Steve Eddy said:
Hello Steve,
You're right, I meant Chapter 2. But please read that chapter again and tell me how you can create a Darlington pair with voltage gain, as you proposed in an earlier post. If you've digested the information in that chapter, you will realize that's an impossibility. You should then be able to answer all of your other questions from that same post.
Best regards,
Charles Hansen
Charles Hansen said:
I didn't propose. I queried. Propositions don't usually end with question marks.
se
Charles Hansen said:
Note they say "3-stage."
So it seems to me that by "feedback" they're referring to feedback around multiple stages. Which is what we'd usually refer to as no GLOBAL feedback around multiple stages and doesn't mean no feedback at all, even local feedback.
How do you know that by "without negative feedback" they're not referring to negative feedback around multiple stages?
Sure.
But second-guessing the meanings of words used in datahseets doesn't get us any closer to that.
Anyway, you and I have already had this discussion.
What I was hoping for was some input from others, hopefully someone who can address this question in a manner a bit more comprehensive than "You're wrong. Go read a book" or "Look what this datasheet says."
se
It seems there's really at least two definitions of feedback that could be used. One defines feedback purely in the context of the physical interconnection of devices (CFB pair and so on). The other is purely mathematical. Looking at the mathematical viewpoint with purely linear circuits with z or y parameters, any circuit with y12 or z12 that's not zero is a feedback circuit in the mathematical sense. So the set of circuits that can be considered "feedback circuits" in the mathematical sense is much larger than the set of circuits that can be considered "feedback circuits" in the physical interconnect sense. Is this some kind of unique, or groundbreaking issue? Absolutely not! By way of analogy, thinking about the English language, the word "you" can mean singular or plural. How do you tell? The meaning must be inferred from the context of the discussion.
The present discussion strikes me as being analogous to an attempt to define "you" unambiguously once and for all as being either singular or plural. This will never happen. So when person A speaks about "feedback circuits" to person B, it's incumbent on both of them to establish the context and meaning - that is, is this the mathematical deifintion of feedback or the physical interconnect definition?
Unfortunately, in the audio world, the definition of "feedback" becomes a political/marketing issue. I suspect the roots of this discussion trace back to this problem. But we are the technical people here.
The present discussion strikes me as being analogous to an attempt to define "you" unambiguously once and for all as being either singular or plural. This will never happen. So when person A speaks about "feedback circuits" to person B, it's incumbent on both of them to establish the context and meaning - that is, is this the mathematical deifintion of feedback or the physical interconnect definition?
Unfortunately, in the audio world, the definition of "feedback" becomes a political/marketing issue. I suspect the roots of this discussion trace back to this problem. But we are the technical people here.
andy_c said:
How 'bout simply the basic physical behavior of the circuit?
Not on my part. I go with what sounds best to me. And the particular Sziklai pair that I've been using in a particular circuit using particular devices sounds better to me than the various Darlingtons I've compared it to.
I don't care whether or not the Sziklai pair has considerably more negative feedback than the Darlington. I'll continue to use what I'm using now until I come across something better.
Just that some time back Charles claimed that the Sziklai pair had considerably more negative feedback than a Darlington pair. And after some discussion of it, I still don't know if that's incorrect or I'm just to dense to see that is correct.
And while I've offered some arguments as to why I feel it is incorrect, I'm ultimately not much more convinced by my own arguments against as by Charles' arguments for.
Which is why I was hoping someone could step up to the plate and say it's either this way or that way and more importantly, explain exactly WHY it's the way it is in more unambiguous terms than what Charles has offered so far.
se
andy_c said:
Hello -
And this leads us back (in an indirect way) to the original poster's question. He wanted a "no feedback" amplifier circuit. Now, we're not sure what he meant by "no feedback", and maybe he didn't really either. If he meant "no *global* feedback", then the circuit that was posted a few pages ago fills the bill.
But on the other hand, so would a "chip" amp with a JFET buffer in front of it. And I don't think that's what he's looking for. Presumably he's trying to get away from feedback because he believes that there may be some (hidden?) problem with using it. Putting a buffer outside the feedback loop of the "chip" amp certainly doesn't solve any problems that may be inherent in that type of approach.
Now, did the posted circuit fill his desires? (I don't know, as we seem to have scared the original poster away....) It also has no global feedback, although there are short feedback loops around both the input and output stages, plus an overall DC feedback loop (with gain).
I don't think there are any definitive answers that will satisfy everyone regarding the "definition" of feedback. For those that are curious and enjoy exploring (as opposed to discussing), I would suggest the following:
Build up the circuit that was posted. It's a fine circuit, and certainly more original than 99% of all the circuits you see out there. Listen to it, measure it, play with it. Then remove the feedback (or at least *my* definition of feedback!) by getting rid of the CFPs in the input stage and disconnecting the Hawksford-style error correction at the output. (One might even add a couple of pairs of output devices to get the distortion back down.) Then let us all know what you think.
In my experience, this type of "zero feedback" amp (again, *my* definition) has a sonic quality that is qualitatively different from an amp that uses feedback (again, *my* definition). Furthermore I would assert that this difference is not just a difference, but a musically meaningful improvement over designs that use feedback.
For other opinions on the sonic merits of zero feedback circuits in a different application, refer to the following threads:
http://www.diyaudio.com/forums/showthread.php?s=&threadid=22206
http://www.diyaudio.com/forums/showthread.php?s=&threadid=16323
Have fun,
Charles Hansen
Has anybody demonstrated thorugh measurements that multi-device-loop feedback causes unexpected signal abnormalities not found on single-device-loop feedback?
