JustMLS Calibration

[fatmarley]

Hi Guys,

I've got JustMLS running through an M-Audio mobilepre USB and a calibrated ECM8000 microphone. I'm pretty sure it's wired up ok, because it seems to be working - Here's how it's wired up:




I'd like to calibrate it properly for two channel mode, but the manual isn't very clear. Here's a quote from the manual:

"A more complex (but more accurate) way to calibrate is to connect the soundcard output
to the right channel via a probe. The mic preamp is also connected to the soundcard
output, it must however be attenuated before. The calibration this time will account for the
transfer function of the mic preamplifier."

I found some more info on the net about calibrating JustMLS, but it's quite basic:

"For calibrating the sound card, I take the speaker output directly to the line in, set the levels, and calibrate at various sample rates and lengths. When you get a good calibration file ("Justmls.cal") make a backup. I've trashed mine a couple of times."

I've drawn a picture of how I think it should be calibrated and just wondered if someone could cast an experienced eye over it for me ?



Does that look right ?


Thank's in advance.

Matt

[fatmarley]

Well i've tried to measure the frequency response of my speakers on three seperate occasions and the results look terrible. I know someone else here was having trouble with Two channel mode in Justmls / MobilePre, so I may just sell it.

Does anyone here successfully use Justmls and if so, what sound card do you use ?

[fjhuerta]

JustMLS can be used with any sound card. I've used with with 3 different ones with no issues.

I checked out your wiring and it seems to be correct. In order to calibrate JustMLS, what I do is - I short the input and output of the soundcard, run the calibration and level procedure, and then connect everything the way you described on your first image.

Are you adjusting the time-of-flight parameter and using the 2 channel input, both on JustMLS? That was my main issue when using JustMLS, and yeah, I got incomprehensible graphs. If you don't set up the correct delay (in meters or miliseconds) and gate the signal (usually around 4 ms) you are going to get ugly results...

[anders]

This is how I do it (on a M-Audio Fast track PRO)

Everything inside the red square is my "jig". The 10 ohm resistor is needed when measuring impedance.

[fatmarley]

I checked out your wiring and it seems to be correct. In order to calibrate JustMLS, what I do is - I short the input and output of the soundcard, run the calibration and level procedure, and then connect everything the way you described on your first image.

Thanks for checking my wiring. At least i'm doing something right

Are you adjusting the time-of-flight parameter and using the 2 channel input, both on JustMLS? That was my main issue when using JustMLS, and yeah, I got incomprehensible graphs. If you don't set up the correct delay (in meters or miliseconds) and gate the signal (usually around 4 ms) you are going to get ugly results...

I've tried calibrating as you described, so I guess the major problem is with my settings.
I always tick the two channel input box and set Channel 1-2 for the results. I also measure the distance from the mic to baffle and enter that in the box, so I guess the problem is with the gate settings.

For a nearfield measurement, say 5mm away from the midbass driver - How long to I set the gate for ?

[fatmarley]

This is how I do it (on a M-Audio Fast track PRO)

Everything inside the red square is my "jig". The 10 ohm resistor is needed when measuring impedance.

Ah, that makes sense. I will have to re-design my jig to be like yours. Thanks very much for the circuit diagram

What a great forum !

[fjhuerta]

Can you post an image of a sample measurement, just to have an idea of what you are looking at?

When doing a nearfield test, I can set the gate as high as I want - the measurement is always clean.

[fatmarley]

Well I redesigned my jig and calibrated at all the sample lengths etc, and the measurement of the tweeter (at 1 meter) looks fine, on the other hand, the midbass driver measurements ( VIFA M21 WG-09- 08 ) look very ragged to me, but I can't find a frequency response graph anywhere on the internet to compare. I did a nearfield measurement of the midbass driver and that looks pretty smooth to me too.

For the nearfield midbass measurement, I used 8000Hz and 32768 samples (the same as the Justmls tutorial)
For the 1 meter tweeter measurement, I used 44,100 with 8192 sample length and I used the same for the 1 meter midbass measurement. I'm wondering if I should experiment with different sample rates and lengths for the midbass driver ?

