comments on thermal distortion?

I'm interested in hearing what others have to say about this site: http://peufeu.free.fr/audio/, and the thermal "memory" distortion ideas Pierre has presented there.

Personally, I think he's done an excellent job with the simulations, although I don't know if I like the solutions he presents as much. In particular, the cascoded CFP diff pair seems rather complex. At some point I might try it, perhaps with the cascode FET repositioned so the gate is connected to the main transistor's emitter instead of the base, and so that the FET lies between the two BJTs that make up the CFP... stability could be a serious issue here though.

When you consider the tiny thermal mass of a transistor's guts, and the time it will take any transient heat buildup to dissipate out through the case, I don't find it the least bit surprising that transistor parameters vary by large amounts. I'm sure the dynamic effects are rather more complex than Peirre's simulations indicate. Perhaps this is a good reason to use transistor pairs like the MAT02, where the thermal coupling is tight, and the pair remains closely matched...

Comments? Flames? Has anyone tried these circuit modifications?
 

Nelson Pass

The one and only
Paid Member
2001-03-29 12:38 am
Certainly there is such a thing as thermal distortion, as
the character of any gain device will change with temperature,
and the dissipation will vary with the signal. Usually this
distortion is quite a bit smaller than the other sources, and
its time constant is sufficiently slow that it is not one of the
more noticeable forms. Think of it as subsonic modulation.

:)
 
Thermal considerations are very important with semiconductors. When very high precision, or very high audio quality is needed, variations of working parameters with thermal effects must be taken in account. Carefully examination of high performance oscilloscope vertical amplifiers is very instructive for the high quality audio amp designer.

I don't agree with M. Pass about long time constants. Input transistors such BC550 have very small physical dimensions, and the collector junction is as close as 1/1000e mm of the emitter junction, temperature variations caused by audio signals in collector junction can easily modulate the gain and the idle collector current in few milliseconds. This phenomenon, amplified by the whole chain of transistors, can cause severe perturbation in the final stage idle current, and therefore transient crossover distortion or other troubles can occur, this is clearly audible.

Furthermore, problems caused by thermal effects in the emitter jonction of the input transistor cannot be canceled by feedback, because this junction is not in the feedback loop... This is the reason why the input stage of an amplifier must be designed with extreme care.

Regards, P.Lacombe
 
maybe slightly off-topic, but still thermally related.

One of the things I had wondered about is the use of an insulating pad under the electrically live tab of a MOSFET when the signal is taken from the drain.

I envision a capacitor to chassis ground the plates of whcih are the MOSFET tab and the heatsink, and the insulator is... well whatever the insulator is (mica, plastic, whatever).

Am I out of my mind here? Or should I let the heatsink float when taking the signal from the drain. (And keep a cage around it to keep curious fingers and beagle noses away!)

Erik
 
Yeah, I''m not so sure the time constants are very long either. A SS bipolar is a miniscule little flake of silicon surrounded by plastic. Plastic happens to be an excellent thermal insulator too... It's not a stretch for me to imagine some high-speed thermal modulation in these devices.

Perhaps one of the reasons the Pass amps sound so good is the larger die size of the devices used, combined with the fact that they are generally run at high bias currents. Thus, the power variances imposed during normal operation are a lesser percentage of the total dissipation. I'm certainly no expert on MOSFET tempcos, but with their larger Vgs compared to BJTs, small temperature induced variations might also be further reduced in their effect...

P.Lacombe:
thanks for the tip. I'll see if I can find some scope schematics to study...
 

BrianL

Member
Paid Member
2002-03-29 5:19 am
USA
Mr. Pass is generally correct. Of course he is, since he's
the "one and only"... ;-)

Note that in virtually every audio amplifier, most stages
are run in class A. Thus there is relatively little modulation
of the power dissipation with signal swing. Output stages
are one exception to this generalization.

Thermal effects can be apparent in monolithic op-amps
but careful design and layout symmetry can minimize and/or
eliminate these.

I believe that Doug Self has an example of thermal distortion
on his web site with actual measurements.

As Nelson said, in a well-designed amp, thermal distortion
effects should be more like third-order or lesser effects,
not primary distortion drivers.
 
thermal distortion

With all respect for Mr Pass, I do not agree with his points.

