Speakers distortion

[Krunok]

Did a little experimenting with sweeps and RTA tones for distortion measurement. I'm not sure we can get there from here.

With both I found moving the Umik 1 even a few inches could alter the distortion significantly. Sometimes 20 db for RTA use. Just for instance using a 780 hz tone and RTA several feet from speakers, I could get .134 % 2nd and .454 3rd, move it six inches resulting in 2.55% 2nd and 6.34% 3rd. Going 6 inches further gave a result of .444% 2nd and .874% 3rd. Obviously at least two are wrong for what the speaker is doing and probably all of them. Peaks and dips for standing waves are a problem.

Similar problems though of smaller magnitude with sweep based distortion measures. And whether you get a lower resulting distortion with long, short or medium length sweeps depended upon where in the room you put the microphone. Getting close to the speaker also didn't seem to help.

Or am I doing this all wrong?

Pretty much the same thing happens when you measure frequency response. For that reason I think maybe it would make sense to move mic while doing RTA as it needs to be done when measuring freq response.

 

[Krunok]

Getting close should give an improvement, but perhaps only significantly so if the speaker is far away from any reflective surfaces and the room is relatively dead. For example, if there’s a 1:2 ratio between the the path lengths of the direct sound to reflected sound, that’s just a 6dB difference, and that’s not even counting subsequent reflections, which will reduce the difference even further. You can imagine that if you hope to benefit from getting close, that ratio will need to be very very low, indeed well within the room’s critical distance at all frequencies of interest, if the measurement isn’t to be dominated by room effects. What makes this probably uniquely more difficult than in-room amplitude response measurement is the comb filter effect. If a harmonic falls on a “tooth” of the comb, it just won’t show up in the measurement, and smoothing can’t fix this. Averaging a number of measurements might help, but it will still very hard to know your measurements reflect reality.

If you can take the speaker outside or have a very large space inside somewhere and can then get the mic quite close there, you’ll get a more reliable result I reckon.

I agree.

Otherwise you’re really just stabbing in the dark IMHO.

I will try to use the same procedure that worked when measuring frequency response. If it wouldn't work I will have to agree with you on that as well.

 

[RayDunzl]

Here it is..

Thank you.

My automated DRC uses a single-point measure.

Seems like I read that paper a couple of years ago.

As I remember, the moving mic averages and the careful single point measure here were similar enough that I discarded any worry about it.

My theory is to apply reasonable corrections to a single point (where I will be most attentively listening now and then) and let the chips fall where they may elsewhere, because they already do, and they will anyway, whatever little adjustments you make.

 

[Krunok]

As I remember, the moving mic averages and the careful single point measure here were similar enough that I discarded any worry about it.

Well, it may work in your room but I have to make 7-9 sweeps in the same chunk of space (is 3D window correct term) where I made RTA MMM (moving mic method) and average them to get simialr resulst with RTA MMM, but a single sweep never worked for me. I still found RTA MMM more reliable than averaged sweep which is no wonder as RTA MMM contains average of more than 70 samples.

Remember that Dirac Live also works with averaged sweep measurement taken at 9 points.

 

[RayDunzl]

Remember that Dirac Live also works with averaged sweep measurement taken at 9 points.

Yes, I tried it (trial) and didn't like it.

Particularly the expense (at least, at the time, which was four or five years ago now (wow, that long?).

Secondarily (or more likely an equivalent reason), that it used the PC as the engine of correction, which left correction for TV, Radio, and CD player out (without further figuring out to get them into the PC). I don't source from the PC, except for test tones, and the occasional playback of something.

 

[Krunok]

Yes, I tried it (trial) and didn't like it.

Particularly the expense (at least, at the time, which was four or five years ago now (wow, that long?).

Secondarily (or more likely an equivalent reason), that it used the PC as the engine of correction, which left correction for TV, Radio, and CD player out (without further figuring out to get them into the PC). I don't source from the PC, except for test tones, and the occasional playback of something.

Sure, I only mentioned it in the context of multipoint sweeps. It is hard to expect that automation can do a better job than a person who knows what he is doing, but it probably can help folks who don't want to trouble themselves with earning how to do it on their own.

Sonarworks room EQ software for which @SIY posted his nice review is also using multi point sweeps.

 

[Krunok]

In the case of loudspeaker measurements, I applied as much rigour as I could when I set up my 801s for on-axis pseudo-anechoic frequency response, as I wanted to achieve a +-1dB variation above 200Hz, which is the lower limit of my measuring capability. I'm reasonably satisfied I did that correctly, but much less confident in the distortion figures I measured as they were totally dependent on room gain as a 3dB boost or cut at any one frequency will double or halve the measured distortion , and 3dB variations in room gain are fairly normal, few domestic rooms have a smoother response than that, but nevertheless, my measurements indicate that there's nothing wrong with what I have.

Are you saing that speakers in-room response is simply not lienar enough to get consistent distortion readings? Not to mention when you add noise..

