It depends on what you mean by "marginally stable design". In this particular case, it was a "conditionally stable design" as discussed above. Of course, a 10 degs phase margin could do the same, but this is not the case here. BTW, the best way to get clipping oscillations, with a conditionally stable amp, is to use a clipping protection circuit that gracefully kills the clipping VAS OLG, hence pushing the conditional stable amp into instabilities. I have a screen shot of such, as well.
Last edited:
to a) there are two kind of gain stages: voltage gain stages (VAS) and current gain stages (buffer stages or phase shifter, only unity or lower gain).
I mean only voltage gain stages
to b) I would say, the disadvantage (downside, as you say) of multiple voltage gain stages is only present at higher frequencies. E. g. for subwoofer applications multiple voltage gain stages is clearly an advantage, because I need high open loop gain for clean and tight low frequency reproduction, especially by heavy weight diaphragm.
Unfortunately, for the midrange (arround 500 - 2000 Hz) I cannot say, whether high or low open loop gain is an advantage. If you have a multi amping or actice crossover system in use, for the midrange it must be individual check out, what deliver the best performance regarded the sonic quality.
But one can clearly say, high open loop gain below 500 Hz and low open loop gain for high frequency aera above 2000 Hz is the royal way.
An amplifier for a full range application (one amp the whole audible aera) is still a compromise solution, regardless of how much effort has been implemented and how expensive it is.
Last edited:
the clear evidence of the accuracy of my statement in post #1 I would need to know, how the THD/IM spectrum looks, if we have the following waveforms:
1) perfect triangle instead of a perfect sine wafe
2) perfect sawtooth instead of a perfect sine wafe
3) a perfect square wave instead of a perfect sine wafe (this spectrum must be similare to that of a clean clipping amplifier without overshooting peaks)
If I know this, I could recognize rspect. guess the source of the distortion spectrum of an unknown amplifier, what design errors were made and from where are this comes.
Perhaps one members can post this typical spectrums
1) perfect triangle instead of a perfect sine wafe
2) perfect sawtooth instead of a perfect sine wafe
3) a perfect square wave instead of a perfect sine wafe (this spectrum must be similare to that of a clean clipping amplifier without overshooting peaks)
If I know this, I could recognize rspect. guess the source of the distortion spectrum of an unknown amplifier, what design errors were made and from where are this comes.
Perhaps one members can post this typical spectrums
I have made some observations through simulation reducing NFB by a resistor shunt in VAS. Some things observed is that a resistive load in VAS does not increase distortion, but when acting Cdom in HF has an increase of distortion (the current is low) the same goes for the output stage as the current drained by Cdom and output stage is low I believe that the distortion is assigned to the loading of the vas capacitive or caused by phase shift.
In principle greater frequency response in open-loop could minimize the effect.
Another observation is that with the lower NFB, the greater the difference between the bases in diff-pair, is an obvious question that the lower gain of the VAS, greater must be the gain of the diff-pair. I believe this is consistent with Article of Bob Cordell that the increase of NFB reduces distortion PIM. I do not know about the distortion PIM, but by my understanding it is attributed to diff-pair.
In principle greater frequency response in open-loop could minimize the effect.
Another observation is that with the lower NFB, the greater the difference between the bases in diff-pair, is an obvious question that the lower gain of the VAS, greater must be the gain of the diff-pair. I believe this is consistent with Article of Bob Cordell that the increase of NFB reduces distortion PIM. I do not know about the distortion PIM, but by my understanding it is attributed to diff-pair.
Member
Joined 2009
Paid Member
I don't know if I can identify a voltage gain stage - what device does this ? It seems to me that they are all transimpedance. You put a voltage differential across the base-emitter / grid-cathode / gate-source and voila, the current flow through the device is modulated. They are all voltage-current amplifiers. If you want a voltage output, you have to flow the current through a load (or use a CCS). And none of them are current operated devices - a BJT responds to the potential applied across the base-emitter and not to a current flow per se - look up in Sze how a BJT works if you don't believe me - but the input impedance of the base is not that high so you have to be able to supply current to support the voltage you are applying.
