Good way to measure transcient distorsion ??
Hello,
As far as I don't have Praxis or Clio, I figured I could measure transcients linear distorsion just with a function generator and a soundcard scope.
I generate a burst that stops and go suddenly from -0 to -144dB, and observe the response at the microhpone on the scope.
It gives me the kind of result attached.
Do you think it is a good way to compare drivers performances ?
Hello,
As far as I don't have Praxis or Clio, I figured I could measure transcients linear distorsion just with a function generator and a soundcard scope.
I generate a burst that stops and go suddenly from -0 to -144dB, and observe the response at the microhpone on the scope.
It gives me the kind of result attached.
Do you think it is a good way to compare drivers performances ?
Attachments
These may interest you:
http://www.markk.claub.net/linear_distortion.htm
http://www.linkwitzlab.com/x-mid_dist.htm
http://www.markk.claub.net/linear_distortion.htm
http://www.linkwitzlab.com/x-mid_dist.htm
Thanks for your answers !
I've already read those two very interesting articles, that's why I'm trying to reproduce this kind of measurement myself.
Unfortunately, my measurement software (Loudspeaker lab 3) has no function to draw the "enveloppe" for the cosine burst, so I'm restricted to using oscilloscope.
Do you know a soft that could do that ??
I've already read those two very interesting articles, that's why I'm trying to reproduce this kind of measurement myself.
Unfortunately, my measurement software (Loudspeaker lab 3) has no function to draw the "enveloppe" for the cosine burst, so I'm restricted to using oscilloscope.
Do you know a soft that could do that ??
You could check out Arta, it does cumulative spectral decay:
http://www.fesb.hr/~mateljan/arta/links.htm
It's not precisely the same as what you're asking for, but I think it will get you the information you want in a form that's useful.
http://www.fesb.hr/~mateljan/arta/links.htm
It's not precisely the same as what you're asking for, but I think it will get you the information you want in a form that's useful.
No, I don't think it is a good method.
The reason is that the behaviour of any linear device is completely described by the impulse response, and if a spectrum is derived from the impulse response (via FFT) the display is probably the most relevant for how we perceive it.
The burst response will contain similar information, but is far more complicated to interpret. The risk is that the system gets optimsed towards something that is very visible for the eye, but not for the ear.
I see methods like these pop up every now and then, and people seem to think that they measure something that the impulse response or frequency response fail to detect, but that is hardly the case. They just provide the same information, more or less obscured.
As for nonlinearities; they will make a change in the response, but at normal levels they will be hard to spot. But this is a differnt issue.
Just my 2 öre...
The reason is that the behaviour of any linear device is completely described by the impulse response, and if a spectrum is derived from the impulse response (via FFT) the display is probably the most relevant for how we perceive it.
The burst response will contain similar information, but is far more complicated to interpret. The risk is that the system gets optimsed towards something that is very visible for the eye, but not for the ear.
I see methods like these pop up every now and then, and people seem to think that they measure something that the impulse response or frequency response fail to detect, but that is hardly the case. They just provide the same information, more or less obscured.
As for nonlinearities; they will make a change in the response, but at normal levels they will be hard to spot. But this is a differnt issue.
Just my 2 öre...
OK, thanks very much for your enlightement.
What kind of graph would you recommend me to evaluate linear distorsion performance (apart from frequency response) ?
Is the impulse response an intepretable graph "as-is" ?
Also, I don't understand how a whole frequency response can be guessed from just one spike. Seems like magic to me lol. Could you explain me ?
Thx !
What kind of graph would you recommend me to evaluate linear distorsion performance (apart from frequency response) ?
Is the impulse response an intepretable graph "as-is" ?
Also, I don't understand how a whole frequency response can be guessed from just one spike. Seems like magic to me lol. Could you explain me ?
Thx !
I think that a large part of the reason that normal music systems do such a poor job recreating the sounds of drums is that drums are just so LOUD compared to what the average system is capable of.
I have a set of speakers and an amplifier that will just barely get me over 100 db. It sounds reasonably good, but there's no way at all that I'd make the mistake that there's a real percussionist in my living room. Somebody really whistling or speaking in a low voice, yeah, but drums? No way.
Drums, when played loudly, can easily peak over 110 or 120 decibels. Most domestic audio equipment just can't produce that kind of peak. If Most recorded music has some degree of compression to bring the quiet sections within 20 to 30 db of the peak, so even if you had a capable sound system, it just wouldn't sound the same.
In short, I don't think that there's anything major missing from the analysis- not anything that would do the DIYer much good. If a driver has good linear and non linear distortion up to its output limit, I think that's all we can really ask for.