Well, when you look at it that way that obviously may indeed be the case.

 

[Cosmik]

Any thoughts on why the JBL M2 is only a 2-way? Would that design benefit from a 3rd driver? It already has the dsp.

Not a speaker I've ever looked at, but I think there was a thread on here somewhere.

As far as I can tell, it's a professional speaker designed to go very loud, so it uses a compression driver and waveguide for high output level and to control dispersion over a wide frequency range.

However, I believe it is generally considered that the direct radiator gives less coloured sound and is practical in less demanding situations e.g. domestic listening. The direct radiator will benefit from more than two ways, and DSP makes it trivial to do. (That's all I'm saying on it, 'cos I know some horn enthusiasts will take exception to that! ).

 

[Cosmik]

Am I missing anyone mentioning the very speaker-specific distortion mechanism known as Doppler?

The major source of frequency intermodulation distortion in audio systems—the loudspeaker

(Are your distortion measurements going to be taking into account intermodulation distortion at all, in fact?)

With a pure tone, the Doppler effect simply adds harmonic distortion: principally, relatively innocuous second harmonic at practicable diaphragm excursions. But with the complex signals of music, where the diaphragm must reproduce lower and higher frequencies simultaneously, the consequences are more serious. In effect, the higher tones are frequency-modulated by the lower tones, giving rise to intermodulation sidebands.

Distortion of the flute was gross at 10mm peak diaphragm displacement and not in the least bit euphonic. On the contrary, Doppler made the sound as harsh as you might expect of a distortion mechanism that introduces intermodulation products. At 3.16mm peak displacement (below Fryer's suggested detectability threshold) the distortion level was obviously lower but still clearly audible; and even at 1mm it could still be heard affecting the flute's timbre and adding "edge."

Everyone who uses a two-way speaker (me included) can take heart from the fact that most music signals are less revealing of Doppler distortion than this special brew. But these findings undermine the view, widely accepted in the last two decades, that Doppler distortion in loudspeakers is not something we should trouble about.

Doppler distortion will be highest with drivers that cover wide frequency ranges with high displacement. Going three-way rather than two- will automatically reduce the problem.

A three-way solution is potentially even better. Three-way speakers bring new design challenges, of course, in particular the need to achieve another perceptually seamless handover between drivers. But from the Doppler perspective, having a crossover for the bass driver at 400Hz or 500Hz is, unquestionably, better.

The 'seamless' bit is the thing that DSP makes trivial.

 

[sergeauckland]

Are you saing that speakers in-room response is simply not lienar enough to get consistent distortion readings? Not to mention when you add noise..

Well, when you look at it that way that obviously may indeed be the case.

Yes, the in-room response will have sufficient variation as to make accurate measurements of distortion difficult. For example, if the room response at 1kHz is 3dB different to 2kHz and/or 3kHz, then the second and third harmonic measurements will be wrong by that amount. It may not matter, if the distortion is low, even +-3dB is still low, but it's not accurate.

S

 

[Krunok]

Yes, the in-room response will have sufficient variation as to make accurate measurements of distortion difficult. For example, if the room response at 1kHz is 3dB different to 2kHz and/or 3kHz, then the second and third harmonic measurements will be wrong by that amount. It may not matter, if the distortion is low, even +-3dB is still low, but it's not accurate.

S

Couldn't that be corrected in a similar way it is done with mic calibration file? If we would have the possibility to enter speakers room response REW would be able to take those variations into account. What do you think?

One more thing: speakers non linear response is one of the very causes of the harmonic distortion. For that reason one might think that your example doesn't really demonstrate non accurate measurement but it's indeed measuring the very cause of the distortion.

 

[sergeauckland]

Couldn't that be corrected in a similar way it is done with mic calibration file? If we would have the possibility to enter speakers room response REW would be able to take those variations into account. What do you think?

One more thing: speakers non linear response is one of the very causes of the harmonic distortion. For that reason one might think that your example doesn't really demonstrate non accurate measurement but it's indeed measuring the very cause of the distortion.

Theoretically, yes you could enter the room's response as a calibration file, but as a room's response changes with any change in furniture, loudspeaker positioning, microphone positioning and possibly other factors I haven't thought of, it would have to be measured each time a measurement is done. In my view it's unnecessary, and better is to use either a near-field measurement where the room's effect is minimised*, or use REW's distortion sweep, which not being a fixed frequency, won't set off room reflections in the same way. The real problem I think with distortion measurements at home is noise limited. If one is doing a measurement at, say, 80dBSPL, and we want to measure accurately distortion of 0.1% (-60dB) then the room has to be quiet enough so the 20dB SPL of distortion isn't swamped by noise. I know that using narrow filtering as a FFT does eliminates a lot of noise outside each band, but still, getting down to -60dB in a normal domestic environment is a challenge. There's a tradeoff between the number of FFT 'buckets', the speed of the sliding tone and the lowest frequency that can be reliably measured, so maintaining accuracy is not easy. I prefer just to be happy that the distortion I measure is low enough not to be of concern even if I don't actually know what the number really is, just that it's not very high.

