Today I was musing on the differences of chip amps compared with discrete ones, and it occured to me that with chip amps the temperature sensing for the biasing of the output transistors would probably be way better. The pn junctions used for temperature sensing could be placed *right next* to the main power devices and as a result they would respond to any temperature variations within milliseconds.
In fact, on low frequency waveforms the output transistor junction temperature on both types of amp could conceivably change up and down during the cycle, and a chip amp biasing may be able to track this in real time, whereas with a discrete amp the bias only follows the average teperature because of the thermal time constant of the power device package, heatsink etc. This slow sensing not being able to follow the junction temperature change would be a source of distortion I think.
In fact, on low frequency waveforms the output transistor junction temperature on both types of amp could conceivably change up and down during the cycle, and a chip amp biasing may be able to track this in real time, whereas with a discrete amp the bias only follows the average teperature because of the thermal time constant of the power device package, heatsink etc. This slow sensing not being able to follow the junction temperature change would be a source of distortion I think.
Circlotron said:
"could", "would" are the key words here. I haven't seen any documentation stating that they actually did that.
Having said that, I am still unconvinced that transistors can fluctuate, within audible spectrum, with current going through it. the smaller ones have considerably less thermal mass so they are the more likely candidate for that but they don't dissipate much heat.
the power transistors, with the amount of thermal heat they dissipate, have considerably higher thermal mass so they aren't likely to react that quickly.
Another way to look at this is the following. If power transistors can be modulated by low frequency signals due to thermal responses, you would expect that device manufacturers to integrate some sort of temperature sensors into the die. Well, we haven't seen that happening to the most part - actually, I don't know of any power transistors with on-die temp sensing devices.
Apparently, people in the know didn't think there is big enough of a market for it. Why is there no demand? maybe thermal modulation isn't such a big deal?
Re: Re: Chip amp vs discrete amp temp compensation of bias.
Sanken have manufactured audio power transistors with inbuilt temperature sensors for some time...
James
millwood said:
Sanken have manufactured audio power transistors with inbuilt temperature sensors for some time...
James
Circlotron said:
Hi Circlotron,
Douglas Self mentions thermal distortion in his books and also has a web article on thermal distortion in IC power amps, see
http://www.dself.dsl.pipex.com/ampins/thermald/thermald.htm
He suggests that IC amps are poorer in this respect.
James
I hope my reply is not too offtopic here.
As a guitarplayer, I used to test poweramps with this method: I connect two inputs to the amp (via mixer if required), one of them a drummachine (because the mic isn't the best for this test) or a bass guitar. The second device is a guitar (via effect device), keyboard, or a singing mic (with a good singer .
). If the drum/bass is not mudulate the secound input, then the amplifier's sound is very good for single device, and I will like it. If the moulation is hard, I will not like it, the sound of single device will very bad, except keyboard or sing, because these two instument well on anything (this is my opinion 
).
I try to understand what is the reason, some poweramps have very high modulation/distortion while testing, some amps haven't. My experience, the amps with biplar transistor ends have high distortion/modulation effect, mosfet/tube amps haven't (or much lower).
Later I read a long article (unfortunately in Hungarian) about the thermal distortion of these devices, and have written, the transistor's have much higher thermal distortion than tubes and mosfets. The article said, the tubes haven't thermal distotrtion too, because the tubes are hot while working (like class-A devices), and the thermal "inpotence" of tubes much higher than bipolars (on class-B mode) which is very good. Shortly it's all what I read about thermal distortion, and I think it' true. I really love mosfet poweramps, much better for me than bipolar ends. Tubes are very good too, but because I like digital effects, tubes not the best for me.
In Engslish I found a very short comparison between bipolar and mosfet devices here:
http://www.arx.com.au/Mosfets.htm
So this is the reason, why I love mosfet poweramps, and the sound for me is really much better than bipolars.
As a guitarplayer, I used to test poweramps with this method: I connect two inputs to the amp (via mixer if required), one of them a drummachine (because the mic isn't the best for this test) or a bass guitar. The second device is a guitar (via effect device), keyboard, or a singing mic (with a good singer .
