Low Sensitivity Speakers Compression Issue.

Trytone wrote: "Perhaps it is not that big a problem, Google "thermal compression" loudspeakers and there is a Stereophile article from 2006, complete with measurements."

I remember that article and it dismissed thermal effects as being inconsequential. Since then I’ve had relevant conversations with two of the top researchers in the field, Earl Geddes and Floyd Toole.

I asked Earl to look at the article and he told me what he found to be wrong with the methodology, the details of which I do not remember. But he felt the measurement method was fatally flawed. (In another conversation he specifically mentioned large voice coil diameter as a characteristic which minimizes thermal compression.)

Later I had a conversation with Floyd Toole on the subject of "thermal modulation", which is rapid-onset (short time constant) thermal compression (as opposed to thermal compression based primarily on the loss of magnet strength as the magnet heats up over time). He said that it is a real phenomenon which has not been adequately studied, and went on to give the example of a three-way speaker he had tested whose midrange driver was compressing peaks on normal program material at normal home audio sound pressure levels by 7 dB!

Finally, some years ago a recording engineer shared with me details of measurements that he had been recording for many years on many loudspeakers, looking specifically at how much they compressed a 20 dB peak. It was not unusual for a loudspeaker to compress a 20 dB peak by 3 or 4 dB, but some exhibited no compression (I don’t recall what SPLs he was using for his tests). He found a general correlation between efficiency and lack of compression on the peaks. There were some moderate-efficiency speakers that defied the trend, and these had unusually robust motors which presumably had very high thermal capacity. I’m not posting the details of what he told me here because he asked me to keep the details confidential, nor have I used some of the terminology that he developed, as he intends to write a paper based on his findings at some point.

Most of my work is in the prosound world, and there the awareness of thermal compression is alive and well. If a prosound speaker is pushed hard, it compresses.

My point being, thermal compression and thermal modulation (short-term thermal compression) are real things, but unfortunately that information is not widely known in the home audio world. It’s almost like a dirty little secret that speaker manufacturers don’t want to talk about. No manufacturer wants to admit that his megabuck exotic-diaphragm tweeters have thermal compression issues that a thirty dollar compression driver doesn’t.

Duke

My point being, thermal compression and thermal modulation (short-term thermal compression) are real things, but unfortunately that information is not widely known in the home audio world. It’s almost like a dirty little secret that speaker manufacturers don’t want to talk about.

an 8% increase in resistance corresponds to a temperature increase of just under 20°C (36°F), and corresponds to a thermal compression of just 0.67dB

Most of my work is in the prosound world, and there the awareness of thermal compression is alive and well.

@kenjit wrote: "the stereophile article says... [thermal compression] is negligible."

Here is the problem with that article: It looked at temperature changes averaged out over time, rather that looking at the instantaneous spike in temperature (and thus voice coil resistance) at the onset of a transient, which is what compresses transients.

From the author’s description of how the measurements were made:

"For the measurements, the DC voltage across the voice-coil was recorded at intervals of 10 milliseconds with a Pico Technology ADC-212 virtual instrument oscilloscope in data-acquisition mode, using Pico’s bundled PicoLog software. The data were then smoothed using a five-second moving average calculation performed in a software utility written for the purpose."

In other words, he was taking the voice coil’s temperature at 10 millisecond intervals, and then smoothing the data across a five-second moving average. So he was NOT measuring the instantaneous spikes in temperature caused by transients, but rather the cumulative effect on voice coil temperature of music spread out over time.

This totally misses the target, because it is those INSTANTANEOUS temperature spikes which are responsible for thermal modulation (compression of transient peaks).

Earl Geddes on the subject:

"The voice coil will heat instantaneously as current is sent through it - there is no time constant. Its mean temperature has time constants but not its instantaneous temperature."

The Stereophile article looked at the MEAN TEMPERATURE, and found the resulting calculated thermal effect to be negligible when averaged out over time.

But the thermal effect which compresses instantaneous peaks would only show up in an examination of the INSTANTANEOUS TEMPERATURE (which might not even be possible to measure precisely).

Earl continues:

"How much resistance change is there from this instantaneous temperature change you should ask. It depends on the amount of metal... Thus the single factor for the instantaneous resistance modulation is the mass of the voice coil conductor."

