problems with acoustic measurements

[jbateman]

About phase, you can put the first marker just before the main pulse instead of 0ms, this will clear the phase plot in a degree.

This can potentially reduce the resolution of the measurement. Here is how it was explained to me by someone smarter:

"Speakerworkshop limits the resolution of the data looking at the sample rate and the time gate. (Apparently this is too conservative) It selects whichever is lower. In your case it is limited with the time gate. The difference between two gate points is what counts. So it helps to move the starting gate at 0 or 0.1
You can check the number of data points (resolution) by right clicking in the dataset and choosing properties. It is under the specifics tab. You can change this number and the graph will look slightly better for crossover modelling. Once you repeat a measurement sw will reset this number according to its algorithm.
To maximise the resolution you have to be careful with what is the bottleneck.

With 44100 sampling rate:
gate 5.805-11.61ms gives 256 points
gate 11.61-23.22ms gives 512 points
gate 23.22-46.44ms gives 1024 points
gate 46.44-92.88ms gives 2048 points

On the other hand, MLS size:
1024 gives 512 points
2048 fives 1024 points etc.

Increasing time gate gradually does not change resolution gradually.
So for example going from 6ms gate to 8ms gate does not make any difference for 44100hz. You need to go beyond
11.61ms for satisfactory resolution (512points and more)."

 

[wintermute]

This is important to understand, Latency setting at Properties is one thing, checking the Measurement.In.L is another.

When you are about to measure a woofer and a tweeter and build a crossover for them, you should make sure you have accurate phase data. This means (among the other things mentioned before) that the time each pulse that will be sent to the driver occurs, is exactly the same for all drivers. This is checked via the peak's time at Measurement.In.L file. You should check the peak's time each time you do an acoustic measurement to be the same.

OK I think I'm getting an idea now why I'm not 100% following, I've just been using the On Axis measurement menu in SW, are you actually doing a raw MLS measurement and then doing an FFT on it to get the response curve??

my measurement.in.l under system folder hasn't changed since Dec last year... (which is odd because I thought I recalibrated recently)... Might be time to start a new clean project and start from scratch

what I have been doing is:

set up mic and speaker
do a measure/pulse response
set the gate time visually based on just measured pulse repsonse
do a measure/freq response/On Axis

I don't have any measurement.in.l files for my individual measurements of drivers only the one in the system folder.....

Tony.

 

[wintermute]

This can potentially reduce the resolution of the measurement. Here is how it was explained to me by someone smarter:

["[/i] [/B]

I remember reading that now I think at one stage I did play with moving the start time but read that explanation, and put it back to zero, I obviously wasn't paying too much attention to phase at that stage

Tony.

 

[jomor]

Yes you are right, I was speaking about putting the 1st marker before the main pulse just to see better the phase, not using it in general.

Yes, I am doing a raw MLS measurement and I apply an FFT to the imulse response manually ( i was using MLSSA ages ago and I guess thats why I m stuck on doing things by hand )

afaik you re not doing anything wrong on the procedure.

each time you do an acoustic measurement, leave the Measurement.In.L file open all the time, and watch if the peak slightly moves left or right, just after the
measurement finishes. If it doesnt, you have a perfect sound card Most soundcards do that, no matter the Latency setting. Latency just adjusts how close the pulse's peak
will be to zero ( 0ms ), but this doesnt mean that the time distance from 0 ms to the pulse will be completely constant at every measurement, sometimes the computer fails to control
the soundcard in time for various reasons.

 

[wintermute]

Thanks jomor,

I just did a loop record of an mls signal as you suggested and watched it. My sound card isn't perfect but its pretty good almost all of the records had the start of the pulse at 0.2ms but occasionaly it started at 0.175ms!! so I think that I don't need to worry too much about this aspect BTW I have zero in the debugging latency box.

BTW I haven't done any acoustic measurements today.... Decided I better measure my Vifa M26WR-09-08's T/S params, as I'm going to my parents place Friday to build my new cabinets for the MTM's and will probably have enough MDF left over for a prototype Sub enclosure

I might not get to try the mods with the mic wand until next Tuesday at the earliest... but I'll post the results once I have them

Tony.

 

[wintermute]

Well I still haven't done the mods to the mic, I decided it was time to stop testing my prototype and just get on and design the final cabinets, and make them

I cut (with the help of my Father and his triton workbench) most of my MDF today

Only one slight oversight, I forgot to check the MDF was really 25mm..... it was 25.5mm...... luckily because of the construction method I have chosen this means that the only panels that are out are the front and rear baffles which are 1mm too short overall I'll just put that at the bottom and it won't be noticable. my internal width will probably be about 1mm less than it should be too, but no big loss.

Cutting acuracy was very good (IMO) at about 1/4 mm maximum variation.

pic of pile of mdf attached

Tony.