Last edited:
I think he explained pretty clearly that he was talking about arrangements with voltage gain >> 1 and those with voltage gain <= 1. Emitter follower (or source follower) is going to be voltage gain basically equal to 1 and is going to effectively amplify current.
You are correct that BJT's operate (at a given temperature) based on the Vbe, but hFE is still about current gain. I don't see, however, how any of this is all that relevant to the original topic.
Member
Joined 2009
Paid Member
Sorry, what I should have said was - "are we sure that the issue is the so called voltage gain stages when surely, any 'stage' represents a non-linear element whether we call it a voltage gain stage or not"
My thinking is that the distortion of each device is related to the how far up and down the transfer curve you go - since the transfer curve is non-linear the more we trace out of the curve the less it can be approximated by a straight line.
Following this thinking - and I may be crazy ! - we should worry about all the stages.
Or is it simply that the VAS suffers from the biggest swing in Vb - Ve and we can neglect the other stages even if they are EF stages ??
My thinking is that the distortion of each device is related to the how far up and down the transfer curve you go - since the transfer curve is non-linear the more we trace out of the curve the less it can be approximated by a straight line.
Following this thinking - and I may be crazy ! - we should worry about all the stages.
Or is it simply that the VAS suffers from the biggest swing in Vb - Ve and we can neglect the other stages even if they are EF stages ??
Ahhh. Now that makes a lot of sense. My expectation would be that you are right and that, if anything, the output stages, which are likely to use more of their transfer curve, vary much more in temperature, and are dealing with a more dynamic load, are generally responsible for more of the distortion than are the earlier stages.
I always thought that strictly speaking BJTs were current controlled devices, but because of the junction drop you may as well consider them power controlled.
Member
Joined 2009
Paid Member
BJTs are quite complicated in detail, it is the distribution of space charge in the base region, as far as I understand, which controls the flow of current through the device. The space charge distribution is primarily controlled by the electric field across the base-emitter.
But the fact you have to be able to supply/source current to the base gets people thinking that it's a current controlled device. As far as I know, there are no current controlled active devices, all known semiconductor and vacuum tube devices are controlled through electric fields (voltage).
The reason I got into this topic myself lately was that I notice how there's all this stuff to learn about common emitter, common collector, common grid, common this and common that - with each variation somehow a mysterious world unto itself. I thought - hang on, the device doesn't know where earth/gnd is, doesn't know that it's in a circuit or that a signal is trying to go from one end to another. Then I realized that all these circuits and topologies with their names is simply a consequence of the history of the subject and if you put yourself in the shoes of the transistor and ask only what voltage difference appears at its control terminals then the rest of it falls away.
(of course it's never that simple, you have to worry about the voltage across the other terminals, but hey, it made an interesting short story!)
But the fact you have to be able to supply/source current to the base gets people thinking that it's a current controlled device. As far as I know, there are no current controlled active devices, all known semiconductor and vacuum tube devices are controlled through electric fields (voltage).
The reason I got into this topic myself lately was that I notice how there's all this stuff to learn about common emitter, common collector, common grid, common this and common that - with each variation somehow a mysterious world unto itself. I thought - hang on, the device doesn't know where earth/gnd is, doesn't know that it's in a circuit or that a signal is trying to go from one end to another. Then I realized that all these circuits and topologies with their names is simply a consequence of the history of the subject and if you put yourself in the shoes of the transistor and ask only what voltage difference appears at its control terminals then the rest of it falls away.
(of course it's never that simple, you have to worry about the voltage across the other terminals, but hey, it made an interesting short story!)
I learned something very useful about BJTs today, thanks mainly to Bigun for actually engaging. If anyone else reads this thread and doesn't know them off-hand maybe they'll check out the Ebers-Moll equations at the bottom of the BJT wiki page too. Far as goes for you Lumba Ogir, if you think of BJT current flow in terms of these equations, so boringly off hand, you may have mentioned it. I might even impress upon you the idea that this knowledge will eventually be so widespread, so fundamental compared to the collective state of human advance, that none of the posts on this entire site will have any but historical value. Until then we'll slowly become more efficient in the way that we propagate information, yet maybe not so slowly as you fear.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.