I agree with Pierre that the thermal time constants can be relatively small in small signal stages and can lead to modulations up into the kHz region.
Also, the Pass amps are in my opinion not representative of the mainstream amps, since their stages are generally run much more in class A than the average amp, also Pass amps are not really in the forefront regarding low distortion, so thermal distortion may be 3rd or 4th order effects in them, but in very low distortion/high precision amps thermal dist may have a much more pronounced effect.

Jan Didden
 
Yes, I'm well aware of D. Self's writings on thermal distortion, I've been reading his work in EW&WW for many years. But, just because a well respected author says it is so, doesn't mean I'm going to take it on blind faith. Mr. Self seems to have the tendancy to occasionally dismiss ideas with a conviction I find a little unsettling. Otherwise, the man's work is tremendously rational, which leaves me puzzled that he isn't a little more open-minded.

Perhaps I neglected to mention that it is not the percentage contribution of thermal distortion to overall THD that I'm concerned with. I am quite certain that thermal modulation is indeed very low on the list of distortion sources. However, in my recent design efforts, I have been looking for a little more challenge, and to branch out and focus on areas which typically receive little attention. I am attempting to address the differences between conventional objective measurements and subjective results. Although I am a degreed BASc (EE), with a very methodical and objective nature, I can't deny what my ears still tell me: transistor amps, especially class-B and class-AB tend not to sound as good as tube or class-A designs. After considerable study, I am now relatively certain that the first-order distortion effects effects have been well studied and pose little threat once treated. In fact, I believe that human hearing is quite insensitive to steady-state distortions, but I have no doubt about the ability of the human ear to discriminate sounds and distortions at extremely low levels, possibly even too low to measure with most test equipment. It seems to me that it is more important to provide stable, uniform distortion behaviour across time and frequency. With this in mind, it is at least worth investigating the possible sonic benefits of reducing short-time thermal distortions.
 
hifiZen said:

[snip]
I am attempting to address the differences between conventional objective measurements and subjective results. Although I am a degreed BASc (EE), with a very methodical and objective nature, I can't deny what my ears still tell me: transistor amps, especially class-B and class-AB tend not to sound as good as tube or class-A designs. After considerable study, I am now relatively certain that the first-order distortion effects effects have been well studied and pose little threat once treated. In fact, I believe that human hearing is quite insensitive to steady-state distortions, but I have no doubt about the ability of the human ear to discriminate sounds and distortions at extremely low levels, possibly even too low to measure with most test equipment. It seems to me that it is more important to provide stable, uniform distortion behaviour across time and frequency. With this in mind, it is at least worth investigating the possible sonic benefits of reducing short-time thermal distortions.

Chad, you are tackling a formidable task; I wish you succes!
As you say, absolute objectivity and methodological approaches are essential.
But, it is already an unproven leap if from the observation that tube amps sound better, you conclude that the ear is sensitive to extremely low level distortions. Actually, you are assuming that the reason for the sound difference is such-and-such and now set out to prove it. No offense meant, but it is exttremely difficult to be unbiased; I speak from experience!

Jan Didden
 
Jan,

It is a formidable task indeed, which I have been eating away at very slowly for some time. Progress is slow and uncertain, and I don't expect to ever reach the end. You're right though. My assumptions about the sensitivity of the human ear are mostly unproven. At present, it is the best explanation I've found for many of my subjective observations... a lot more actually than just "tube amps sound better", since it isn't necessarily true in all cases or respects. Transistor amps have some subjective advantages to my ear. But, I'm fairly confident that others have experienced sonic differences between circuits which only have very minor differences such as coupling capacitors. To me, this indicates impressive discrimination of some as-yet elusive properties, not readily apparent in most measurements and therefore likely a very subtle low-level effect.

At any rate, I do endeavour to remain open-minded and objective, though I do have certain biases which emerge as I give greater thought to a subject. My biases are still flexible though if sufficent evidence or a persuasive argument comes along, or so I'd like to believe. ;)
 
thermal distortion

Chad,

What is you opinion on the double blind testing that is often advocated like with the ABX box? I never had the chance to attend one, unfortunately. On the face of it it seems totally objective, but I have read reports that challenged that, on the grounds that there are always subtle clues that gave away which equipment was playing at a certain time..

