related problem to my wiring or incompetence. If I measure the driver resistance with a meter, for example, both W16 woofers yield the same 6.0 Ohms.

[JonW]

A lot of discussion here I've missed out on. Good measurements are not an option, and the variation you see is very diagnostic of issues with accuracy and precision.

I don't know if you can afford to spend a little more on gear, but there's nothing like a dedicated tool designed for the job.

My old CLIO system does woofer measurement accurately and consistently, and so does this:

http://woofertester.com/wt2product.htm

Runs fine in WIN7 and VISTA.


I've measured bazillions of SS tweeters, and their impedance has always been very close to factory spec.

without good measurement data, it's GIGO. Doesn't have to be a $2K instrumentation mic, but must be a solid preamp and basic mic, PREFERABLY with some way to establish either acoustic reference levels at the mic, or measure and verify the drive levels right at the speaker terminals. (I use an HP digital scope to confirm drive levels with sine chirps that Praxis or Fuzzmeasure uses). Good ole 2.83 VRMS. Works out to 4V Peak; really easy to monitor at 2V/div.

Best practices mean making all measurements on all drivers without touching any level controls; then you're guaranteed having data with the correct relative level calibration.

Yes, agreed. Equipment dedicated to the job is a good idea. My SPL measurements are all fine. It’s only impedance that is the thorn in my side. The drivers all measure the same resistance with a meter so I’m pretty sure that the factory specs are fine and it’s a problem with me. I can’t blame the hardware- as much as I’d like to. It’s the (very) soft-ware between my ears that is causing the troubles.

Ryan (-k-) offered to mail me his unused Woofer Tester. What a swell guy. So that’s on the way. After messing around with JustMLS so much, Woofer Tester looks too easy to be true. If it works as advertised, it’s a worthwhile investment. Then I’d use JustMLS for SPL measurements and Woofer Tester for impedance. It sounds excellent. We’ll see what happens...

[5th element]

Thanks again for all the help. Interesting. That sounds like a strength of ARTA. I had trouble getting JustMLS to work well on my first project. But I got it, eventually. And now taking FR measurements seems to go smoothly. In fact, for this project I was surprised at how well it went, given that it had been a few years since I had done it last.

The one thing that JustMLS is well set up for is making FR measurements. Beyond the things you need for basic loudspeaker stuff though it starts to show its limitations. The version I've got is pretty old and I don't know if the ability to save impulse files has been added as a recent feature, if it has this then it'd be almost perfect for making measurements for 99% of the stuff you'd want for most loudspeaker design.

ARTA on the other hand is a different kettle of fish entirely, I used it extensively with only the free version and didn't realise how much I'd actually love the ability to save the impulse files. The same goes for saving data in STEPS and LIMP in its native format. This is obviously different to simply exporting data, as that data is largely processed, whereas saving the raw data gives you many more options later on. Kind of like being given a jpeg of a picture vs being given a photoshop file that contains all the various layers and image info, giving you the ability to manipulate the image right at the basic level.

One of the really nice things that the paid version adds is the ability to create polar maps and sonograms, the latter are especially useful for displaying how a loudspeaker behaves off axis and is more useful if you've got an interest in wave-guides as it helps to show where you're best suited to place the xover for a smooth directivity match.

Having used JustMLS as my main measurement program, I knew what I was doing there and when I first tried ARTA I found it very intuitive. The UI is logically designed so that you can pretty much find the functions you're after without even needing to look at any help files. Of course if you're not quite up to speed on the specific nomenclature associated with this stuff, then you might struggle a little, even if you know what it is you're doing, if you don't know a functions specific name you'll get stuck sometimes.

I did write a guide for ARTA that will walk you through the basics with regards to making FR measurements.

http://www.mediafire.com/file/7dsikmz44lnf7br/measurement_guide.pdf

That should hopefully give you a decent idea of what it is you need to do with the program to get where you want to go.

I’m now guessing that it’s some reference/offset issue compounded by a bad connection somewhere. When I took out the measurement rig for the third time, the measurements looked more consistent, although still with a shifted baseline. I can’t say that with too much confidence, however, because I stopped saving files and trying different drivers when I could not get the baseline correct.

If things are starting to look a little more consistent with reuse then this does hint at the possibility of a bad connection as physically manipulating things often helps with poor contacts.