I will make sure I get some screen shots when I next set-up my measuring equipment. I did save the measurements in Justmls using the 'EXPORT' button, but i'm not sure how to put them in LspCAD. I've just printed the LspCAD manual, so hopfully everthing will become clear after i've read it.

[fatmarley]

Well, i've been doing a bit of searching and it looks like I need to either measure in a bigger room or wait for a nice day (it's very windy today) and set up my gear outside.

The ceiling is 2.4 meters high and the width of the room is 3.1 meters at the narrow point (chimney breast), so I guess room reflections are causing the ragged frequency response in the midrange. I'm sure I read that a ragged response caused by room reflections will also mess up the phase.

[JonW]

Hi Matt,

From your PM’s, I thought that it might be best to just discuss things here in your thread. I know that these questions will come up again, for others.

And, really, the first thing I should say is that I found the biggest pain about designing a speaker to be getting a functional measurement setup. Nothing frustrated me more. So keep at it. I got through it. And if I can… No, don’t sell the MobilePre setup. I got it working fine. With a little help and trials, you’ll get there.

Yes, the first diagram you sent to me looks exactly like what I have for my setup. You’re good there, I think. It was this one:



I do think that you should post some pics of what type of measurements you’re getting. We can help more if we see what you are seeing.

One thing that comes to mind is make sure that the knobs on the MobilePre are turned up enough, but not too much. Make sure you’re getting a signal such that the LED’s light up, the signal level is high in JustMLS. But not too high. Some other things that come to mind…

For measuring at 1 meter, I used a sample rate of 48,000, MLS length of 8192, and a 7 ms gate. It sounds like you’re close to those, so your parameters should be fine.

Maybe try this: Measure the tweeter at 1 meter, with the mic pointing directly at the tweeter center and use those parameters. The tweeter response should look good but the phase might not. Then adjust the offset by a few cm in either direction to whatever is needed to get the phase to look pretty. Then move the mic down to point at the woofer center, still 1 meter away. Do not change anything on the computer. Take another measurement. Don’t worry (and don’t change) the phase. You should be good to go.

Shots of my T and W measurements can be seen in post 444 here:

https://www.htguide.com/forum/showthread.php?t=25031&page=13&pp=35

(I can’t get into my photo-bucket account now, for some reason, so links instead of pics will have to suffice.) Those measurements are after I got all the kinks worked out and in the final cabinet with lots of facets cut to smooth the response. My first measurements did not look that nice.

I did my measurements at 1 meter and I think that they came out OK. I put the speaker up on a table so that the drivers were about half way between the ceiling and floor. Sure, outside or in a chamber would be ideal. But I think that things were OK this way. The 7 ms gate cuts off a lot of the room bounce. That’s what it’s there for.
A pic of, roughly, how I did my measurements can be seen in the first post here:

https://www.htguide.com/forum/showthread.php?t=25031&page=5&pp=35

By the way, what’s the project you’re making?

I hope that helps. With more info and screen shots we might be able to help more. Good luck!

-Jon

[fatmarley]

Hi Matt,

From your PM’s, I thought that it might be best to just discuss things here in your thread. I know that these questions will come up again, for others.

Hi Jon,

It's great to have your support. Your thread is a real inspiration to me.

And, really, the first thing I should say is that I found the biggest pain about designing a speaker to be getting a functional measurement setup. Nothing frustrated me more. So keep at it. I got through it. And if I can… No, don’t sell the MobilePre setup. I got it working fine. With a little help and trials, you’ll get there.

Yes, frustrating is a word that comes to mind. Although I think i've got my set-up sorted now (thank God). I think I need to experiment more with settings in JustMLS though.

I do think that you should post some pics of what type of measurements you’re getting. We can help more if we see what you are seeing.

Ok, i've managed to upload some screen shots, but i'm not sure if they will tell you much. I will get some better screen shots (using the Windows screen print function) when I next get the measuring gear set-up.

One thing that comes to mind is make sure that the knobs on the MobilePre are turned up enough, but not too much. Make sure you’re getting a signal such that the LED’s light up, the signal level is high in JustMLS. But not too high.

I've been turning the knobs up until the red LED lights and then back down until it goes out and a solid green LED light stays on.