 

[wintermute]

and a clamped together box

note that the internal baffles between the two chambers haven't been cut yet....

Tony.

 

[tcpip]

Okay, friends, I think it's now my turn to suffer and sweat. I'm using SW and a Panasonic and trying to get some SPL measurements, and I'm getting nowhere. I'll tell you what I have checked:

• The mic, cable, and sound card works. I recorded my voice and played it back.
• The sound card works with SW, because I have used it for T/S parameter measurements.
• When I try to do a loop record of a 1KHz sine wave using my setup, it works fine. I can see sine waves in the graph, and adjusting my preamp's volume control changes the amplitude. This means, I guess, that SW has no problems emitting a sine wave and getting it back through my mic.
• When I try to do a frequency measurement using this setup, it doesn't work.

Basically, I get a very low amplitude signal, very near 0dB. And it's just a random bunch of squiggles. The speaker hisses quite loudly, but I get no graph.

The strange part of all this is that even though I fail to get an SPL graph, I can see SW's VU meter showing good readings. I tried "measure->frequency->nearfield" and played with my preamp's volume control, and I could get amplitude readings of about +/-20,000 or more. However, I never got anything which looked like a frequency curve. I'm only getting tiny squiggles.

I tried playing with the sample rate and sample size, and I get meaningless squiggles with all settings. I don't have one of those graphs right now to show you, so I thought I'd first ask. If I get around to taking some more readings in a few hours, I'll post a graph.

Any suggestions to tell me what I'm doing wrong? This is my first attempt at doing SPL readings, so I guess I'm doing something really silly. I had done something very simple and silly even for impedance measurements (very small measuring time window, identified with jbateman's help here). I guess something equally silly is at work here. Please help.

In the meantime, I'm re-reading the old stuff in that earlier thread. I'll re-try my measurements after increasing the duration between the time markers. Currently, I think the markers are about 10msec apart. I'll set them to 30msec apart, as per what jbateman had said here. Let's see....

 

[tcpip]

Got some graphs

I did some measurement attempts, and got some graphs. The one below is what I got when I tried "Measure->frequency response->nearfield":

An externally hosted image should be here but it was not working when we last tested it.

And this is what I get from "Measure->Pulse response":

An externally hosted image should be here but it was not working when we last tested it.

I used sample rate of 44.1K, and sample size of 256Ksamples.

What am I doing wrong?

 

[wintermute]

Hi TCP/IP,

It looks to me like you actually are recording the loop back (not the mic) and that the level is too high causing some peaking (I could be completely off the mark here).

In SW goto options/preferences/acoustic tab and make sure you have reference and mic set to the correct channels (and above all that they aren't both set to the same channel!)

That's all I can think of at the moment

If you aren't using a mic preamp (ie just have a mic plugged into the sound card direct) you can't use the reference channel and need to change it to none.

Tony.

 

[tcpip]

In SW goto options/preferences/acoustic tab and make sure you have reference and mic set to the correct channels (and above all that they aren't both set to the same channel!)

Currently, in that dialog box, I have "Mic" set to "right" and "Ref" set to "left". I know that setting "mic" to "left" didn't work. So I left it as it is.

If you aren't using a mic preamp (ie just have a mic plugged into the sound card direct) you can't use the reference channel and need to change it to none.

I don't have a reference input, and my "ref" setting in the dialog box is set to "left", not to "none". Maybe this is it! I'll try with "ref" set to "none" and I'll let you know. If this works, I owe you a crate of beer.

And yes, I have the mic plugged directly into the mic input of my sound card, without any jig or reference loopback channel.

 

[tcpip]

You did it, wintermute!

My SPL measurement is working, and I am getting something other than flat 0dB curves! Thanks! When are you coming to Bombay next? The beer awaits!

Now for the curves. I decided to inaugurate the measurement system by taking near-field measurements of the three drivers of my Wharfedale speakers at home. These were taken using "Measure->Frequency response->Nearfield", with the mic+wand hand-held, shaking slightly, about a quarter-inch in front of the diaphragm/cone/dust-cap of each driver. The sample rate was 44.1KHz, sample size was 256Ksamples, and repeat count was 5. I thought near-field measurements are the simplest to get right, so I thought I'd try this first.

See what I got. The tweeter is:

An externally hosted image should be here but it was not working when we last tested it.

The midbass is:

An externally hosted image should be here but it was not working when we last tested it.

The woofer is:

An externally hosted image should be here but it was not working when we last tested it.

And the port is:

An externally hosted image should be here but it was not working when we last tested it.

The Wharfedale Pi40 is a mid-fi speaker, with a 2.5-way design. The electrical xo is 2-way, but the two bass drivers handle different ranges of frequency because one is in a small sealed chamber and the other is in a larger BR chamber. When I took the measurements, all drivers were emitting sound, irrespective of which driver I was reading with the nearfield mic placement.