* If the loudspeaker is set up so that the SPL at 1m is, say, 80dB, then at a few mm from the driver the SPL will be a lot higher, so reducing the effect of the room's background noise and reflections.

S.

 

[Krunok]

Theoretically, yes you could enter the room's response as a calibration file, but as a room's response changes with any change in furniture, loudspeaker positioning, microphone positioning and possibly other factors I haven't thought of, it would have to be measured each time a measurement is done. In my view it's unnecessary, and better is to use either a near-field measurement where the room's effect is minimised*, or use REW's distortion sweep, which not being a fixed frequency, won't set off room reflections in the same way. The real problem I think with distortion measurements at home is noise limited. If one is doing a measurement at, say, 80dBSPL, and we want to measure accurately distortion of 0.1% (-60dB) then the room has to be quiet enough so the 20dB SPL of distortion isn't swamped by noise. I know that using narrow filtering as a FFT does eliminates a lot of noise outside each band, but still, getting down to -60dB in a normal domestic environment is a challenge. There's a tradeoff between the number of FFT 'buckets', the speed of the sliding tone and the lowest frequency that can be reliably measured, so maintaining accuracy is not easy. I prefer just to be happy that the distortion I measure is low enough not to be of concern even if I don't actually know what the number really is, just that it's not very high.

* If the loudspeaker is set up so that the SPL at 1m is, say, 80dB, then at a few mm from the driver the SPL will be a lot higher, so reducing the effect of the room's background noise and reflections.

S.

But in-room response affects distortion, correct? if you measure near field you will probably get a better picture of the distortion your speaker would have in anechoic environment but that is not really what you are listening when sitting in your chair at LP, right?

 

[andreasmaaan]

But in-room response affects disortion, correct? if you measure near field you will probably get a better picture of the distortion your speaker would have in anechoic environment but that is not really what you are listening when sitting in your chair at LP, right?

I'm not sure this has been investigated in respect of distortion specifically. But my hunch is that, much like when we "hear" a speaker's frequencies response our ears separate the direct sound from the reflections quite well (at least above a certain frequency), we "hear" a speaker's distortion similarly.

 

[Krunok]

This is what I got when measuring distortion the very same way I'm measuring frequency response:



There was 70-80 samples per each measurement and as you can see the numbers are much more consistent.

I was moving mic in a pattern similar to Diac Live points, 50cm left and right from the center position, 15cm up and down and 25cm in front and backwards.

 

[Krunok]

I'm not sure this has been investigated in respect of distortion specifically.

I was referring to this: "For example, if the room response at 1kHz is 3dB different to 2kHz and/or 3kHz, then the second and third harmonic measurements will be wrong by that amount. "

I agree with that, but I would say that third harmonic measurement will be affected by that ammount, not wrong.

 

[sergeauckland]

But in-room response affects distortion, correct? if you measure near field you will probably get a better picture of the distortion your speaker would have in anechoic environment but that is not really what you are listening when sitting in your chair at LP, right?

The room's response will affect what you measure (and of course what you hear) but won't change the loudspeaker's intrinsic distortion. As with any measurement, the answer you get depends on the question you ask. Do you want to know what distortion your loudspeaker generates or what distortion exists in the room? I'm more interested in the loudspeaker, others may be more interested in the in-room result.

S.

 

[andreasmaaan]

The room's response will affect what you measure (and of course what you hear) but won't change the loudspeaker's intrinsic distortion. As with any measurement, the answer you get depends on the question you ask. Do you want to know what distortion your loudspeaker generates or what distortion exists in the room? I'm more interested in the loudspeaker, others may be more interested in the in-room result.

S.

Sure, I'm also more interested in the loudspeaker. But I was suggesting that what we "hear" may not be as simple as the (steady-state) distortion of the loudspeaker + room at the listening position, either.

 

[Krunok]

The room's response will affect what you measure (and of course what you hear) but won't change the loudspeaker's intrinsic distortion. As with any measurement, the answer you get depends on the question you ask. Do you want to know what distortion your loudspeaker generates or what distortion exists in the room? I'm more interested in the loudspeaker, others may be more interested in the in-room result.

S.

Ok, fair point!

How do new numbers look to you?
They obvously aim to represent what distortion exists in the room, in a same manner frequency response measurement taken this way aims to represent response at LP.

 

[andreasmaaan]

I was referring to this: "For example, if the room response at 1kHz is 3dB different to 2kHz and/or 3kHz, then the second and third harmonic measurements will be wrong by that amount. "

I agree with that, but I would say that third harmonic measurement will be affected by that ammount, not wrong.

Maybe. But you'll still (more or less) perceive the direct sound as distinct from the sound field, at least in terms of spectral balance. I suspect you may perceive the direct sound quite distinctly from the sound field in terms of distortion, too.