For measuring at 1 meter, I used a sample rate of 48,000, MLS length of 8192, and a 7 ms gate. It sounds like you’re close to those, so your parameters should be fine.

Ok, I'll try 48,000 and 7 ms gate and will report back...

Maybe try this: Measure the tweeter at 1 meter, with the mic pointing directly at the tweeter center and use those parameters. The tweeter response should look good but the phase might not. Then adjust the offset by a few cm in either direction to whatever is needed to get the phase to look pretty. Then move the mic down to point at the woofer center, still 1 meter away. Do not change anything on the computer. Take another measurement. Don’t worry (and don’t change) the phase. You should be good to go.

Cool, something else to try. I've got to admit, I didn't really take much notice of the phase, because I was more interested in getting an accurate frequency response. IIRC the phase should go through the passband just the once (not zig zagging through, like some of my measurements).

Shots of my T and W measurements can be seen in post 444 here:

https://www.htguide.com/forum/showthread.php?t=25031&page=13&pp=35

(I can’t get into my photo-bucket account now, for some reason, so links instead of pics will have to suffice.) Those measurements are after I got all the kinks worked out and in the final cabinet with lots of facets cut to smooth the response. My first measurements did not look that nice.

I did my measurements at 1 meter and I think that they came out OK. I put the speaker up on a table so that the drivers were about half way between the ceiling and floor. Sure, outside or in a chamber would be ideal. But I think that things were OK this way. The 7 ms gate cuts off a lot of the room bounce. That’s what it’s there for.
A pic of, roughly, how I did my measurements can be seen in the first post here:

https://www.htguide.com/forum/showthread.php?t=25031&page=5&pp=35

I'll have to read through your thread again. It's full of little nuggets of helpful information.

By the way, what’s the project you’re making?

It's based on the same midbass driver that was used in the old Snell type K. JPW and Heybrook HB1 used the same driver, but aren't supposed to be as good as the Snells. It's an 8", treated, paper cone with a foam surround and has quite high sensitivity (90db), compared to most drivers designed for sealed box loading. It also has quite an extended frequency response for an 8" driver. It was crossed over to a 19mm tweeter in some comercial designs (IIRC the Snell K had a 19mm tweeter)

The reason the Snell k betters JPW and heybrook must be because of the tweeter and crossover. My idea uses a Monacor DT300 tweeter with it's WG300 waveguide. I'll upload some photos in a bit, but it's only a prototype so not that pretty.

I hope that helps. With more info and screen shots we might be able to help more. Good luck!

-Jon

Ok, I'll upload them now

[fatmarley]

Here are a couple of pics of the speaker. The one on the left is a Proac studio 100 (nice speakers)





And here is a shot of the 1 meter midbass measurement:



Here's the nearfield measurement:

[JonW]

Hi Matt,

Those pics will help (I think). To be honest, I cannot pinpoint the immediate problem but I can take a stab at it.

As long as you have everything hooked up correctly, perhaps you are not getting enough signal/sound to the mic and computer. The far field plot looks really noisy. Like perhaps you’re just recording noise and not the output of a driver. You know how you adjusted the MobilePre knobs to get a good signal, but not get into clipping? You also need to do the same thing in the JustMLS program. There are level meters in there. Make them read a good, high signal without clipping.

The mid drops off above ~5,000 Hz as we might expect. But we might also expect it to drop off below ~200 Hz, which it does not seem to be doing. Not that measurements are valid below ~200 Hz but if you look at the example plot I linked to, you could see a drop off in my woofer.

The magnitude of your signals versus drop offs look about right, I guess.

It would be helpful to see the tweeter measurements as well. We will have a better expectation for the tweeter response. It should be flat well above 10,000 Hz and drop off below ~1,000 Hz, all well into the range of good measurements.

Alternatively, you could have the noisiest, least flat driver ever, but I doubt it’s that.

Another alternative is that your near field looks more like we might expect and so you need to adjust parameters for the 1 meter distance. Except those dips at ~1.1k, ~2.2k, 3.3k are pretty extreme. (1.1 -> 2.2 -> 3.3 may tell us something.) My first thought there was maybe metal cone resonance, but you said it’s paper. I really don’t know.