The same speaker has been measured in lab tests by What Hi-Fi a few years ago (yes, I've read some of those magazines with great zeal at one time), and they remarked about how flat the freq response was. It apparently didn't even have the BBC dip, thus resulting in a slightly "forward" sound. Whatever the case may be, I don't think "correct measurements" for this speaker will give these 20dB peaks and troughs.

Are these graphs "right"? Why is the tweeter's graph so like a sawtooth? Is some sort of comb filtering happening somewhere? Is this the best curve one can hope to get when all drivers are emitting output simultaneously, even though I'm doing nearfield?

I tried a "gated" measurement of the midbass unit, with 10ms between start and end marker, and this is what I got:

An externally hosted image should be here but it was not working when we last tested it.

I tried smoothing all the graphs using 1/8-octave smoothing, and I got these. The first is the tweeter:

An externally hosted image should be here but it was not working when we last tested it.

The midbass:

An externally hosted image should be here but it was not working when we last tested it.

The woofer:

An externally hosted image should be here but it was not working when we last tested it.

And the port:

An externally hosted image should be here but it was not working when we last tested it.

I tried combining the nearfield graphs of all four sound emitters (including the port), just to see what the resultant graph looks like. (I know nearfield curves can't be combined just like that to generate anything meaningful.) The resultant graph is:

An externally hosted image should be here but it was not working when we last tested it.

Am I doing something wrong? Is this what I should get?

I also tried a "pulse response" of one of the drivers. I got nothing in the graph, even though the VU meter showed signals being recorded.

Desperately need guidance and comments. Please help.

 

[wintermute]

Glad you have made some progress

ok first thing is I think you are overloading the mic, those response curves look really furry, they do that when the mic (or the preamp) is clipping. try reducing your drive level, I do my nearfield tests quite close (about 4mm) and use a fairly low drive level, much less than 1W).

second you really need to use some sort of stand to make sure the mic is staying completely still, I use a camera tripod and use a cable tie to hold the mic wand on it.

3rd what are you using for a mic cable?? the big peaks and drops look very much like something I was seeing when I joined to peices of microphone cable together, they don't appear when I use a single length of mic cable. The first nearfield measurements picture at the beginning of this thread has had NO smoothing applied it is as it came off the mic.... now these are fairly high end woofers so I'm not saying yours should be that smooth, but note that it isn't furry

4th nearfield isn't really usefull on tweeters It's only good up to 200-300Hz (after this its accuracy drops off).

5th you really need to look at the pulse response and set the markers based on it.... if you have a normal room you are probably only going to be able to get 3ms max of data, possibly a bit more if you can suppress floor bounce using egg crate foam or similar. you can use a larger gate time but you will start to see room effects in the measurements. Try chaging the graph Y axis settings to +- 800 (instead of 32K) and 10ms for the maximum on the x axis, you should see the pulse properly then, you can fine tune from there.

6th I just noticed all drivers were playing when you did these measurements, you really need to disconect the others and do one at a time otherwise you will have no idea what is interference from the other drivers (possibly none) and what is the freq response of the driver you are looking at. Also if you are testing with the crossover in place it is going to give you very different curves to without (one would hope )

I was going to say something else but I forgot, but I think that is enough for you to think about for a while

Not sure when I will be in Bombay, but I will be sure to look you up

Tony.

 

[tcpip]

Glad you have made some progress

Ditto. About me, I mean.

ok first thing is I think you are overloading the mic, those response curves look really furry, they do that when the mic (or the preamp) is clipping.

I'll try this. My levels are between 17-21K now, as per SW. There was no mention of fur problems in the Unofficial Manual... they kept on reminding one to increase the volume to a high enough level. I'll try a lower level and see.

second you really need to use some sort of stand to make sure the mic is staying completely still, I use a camera tripod and use a cable tie to hold the mic wand on it.

I knew I'd certainly not get anywhere for accurate measurements without a stand... this round was just "to see what I get". However, your idea of a camera tripod is brilliant.. I'd never have thought of it. I was going to ask my friend to invest in a mic stand of the kind they use on stage.

3rd what are you using for a mic cable?? the big peaks and drops look very much like something I was seeing...

I am using two-core-plus-shield thin cable. I'm using the two cores for connecting to the two pins of the mic capsule. At the sound card end, the shield is connected to one of the cores. I have no joins in my mic cable... it's one continuous piece.

The first nearfield measurements picture at the beginning of this thread has had NO smoothing applied it is as it came off the mic....

Don't make me jealous. I have a hard enough time as it is.

4th nearfield isn't really usefull on tweeters It's only good up to 200-300Hz

So I have to use gated for tweeters? I didn't know that. Okay, will try gated then.