In other words the brute force approach - voice coil thermal mass - is what matters most (for a given wattage input) if the goal is minimizing thermal modulation, as others have said in this thread.

Earl again:

"The longer term rise in the voice coil temperature tends to be on the order of several milliseconds [which is what the Stereophile article was looking at]. The instantaneous modulation is what I am curious about as it has not really been looked at in any detail that I know of."

So the Stereophile article was not looking at the thing that really matters.

Earl is a leading peer-reviewed expert on the subject of distortion and distortion perception, and he invented a voice coil which has negligible increase in resistance as its temperature rises. So he is a genuine expert in this field.

Duke

"If memory serves, the late, great, Siegfried Linkwitz, had some very interesting oscilloscipe traces, using tone bursts, and it was in fact quite remarkable how fast thermal compression set in."

That makes sense.

"Another thing, there’s other types of compression, which results in easy to measure frequency response changes."

Yes, the real world is a complicated place. Different drivers in a system will often have different thermal characteristics, such that the tonal balance changes with level. This change in tonal balance can be small or large.

Is there a correlation between loudspeaker thermal characteristics and perceived sound quality?

Again quoting Earl Geddes:

"I started measuring thermal responses in loudspeakers and found that this did seem to correlate better to perceived sound quality than frequency response[!] or THD."

To the best of my knowledge Toole did not include thermal characteristics in his study of the correlation between objective measurements and subjective preference, so that is a very intriguing statement.

Duke

IF thermal modulation (compression of peaks due to the instantaneous voice coil heating and resulting resistance at the onset of a transient) is a real thing, THEN there is an interesting implication about amplification choice:

What if we used an amplifier that did not deliver proportionally LESS power as the voice coil resistance (and therefore load impedance) rises? This would make the loudspeaker/amp combination relatively immune to thermal modulation effects.

Zero (or very low) global negative feedback amplifiers (typically tube amps but sometimes solid state) tend to deliver approximately constant power, rather than approximately constant voltage, and therefore are far less susceptible to thermal modulation effects.

Here is a link to a paper on the subject:

http://www.atma-sphere.com/Resources/Paradigms_in_Amplifier_Design.php

Disclaimer: I’m a dealer for the amps made by the paper’s author, Ralph Karsten.

Duke

Erik wrote: "By the way, I’ve never seen anything about a single peak causing thermal compression, or that only the peaks are then compressed, or that everything is fine BUT the peaks due to thermals."

If, as Earl Geddes says, the heating is instantaneous (even if we can’t measure it instantaneously), then it makes sense that the peaks would be affected the most.  And I never said only the peaks are compressed, but if there is compression, isn't that where it would be most noticeable?

And if, as Earl and Floyd Toole both say, thermal modulation has not been looked at in detail yet, it’s not surprising that you’ve never seen anything about it. And unless some big company with deep pockets decides it’s worth studying in detail, that’s unlikely to change much. Well perhaps my recording engineer friend will get his paper written and published, as apparently he has been documenting compression of peaks for decades (again, I have refrained from posting the details he shared with me because he considered that to be proprietary information).

Duke

"Still, this is not the only thing to look for in a speaker..."

Totally agree.

Duke

There is not trade off between high sensitivity and natural tone.

Yep, materials do not heat up instantly when power is dissipated.

There is a lot of mis conception and misunderstanding about dynamic range

georgehifi,

Horn colouration and efficient drivers used used outside of their linear range because of low order xovers, are "not natural" in their "tone".

Says you...again.

"Cheers", John

Horn colouration and efficient drivers used used outside of their linear range because of low order xovers, are "not natural" in their "tone".

Horns can be made with very little coloration and certainly get the nod in terms of efficiency. If you are in love with low powered amps there is no better solution. However, I don't like them. I do not like them because they are point source radiators and do not broadcast a realistic sound stage. I think that most well designed drivers run within their established parameters are not going to compress anything. 

do not broadcast a realistic sound stage.

Tell that to owners of the Fairchild compressors like the model 670!! It uses a light bulb as part of its compression scheme. You already know how light the wire is in a voice coil. That it could heat up quickly enough to actually compress a transient peak should come as no surprise. Of course, a lot depends on what is meant by 'instantly' :)