That test will identify ALL nonlinearities as well as time invariances and it is very diffcult to sort out which is which. The use of multiple impulses for analysing nonlinearities is well know and covered in Schetzen's text on nonlinear systems. It is assumed that the system in time invarient however because if it is not then the time variance gets all mixed up with the nonlinearity.
Finding time variance in a system that is also nonlinear is very tough sorting out one from another.
I have tests that work and I can measures significant differences between systems, its just that there is no data history and nothing to hang the significance of the number on to. Basically its like THD - the numbers are easy to get, its making anything meaningful out of them.
The assumption is that non-linear distortion products ARE audible. All the data that I know of says that they are not, making the assumption wrong and hence the statement meaningless.
Perhaps true, my wife certainly says this. But the impact of short term loud levels on hearing loss is unknown. All of the studies are for long term exposure, not minutes or at most an hour. Worst case is a movie, like GI-Joe, which blasts away for more than two hours at fairly high levels much of the time. But music is not likely to be a problem - unless you go to extremes.
"This speaker is capable of reproducing realistic sound levels which can lead to hearing damage. Hear responsibly."
Yes, it will test all nonlinearities, but significant compression would have to show itself in the magnitude of the impulse peak very clearly. If, for example, there were a 3dB compression of transient peaks, the when scaling up the input at some point you would see the scale factor of the output being only 0.707 of the input. In other words, if at low levels the impulse had a peak value of 1 unit and the high level was 1000 (60 dB higher) input, then a 3dB compression would show an impulse peak of 707. Since the suggested test uses an impulse (not an MLS signal) the energy is concentrated at high frequency and excursion of the driver should remain small (within it's linear range) even at high levels.
However, while it is true that the test may not isolate thermal effects, it would show if any compression existed and/or the limits and degree of such compression. The single impulse will test for linearity with amplitude and the repeated impulse will check for time invariance.
The bottom line is that such testing will reveal compression even if it does not isolate the source precisely. Before worring about eliminating compression first it has to shown to be a problem.
As I said I have a test that I believe works for what I am looking for, and I don't think that what you propose would work as well as I will discuss below.
Yes, you are correct that the loop needs to be closed. But loops always have to start somewhere. As with Horn "harshness" it was apparent that there was a subjective effect and that it was not simple or shown by classic tests. It was said to be nonlinear distortion, but further investigation showed that this was not the case, at any rate, these things usually do start with some ill-defined subjective effect and then proceed to better quantification and eventually the loop gets closed. Its almost closed on Horn "harshness", but it started wide open.
That's what has happen with dynamics. Many here say the standard - its nonlinearity - I don't buy it. I do believe that "dynamics" is a real effect and it's certainly not quantified or understood. Toole makes no mention of it whatever and has never studied it, but I do believe that there is something there. So I did some simple tests awhile back that did show some significant results. These were much like you suggest, but a little different. I used noise over a 20 second playing at very high levels so that significant heating would occur, then calculated the impules response from that at different times. This has the advantage of ignoring SOME nonlinearities - but not all. It showed marked differences in different systems, with some systems going all to he77 and others staying fairly stable.
I am now using random modulated noise and looking not at anything spectral, but cross correlating the signal power with the output power. This gives me time domain data which can be used to calculate the theraml time constants for various parts of the driver. The results for this test are premature, but it is certain that they are not "null". There is useable data - it just needs to be mined.
Well, Mr. Klippel, demonstrates this. When I asked specifically what BL shape it was, he explained it very clearly. Later when I used a pair of speakers with much better linearity such was not audible under normal listening conditions, and listened in a very dead room, I needed no convincing that there was something I had to work on.
Last edited:
The BL and Km curves will also have direct effect. Much more than thermal compression in real everage home listening environment. Just measure your speaker driver temeprature while you are doing your daily listening and log it for a period of time and you get reality, they don't go as high as such with loud noise.
Of course it is possible to prove thermal compression exists with noise testing, but it has no relation with what is realistically happening in the home listening environment. Maybe noise testing is applicable in the PA aplication.
Last edited:
Read this along with some other stuff today on ATCs website. I had read it before but I was looking for something else and it seemed relevant.
"Dynamic Range
Dynamic range is a complex issue. The dynamic range of a driver is primarily a
function of the voice coil operating temperature and total magnetic flux. As the
voice coil temperature rises, its electrical resistance increases, requiring more
power to maintain a given sound pressure level. This effectively compresses the
dynamics of the music, often leading to a blurred or undefined sound as the
speaker heats up. This is an insidious form of distortion because it isn’t initially
present and it increases slowly as the driver heats up. At the same time our
hearing is constantly changing throughout the workday, so it’s difficult to know
just how much power compression is present. Many loudspeakers exhibit a
significant amount of Power Compression, and it’s demonstrable. Just ask the
producer who’s been off on a half day coffee break to listen to a mix that you
worked on with him earlier in the day. If he asks “What did you change?” he’s
probably hearing the effect of power compression. ATC SL drivers address this
issue at the design stage so that there will be minimal driver heating, and hence
minimal power compression.