So do you not use the voltage probes? I thought that you then risk damaging the sound card. Whebn using LIMP, by the time that I hooked up the amp, probes, etc. it was about as complex of a system as with JustMLS.

No I don't. You do risk damaging the sound card, but you do have to get pretty careless to go that far. Most decent sound cards have opamps at their inputs and most opamps can accept +-15 volts on their inputs without this causing any damage. 1 watt into an 8 ohm load = 8 volts peak to peak, or +-4 volts. Clearly you're not going to damage stuff under normal controlled testing, the only thing that would cause damage would perhaps be turn on thumps etc, or any DC offset associated with half bridges in class D amplifiers (yes that's what got me!). So yeah I blew one input on my sound card, but a replacement opamp cost less than $1 and about 10 minutes of my time.

From the input level stand point though most line inputs clip when they have around 5.6v peak to peak on them. This is plenty for impedance measurements and providing you keep the volume control on the amplifier turned down, you shouldn't run into problems. I mean before I got careless and fried the input with the class D amp, I had hooked up all sorts of prototype amplifiers up to the sound card without any protection and I'd never done any damage. I'd measured all sorts of class D amps too without trouble, it's when I was in a rush and not thinking, that I didn't remember the DC blocking cap and poof, dead input. If you're paying attention then you shouldn't fry the inputs. If you get sloppy then anything can happen

I can appreciate how you wouldn't want to risk it though, I only omitted the protection because I knew how to fix it if it ever went wrong. It might be worth trying it without the protection, just once, to see if that is causing the problems, at least then you will know where to trouble shoot.

Hmmm... Interesting. I’m using a Yamaha reciver (RX-V659) as a prepro/volume knob and an Adcom GFA-555 amp. The amp can be bridged but I did not have it in bridged mode. I also treid skipping the Adcom and just using the amp inside the receiver. Same results. Neither the Yamaha nor the Adcom have a third grounding plug on their electric cords. No ground lift switches that I know about.

So the entire system is floating/double insulated then. You wouldn't need a ground lift if there's no connection to safety earth.

I mean one input channel of the sound card measures the voltage on the output of the amplifier and one measures the voltage at the division between the reference and the loudspeaker. This compares the output voltage of the power amplifier, which remains constant with frequency, vs the voltage present at the division, which varies with frequency as the loudspeakers impedance varies. If the measuring software knows the reference resistors value and the two measured voltages on the left and right input, then it can calculate what the loudspeakers impedance is.

What happens to be driving the loudspeaker/test set up, isn't important in the least providing that it can drive the test set up with a decent amount of oopmh. What you're really after is for the voltage measured, by the channel connected to the amplifiers output, to be just below the level required to clip the input. In other words monitor the input levels in whatever software you are using, increase the voltage on the amplifier and stop just before the input clips. This will maximise the signal to noise ratio and hopefully help to get you better results.

ARTA/LIMP/STEPS all have a little meter in the bottom right hand corner that basically reads -XdB L/ -XdBR or something. These are referenced to 0dB which represents the full digital signal level, so anything less than this gets a negative going number, ie how many dB you are away from clipping the inputs. If you pay careful attention to this number, when setting the volume levels, then you should never overload the inputs. If you're using LIMP one channel will read higher (closer to zero) than the other. I would adjust the volume till that one reads around -5dB. Most ADCs perform slightly better a few dBs away from their maximum input level.

Good to know. I tried LIMP with the sound card only, no amp, and it didn’t work. Then I realized that I had not connected all the negative wires as needed. By that point, I already had the receiver, amp, and probes hooked up. Maybe it would work with only the sound card and a resistor. Hmmm...

This is what I would try next. There should be virtually no chance of damaging anything if you're using the sound cards output. It should be inherently protected from over current/short circuits and the headphone output especially, should be absolutely fine with 32ohm loads. Some headphones even dip lower than this, so providing you're using a decent sized reference resistor (25 ohms or so), then you should't have any problems whatsoever. Do bare in mind here though, that when using the output of the sound card, it might not be able to deliver enough current to give -5dB on the input when driving the loudspeaker + reference. If you're using close to 100 ohms though, as LIMP recommends, then you shouldn't have a problem with that.