I have an audio friend who has been brought up with not very good sound, you know, the average muddled AM/FM radio sound.
He is now older, has money to spare and constantly swaps and buys equipment, really high-end, to find his elusive ideal. But everytime he has a system that increases the transparancy and resolution, he rejects it as "too sharp", because it goes against what he is used to. We are constantly debating it, but I can't get him to see the light (well, *my* light at least)! He really has a problem, because evertime he upgrades, he rejects it, so he is constantly going around in circles which is very frustrating to him. That's the power of conditioning!


Jan
 
Hi hifiZen

I don't want to discourage you, quite the opposite, but I've done some subjective audio work before and you're biting off a big chunk.

There are two parts to what you're trying to do as I understand it. Firstly you have to establish if there is a difference between the steady state distortion performance of amps and their dynamic distortion performance. That's something nobody I'm aware of has nailed down but I think it's probably the easy bit.

The hard part will be to determine objectively if human hearing is sensitive to dynamic distortion in a different way to it's sensitivity to steady state distortion. Humans are very good at adapting to and disregarding steady state stimulus off all sorts, taste, smell, touch, sight and hearing. (This is why the emergency services use modulated noise signals and flashing lights.) So it makes intuitive sense that we might be more sensitive to dynamic distortions, but proving it is another matter.

Best of luck with it and I'm sure you'll get lots of helpful comments from this forum.
 
Class A does not eliminate thermal distortion

because thermal distortion is based on thermal load, which itself is defined by the product of voltage and current. For a given voltage swing, the "power swing" is even greater, when the idle current is bigger.

The true reason, Class A in some cases sounds better, comes from improved linearity and better gain bandwidth product at that operating point. This you can immediately confirm for yourself, just listen to your DIY amp in Class A mode, is it better in the bass or in the treble region than the same amp in Class AB mode ;-)

Of course, low feedback designs need class A in the output section for a minimum of linearity.

regards,
Hartmut from Munich
 
Re: Class A does not eliminate thermal distortion

hifidaddy said:
[snip]
The true reason, Class A in some cases sounds better, comes from improved linearity and better gain bandwidth product at that operating point. [snip]
regards,
Hartmut from Munich

Sorry Hartmut from Munich, this is a contradictionary statement. If the *true reason* is as you state, why does it only sound better in some cases?

Jan Didden
 
Re: Re: Class A does not eliminate thermal distortion

janneman said:


Sorry Hartmut from Munich, this is a contradictionary statement. If the *true reason* is as you state, why does it only sound better in some cases?

Jan Didden

because of thermal distortion, which can make a bad bass. Exceptions are power amps where the current is controlled by local feedback or in a small feeback loop, e.g. a CFP like Hiraga Le Monstre or similiar.

The Hiraga is a case where Class A sounds better than Class AB. The Naim power amps (schematics see in another recent thread) are cases where class AB sounds way better than class A.

To paraphrase my above post: a Class A power amp may sound better due to better bandwidth and linearity, but also may sound worse due to other reasons.

You did not explain why you think that my statement is a contradiction. Please do.

regards,
Hartmut
 
Re: Class A does not eliminate thermal distortion

hifidaddy said:
because thermal distortion is based on thermal load, which itself is defined by the product of voltage and current. For a given voltage swing, the "power swing" is even greater, when the idle current is bigger.

The true reason, Class A in some cases sounds better, comes from improved linearity and better gain bandwidth product at that operating point. This you can immediately confirm for yourself, just listen to your DIY amp in Class A mode, is it better in the bass or in the treble region than the same amp in Class AB mode ;-)

Of course, low feedback designs need class A in the output section for a minimum of linearity.

regards,
Hartmut from Munich

Well, if you say that class A mode sounds better because of better GBW product and better OP lineariry, why does it sound better only most of the time? Does class AB sometimes have better GBW product and OP linearity than class A? If yes, then what's the point?
You also state that it immediately comes clear if you listen to class A mode versus class AB mode on your DIY amp. That means *always*, doesn't it. Yet you also state *in some cases*. Which is it?

BTW, I don't agree to the blanket statement that low feedback needs necessarily be in class A. I agree that there may be a trade-off in feedback level versus quiescent current level, but it is more suble than your black-and-white statement.

Cheers, Jan Didden
 
Re: Re: Class A does not eliminate thermal distortion

Jan,

I adressed most of your questions in the later email. Please read carefully.