Equally important to a speaker’s dynamic range is the mechanical integrity of the
frame, magnet, diaphragm, and suspension structure. Finally, the amplifier
driving the loudspeaker must deliver sufficient power over the full dynamic range
of the program material to sustain the required sound level."
http://site.lasvegasproaudio.com/ATC_Engineering_Goals_and_Approaches.pdf
There was some other stuff over my head about eddys in magnets and how this can mess things up as well.
Also one thing I keep coming across over and over is that passive crossovers can add yet another layer of unpredictability in terms of the energy being unevenly converted to heat and changing the filter properties. So I would hope that a test would try to take that variable out of the way by using DSP or at least a line level active crossover.
"Dynamic Range
Dynamic range is a complex issue. The dynamic range of a driver is primarily a
function of the voice coil operating temperature and total magnetic flux. As the
voice coil temperature rises, its electrical resistance increases, requiring more
power to maintain a given sound pressure level. This effectively compresses the
dynamics of the music, often leading to a blurred or undefined sound as the
speaker heats up. This is an insidious form of distortion because it isn’t initially
present and it increases slowly as the driver heats up. At the same time our
hearing is constantly changing throughout the workday, so it’s difficult to know
just how much power compression is present. Many loudspeakers exhibit a
significant amount of Power Compression, and it’s demonstrable. Just ask the
producer who’s been off on a half day coffee break to listen to a mix that you
worked on with him earlier in the day. If he asks “What did you change?” he’s
probably hearing the effect of power compression. ATC SL drivers address this
issue at the design stage so that there will be minimal driver heating, and hence
minimal power compression.
Equally important to a speaker’s dynamic range is the mechanical integrity of the
frame, magnet, diaphragm, and suspension structure. Finally, the amplifier
driving the loudspeaker must deliver sufficient power over the full dynamic range
of the program material to sustain the required sound level."
http://site.lasvegasproaudio.com/ATC_Engineering_Goals_and_Approaches.pdf
There was some other stuff over my head about eddys in magnets and how this can mess things up as well.
Also one thing I keep coming across over and over is that passive crossovers can add yet another layer of unpredictability in terms of the energy being unevenly converted to heat and changing the filter properties. So I would hope that a test would try to take that variable out of the way by using DSP or at least a line level active crossover.
Last edited:
Yes, I have done similar tests but used an MLS pulse. Since MLS has much more low frequency content than an impulse doing the same test, one pule and multiple pulses, show more a result of longer term heating of the VC. We discussed the differences between short term transient heating and more or less long time constant increases in temperature some time ago in another thread. But that isn't "dynamic compression" in my view since it results in an overall reduction in level. Once the VC reaches a steady state temperature there is still the transient effect superimposed on it.
I also did some tests several years ago on a fairly cheap driver in an interesting way. It was a 5" midrange driver with passive 2nd order LP filter. As you know the Q of such a filter would be sensitive to the driver impedance. So by making response measurement of the voltage at the terminals of the speaker with different levels of MLS signals I was able to see changes in the filter response. initially I believed these were VC heating effects and reflected an increase in Re. However, when trying to match the result with modeling I found that increasing Re didn't come close to the observed effect. Some one else in the discussion group where we were talking about this suggested that, since excursions were getting high at the levels I was testing, perhaps the observed effects were not do to Re(T) but rather Le nonlinearity. When I went back to my models I was able to come very close the the measured result by modeling the Le(x) effects. So maybe the thermal effects are not the necessarily the cause of dynamic compression, but Like George suggested, could be related to BL nonlinearity, or even Cms nonlinearity or anything else that reduces or acts in opposition to the motor force.
Yes, we did discuss this long ago and Michael concluded that it was a non-issue, which I never did accept, but it died and I just let it. I never agreed with the conclusions of that thread. As you well know, sometimes you just have to let people walk away thinking the wrong thing since 
gets tiring.
I discount the whole nonlinear thing for many reasons, the biggest being my and many others that have shown that nonlinearities in loudspeakers are just not that audible. On a poor driver design nonlinearity could certainly be an issue, but there is still a dynamics issue when these are cured. I believe the test that you describe makes perfect sense, but a driver with a shorting ring won't exhibite this kind of effect. I only use drivers with shorting rings.
I have no doubt that there could be the two effects in a driver that does not control for nonlinearity, but the nonlinearity issues are, for the most part, a thing of the past and all that is left IMO is the thermal stuff.
As I said, I can show that the effect is there, I just havn't correlated it with subjective impressions. On the other hand all attempts to correlate nonlinearity to subjective impressions have failed - and thats been tried on many ocasions. Many people just don't want to accept that this is true. Even Klippel has tried to do this kind of correlation and failed. What he uses for his audible demonstrations are gross nonlinearites the type of which no competent designer would ever use. Its not a valid demo.
Since the level changes because of temperature and the temperature changes because the signal is dynamic, the level changes are dynamic as well. Thats "dynamic compression" in my book.
gets tiring.I discount the whole nonlinear thing for many reasons, the biggest being my and many others that have shown that nonlinearities in loudspeakers are just not that audible. On a poor driver design nonlinearity could certainly be an issue, but there is still a dynamics issue when these are cured. I believe the test that you describe makes perfect sense, but a driver with a shorting ring won't exhibite this kind of effect. I only use drivers with shorting rings.