). If the drum/bass is not mudulate the secound input, then the amplifier's sound is very good for single device, and I will like it. If the moulation is hard, I will not like it, the sound of single device will very bad, except keyboard or sing, because these two instument well on anything (this is my opinion ).I try to understand what is the reason, some poweramps have very high modulation/distortion while testing, some amps haven't. My experience, the amps with biplar transistor ends have high distortion/modulation effect, mosfet/tube amps haven't (or much lower).
Later I read a long article (unfortunately in Hungarian) about the thermal distortion of these devices, and have written, the transistor's have much higher thermal distortion than tubes and mosfets. The article said, the tubes haven't thermal distotrtion too, because the tubes are hot while working (like class-A devices), and the thermal "inpotence"
of tubes much higher than bipolars (on class-B mode) which is very good. Shortly it's all what I read about thermal distortion, and I think it' true. I really love mosfet poweramps, much better for me than bipolar ends. Tubes are very good too, but because I like digital effects, tubes not the best for me.In Engslish I found a very short comparison between bipolar and mosfet devices here:
http://www.arx.com.au/Mosfets.htm
So this is the reason, why I love mosfet poweramps, and the sound for me is really much better than bipolars.
Re: Re: Chip amp vs discrete amp temp compensation of bias.
douglas reasoned or speculated that the rise in thd is due to thermal distortion. he didn't really set out to prove it.
If he is right, circiltron's point on biasing is correct.
douglas did say those such thermal distortion doesn't exist in discrete amps, 🙂.
nemestra said:
douglas reasoned or speculated that the rise in thd is due to thermal distortion. he didn't really set out to prove it.
If he is right, circiltron's point on biasing is correct.
douglas did say those such thermal distortion doesn't exist in discrete amps, 🙂.
StratoFan said:
some of it is true for some of the mosfets working in certain regions. For example, irf mosfets do exhibit positive thermal effect, but at very high current. Otherwise, it has negative thermal coefficient, much like the bjts.
My opinion is that any transistor in the amplifier chain can thermally fluctuate, and cause distortion in the audio spectrum. And it is something that should be adressed in the design, for sound quality.
The thermal distorsion can be kept out of the voodoo zone, by observing THD curves. If the 2nd and 3rd harmonics rise towards lower frequencies, you have thermal distortion.
Especially noticable on datasheets with non feedback designs, but for sound quality probably noticable in any amplifier with this type of distortion.
Bias fluctuation are obviously not so harmful as amplifier stage fluctuations.
Bigger (slower) BIAS feedback parts are probably slightly better, because they maintain the BIAS at some average point instead of otherwise fluctuating (in or out of phase) with the power device.
The thermal distorsion can be kept out of the voodoo zone, by observing THD curves. If the 2nd and 3rd harmonics rise towards lower frequencies, you have thermal distortion.
Especially noticable on datasheets with non feedback designs, but for sound quality probably noticable in any amplifier with this type of distortion.
Bias fluctuation are obviously not so harmful as amplifier stage fluctuations.
Bigger (slower) BIAS feedback parts are probably slightly better, because they maintain the BIAS at some average point instead of otherwise fluctuating (in or out of phase) with the power device.
Konnichiwa,
That is indeed the case if the Chip is well designed.
HOWEVER, you pay a price for this, namely that the Input Stages are more thermally modulated.This too can be solved by apropriate design of course, rasulting in an overall better amplifier on a chip than many discrete Circuits. As always,it is how you do it that counts, not what you do.
Yes, at low frequencies this becomes observable, very much so in fact.
Sayonara
Circlotron said:
That is indeed the case if the Chip is well designed.
HOWEVER, you pay a price for this, namely that the Input Stages are more thermally modulated.This too can be solved by apropriate design of course, rasulting in an overall better amplifier on a chip than many discrete Circuits. As always,it is how you do it that counts, not what you do.
Circlotron said:
Yes, at low frequencies this becomes observable, very much so in fact.