I have a set of speakers and an amplifier that will just barely get me over 100 db. It sounds reasonably good, but there's no way at all that I'd make the mistake that there's a real percussionist in my living room. Somebody really whistling or speaking in a low voice, yeah, but drums? No way.
Drums, when played loudly, can easily peak over 110 or 120 decibels. Most domestic audio equipment just can't produce that kind of peak. If Most recorded music has some degree of compression to bring the quiet sections within 20 to 30 db of the peak, so even if you had a capable sound system, it just wouldn't sound the same.
In short, I don't think that there's anything major missing from the analysis- not anything that would do the DIYer much good. If a driver has good linear and non linear distortion up to its output limit, I think that's all we can really ask for.
hi Joe - I can get some tiny semblance of drums using original or modified Karlson as cone excursion on the strongest passages of say Steve Clarke's "solo Drums" cd will give about 1/8" p-p cone excursion for clip lite on my amp (~400W 8 ohm) - a B&C 15" in 3.6 cubic foot reflex will move quite a bit further, and has small signal maybe 6-8dB less. My Khorn with 150-200W tube makes noise but somewhat muddy in articulation - could be its nature.
a Ciare CH250 at 7 inches can't play a drum and its moving lots more back and forth than my midbass couplers with lots of power - lol
seems to me Karlson type usally reduce sideband with 2-tone tests by around 10dB - - sine at 50Hz (prettly close to F10 outdoors) on an 8 cubic foot coupler with 103g mms 18 ran ~2% HD
Tom Danley's Unity horn is compact and should have low distortion from 300 up
Beta 15CX - clean but weak motor - regular 1951 degisn Karlson
http://img486.imageshack.us/img486/5083/dop151ws9.jpg
little midbass coupler built by Acoustic - might be fun with a tapped horn sub and xover ~80 - its loud with EV15L
http://img373.imageshack.us/img373/8564/115bk20vrmswjs9.jpg
coupler with 18 which seems to play pretty well
http://img483.imageshack.us/img483/2388/nowayce7.jpg
a Ciare CH250 at 7 inches can't play a drum and its moving lots more back and forth than my midbass couplers with lots of power - lol
seems to me Karlson type usally reduce sideband with 2-tone tests by around 10dB - - sine at 50Hz (prettly close to F10 outdoors) on an 8 cubic foot coupler with 103g mms 18 ran ~2% HD
Tom Danley's Unity horn is compact and should have low distortion from 300 up
Beta 15CX - clean but weak motor - regular 1951 degisn Karlson
http://img486.imageshack.us/img486/5083/dop151ws9.jpg
little midbass coupler built by Acoustic - might be fun with a tapped horn sub and xover ~80 - its loud with EV15L
http://img373.imageshack.us/img373/8564/115bk20vrmswjs9.jpg
coupler with 18 which seems to play pretty well
http://img483.imageshack.us/img483/2388/nowayce7.jpg
I think that horns stand a decent chance at reproducing drums in the home, since their efficiency (if matched with decent power handling) will allow them to reach high SPLs.
Beyond just playing loudly, drums need accurate reproduction with low linear and nonlinear distortion. They provide signals that are a bit like impulses and noise bursts, and the period after the initial attack can lay bare any lingering coloration.
I had some JBL TR225 speakers (about 95 db/w/m) with 400 watts behind them, and they really didn't provide a convincing illusion of drums. I had them side by side with my acoustic drum set, and they were capable of amplifying guitars or vocals loud enough to keep up with my playing- but drums on recorded music weren't convincing. That was long before I got my SPL meter, so maybe that had something to do with it- maybe they weren't as loud as I thought. More likely, though, the cheap horns, compression drivers, and paper cone woofers, just didn't have the cleanest response. The woofers barely had crossovers on them, and probably rang a bit. The who set of speakers cost less than some really premium compression drivers... I never measured, but I don't think we were getting the best performance out of them.
Interestingly enough, I heard the Beolab 5, and it had a certain something that got drums a bit closer than most speakers i've heard. It could have been the in-room reverberation reminded me of how drums sound live- because with their compliment of drivers I really don't think they're breaking any SPL records.
I'd be far more interested to hear how something like the Orion does with drums.
Geez, this seems to have strayed fairly far off of the original topic. I hope it was at least interesting for someone.
Beyond just playing loudly, drums need accurate reproduction with low linear and nonlinear distortion. They provide signals that are a bit like impulses and noise bursts, and the period after the initial attack can lay bare any lingering coloration.
