Thanks, Jbateman and Tony. I'm reading your postings carefully, and will make notes, reboot into Windows, and try out the calibrations. I'll see if I can get rid of the fuzz in the impedance graph. 16ksamples seems good... I'll try that.
I really hope I can do my work without an amp, because I don't have any amp at home other than the commercial power amp in my main music listening system, and having to work sitting in a mess of cables around that audio rack will really slow me down. Let's see...
I'll certainly keep you posted, and thanks a ton for your help. You can be sure I'll post more questions as I go along.
Tarun
I really hope I can do my work without an amp, because I don't have any amp at home other than the commercial power amp in my main music listening system, and having to work sitting in a mess of cables around that audio rack will really slow me down. Let's see...
I'll certainly keep you posted, and thanks a ton for your help. You can be sure I'll post more questions as I go along.
Tarun
Some impedance measurements
Dear Jbateman, Jomor and others,
I did some quick measurements of two drivers last night, more to test my method than to actually get accurate data. The first is for a Peerless India woofer with Kevlar cone, 6.5in:
Moving forward to estimating T/S parameters, this is what SW reported. Note that the Vas value has been entered by me manually... it was not deduced by SW, since I didn't do the impedance-in-box step:
I did the same for another local (Bolton) 6.5" poly-cone woofer:
I'm not using a power amp for these measurements. Are there any obvious mistakes or problems in my method? Am I getting too few data points on the lower end of the freq scale? Also, when I repeat the same measurements multiple times, I am getting quite different values. In particular, if I switch from 65ksamples to 256ksamples, I once got a big jump in Qts value from 0.38 to 0.27 or so. Why are these values changing like this?
Please give me any comments/tips.
Dear Jbateman, Jomor and others,
I did some quick measurements of two drivers last night, more to test my method than to actually get accurate data. The first is for a Peerless India woofer with Kevlar cone, 6.5in:
An externally hosted image should be here but it was not working when we last tested it.
Moving forward to estimating T/S parameters, this is what SW reported. Note that the Vas value has been entered by me manually... it was not deduced by SW, since I didn't do the impedance-in-box step:
An externally hosted image should be here but it was not working when we last tested it.
I did the same for another local (Bolton) 6.5" poly-cone woofer:
An externally hosted image should be here but it was not working when we last tested it.
An externally hosted image should be here but it was not working when we last tested it.
I'm not using a power amp for these measurements. Are there any obvious mistakes or problems in my method? Am I getting too few data points on the lower end of the freq scale? Also, when I repeat the same measurements multiple times, I am getting quite different values. In particular, if I switch from 65ksamples to 256ksamples, I once got a big jump in Qts value from 0.38 to 0.27 or so. Why are these values changing like this?
Please give me any comments/tips.
Hi again, imho the lowest frequency ripples are from the sound card itself. I used to get strange peaks at very low frequencies when i was using my AC97 onboard sound card. I later switched on a sound baster live which gave me better results. Here is an impedance curve of my B&W DM601 woofer (in the bass reflex box) using a SoundBlaster Live card:
I m not saying its your sound cards' fault, but keep it in mind that this could be the case.
I use 256k sample SIZE and many repeat counts to get more accurate results for impedance measurements. It is important to have a good quality sound card, especially for calculating t/s parameters of drivers with low fs, since most soundcards behave quite strange at very low frequencies. You already know it of course but sometimes it skips our mind, its very important to check the VU meters for non-clipping levels when measuring. Also using a sample RATE above 48KHz gives strange results on some sound cards, so try experimenting with the sample rate too, just to be sure.
An externally hosted image should be here but it was not working when we last tested it.
I m not saying its your sound cards' fault, but keep it in mind that this could be the case.
I use 256k sample SIZE and many repeat counts to get more accurate results for impedance measurements. It is important to have a good quality sound card, especially for calculating t/s parameters of drivers with low fs, since most soundcards behave quite strange at very low frequencies. You already know it of course but sometimes it skips our mind, its very important to check the VU meters for non-clipping levels when measuring. Also using a sample RATE above 48KHz gives strange results on some sound cards, so try experimenting with the sample rate too, just to be sure.
Real smooth. I'll do some more experiments tonight and report.jomor said:
Yes, I've been told to use a lowish sample rate and a significant repeat count. Therefore I'll do that and see what I get. Your graph is certainly much cleaner at the low end than mine.