This is about as simple as it gets and if you still can't get good impedance data like this, then I'd be scratching my head too. For what its worth, I've had excellent results from using the sound cards output at one point in time, so I know that this can work very well.

[cjd]

I should point out that I do not use an amp for impedance measurements, nor do I use voltage probes. I used the wiring illustrated in the ARTA measuring box from their site on page 3, etc.

[Silversmoky]

I could be totally wrong, but doesn't that little blip in Jon's impedence measurements around 60 hz point to bad connection or the like? Whenever I have experienced an impedance measurement like that I have always traced it back to a bad connection, not totally faulty but loose enough to cause errors. Check the connection right at your soundcard. Try securing it as you are taking your measurement and see if anything changes or even sounds slightly different, maybe a little static sounding.

[cjd]

60hz suggests mains leakage somewhere but possibly not.

[5th element]

It also suggests the presence of a ground loop, but as his setup is apparently all double insulated this isn't going to be possible, at least not in the traditional sense. As the laptop will use switching supplies and can be run entirely from its own internal battery, it's most likely that this is coming from the amplifier as some form of induced cross-talk from the main transformer. This isn't uncommon and will be improved by keeping the SnR ratio of the measurements high. Of course if the problem exists before the volume control in the amp, then the 60Hz noise will increase at the same rate as the signal.

[JonW]

The one thing that JustMLS is well set up for is making FR measurements. Beyond the things you need for basic loudspeaker stuff though it starts to show its limitations. The version I've got is pretty old and I don't know if the ability to save impulse files has been added as a recent feature, if it has this then it'd be almost perfect for making measurements for 99% of the stuff you'd want for most loudspeaker design.

ARTA on the other hand is a different kettle of fish entirely, I used it extensively with only the free version and didn't realise how much I'd actually love the ability to save the impulse files. The same goes for saving data in STEPS and LIMP in its native format. This is obviously different to simply exporting data, as that data is largely processed, whereas saving the raw data gives you many more options later on. Kind of like being given a jpeg of a picture vs being given a photoshop file that contains all the various layers and image info, giving you the ability to manipulate the image right at the basic level.

Right. I follow you. My needs are pretty simple with regard to measurements so JustMLS has been fine. I’m not sure if you can save the impulse files. You can see the impulse blip but I’ve never tried to save it. I would guess that you might be able to export it to a text (x,y) file. But I can see why you are saying that’s useful, although I’ve not felt the need to do it myself.

One of the really nice things that the paid version adds is the ability to create polar maps and sonograms, the latter are especially useful for displaying how a loudspeaker behaves off axis and is more useful if you've got an interest in wave-guides as it helps to show where you're best suited to place the xover for a smooth directivity match.

It’s funny that you mention the polar maps and sonograms. As far as I can tell, my non-professional version of LspCAD does not have those features. And I have wanted to have them. But more to see how crossover changes influence the plots- out of curiosity and wanting to learn more. The absence has not been a major detraction from my projects, though.

Having used JustMLS as my main measurement program, I knew what I was doing there and when I first tried ARTA I found it very intuitive. The UI is logically designed so that you can pretty much find the functions you're after without even needing to look at any help files. Of course if you're not quite up to speed on the specific nomenclature associated with this stuff, then you might struggle a little, even if you know what it is you're doing, if you don't know a functions specific name you'll get stuck sometimes.

I did write a guide for ARTA that will walk you through the basics with regards to making FR measurements.

http://www.mediafire.com/file/7dsikmz44lnf7br/measurement_guide.pdf

That should hopefully give you a decent idea of what it is you need to do with the program to get where you want to go.

Got it. It sounds like for someone starting from scratch, ARTA might be a very good program to start with. LspCAD/JustMLS has been OK for me. Steep learning curve and this annoying impedance issue. Otherwise, I do like it.

If things are starting to look a little more consistent with reuse then this does hint at the possibility of a bad connection as physically manipulating things often helps with poor contacts.

Makes sense. I’m not sure if it’s the case, but maybe.

No I don't. You do risk damaging the sound card, but you do have to get pretty careless to go that far. Most decent sound cards have opamps at their inputs and most opamps can accept +-15 volts on their inputs without this causing any damage. 1 watt into an 8 ohm load = 8 volts peak to peak, or +-4 volts. Clearly you're not going to damage stuff under normal controlled testing, the only thing that would cause damage would perhaps be turn on thumps etc, or any DC offset associated with half bridges in class D amplifiers (yes that's what got me!). So yeah I blew one input on my sound card, but a replacement opamp cost less than $1 and about 10 minutes of my time.