I had some JBL TR225 speakers (about 95 db/w/m) with 400 watts behind them, and they really didn't provide a convincing illusion of drums. I had them side by side with my acoustic drum set, and they were capable of amplifying guitars or vocals loud enough to keep up with my playing- but drums on recorded music weren't convincing. That was long before I got my SPL meter, so maybe that had something to do with it- maybe they weren't as loud as I thought. More likely, though, the cheap horns, compression drivers, and paper cone woofers, just didn't have the cleanest response. The woofers barely had crossovers on them, and probably rang a bit. The who set of speakers cost less than some really premium compression drivers... I never measured, but I don't think we were getting the best performance out of them.
Interestingly enough, I heard the Beolab 5, and it had a certain something that got drums a bit closer than most speakers i've heard. It could have been the in-room reverberation reminded me of how drums sound live- because with their compliment of drivers I really don't think they're breaking any SPL records.
I'd be far more interested to hear how something like the Orion does with drums.
Geez, this seems to have strayed fairly far off of the original topic.
I hope it was at least interesting for someone.Hmmm what you say about drums sounds very interesting to me.
When I wrote down the requirements for my new system, I wanted it to be able of 110dB/1m SPL from 40Hz and up.
I had in mind some safety margin for excursion (and distorsion), power handling during "home parties", but I didn't think about dynamics.
But isn't there a measurement that shows dynamic limitations ? Apart from mathematical limits of course.
When I wrote down the requirements for my new system, I wanted it to be able of 110dB/1m SPL from 40Hz and up.
I had in mind some safety margin for excursion (and distorsion), power handling during "home parties", but I didn't think about dynamics.
But isn't there a measurement that shows dynamic limitations ? Apart from mathematical limits of course.
freddi said:
I think you are asking the wrong question here:
-You should ask whats wrong with the speaker that does make a commercial CD to sound like a real drum, and not ask why they dont...
You have a lot of (both intentional and unintentional) compression and limiting going on in the recording chain, all the way from the microphones to the finished CD. (And thats a good thing for music aimed at the average consumer with their vastly different speakers)
Try a couple of good condensermicrophones on a good recordingdesk instead, and connect your speakers directly to that, then they should sound real.
youyoung21147 said:
I would recommend the frequency response, and possibly the phase response in cases when it is important. For visual interpretation, that is. If the linear properties of the sytem is what is important, there is no need for anything else.
The impulse response in itself is largely as bad at the bursts, for direct evaluation "by eye".
youyoung21147 said:Also, I don't understand how a whole frequency response can be guessed from just one spike. Seems like magic to me lol. Could you explain me ?
Thx !
Yes, it might seem as magic, but since the impulse contains all frequencies, all that is needed is a good estimate of the impulse response and a fourier transform.
The "good estimate" can be measured in several ways, and one of the worst ways is actually by exciting the system with an impulse (because it is hard to put enough enery in an impulse, thus SNR will be too poor). It is better to excite the system with some other signal, like a sweep or an MLS sequence and let the computer calculate the impulse response from that.
Svante : you seem to know way more than me on the subject.
But from what I can see, and according to my experience, some drivers which have a very very linear frequency response may very well have a poor CSD and a poor response on tone bursts (slow decay), while some not very linear drivers have fast decay (which is rare).
Often this is traduced by a lack of detail and some "roundness". Is there an explanation to this apart from energy storage ?
When I capture impulse response and compute CSDs, I always use MLS which allows gating and gives cleaner results (as you said, better SNR).
454Casull : oh yes, unfortunately I will never be able to get 110dB at the listening point (except if the room rings like a bell lol). But I figured 110dB was already a correct value which does not imply the use of public adress drivers for the tweeter and midrange, which don't sound quite as clean.
But from what I can see, and according to my experience, some drivers which have a very very linear frequency response may very well have a poor CSD and a poor response on tone bursts (slow decay), while some not very linear drivers have fast decay (which is rare).
Often this is traduced by a lack of detail and some "roundness". Is there an explanation to this apart from energy storage ?
When I capture impulse response and compute CSDs, I always use MLS which allows gating and gives cleaner results (as you said, better SNR).
454Casull : oh yes, unfortunately I will never be able to get 110dB at the listening point (except if the room rings like a bell lol). But I figured 110dB was already a correct value which does not imply the use of public adress drivers for the tweeter and midrange, which don't sound quite as clean.
youyoung21147 said:
Ok, so CSD is one of the methods that i half like
They are in essence frequency response measurements and might actually help in visualizing resonant behaviours.
On the other hand, these resonances are visible in a simple frequency response measurement too, given that it is done with a sufficiently narrow band analysis. Frequency response measurement software today offer too many smoothing options IMO, and it is very easy to obscure resonances like those you describe by using too much smoothing. In particular if you "want to sell that driver".
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.