Also, keep in mind that I really have close to no idea what I’m talking about. You’ll be better off if you get help from someone more knowledgeable. Which means pretty much any idiot walking down the street.

Sounds like a cool, fun design. Good luck!

-Jon

[fatmarley]

As long as you have everything hooked up correctly, perhaps you are not getting enough signal/sound to the mic and computer. The far field plot looks really noisy. Like perhaps you’re just recording noise and not the output of a driver. You know how you adjusted the MobilePre knobs to get a good signal, but not get into clipping? You also need to do the same thing in the JustMLS program. There are level meters in there. Make them read a good, high signal without clipping.

I set the levels for the tweeter, so that the signal bars in JustMLS are in the yellow, but I think I made the mistake of adjusting the midbass driver to be in the yellow too - but I can't see how that would have given the ragged response. If all goes well, I will have a free day this thursday to try some more measuring.

The mid drops off above ~5,000 Hz as we might expect. But we might also expect it to drop off below ~200 Hz, which it does not seem to be doing. Not that measurements are valid below ~200 Hz but if you look at the example plot I linked to, you could see a drop off in my woofer.

I wonder why it's not dropping off below 200 Hz ? Does anyone here have any idea ?

It would be helpful to see the tweeter measurements as well. We will have a better expectation for the tweeter response. It should be flat well above 10,000 Hz and drop off below ~1,000 Hz, all well into the range of good measurements.

For some strange reason i've lost the tweeter measurement, which is a shame because it was my best measurement. I'll post some more screen shots when I next have a go at measuring ( hopfully they will be better than these ones )

Alternatively, you could have the noisiest, least flat driver ever, but I doubt it’s that.

The drivers were kindly given to me, but the foam surrounds were rotten and I had to re-foam them. I cant see that this would cause the ragged response, but I guess it's possible.

Also, keep in mind that I really have close to no idea what I’m talking about. You’ll be better off if you get help from someone more knowledgeable. Which means pretty much any idiot walking down the street.

I think you're being a rather modest. You did really well and obviously put a lot of effort and thinking into your design.

[JonW]

I set the levels for the tweeter, so that the signal bars in JustMLS are in the yellow, but I think I made the mistake of adjusting the midbass driver to be in the yellow too - but I can't see how that would have given the ragged response. If all goes well, I will have a free day this thursday to try some more measuring.

OK, then my current best guess is that, perhaps, you have too much signal, rather than too little? It's possible that you're clipping somewhere. That could account for the lack of drop off below 200 Hz. And also maybe the big nulls at 1.1, 2.2, 3.3 kHz. Those 1.1, 2.2, 3.3 kHz nulls are probably telling us something, but I'm not sure what.

Do what I said in my prior post: Get a good tweeter measurement and save it. Then move the mic to the woofer but don't change anything else. Take the woofer measurement and save it. The signal intensity on the MobilePre, the bars in JustMLS, and the output levels of the plot may all be a little lower than the tweeter (or not) but that's OK.

One other thought: Did the calibration work out OK? If it's not calibrating properly, things could look funny. My memory here is fuzzy, but you should recognize a good calibration by pretty flat plots, with occasional noise but it all looks like electronics noise, not microphone-measured noise. Although your plots don't look fine except that the mic is not calibrated. So I don't think that's the problem.

I don't remember if I ever got the 2 channel measurements to work well or not. I might have stuck with 1 channel... or not. But it all worked out OK, in the end.

Thanks for the kind words. I feel like there are people out there who could help you better than I. Unless one of them starts typing, you're stuck with my help. :P

[fatmarley]

OK, then my current best guess is that, perhaps, you have too much signal, rather than too little? It's possible that you're clipping somewhere. That could account for the lack of drop off below 200 Hz. And also maybe the big nulls at 1.1, 2.2, 3.3 kHz. Those 1.1, 2.2, 3.3 kHz nulls are probably telling us something, but I'm not sure what.