I'm curious about one thing: how much does the Fs and Qts change due to driver break-in? What percentage change have you seen in these values due to break-in? I'm talking about some serious break-in, for at least 8 hours at fairly high SPLs.
tcpip,
Earlier I suggested that you use a 16k sample size. But that was mainly for the type of measurements you were doing at that time...line level response and impulse tests.
Now that you're doing woofer impedance, a large sample SIZE will give more resolution at the lower frequencies.
And jomor is quite correct that different sound cards can give better or worse results. This is why I recommended using a small power amp...to isolate the sound card from the driver.
As far as getting different results each time, do as jomor said and use several repeat counts for the measurement. This will help average out the variables. Also, for checking woofers, it's a good idea to have the driver clamped or mounted to something sturdy, as vibration during the test can throw off the results.
It sounds like you are just measuring free air impedance, then ESTIMATING PARAMETERS. You will have to eventually learn how to do the ADDED MASS or SEALED BOX method for checking T/S parameters. This will further improve your results.
Typically you would MEASURE FREE AIR, then MEASURE with ADDED MASS, then ESTIMATE parameters. When the estimate is done, there will be a graph that shows the impedance of the driver based on it's calculated parameters. If this graph looks very close to the initial free air graph, then you know the estimated parameters are correct (bottom left chart).
So far it seems you are making progress, which is a good sign.
Earlier I suggested that you use a 16k sample size. But that was mainly for the type of measurements you were doing at that time...line level response and impulse tests.
Now that you're doing woofer impedance, a large sample SIZE will give more resolution at the lower frequencies.
And jomor is quite correct that different sound cards can give better or worse results. This is why I recommended using a small power amp...to isolate the sound card from the driver.
As far as getting different results each time, do as jomor said and use several repeat counts for the measurement. This will help average out the variables. Also, for checking woofers, it's a good idea to have the driver clamped or mounted to something sturdy, as vibration during the test can throw off the results.
It sounds like you are just measuring free air impedance, then ESTIMATING PARAMETERS. You will have to eventually learn how to do the ADDED MASS or SEALED BOX method for checking T/S parameters. This will further improve your results.
Typically you would MEASURE FREE AIR, then MEASURE with ADDED MASS, then ESTIMATE parameters. When the estimate is done, there will be a graph that shows the impedance of the driver based on it's calculated parameters. If this graph looks very close to the initial free air graph, then you know the estimated parameters are correct (bottom left chart).
An externally hosted image should be here but it was not working when we last tested it.
So far it seems you are making progress, which is a good sign.
Yes, last night I spent a lot of time experimenting with sample rate and sample size. I learned that setting a low sample rate puts an upper cutoff point on the freq range SW analyses for. This is good for woofer measurements, I thought, because I can use 12Ksamples/sec and thus limit my upper freq limit to 6K. On the other hand, increasing the sample size to 128ksamples or 256ksamples gives me finer granularity of data points, it seemed.jbateman said:
I also tried playing with repeat count. I was under the impression that SW would average out multiple iterations and give me a smoother curve if I used higher repeat counts. I found that it was the opposite. I got more fuzz and squiggles if I used a repeat count of 5. If I dropped it to 2, I got a somewhat smoother response. Am I right, or am I making a mistake? Anyway, I took a lot of readings (four midbass units, two tweeters) last night this way.
For the tweeters, I needed to go high, so I increased the SR from 12ksamples/sec to 44100ksamples/sec, so that I could plot high. That seemed to work well.
Good suggestion. I didn't do this. Will figure something out next time. Is there any way to build a simple, reusable frame or some such thing which will involve less work than actually cutting out a circular hole in a baffle board for each driver I encounter?
Yes, you're right, I was doing just free-air followed by estimate-params. I was under the impression that this approach will give me all the parameters accurately other than Vas. Am I not right?
Today, I hope to do sealed-box, and then re-run the estimate-params to get the Vas readings. I somehow don't feel comfortable doing the added-mass approach because I don't know how to stick a heavy coin on the dust-cap in a clean and removable way without damaging the speaker. Any suggestion?
I'll post some graphs, both free-air and estimated, for the readings I did last night. As I said, they don't include Vas readings. You can tell me whether I'm getting useful data.
Jomor and you have been total pillars of help and support online, and Angshu has been on the phone with me from Delhi, giving me voice-based support. 🙂 Without all of you, I'd be nowhere.
Some strange free-air readings
I'll show you some graphs of free-air impedance I got, and the estimated plots SW generated for them. In one case, the estimate seems good, in another case, quite questionable.