From the input level stand point though most line inputs clip when they have around 5.6v peak to peak on them. This is plenty for impedance measurements and providing you keep the volume control on the amplifier turned down, you shouldn't run into problems. I mean before I got careless and fried the input with the class D amp, I had hooked up all sorts of prototype amplifiers up to the sound card without any protection and I'd never done any damage. I'd measured all sorts of class D amps too without trouble, it's when I was in a rush and not thinking, that I didn't remember the DC blocking cap and poof, dead input. If you're paying attention then you shouldn't fry the inputs. If you get sloppy then anything can happen

I can appreciate how you wouldn't want to risk it though, I only omitted the protection because I knew how to fix it if it ever went wrong. It might be worth trying it without the protection, just once, to see if that is causing the problems, at least then you will know where to trouble shoot.

Interesting. Thanks.

So the entire system is floating/double insulated then. You wouldn't need a ground lift if there's no connection to safety earth.

I mean one input channel of the sound card measures the voltage on the output of the amplifier and one measures the voltage at the division between the reference and the loudspeaker. This compares the output voltage of the power amplifier, which remains constant with frequency, vs the voltage present at the division, which varies with frequency as the loudspeakers impedance varies. If the measuring software knows the reference resistors value and the two measured voltages on the left and right input, then it can calculate what the loudspeakers impedance is.

What happens to be driving the loudspeaker/test set up, isn't important in the least providing that it can drive the test set up with a decent amount of oopmh. What you're really after is for the voltage measured, by the channel connected to the amplifiers output, to be just below the level required to clip the input. In other words monitor the input levels in whatever software you are using, increase the voltage on the amplifier and stop just before the input clips. This will maximise the signal to noise ratio and hopefully help to get you better results.

ARTA/LIMP/STEPS all have a little meter in the bottom right hand corner that basically reads -XdB L/ -XdBR or something. These are referenced to 0dB which represents the full digital signal level, so anything less than this gets a negative going number, ie how many dB you are away from clipping the inputs. If you pay careful attention to this number, when setting the volume levels, then you should never overload the inputs. If you're using LIMP one channel will read higher (closer to zero) than the other. I would adjust the volume till that one reads around -5dB. Most ADCs perform slightly better a few dBs away from their maximum input level.



This is what I would try next. There should be virtually no chance of damaging anything if you're using the sound cards output. It should be inherently protected from over current/short circuits and the headphone output especially, should be absolutely fine with 32ohm loads. Some headphones even dip lower than this, so providing you're using a decent sized reference resistor (25 ohms or so), then you should't have any problems whatsoever. Do bare in mind here though, that when using the output of the sound card, it might not be able to deliver enough current to give -5dB on the input when driving the loudspeaker + reference. If you're using close to 100 ohms though, as LIMP recommends, then you shouldn't have a problem with that.

This is about as simple as it gets and if you still can't get good impedance data like this, then I'd be scratching my head too. For what its worth, I've had excellent results from using the sound cards output at one point in time, so I know that this can work very well.

Cool. Thanks again for all the details.

[JonW]

I could be totally wrong, but doesn't that little blip in Jon's impedence measurements around 60 hz point to bad connection or the like? Whenever I have experienced an impedance measurement like that I have always traced it back to a bad connection, not totally faulty but loose enough to cause errors. Check the connection right at your soundcard. Try securing it as you are taking your measurement and see if anything changes or even sounds slightly different, maybe a little static sounding.

60hz suggests mains leakage somewhere but possibly not.

It also suggests the presence of a ground loop, but as his setup is apparently all double insulated this isn't going to be possible, at least not in the traditional sense. As the laptop will use switching supplies and can be run entirely from its own internal battery, it's most likely that this is coming from the amplifier as some form of induced cross-talk from the main transformer. This isn't uncommon and will be improved by keeping the SnR ratio of the measurements high. Of course if the problem exists before the volume control in the amp, then the 60Hz noise will increase at the same rate as the signal.

OK, that’s all interesting. A bad connection or a ground loop. I can live with the 60 Hz blip. As long as whatever is causing the blip is not also causing the baseline shift.