I guess that could be possible, but I don't let anything go into the red. Just had a thought though - since changing my probes to the way Anders does it for calibrating, the left channel in JustMLS goes all the way up to yellow, but the right channel only goes half way up (stays in the green)

Do what I said in my prior post: Get a good tweeter measurement and save it. Then move the mic to the woofer but don't change anything else. Take the woofer measurement and save it. The signal intensity on the MobilePre, the bars in JustMLS, and the output levels of the plot may all be a little lower than the tweeter (or not) but that's OK.

Yes I will definitely try that.

One other thought: Did the calibration work out OK? If it's not calibrating properly, things could look funny. My memory here is fuzzy, but you should recognize a good calibration by pretty flat plots, with occasional noise but it all looks like electronics noise, not microphone-measured noise. Although your plots don't look fine except that the mic is not calibrated. So I don't think that's the problem.

My plots look pretty rough. I will take some screen shots and upload them here.

I don't remember if I ever got the 2 channel measurements to work well or not. I might have stuck with 1 channel... or not. But it all worked out OK, in the end.

If I can't get a good measurement, I will try it in single channel mode.

Thanks for the kind words. I feel like there are people out there who could help you better than I. Unless one of them starts typing, you're stuck with my help. :P

I'm DOOMED!!!

[JonW]

Yeah, give a shot to the single channel measurement. I did mine about 3 years ago, so my memory is, umm, a tad fuzzy. But I may have ended up getting better results with 1 channel. Even though, in theory, 2 channel should be better.

Then again, I may have done it 2 channel. Who knows? Did I ever build a speaker...? Not really sure. Those are funny looking boxes that make sound, right? Like to make all the sounds in your head come from somewhere else, right? Who am I and what am I doing here? :W

Doomed... indeed, you are doomed, sir. :P

[fatmarley]

Oh, no! Sounds like your memory is as bad as mine

[fatmarley]

Finally had a chance to try some more measuring.

Here's the tweeter at 1 mtr:



Here's the 1 meter midbass measurement:



This one is 5mm from the midbass:



This one is nearfield and farfield just before pressing the 'Merge' button:



I had to set the delay to 220 to get the phase to line up.

[fatmarley]

This is my first attempt at a crossover. It's first order electrical for the woofer and tweeter. I want to keep the crossover as simple as possible because i've heard a complicated one can suck the life out of the speaker.



Does anyone know what would happen if I used a second order electrical for the woofer and kept the first order on the tweeter ? Just wondered if it would help fill in the notch between 1000Hz and 4000Hz...

[fatmarley]

Here are all the graphs from LspCAD:

[fatmarley]

Forgot to mention why I was getting bad measuring results - I replaced the 20watt per channel Teac CR-L600 amplifier with a 60watt per channel Onix OA21 and also replaced the 47R resistors in the the probes for 470R, re-calibrated, and everything was fine. I can see why people don't recommend those low powered T-amps (Assuming it was a lack of power that caused the problem)

[fatmarley]

One more picture - This is my measuring set-up:

[JonW]

Hi Matt,

Your 1 meter measurements look excellent! Congrats. :T The most annoying part of speaker building is now behind you.

I’m not sure why the 5 mm measurement looks funny. But the 1 meter T and W look great, so go with that.

Crossover philosophy can be very personal. So take my thoughts with a grain of salt. I have only 1 complete project under my belt. But for whatever they are worth…

I think that you could have the frequency response flatter. I might try to knock down the hump at ~700 Hz and bring up the dip at ~2,000 Hz, to be on level with the tweeter at 4,000-15,000 Hz. Although a BBC dip can be pleasant. As can a lot of bass, but my guess is that the bass hump will sound good at first but bloated after a lot of listening.

Try to get the phases of the 2 drivers to match up better. It’s most important for the frequencies at which both drivers are playing and can be heard. Let’s say ~500 - 5,000 Hz, roughly. Below, as an example, I have a pic of the phase overlap for the Spassvogels. I got them to overlap pretty decently. For me, the speaker really came alive when I got the phase under control. That’s what jumped it up a level to “wow.”

If you're up for it, also take measurements at 30 degrees off axis. After you have a crossover that looks good, plug in the 30 degree measurements. The tweeter response will drop above ~10,000 Hz. But if everything else looks good, you're in good shape for off axis issues. My guess is that many people skip this part.