Here is the first graph, free-air impedance of a midbass driver:
And here is the estimated theoretical curve which SW generated from it:
Now this one is for a tweeter I measured (a Peerless India fabric dome):
And here is the theoretically generated curve from SW for it:
If you see the two curves for the tweeter, they don't seem to match well. The curve shape is different in the region of the resonant frequency, plus the absolute value of impedance in the flat portion is different for the two graphs.
What am I doing wrong? Can you give me some tips?
I'll show you some graphs of free-air impedance I got, and the estimated plots SW generated for them. In one case, the estimate seems good, in another case, quite questionable.
Here is the first graph, free-air impedance of a midbass driver:
An externally hosted image should be here but it was not working when we last tested it.
And here is the estimated theoretical curve which SW generated from it:
An externally hosted image should be here but it was not working when we last tested it.
Now this one is for a tweeter I measured (a Peerless India fabric dome):
An externally hosted image should be here but it was not working when we last tested it.
And here is the theoretically generated curve from SW for it:
An externally hosted image should be here but it was not working when we last tested it.
If you see the two curves for the tweeter, they don't seem to match well. The curve shape is different in the region of the resonant frequency, plus the absolute value of impedance in the flat portion is different for the two graphs.
What am I doing wrong? Can you give me some tips?
You can use the right click - Add.. command on a graph to put both plots on the same graph for direct comparison.
You can't compare in different windows easy, especially when the vertical axis ragnes are different from graph to graph
🙂
You can't compare in different windows easy, especially when the vertical axis ragnes are different from graph to graph
🙂
Thanks. I didn't know this. This will be very useful.jomor said:
I think I should now return to the Unofficial Manual and complete reading it. 🙂
Your woofer measurement looks great.
When you raised the SAMPLE RATE for the tweeter measurement, you should have lowered the SAMPLE SIZE...that's part of the reason there is so much noise on the plot. To be honest, I have never used the ESTIMATE PARAMETERS on a tweeter...there's really not any use for that in the real world.
In the future if you need to remove some of the noise/variations on a plot, you can select TRANSFORM/SMOOTH, then check one the choices offered.
As far as securing the woofer, some people will slide to tables togeher, with a space between them, then use spring clamps to hold the drivers edge to the tops...driver pointed upward.
I prefer just cutting a hole in a large board, bolt driver in place, then hold board firmly to the ground...driver pointed horizontal. This is really only necessary on large drivers with low Fs.
When you raised the SAMPLE RATE for the tweeter measurement, you should have lowered the SAMPLE SIZE...that's part of the reason there is so much noise on the plot. To be honest, I have never used the ESTIMATE PARAMETERS on a tweeter...there's really not any use for that in the real world.
In the future if you need to remove some of the noise/variations on a plot, you can select TRANSFORM/SMOOTH, then check one the choices offered.
As far as securing the woofer, some people will slide to tables togeher, with a space between them, then use spring clamps to hold the drivers edge to the tops...driver pointed upward.
I prefer just cutting a hole in a large board, bolt driver in place, then hold board firmly to the ground...driver pointed horizontal. This is really only necessary on large drivers with low Fs.
I'll remember this, thanks. In fact, I'm tempted to re-do the tweeter plots.jbateman said:
I guess I just wanted the system to confirm the Fs which I was seeing visually. 🙂
Cool. I didn't know about this. Will use it.
Also, in an earlier post, I'd said fuzz seems to increase with repeat count. I was wrong. Last night's experiments seemed to show that repeat count actually helps smooth out things.
Thanks. I'll keep the questions coming. 🙂
I would not smooth an impedance curve, since it might bring errors in the T/S calculation.
Regards.
Claudio
Keep in mind that the smooth function is more for appearance and not really helpful when calculations are involved. The Impedance Curve is necessary to calculate crossover components and smoothing it will llikely introduce error into your calculations. If you are getting funky results, It would be best to check your cables and connections. Smoothing is nice when you want to review your results but is not really helpful if you want accurate calculations.
I would heed Claudio's guidance.
Jay
I would heed Claudio's guidance.
Jay
claudio said:
Point taken. Better to get them right than get them pretty.JMB said:
I think Jbateman too was telling me about smoothing more as a way to tell me "such facilities also exist" than as a message that I should use it for my parameter estimation process.
Thanks, guys. 🙂
- Status
- Not open for further replies.