Pano, these are the kinds of things (thermal distortion) that happens mostly on transient basis, just like fuses. Real time audio amp designers like Thorsten, Wavebourn, and me, for example, take this into consideration. I previously explained why I use 'heavy' Class AB1, with thermal grease, extra quality transistor pads, and BIG heat sinks (when I possibly can). I not only want to REDUCE thermal resistance, but also INCREASE the thermal capacitance in order to reduce any momentary internal temp changes inside the active devices. Another problem is the fact that OPTIMUM Re reduces as Iq increases, and .33 ohms is almost useless, except for my cheapest stuff. I usually use .15 -.22 ohms per transistor (up to 18). This makes the chance for thermal distortion to increase, but it is necessary for the most linear transfer function across the transition from Class A to Class B, which can be very problematic. That is why I test for 7th and 9th harmonic distortion there, principally.
Let's see if we can hit this moving target.
No, he's correct about that- Kelvin is the right way to think about proportionate temperature changes. However, there is still the question of thermal inertia, both from the mass of the silicon (power transistors have decent size chips in them!) and all the thermal resistances and capacitances. Again, if thermal distortion were a real problem, you'd see a rising distortion with lowering of frequency (we're talking below the dominant pole). This would especially be true of amps biased toward class B because of the larger variation in instantaneous current. This effect plain and simply isn't observed. Or when it is, the cause is something else (see for example figure 7 in Self's "Distortion in Power Amplifiers 6: the remaining distortions").
It will return at 75° at about the same time it needed to go from
75 to 125°.....
Also , when power is increased , the die temperature doesnt increase
as fast as you re supposing as there s a transient thermal resistance
that come into play , so although it rise quite rapidly , the phenomenon
is quite slowed by thermal inertias.
Indeed, the exemple of a power transistor device dissipating 50W seems to me
a quite inadequate exemple as no serious designer would push a device
at so high permanent dissipation knowing the current state of technology..
The following datasheet has a transient thermal resistance graph.
Only if the source of energy input and the path of energy release have the same time constants. I don't think they do. That doesn't invalidate your main point, however.
Thanks John, that's the sort of thing I'm curious about, just don't know where to look.
I did some thinking about it and see Thorsten's point. Although Kelvin is "arbitrary" too. The absolute zero makes sense, but the rest of the scale is just as arbitrary as any other. Make not make a difference, tho, if the thermal distortion is linear with degrees Kelvin, as I think TL was pointing out.
It's probably a mistaken intuition on my part, but I've held it so long it's hard to shake. 😱 I'd have to see proof, or find it myself, that temperature does not matter, so stuck in my brain is the idea.
It's not arbitrary- if you plug a temperature into a Boltzman equation (or any of the equations based on Boltzman, e.g., Ebers-Moll), it HAS to be absolute temperature. Differences, sure, it doesn't matter if you use C or K, but ratios, yes, you have to use absolute (Kelvin or Rankine).
The zero point isn't arbitrary, but since the magnitude is based on Celsius, it's just as arbitrary as the Celsius scale. Not really on topic tho, so I'll drop it.
But when you ratio things, it doesn't matter which scale you use as long as it's absolute since any scaling is constant and gets ratioed out to one. You could just as easily use Rankine (Fahrenheit) and get exactly the same answers. Ditto using non-Kelvin/Celsius in any other kinetic calculation (as long as you use the right units for the Boltzman constant).
Use anything other than absolute temperature (regardless of scale) and... TILT.
Use anything other than absolute temperature (regardless of scale) and... TILT.
Magnitude is arbitrary in the sense that if you measure temperature on some other linear scale from absolute zero then your value for Boltzmann's constant 'k' changes too and exactly cancels your temperature rescaling. Maybe we should really be measuring temperature in terms of kT in eV.
The point about syllabic rate thermal changes seems valid to me. The LF rolloff might stop us from directly measuring subsonic stuff, but a subsonic envelope on an audio signal might show up.
The point about syllabic rate thermal changes seems valid to me. The LF rolloff might stop us from directly measuring subsonic stuff, but a subsonic envelope on an audio signal might show up.
SY, please find in datasheets temperature dependence of parameters of transistors and check how stable they are. You will see that OL gain and non-linearities are modulated by temperature. Not only in output devices, but in input and VAS devices as well. However when you measure static parameters of amps they more depend on feedback resistors, but in dynamics they change. Distortions are modulated by signal envelope since temperatures of crystals depend on power dissipated by them. It means one mechaism of dynamic distortions in transistor amps is obvious.
If you insist on LF distortions, measure it without negative feedback and see for yourself. People who designed transistor amps know that it is not so easy like measure balance of vacuum triode LTP or for Concertina. No kidding, SY! 😀
And things in discrete design are more complicated than in IC, where say diamond buffer works as designed because it's transistors made on the same die in the same process, and thermally track each other.
If you insist on LF distortions, measure it without negative feedback and see for yourself. People who designed transistor amps know that it is not so easy like measure balance of vacuum triode LTP or for Concertina. No kidding, SY! 😀
And things in discrete design are more complicated than in IC, where say diamond buffer works as designed because it's transistors made on the same die in the same process, and thermally track each other.
Last edited:
That's not the point I was arguing- Ebers-Moll (for example) is correct. The question is, does that temperature get modulated by audio signals to cause distortion? And the evidence says no. That's very different than slower thermal heating from continuous (or nearly so) high power operation, where the heat sinking and thermal resistances all come into play.
Temperature could play a role in SS like screen voltage stabilisation in pentode amps: increased IM at subsonic difference frequencies. In both cases you have an OL gain which depends on the recent signal envelope. The effect should be calculable, but whether it is audible is another matter. Temperature effects do play a role in the IM performance of RF linear amps, although the chip size may be smaller.
SY, I have such evidences every time I design solid state gear. I can't even design it without thermal regimes in mind. 😀
Try it and check for yourself. You even do not need to design! Take any good solid state amp, shunt feedback loop by AC, and see for yourself what you get as the result.
Anatoliy, I'm not arguing that steady-state temperature has no effect on distortion. Of course that's a major consideration in design, at least if you want your devices to refrain from transmuting between silicon and carbon. My argument is that temperature modulation with signal is a negligible source of distortion.
Precisely. For chemists and physicists, the more common term is "heat capacity," but the analogy to capacitance is useful.
My argument is opposite: temperature modulation is one of major attributes to dynamic distortions that are hard to measure in traditional static measurements of amps with deep feedbacks. Dynamic distortions add artificial colorations to the sound amplified by amps that measure very well.
Edit: do you think Stuart I am crazy, when instead of using complementary source followers load one follower on a current mirror? Do you think I am crazy when bootstrap this follower by another one loosing extra power? Do you think I am crazy using 12 transistors for current mirror instead of 2? No, I am not. Such a way I take care of dynamic distortions right in place, instead of changing their outcome by feedback as others do.
Last edited:
Hi,
A third attempt to get at least the basic idea across...
Your intuition is not mistaken, but the answers do not derive from the (quite real) thermal memory distortion.
The nature and mechanics of why "hot gear" subjectively sounds better are simply not amenable to debate in this forum, so I'll leave things at that.
Ciao T
A third attempt to get at least the basic idea across...
Your intuition is not mistaken, but the answers do not derive from the (quite real) thermal memory distortion.
The nature and mechanics of why "hot gear" subjectively sounds better are simply not amenable to debate in this forum, so I'll leave things at that.
Ciao T
- Status
- Not open for further replies.