Also if you're up for it, you might want to try to get rid of the impedance hump at 1500 Hz. I found that it helped the sound a bit. Here too, I think that many people skip this part. See below:

Regarding simple crossovers sounding better, I’m not buying it. If there are issues to address (peaks in frequency, poor phase overlap, impedance jumps, etc.) it might sound better if you address them in the crossover. Then again, some people prefer the simplicity of a single driver speaker.

In the Spassvogel crossver, I have 2 LCR filters. Looking at the crossover, it may seem overly complex and as though the sound is going to be sucked out of the speaker. Some people have said it looks like that. However… I clipped together test crossovers and listened to things both with and without each LCR filter. I preferred the sound of having one by a significant margin and the other by a small bit. They are there for a purpose and they seem to help. Listening impressions on the LCR’s are here in post 459:

https://www.htguide.com/forum/showthread.php?t=25031&page=14&pp=35

But don’t take my word for it. Try it for yourself. Clip together parts for the simple crossover. Have a listen. Then do the same for a more complex crossover. Which do you prefer? Go with whichever you like better.

Excellent progress, Matt. Good luck! :T

[fatmarley]

Thanks for the encouragement Jon. I bought some crocodile test leads after reading your thread. I thought it was a good idea for experimenting with different components.

I remember someone saying getting the phase right is vital, but I don't know what to do to improve it. I don't suppose you can remember ? I honestly haven't got a clue.

I took some some measurements at 15, 30, 45, and 60 degrees (you can just see the protractor that I printed, under the speaker stand) I put a nail in the base of the stand, in line with the tweeter. The girlfriend would kill me if she found out I put a hole in the table lol

[JonW]

Hi Matt,

I remember someone saying getting the phase right is vital, but I don't know what to do to improve it. I don't suppose you can remember ? I honestly haven't got a clue.

You know how, in LspCAD, you have the crossover. You change the value of a component and then look to see how it changes the frequency response? Same thing, just look at the phase plot. Change component values (or even the presence or absence of components) to get the phase looking good. Most likely you may arrive at a point where you can have either the phase look better or the frequency response. Frustrating. Pick your poison. Have a listen each way and see what you think. Maybe you'll get lucky and the good phase and good frequency response and good impedance all coincide. But if it gets too frustrating, have a drink, relax, and remember that it’s not nearly as frustrating as trying to get good measurements.

Also check for the reversed tweeter null. If it's deep, you're onto something good.

The girlfriend would kill me if she found out I put a hole in the table lol

You're lucky that the ladies pretty much never stop by here. So I think that your secret is safe.

[fatmarley]

I couldn't get the phase to line up with a first order crossover, so I tried a second order.

The frequency response isn't perfect, but the phase looks far better than before.



Do you think the phase is good enough ? I'm going to clone this attempt and keep trying to improve it. I wonder if a zobel would help or make things worse ?

[CraigJ]

Unless one of them starts typing, you're stuck with my help. :P

Matt,

You are in very good hands. Welcome back Jon, you've been missed.

Craig

[MikeSeddon]

Phase Response Target?

Hi,
I am glad this thread came up because I have been trying to understand what I should be targeting for the phase response when modeling a speaker.

I have attached a modeled speaker with Frequency and Phase Response for a 2nd order and a 3rd order crossover. Both model well in the frequency response and have good nulls when inverting the phase of one of the drivers but I do not know what to aim for in the phase response.

Maybe you could have a look and critique these modeled responses and maybe point me to somewhere where I can do some further reading.

Unfortunately I am overseas for an extended period at the moment and do not have access to my Loudspeaker Design book. I am modeling with the LspCAD 6.37 Pro Demo.

2nd Order:

Images not available

3rd Order:

Images not available

Edit: Added an additional 3rd Order for comparison:

Images not available

Thanks,
Mike

[fatmarley]

Matt,

You are in very good hands. Welcome back Jon, you've been missed.

Craig

I really appreciate Jons (and the other guys) help. I would have struggled, if not given up, without their input.

[fatmarley]

Mike,

Going by what Jon says and his example. It looks like you need to tidy up the phase a bit more.

Here's my best attempt (second order) :





I've had better phase, but the frequency response was quite bad.