[JonW]

In related news, Ryan’s Woofer Tester had arrived. If things go well, I will get a chance to try it out tonight. It’s this teeny little thing, looking too simple to be true for what I’m after. We shall see.

[Face]

Soon you'll be wondering why you bothered with anything else...

[JonW]

Soon you'll be wondering why you bothered with anything else...

So true. See my next post.

[JonW]

Success! Woofer Tester did the trick. It’s amazing. Big thanks to Ryan (-k-). How much do you want to sell this thing to me for, old buddy, old pal? Making the plots below and writing this post will take longer than installing the software and collecting all the data. Wow.

There is a little bad news but let’s start with the excellent news. Check out how nice all the woofers look. There are no frequencies on the x-axis owing to my graphing program not liking the format of the file from Woofer Tester. But you get the idea.

Image not available


And here are 3 of the 4 tweeters. Super.



Now for the bad news. One of the tweeters looks like this:


An odd hump at lower frequency than the main peak. And the main peak is ~1.5 Ohms lower than that of the other three tweeters. Here is a plot of all four tweeters overlapped.



That hump shows up for this particular tweeter in both Woofer Tester and my older data from JustMLS. Any thoughts on what’s going on with this tweeter? I bought all the drivers a few years back, when I started making the cabinets. So there can’t be any return or warranty left on the thing. Am I about to order a new one from Madisound?

[JonW]

Here are the SPL/FR plots for all four tweeters. The bad one is in black. It looks very similar to the other three in this regard. Hmmm...

Image not available

[---k---]

Progress! I'm glad the Woofer Tester worked so well for you. It is a nice little tool. And remember, it works on the completed boxes too. I left my original CD in the box, but iirc they have some updated software that fixed some issues and is better. Might want to grab that if you didn't. And, you can save and overlay multiple impedances in the WT software and then export a JPG graph. Don't have to export the data and plot in, what is that Mattlab? Unless you have some other reason.

As for the tweeter, I bet you've had those so long that Madisound (I think) wouldn't be willing to trade it. But, you never know and might want to give them a quick call. They surprised me once with a refund.

[5th element]

Load up STEPS and do some distortion sweeps at say 2.83vrms from 500-20,000Hz. Do this with a good and then the bad tweeter and see what's what. If you've got rubbing it will show up clear as day. If there isn't any rubbing then there's two things I can think of. The first is that the ferrofluid is causing problems and the second is that a foam plug inside might have moved/come loose. The ferro might sort itself out after some use and the foam should be fixable providing you can remove the dome assembly.

[JonW]

Progress! I'm glad the Woofer Tester worked so well for you. It is a nice little tool. And remember, it works on the completed boxes too. I left my original CD in the box, but iirc they have some updated software that fixed some issues and is better. Might want to grab that if you didn't. And, you can save and overlay multiple impedances in the WT software and then export a JPG graph. Don't have to export the data and plot in, what is that Mattlab? Unless you have some other reason.

As for the tweeter, I bet you've had those so long that Madisound (I think) wouldn't be willing to trade it. But, you never know and might want to give them a quick call. They surprised me once with a refund.

Yeah, thanks very much! What do they say about having the right tool for the job? Actually, I would rather not update the software. I do all my speaker work on a retired, ~12 year old computer. Any updates or software from this present decade could easily exceed the capabilities of the laptop. If WT works now, I won’t touch it. Which leads me to ask: Do you want to sell the WT? If you’ll use it, I’ll send it back when this project is done. But if you don’t think that you’ll ever use it, let me know.

Yes, the impedance data that I collected last night was all in the completed cabinets. It’s very interesting to see how the W16 versus the ER15 woofers yield different tunings. But that is a topic that I will discuss at another point.

That’s Kaleidagraph for the graphing program. It’s what I use at work. Anything is better than having to use something from Microsoft.

On the tweeter, maybe I will call Madisound. Good idea. More info below...

Load up STEPS and do some distortion sweeps at say 2.83vrms from 500-20,000Hz. Do this with a good and then the bad tweeter and see what's what. If you've got rubbing it will show up clear as day. If there isn't any rubbing then there's two things I can think of. The first is that the ferrofluid is causing problems and the second is that a foam plug inside might have moved/come loose. The ferro might sort itself out after some use and the foam should be fixable providing you can remove the dome assembly.

