Pavel's MOSFET Follower - No Darlington Mod

Status
Not open for further replies.
[snip]Emitter resistors - why an explosion ? if thermal runaway is controlled through other means what purpose do these resistors serve ?

If you can control thermal stability with other means you wouldn't need them for that, but it would be interesting to see how you would do that!

BTW They (emitter resistors) also have an influence on the nature of the cross-over (non)linearity as they provide some local feedback of course.

jd
 
Jan,

OTOH the emitter resistors greatly contribute to compressive losses at higher amplitudes, which by Fourier analysis generates the higher order, odd harmonics.

Can't have your cake and eat it too..... bugger.....

Hugh

Compressive losses? You mean "make the amp clip slightly earlier"?
Let's see. At 100W RMS into nominal 4 ohms there's a peak current of 7 amps, through 0.1 ohms Re that's a loss of a whopping 0.7 V out of a typical supply voltage of 40 or 50V. 😉 .

Things are never as simple as they look. You may find that with Re the amp clips slightly earlier, but softer, which by Fourier analysis generates less higher harmonics.

jd
 
I had a long answer to this and lost it.

Ah well. Yes, call it what you like. I use 0.5R, not 0.1R, the compressive loss, clipping earlier, is much stronger, 3.5V at 7A in fact, plus the increased Vbe loss at large current.

The 'compressive loss' is large signal distortion, ameliorated but only reduced (by loop gain) by global nfb.

The semantics are secondary, however, to the mechanism. Do you understand my point, Jan, that is the question, and if so, in your view, is it true or false?
 
I had a long answer to this and lost it.

Ah well. Yes, call it what you like. I use 0.5R, not 0.1R, the compressive loss, clipping earlier, is much stronger, 3.5V at 7A in fact, plus the increased Vbe loss at large current.

The 'compressive loss' is large signal distortion, ameliorated but only reduced (by loop gain) by global nfb.

The semantics are secondary, however, to the mechanism. Do you understand my point, Jan, that is the question, and if so, in your view, is it true or false?

Yes I think I see your point, it's just that I need to get used to these innovative terms for the same mechanism. Compressive loss normally refers to gain changes through the signal cycle, most apparent with tube gear as I'm sure you know. I get confused if people use one term for another phenomenon and call it 'just semantics'. 😉

jd
 
Well, Jan, that would be because I taught myself all this stuff by doing it, and have not the engineering terms for it....... I'm terribly sorry if I confused you, but you are in excellent company, most engineers haven't a clue what I'm on about, and those who do just get angry.

Have a great weekend!

Hugh
 
Well, Jan, that would be because I taught myself all this stuff by doing it, and have not the engineering terms for it....... I'm terribly sorry if I confused you, but you are in excellent company, most engineers haven't a clue what I'm on about, and those who do just get angry.

Have a great weekend!

Hugh

Good company indeed! I am self-taught as well. Maybe that's why I try to be as concise as possible, to avoid confusion and misunderstanding.
The issue at hand is well explained here: Gain compression - Wikipedia, the free encyclopedia

Edit: BTW This sort of thing happens to me a lot too. You try to explain something in different words, but unfortunately, most different words are already taken. 😉 .

Great weekend? We have snow here, looks great indeed!

jd
 
Last edited:
It appears that Pavel likes less H2 and relatively more H3 (albeit all distortions minimised to levels practically achievable) and Wavebourn prefers more even order distortion!

Pavel you have not commented about R27.

Samuel,

The R27, with 0.001R value, was added to the circuit only for simulation purposes, as a current sensor.

I would not say I like less H2 and relatively more H3. When we speak about the displayed class A CFP power diamond buffer, please take into account that that H2 and H3 are about -100dB and more below fundamental for output amplitudes of several Volts (i.e. completely inaudible), and higher harmonics are -120dB or -130dB and less, with 5th, 6th etc. invisible. For the SE mosfet follower, it is a different story; H2 would be about -65 dB, H3 -80dB and H4 -100dB for several volts of output amplitude. No way it has less distortion in any of the harmonics.

The weakest point of the SE follower, feeded by CCS, is the limited output current capability. Anytime the speaker needs more current than the CCS is able tu supply, harsh clipping takes place. Believe me, for serious amplification purposes this circuit is unusable.

Regards,
 
Regarding 0.1R resistors - lot of misinformation has been said here.

In a real circuit, they are absolutely needed to keep thermal stability. There is no way how to make thermal feedback that accurate to omitt these resistors.