I have no doubt that there could be the two effects in a driver that does not control for nonlinearity, but the nonlinearity issues are, for the most part, a thing of the past and all that is left IMO is the thermal stuff.
As I said, I can show that the effect is there, I just havn't correlated it with subjective impressions. On the other hand all attempts to correlate nonlinearity to subjective impressions have failed - and thats been tried on many ocasions. Many people just don't want to accept that this is true. Even Klippel has tried to do this kind of correlation and failed. What he uses for his audible demonstrations are gross nonlinearites the type of which no competent designer would ever use. Its not a valid demo.
Since the level changes because of temperature and the temperature changes because the signal is dynamic, the level changes are dynamic as well. Thats "dynamic compression" in my book.
Last edited:
Geddes if you win you still lose.
Because the best way to compensate for rising resistance of the voice coil is electronically - not by making the driver bigger. unless the voice coil gets vaporized its big enough.
Since you don't seem to have the capability to develop appropriate DSP to accomplish this - why do the hard work of research just so that EAW can implement the solution in their products ?
if i were you i would first evaluate the feasibility of solving the problem before asking whether it exists.
also if the effects are audible then aluminum voice coils should sound better than copper ones of same weight.
Because the best way to compensate for rising resistance of the voice coil is electronically - not by making the driver bigger. unless the voice coil gets vaporized its big enough.
Since you don't seem to have the capability to develop appropriate DSP to accomplish this - why do the hard work of research just so that EAW can implement the solution in their products ?
if i were you i would first evaluate the feasibility of solving the problem before asking whether it exists.
also if the effects are audible then aluminum voice coils should sound better than copper ones of same weight.
Last edited:
I guess I take a more simplistic view of things. I find it usually serves me well. The system is either linear of it isn't. IM and THD are just measurements which show that a system has nonlinearities. They don't say anything about the source or true behavior of these nonlinearities. Thermal effects introduce nonlinearities. If a rapid increase in temperature results in an increase in Re that reduces the current flow through the VC at the peak of the impulse then the motor generated force is curtailed form reaching the expected maximum. The motor force behaves nonlinearly. The driver doesn't reach its expected max acceleration and there is a loss of dynamic response. But by the same token some driver manufactures used shorted turns in the VC to prevent over excursion. Thus, even if the VC temp was perfectly constant, an input that would push the driver toward max excursion would encounter a strong restraining force with result that the net force accelerating the driver is reduced at the peak in very much the same way as a thermal effect.
However, I think there is also another piece of the puzzle that has not been discussed. It is fine to use pro drivers in a high efficiency home system to achieve wide dynamics with moderate amplifier power. Since reduced power is required, the logical argument follows that there is reduced VC heating and therefore Re(T) effects should be small. But is this really the factor? Take those same drivers in a pro system at a live, amplified concert event. My experience has been that such live events seem to preserve the dynamics, but it is hardly because low power is applied. Certainly more power is dissipated in the VC in under these conditions than ever is in a home "hifi" system of moderate efficiency. And there is the simple point that a system with 96dB sensitivity can do with 50 watts what a system with 86dB sensitivity would require 500 watts. 96dB powered by 50 w can reach linear peaks of 113dB. that 86dB system with a typical 100 W amp would struggle to get to 106dB.
{edit} and if you want to eliminate the VC heating problem all together use constant current amplification as opposed to constant voltage.
However, I think there is also another piece of the puzzle that has not been discussed. It is fine to use pro drivers in a high efficiency home system to achieve wide dynamics with moderate amplifier power. Since reduced power is required, the logical argument follows that there is reduced VC heating and therefore Re(T) effects should be small. But is this really the factor? Take those same drivers in a pro system at a live, amplified concert event. My experience has been that such live events seem to preserve the dynamics, but it is hardly because low power is applied. Certainly more power is dissipated in the VC in under these conditions than ever is in a home "hifi" system of moderate efficiency. And there is the simple point that a system with 96dB sensitivity can do with 50 watts what a system with 86dB sensitivity would require 500 watts. 96dB powered by 50 w can reach linear peaks of 113dB. that 86dB system with a typical 100 W amp would struggle to get to 106dB.
{edit} and if you want to eliminate the VC heating problem all together use constant current amplification as opposed to constant voltage.
Last edited:
excellent point which gets consistently overlooked.
absolutely WRONG. they say EVERYTHING.
good job destroying Geddes once again. his premise that an ENORMOUS nonlinearity due to BL curve is irrelevant but a miniscule nonlinearity due to VC heating is paramount seems silly.
i understand Geddes you are trying to say that one distortion is time invariant and one is time dependent - but that is not true as well. IMD is VERY MUCH time dependent. it depends on two tones coinciding in time.
exactly. VC heating is only one of many factors influencing dynamics and probably not a very important one. phase & frequency response as well as directivity and room treatment are probably more important.
oh snap ! i forgot about that 🙂 well, it still favors EAW over Geddes 🙂
- Status
- Not open for further replies.