Thanks. Taking distortion measurements is a good idea. Although I must admit that, at this point, I’d rather just buy a new tweeter than learn a whole new measurement system.

Now that you mentioned rubbing and parts coming loose, it jogged my memory... When I got the tweeters, one of them kind of came apart in my hands. As I was taking it out of the shipping box, the top flange came off. Kind of like the lid of a jar falling off because it was unscrewed. I just screwed it back on. But I’ll guess that something is in the wrong spot now. If I were to take the tweeter apart intentionally, what am I looking for to know if everything is (or is not) in the right place?

[5th element]

Err, well tweeters, if decently made, usually come with physical parts that align themselves via design, so that voice coils remain centered etc.

Now depending on how the tweeter ended up in pieces this could have done a number of things.

1) Permanently damage part of the device so that no matter how you put it back together it just wont go.
2) Something is dislodged or misaligned inside, the foam/felt pad that usually covers the pole vent (sits behind the dome) is usually fastened down in way that could come loose if the tweeter was dropped by mistake. If it has come loose and is touching the dome from behind it will be obvious when/if you remove the dome. It's an easy fix also.
3) If the centering assembly for the voice coil was damaged or knocked a little out of place then screwing the face plate down afterwards would keep it held that way. Sometimes all it takes is removing the faceplate and recentering the parts and screwing the faceplate back to get the alignment back. The tolerances/distances between parts in a well made tweeter are usually very tight. Knock things out by 0.2mm and it could cause things to rub.

Now the way in which the tweeter is designed, such as to hold the voice coil centered, or to make replacing the dome/coil easy can be quite different from on tweeter to the next. SEAS often use large plastic assemblies that fit over the entire motor/magnet structure, Scan used to use a simple plastic ring that was glued in place, in some of their recent tweeters they use pegs machined into the top plate that the assembly slots over. The SEAS method and the peg method do not allow the voice coil to be rotated and only tend to allow the assembly to be placed in one specific orientation. The plastic ring method does allow you to rotate the dome/coil though, it is less precise than the other two but it does give you some room to play with, ie if the coil is rubbing rotate it 90 degrees and see if it sorts it. The best way to do this is when playing a sine wave through the tweeter, if it's rubbing, it will be very obvious when the rubbing stops, so rotate and hope that happens at some point.

Hopefully this is fixable.

STEPs is not hard to use. Put the mic 30cm or so from the driver, play a test sine wave through the loudspeaker and measure the voltage with a multimeter. Set the volume till the meter reads 2.83vrms and adjust the gain of the mic preamp so that nothing is clipping. Hit start and let it do its thing. Due to the varying sensitivity of the driver vs frequency you might get clipping occurring at some frequencies, turn things down (usually the preamp gain) if this happens and go again.

You do not need to calibrate or reference anything in STEPs to make it work well. Just make sure nothing is clipping and you're good to go.

[Silversmoky]

Now those charts are more like it Jon! Glad the woofer tester worked for you. The easier the better sometimes right? Stinks about that tweeter though. If you wanted to send it to me I could run some distortion tests on it if you don't figure out what's wrong and get it fixed. I have a single good tweeter of the tweeters you are using that measures excellently. Could get a good comparison to see really how bad it is. Glad you are up on running on these!

[JonW]

Thanks, folks. That’s all very helpful, once again. And it gives me a quite easy plan to follow:

-Remove the bad tweeter from the cabinet
-Record an impedance sweep
-Take the tweeter apart and put it back together, gently
-Collect another impedance sweep. If it looks good, I’m done
-If not, take it apart and reassemble it a few more times
-Take more impedance sweeps
-If it ever looks good by impedance, I’m done.
-If I can’t get it to look like the other tweeters… just order a replacement.

The tweeter is only a minor setback in the project. This won’t be much effort to fix. Not like the trouble with measurements. And if it’s not fixable, a new tweeter will solve the problem. I just need to be careful that I do not strip the wood of the cabinet. This small tweeter only takes tiny #4 screws. So removing and replacing the tweeter from the cabinet is not something that I want to do too many times.

Silversmoky-
Thanks for the offer! If I can fix it, I’m good. But if not and I order another one, I can send you the bad tweeter to keep, if you’d like.

I’ll post what happens…

[---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.