[fatmarley]

Right - I've got three or four 2nd order crossovers shortlisted now. They all have good phase and impedance and the frequency response is pretty flat on all of them (well, as flat as I can get without lots of notch filters).

I've ordered loads of components, so I can experiment. Must of spent around £300, so I hope it's going to be worth it :E

Here's the graphs of one of them:







I'll report back with my opinions when I've had a good play with the various crossovers...

[JonW]

Hey Guys,

Sorry for the silence. The kid, work, etc. sometimes get in the way of speakers. It’s no excuse, I know.

Matt,

You are in very good hands. Welcome back Jon, you've been missed.

Craig

Craig!!

How’s it going! Yeah, it’s good to be back a little bit. I still don’t have enough time to pick up my half finished speaker project. But at least I can read and type a little. What are you up to? What’s the latest project?




Matt-

You’re coming along wonderfully! The phase in post 27 looks excellent. Don’t change it.

For post 32, I’d say that your phase still looks super. The impedance also looks excellent. I’ve seen better. But I’d say that most are worse. So you’re in great shape for phase and impedance.

For the frequency response plot, I don’t like the dips at 300, 500, 7,000, and 13,000 Hz. But, eh, they’re in the raw driver measurements. So I don’t think that you can do too much about them.

The only thing in there that I might suggest changing at this point is the woofer level. Can you knock it down about 2 db? Alternatively bring up the tweeter level? If you have a tweeter padding resistor in the circuit, try decreasing the value by ~2 ohms. I’m thinking that the extra output between 400 and 1000 Hz may sound a little off after long listening. But it might be fine. I doubt it will sound bad right away.

Also, try reversing the tweeter polarity in LspCAD and see if it models a good null (20+ db deep or more).

Your next step is, I would say, the correct one: Buy some parts, clip things together, and have a listen. I found that it helped when I did that. I also changed component values a bit and listened some more. For example, if there was a 10 uF cap, I tried it with 8 and 12. I did that with every part of the crossover. Some of the values changed. And it sounded better, overall. Interestingly, when I plugged all those new values into the crossover, the plots of frequency response, phase, and impedance barely changed. LspCAD is wonderful, as is a crossover model. But at some point you have to use your ears. Besides, what one person really likes, another may not.

[JonW]

Hi,
I am glad this thread came up because I have been trying to understand what I should be targeting for the phase response when modeling a speaker.

I have attached a modeled speaker with Frequency and Phase Response for a 2nd order and a 3rd order crossover. Both model well in the frequency response and have good nulls when inverting the phase of one of the drivers but I do not know what to aim for in the phase response.

Maybe you could have a look and critique these modeled responses and maybe point me to somewhere where I can do some further reading.

Unfortunately I am overseas for an extended period at the moment and do not have access to my Loudspeaker Design book. I am modeling with the LspCAD 6.37 Pro Demo.

Hi Mike,

Welcome to htguide. And what a way to start here- with a partially complete project. :T

Like Matt and I have been discussing, try to get the phase to overlap at the frequencies where you will be able to hear both drivers playing. About 500-5000 Hz.

For your 2nd order example, your phase looks super. As does your frequency plot. Is it possible to drop the woofer level by about 2 db in the 500-1000 Hz range? Probably not a big deal at all. It’s the only thing I can see there even worth mentioning. This crossover looks great.

For your first 3rd order, the frequency response plot looks about as good as the 2nd order. The phase, however, isn’t great. The 2 drivers are not playing well together above 3000 Hz.

For the additional 3rd order, I like this frequency output the best. Smooth and even. The phase looks better than the other 3rd order but not as good as the 2nd order.

I hope that helps. I’m just armchair quarterbacking here. And I don't even claim to do that well. Good luck! :T

[MikeSeddon]

Thanks for the reply Jon and Matt.

Matt:
Have you tried the crossover emulator function in LspCAD? Because you are using a two way crossover it means that you can use the LspCAD PRO demo and try the crossover emulation (even though you cannot save the design). That is a very cool feature of LspCAD which would still allow you to tweak even though you have already spent some serious cash on crossover testing components.

When I tried it I found it very cool as you can change things in real time even on quite an old computer. If you use the saved crossover configurations (the S1 through S8 buttons) you can switch them whilst the music is playing.