The second important reason is that in case one gets into UHF oscillations, these resistors would save the circuit, and also when crossconduction occurs. Voltage drop across them adds to the Vbe of Q3, Q4, Q9, Q10, so the output stage close.

Their 0.1R impedance is that low, that damping factor is well above 100, which is enough.
 
Last edited:
Regarding 0.1R resistors [...] they are absolutely needed to keep thermal stability. There is no way how to make thermal feedback that accurate to omitt these resistors.

Perhaps it doesn't have to be accurate, it just has to be 'safe'. What I envisage is a -ve thermal compensation that reduces current as temperature increases; it is over-compensated. It isn't perfect in that it wouldn't be a critically damped control system.

The second important reason is that in case one gets into UHF oscillations, these resistors would save the circuit.

I do worry about oscillations with Sziklai (they have that reputation don't they) but I haven't ever read that circuits were lost on this basis, only that they ruin the sound ?

when crossconduction occurs.

How will cross conduction happen in Class A ?
 
How will cross conduction happen in Class A ?

Easily. Cross-conduction is a case of turn-off times. This is set by transistor parameters and charge withdrawal by 100R B-E resistors at ouput transistors. Just put full-scale 100kHz to the input and you will see.

The second case is oscillations, as Jan has said correctly.

Dear Sir, everything I write here is based on real life measurements on real circuits. I have blown several couples of 2SA1302/2SC3281 in this circuits some 7 years ago, when I repeatedly tried to operate it without those 0.1R resistors. If you have your real-life experience with the same circuit, please come in.

Regards,
 
Easily. Cross-conduction is a case of turn-off times. This is set by transistor parameters and charge withdrawal by 100R B-E resistors at ouput transistors. Just put full-scale 100kHz to the input and you will see. The second case is oscillations, as Jan has said correctly.

This is useful, I had certainly not even considered cross-conduction, assuming it to be irrelevant in Class A. I shall include the resistors.

Dear Sir, everything I write here is based on real life measurements on real circuits. I have blown several couples of 2SA1302/2SC3281 in this circuits some 7 years ago, when I repeatedly tried to operate it without those 0.1R resistors. If you have your real-life experience with the same circuit, please come in.

Regards,

I'm here to learn ! - mind you, I didn't blow up any devices yet, so it seems I am already 'in' and ahead of you 😉
 
Bigun,

I understand fully well why you want to get rid of those Re's. Hundreds of smart engineers have tried that for the last 50 years or so, and (with a very few exceptions) concluded it's a bad idea.

Now, I would be the firsat to say that that is not a reason to stop trying; tradition is a bad counseler (just like authority and religion). But it means it's really a hard problem! You really need to do a lot of work to find the one solution all those others have missed. Are you willing to spend the next few years on that? 😉

jd
 
Samuel,

The R27, with 0.001R value, was added to the circuit only for simulation purposes, as a current sensor.

I would not say I like less H2 and relatively more H3. When we speak about the displayed class A CFP power diamond buffer, please take into account that that H2 and H3 are about -100dB and more below fundamental for output amplitudes of several Volts (i.e. completely inaudible), and higher harmonics are -120dB or -130dB and less, with 5th, 6th etc. invisible. For the SE mosfet follower, it is a different story; H2 would be about -65 dB, H3 -80dB and H4 -100dB for several volts of output amplitude. No way it has less distortion in any of the harmonics.

What is good, distortions go down when sound level goes down. It makes the amp inaudible.

The weakest point of the SE follower, feeded by CCS, is the limited output current capability. Anytime the speaker needs more current than the CCS is able tu supply, harsh clipping takes place. Believe me, for serious amplification purposes this circuit is unusable.

I would say, for serious amplification purposes it have to dissipate serious power.

When it oscillates....?

Exactly. And when too high frequency signal is applied to input.
 
Last edited:
I understand fully well why you want to get rid of those Re's. [...]Are you willing to spend the next few years on that? 😉

Well considering that I started this hobby last January and have managed to build only two amplifiers and still understand so little - then no way I will spend the next few years on these two resistors 🙂

However, I will have something built this winter, I need the heat !
 
Well considering that I started this hobby last January and have managed to build only two amplifiers and still understand so little - then no way I will spend the next few years on these two resistors 🙂

However, I will have something built this winter, I need the heat !

Only last January? Man, you're on a fast track! I started to worry about these resistors ohhh, probably after 5 years or so. I'm a slow starter I guess.

jd
 
Status
Not open for further replies.