Sayonara
Konnichiwa,
He figured non was needed. The point is so obvious even Dubya would get it without long explanations.
Do you always have to be contratrary for the sake of being contrary? I have to assume that you do have sufficient intelligence to understand electronic basics and thus cannot accept the explanation that you are too stupid to understand BTW.
It would really to raise the generel S/N in this and other Forums if you simply could cut out the constant and consitent saying Nay, to each and everything that cannot be covered by Ohms law alone unless accompanied by dissertation length and strength proof where really obvious stuff is concerened.
Sayonara
millwood said:
He figured non was needed. The point is so obvious even Dubya would get it without long explanations.
Do you always have to be contratrary for the sake of being contrary? I have to assume that you do have sufficient intelligence to understand electronic basics and thus cannot accept the explanation that you are too stupid to understand BTW.
It would really to raise the generel S/N in this and other Forums if you simply could cut out the constant and consitent saying Nay, to each and everything that cannot be covered by Ohms law alone unless accompanied by dissertation length and strength proof where really obvious stuff is concerened.
Sayonara
millwood: Because for other distorsion effects they would increase with frequency because of capacitive effects. The only effect that could cause increasing THD at lower frequencies is thermal. This is owing to the fact that the crystal (or crystals) of the semiconductors have a thermal 'weight' that filter out temperature changes at high frequencies. I don't know the technical 'physics language' name of this thermal property in english.
At lower frequencies the constant 'thermal weight' of the crystal has less stabilizing power, thus causing the temperature on the chip to fluctuate, or in other words become modulated with signal, and cause distortions.
Who is Douglas?
At lower frequencies the constant 'thermal weight' of the crystal has less stabilizing power, thus causing the temperature on the chip to fluctuate, or in other words become modulated with signal, and cause distortions.
Who is Douglas?
Lars Clausen said:
what about semiconductor substrate resonnace?
also, why would it impact just 2nd /3rd harmonics? what's so magic about them?
the "thermal weight" you were talking about is called thermal mass. it include the thermal mass of the die, and casing and heatsinks, etc. some companies publish thermal response curves for their devices. I also recall that when we talked about thermal distortion last time, someone mentioned a mil spec article and one by Mueller on thermal distortion.
Peter Daniel said:
Peter, read and read and read before asking, especially the very link is presented on the same page.
🙂
Maybe I should refrase my question (or remark rather). I've noticed many times certain people reference to DS, however, I fail to understand why he would be treated as some sort of authority in (high end) audio?
Peter Daniel said:
I am not sure if anyone in this thread was / is referring to him as any authoritative figure in audio, high-end or not.
However, I would venture to say, from what I have read / heard about his work, he knows a lot more about audio than a lot of us, especially those "experts" here do.
I certainly look to him as a source of knowledge.
You are saying that chip amps benefit from close proximinty of there power transistors to bias control, and discrete amps suffer from temp distortion due to time/heat transfer of discrete's.
This makes sense, yet
Nemestra stated: Douglas says chip amps are poorer in this respect.
Were talking about thermal distorion in chip amps and that Douglas believes chip amps are worse?
But then Millwood retorts with
If douglas is right then by menestra quote then circiltron's point would be incorrect, the chip amps would not benifit from thermal tracking.
Yet Millwood states circiltron would be right if Douglas is right on thermal tracking.
I'm confused, Who's right?
Easyamp said:
how about both? 🙂
Douglas asserted that because of the close proximity between the input and output stages in a chip amp, the thermal coupling between the output and input stages will induce distortion, which he claimed is backed by his distortion chart.
If indeed, close proximity induced thermal coupling can input the input stage, it will likely impact the biasing circuitry in the chip amp as well. So any thermal and electrical fluctuations in the output stages will be compensated by corresponding and resulting changes in the biasing stage. That is precisely the point that circletron advocated at the beginning of this thread.
So they are talking about the same thing (thermal coupling), but from two different angles: Douglas on its impact on the input stage, and circletron on its impact on the bias mechanism.
so both are right, for different reasons.
- Status
- Not open for further replies.