My problem was that I completely stuffed up the measurement of the speaker frequency/phase and therefore ended up with a speaker which modeled flat but sounded messed up. Rubbish in = Rubbish out.

Jon:
Thanks for your feedback.

I have been researching about idea phase response on the web but haven't found that much. I have found this page Aspen VSonics which spouts some theory about phase but I haven't heard this theory before.

For good imaging performance, the driver phase differential should be less than 5 degrees at crossover, here it is less than 0.1 degree

Interesting, I have found that most of the more popular DIY designers do not post phase plots alongside the frequency response.

I think that second order makes sense for this design as both drivers are capable of performing over very large frequency ranges (which is why I chose them). Unfortunately the little W12CY003 (the 4.5" Seas Nextel driver) has quite a severe peak at 900Hz which, according to Troels Gravesen, is caused by the surround resonance.

I have modeled one more crossover with particular focus on flat frequency response and very close phase matching at the crossover point (which has reduced to ~2300Hz).

Images not available

And here is the speaker itself - It is using 22mm bamboo plywood which is very expensive in Australia but I think really nice for speaker boxes.

Image not available

Thanks for your help. I think that passive crossover design is a dying art now that digital sound processing and switching amplifiers solutions are becoming relatively cheap.

Cheers,
Mike

[fatmarley]

Hey Guys,

Sorry for the silence. The kid, work, etc. sometimes get in the way of speakers. It’s no excuse, I know.

Ok, we'll let you off this time

Matt-

You’re coming along wonderfully! The phase in post 27 looks excellent. Don’t change it.

For post 32, I’d say that your phase still looks super. The impedance also looks excellent. I’ve seen better. But I’d say that most are worse. So you’re in great shape for phase and impedance.

Cool ! I'm doing something right for once.

For the frequency response plot, I don’t like the dips at 300, 500, 7,000, and 13,000 Hz. But, eh, they’re in the raw driver measurements. So I don’t think that you can do too much about them.

Yea I know what you mean. I found a cure for that

The only thing in there that I might suggest changing at this point is the woofer level. Can you knock it down about 2 db? Alternatively bring up the tweeter level? If you have a tweeter padding resistor in the circuit, try decreasing the value by ~2 ohms. I’m thinking that the extra output between 400 and 1000 Hz may sound a little off after long listening. But it might be fine. I doubt it will sound bad right away.

The graph above is flatter between 400 and a 1000Hz, but the high frequencies roll off a little. I'll see what sounds best and report back.

Also, try reversing the tweeter polarity in LspCAD and see if it models a good null (20+ db deep or more).

Just tried that and I have a good null

Your next step is, I would say, the correct one: Buy some parts, clip things together, and have a listen. I found that it helped when I did that. I also changed component values a bit and listened some more. For example, if there was a 10 uF cap, I tried it with 8 and 12. I did that with every part of the crossover. Some of the values changed. And it sounded better, overall. Interestingly, when I plugged all those new values into the crossover, the plots of frequency response, phase, and impedance barely changed. LspCAD is wonderful, as is a crossover model. But at some point you have to use your ears. Besides, what one person really likes, another may not.

I've got the crossover boxes wired-up ready. All i'm waiting for is the inductors.

Here's the caps/resistors and crossover boxes:

[fatmarley]

Matt:
Have you tried the crossover emulator function in LspCAD? Because you are using a two way crossover it means that you can use the LspCAD PRO demo and try the crossover emulation (even though you cannot save the design). That is a very cool feature of LspCAD which would still allow you to tweak even though you have already spent some serious cash on crossover testing components.

When I tried it I found it very cool as you can change things in realtime even on quite an old computer. If you use the saved crossover configurations (the S1 through S8 buttons) you can switch them whilst the music is playing.

My problem was that I completely stuffed up the measurement of the speaker frequency/phase and therefore ended up with a speaker which modeled flat but sounded messed up. Rubbish in = Rubbish out.

No, I haven't tried the emulator. Sounds cool ! I'll give it a try when I have some spare time.
I did try the Optimiser, but found it usually gave rather strange results, so I didn't bother with it in the end.