Try chaging the graph Y axis settings to +- 800 (instead of 32K) and 10ms for the maximum on the x axis, you should see the pulse properly then, you can fine tune from there.

Will do this. Let me see what I find with the increased sensitivity.

6th I just noticed all drivers were playing when you did these measurements, you really need to disconect the others and do one at a time otherwise you will have no idea what is interference from the other drivers (possibly none) and what is the freq response of the driver you are looking at.

Actually, I was hoping that the effect of neighbouring drivers would be very mild because I was doing near-field measurements. But I guess I was being naive.

Not sure when I will be in Bombay, but I will be sure to look you up

Where in Australia do you stay? Who knows... I can try doing a home delivery of my promise. Not that I have any immediate plans.

 

[tcpip]

about 3 or 4 mm... I've tried varying it from about 2mm to 10mm but it doesn't seem to make any difference (apart from absolute levels).

Does this mean you can get good SPL from 3mm distance up to 20KHz using nearfield techniques? Or can one do this only upto 1K or so?

 

[wintermute]

Hi TCP/IP,

even though the levels in SW are ok, it doesn't mean that the mic capsule itself isn't being overloaded I have all sorts of troubles getting the levels to match, and often I resort to something between 7 and 15K

ahhhh I just remembered what the other thing was is your sound card capable of sampling greater than 44.1Khz?? go into options/preferences/general and make sure that both 48Khz and 96JKhz sample rate checking are ticked, close SW and open again and see if you can set the sample rate higher. Some cards that support 48Khz and higher have really bad performance at 44.1Khz, I know my Audigy II ZS does!! It seems to work best at multiples of 12Khz.

The effect of neighboring drivers may well be minimal (depending on how far apart they are mounted), but I wouldn't rely on it

I'm just happy to help TCP/IP if you are here in Aus or I'm in India you can buy me a beer, other than that the fact that I helped you get going is reward enough I like to get help when I need it, and in return I like to give help when I can

Tony.

 

[soongsc]

When you do near field measurements, try increasing the repetition count on measurements and reduce mic preamp gain (if you're using the Wallin preamp, there are two gain settings). This will make sure that you don't have signal saturation and eleminate noise effects.

 

[wintermute]

Does this mean you can get good SPL from 3mm distance up to 20KHz using nearfield techniques? Or can one do this only upto 1K or so?

Ummmm basically what happens is that as freq goes up the relative SPL goes down, so at 1Khz it may be down say 3db (from actual), at 5 Khz it may be down 6db (note these are figures I just pulled out of the air and probably bare no resemblance to reallity).

I find that the nearfield measurements seem to mirror the gated ones except that I do notice the trailing off at higher freqencies, I found that nearfield was also useful for testing cabinet reflections (and the effect that different damping materials had on them) even at frequencies up to about 1Khz.

Dickason states that nearfield is only usefull up to 200Hz or so. If you look at the freq response curve at the begining of this thread, you will see that it is tapering off from 200Hz up (with some peaks starting at 700 or so hz which are cabinet reflection related).

Tony.

 

[tcpip]

even though the levels in SW are ok, it doesn't mean that the mic capsule itself isn't being overloaded I have all sorts of troubles getting the levels to match, and often I resort to something between 7 and 15K

Will do. This evening itself.

ahhhh I just remembered what the other thing was is your sound card capable of sampling greater than 44.1Khz??

The card I have is a low-end relative of your Audigy: I use the SB Digital Music USB. It's a 2-in, 2-out, plus mono-mic, plus stereo earphone, USB sound module. Very inexpensive. It has worked well (without any amp) for impedance measurements, and I'm hoping it'll do SPL too.

I'm just happy to help TCP/IP if you are here in Aus or I'm in India you can buy me a beer, other than that the fact that I helped you get going is reward enough I like to get help when I need it, and in return I like to give help when I can

Hey, don't take me all that seriously... I was only joking. (The beer offer in Bombay is dead serious, though. ) I have found offers of beer to be a good ice-breaker, and I've made friends on this forum who have carried forward in real life very happily. This list includes "gjo", "corbato", "angshudas", "variac", and many others.

And in case you wake up in a cold sweat at night worrying that this mad chap from Bombay will come haring after you with a crate of beer, you can relax. I usually don't work that hard to keep my promises.

 

[tcpip]

I find that the nearfield measurements seem to mirror the gated ones except that I do notice the trailing off at higher freqencies, I found that nearfield was also useful for testing cabinet reflections (and the effect that different damping materials had on them) even at frequencies up to about 1Khz.

How do you do this? I can imagine fixing a mic to a speaker wall using SuperGlue, but that'll be a one-shot test where I have to say goodbye to the mic after the test. What do you do? Will just holding the mic a couple of mm away from a speaker wall allow me to measure panel vibrations?