But especially for you, I will paraphrase my pre last post again:

Of course Class A has always better linearity and GBP than Class AB operation. Better GBP generally transforms into better treble detail and better linearity into better midrange neutrality. It is immediately detectable, that Class A operation is good for treble qualities, when changing the operating point from class AB to A.
But whether Class A operation is beneficial sonically over the whole frequency range is another subject. Thermal distortion comes into play. Bass might get worse, if the operating points are not controlled. And don't forget: the power supply must be good for Class A.

OK ? Did you get me ?

Caveat: The above statement does not address finer details of topology engineering like stability margins of the amplifier, when GBP of one component is changed. A power amp with changed GBP may oscillate, sound worse, and even damage itself and the driven loudspeaker.

Now your question on low feedback and class A: without class A, you would have crossover distortion and no feedback enough to cancel that out. Isn't that simple ?

regards,
Hartmut
 
Re: Re: Re: Class A does not eliminate thermal distortion

hifidaddy said:
Jan,

I adressed most of your questions in the later email. Please read carefully.

But especially for you, I will paraphrase my pre last post again:

Of course Class A has always better linearity and GBP than Class AB operation. Better GBP generally transforms into better treble detail and better linearity into better midrange neutrality. It is immediately detectable, that Class A operation is good for treble qualities, when changing the operating point from class AB to A.
But whether Class A operation is beneficial sonically over the whole frequency range is another subject. Thermal distortion comes into play. Bass might get worse, if the operating points are not controlled. And don't forget: the power supply must be good for Class A.

OK ? Did you get me ?

Caveat: The above statement does not address finer details of topology engineering like stability margins of the amplifier, when GBP of one component is changed. A power amp with changed GBP may oscillate, sound worse, and even damage itself and the driven loudspeaker.

Now your question on low feedback and class A: without class A, you would have crossover distortion and no feedback enough to cancel that out. Isn't that simple ?

regards,
Hartmut

Hartmut,

Yes, I got it. Rereading your post I now see the connection between the thermal distortion issue and your statements which wasn't immediately clear to me at the start.

So, I got it, but I don't agree. I don't see that class A *a priori* has larger GBW and better linearity. At least one researcher (Self) has shown that well designed AB can have at least the same linearity as class A. You seem to regognize it as well in your statement on the NAIM (class AB better than A). And GBW is a function of the devices and topology. I don't see why GBW should improve with higher Iq.

Your Caveat I can follow, no discussion on that.

My question on low feedback and class A (which wasn't really a question): Of course I understand that lower feedback requires more effort in the output stage to get the same quality. What I tried to point out, apparently without succes, is that it isn't either/or. I can imagine a low feedback with a well designed AB stage that sounds better than another class A amp with higher feedback. So many things influence the end result. I mean, what is low feedback? 10dB? 20dB? Can you accept that a well designed AB with 15dB feedback *may* sound better than a not-so-hot class A with 25dB feedback? So, to answer your question: no, it isn't simple.

I think my response to your post was mainly triggered because I have learned to distrust sweeping statements like: *everyone knows that*, or *it always is such and such*, or *the real truth is*.

{Note: this doesn't mean I distrust you personally; I don't know you and have no opinion on you as a person, so please don't take offense}

Cheers, Jan Didden
 
GBW, Naim amps

Jan,

GBW is a function of Ic with its maximum at app. 1/10th of the max Ic when regarding bipolar transistors, and a function of forward transcondunctance vs. input and miller capacitance taking in account any parasitic resistances for FETS. As the forward transconductance is a function of Id, then, regardless whether MOSFET or bipolars in the output, the output devices are faster when driven in class A. I was talking about exactly that.

Naim amps do not have lots of global feedback, their 8 Ohm damping factor is about 26. In my opinion, their relative low optimum point of idle comes from bad stability of idle current. Have in mind that that Naim amps are not known for good treble reproduction. Though I think their optimum point is more at 40mAmps and not at 10 or even zero.

I have seen the plots of output resistance of some class A amps like the Hiragas (they are in some L'Audiophile issue, where Mr Perrot of Hephaistos Laboratories made comparisons of output stages. Mr Perrot is the designer of the Lavardin amps, BTW), and these plots show a step where one of the outputs shuts off. IIRC, these power amps have about 0.5 Ohms output resistance in the Class A region, and 1 Ohm outside, and this is what I call low negative feedback.

regards,
Hartmut