Subjective and placebo stuff should never be messed with technical and scientific matters. There isn't music inside amplifiers, there are only current and voltage waveforms and recordings are nothing but waveforms too
I think it would be a waste of time to mantain what is intended to be a technical discussion taking as its starting point a subjective criteria
Subjective and placebo stuff should never be messed with technical and scientific matters. There isn't music inside amplifiers, there are only current and voltage waveforms and recordings are nothing but waveforms too
I think it would be a waste of time to mantain what is intended to be a technical discussion taking as its starting point a subjective criteria
Eva,
the whole problem is to find out exact correlation between measurement results and sonic (subjective) results, this has not been exactly possible yet. Human perception is unfortunatelly extremely sensitive for another kinds of distortion than that clearly shown by objective now-a-days measurements. You have to be both scientist and experienced amps designer/listener to find out what from objective measurements is important, what not and what we do not know yet. As long as music is evaluated by our ears subjective factor will be important.
Pavel
the whole problem is to find out exact correlation between measurement results and sonic (subjective) results, this has not been exactly possible yet. Human perception is unfortunatelly extremely sensitive for another kinds of distortion than that clearly shown by objective now-a-days measurements. You have to be both scientist and experienced amps designer/listener to find out what from objective measurements is important, what not and what we do not know yet. As long as music is evaluated by our ears subjective factor will be important.
Pavel
Darlington vs. Sziklai
At risk of confusing people, I am going to succumb to the temptation to answer why a Darlington has only about as much feedback as a single emitter follower.
It is true that the feedback in the emitter follower is between the base and emitter. But it is essentially a transresistance (current in the emitter becomes voltage difference between base and emitter with a very high gain) feedback INSIDE the transistor, and the signal current present at the base terminal is attenuatde by a factor of beta.
So the current signal present that drives the intrinsic feedback in the first stage of the darlington is something like 1/20 to 1/100 of what you see in the output stage itself, so the first stage maybe could perhaps be considered to stiffen the feedback by about 1% to 5%.
The gain of the emitter follwer is related to the ratio of the current in the follower which determines its output impedance, and the load impedance.
On the other hand, a Sziklai has high gain voltage feedback of two common emitter stages in a loop, so the small signal gain in the first stage is only limited by the Early voltage of the transistor used assuming you can use high impedance current source loads ... The second stage has low gain because its intrinsic output impedance is high, and the load Z low. I suppose that I'm too lazy right now to do calculations to verify that there is numerical sense in what this paragraph says ... maybe there isn't.
I hope this helps, but I worry that it won't; I probably didn't put enough time into writing it.
good night folks
...
At risk of confusing people, I am going to succumb to the temptation to answer why a Darlington has only about as much feedback as a single emitter follower.
It is true that the feedback in the emitter follower is between the base and emitter. But it is essentially a transresistance (current in the emitter becomes voltage difference between base and emitter with a very high gain) feedback INSIDE the transistor, and the signal current present at the base terminal is attenuatde by a factor of beta.
So the current signal present that drives the intrinsic feedback in the first stage of the darlington is something like 1/20 to 1/100 of what you see in the output stage itself, so the first stage maybe could perhaps be considered to stiffen the feedback by about 1% to 5%.
The gain of the emitter follwer is related to the ratio of the current in the follower which determines its output impedance, and the load impedance.
On the other hand, a Sziklai has high gain voltage feedback of two common emitter stages in a loop, so the small signal gain in the first stage is only limited by the Early voltage of the transistor used assuming you can use high impedance current source loads ... The second stage has low gain because its intrinsic output impedance is high, and the load Z low. I suppose that I'm too lazy right now to do calculations to verify that there is numerical sense in what this paragraph says ... maybe there isn't.
I hope this helps, but I worry that it won't; I probably didn't put enough time into writing it.
good night folks
...
PMA :
If the input to the amplifier is a waveform and the output is also a waveform [proportional to the input and with some error], then, why do we need such auditive correlations?
How could the error signal on the output of the amplifier be evaluated by means of listening tests without even knowing how the input signal 'sounds'? I think this is anti-scienfitic and a waste of time
Are we talking about linear signal amplification or about cognitively pleasing the user of the amplifier? [by cognitively I mean not only hearing but any sort of perception or ilusion]
We hear just what we want to hear
If the input to the amplifier is a waveform and the output is also a waveform [proportional to the input and with some error], then, why do we need such auditive correlations?
How could the error signal on the output of the amplifier be evaluated by means of listening tests without even knowing how the input signal 'sounds'? I think this is anti-scienfitic and a waste of time
Are we talking about linear signal amplification or about cognitively pleasing the user of the amplifier? [by cognitively I mean not only hearing but any sort of perception or ilusion]
We hear just what we want to hear
Eva,
not so. I have collected enough examples in practice with listeners not told about technical solutions. Just 2 examples:
1. Audio Buffer circuit. You can find it on my www (+ spectra measurement). This is a buffered signal transfer through terminated cable (50R series impedance + 50R terminating impedance). You will measure no THD change. Every listener can hear considerable difference in tougher bass, non-harsh mids and overall clarity.
2. The preamp with and without HF filtering in every stage. Again - no change in THD and IMD. Considerable difference in listenning tests with people not knowing the difference in technical solution.
My aim ist to find why, not to struggle for objective or subjective, which is only the way to hell 😉
not so. I have collected enough examples in practice with listeners not told about technical solutions. Just 2 examples:
1. Audio Buffer circuit. You can find it on my www (+ spectra measurement). This is a buffered signal transfer through terminated cable (50R series impedance + 50R terminating impedance). You will measure no THD change. Every listener can hear considerable difference in tougher bass, non-harsh mids and overall clarity.
2. The preamp with and without HF filtering in every stage. Again - no change in THD and IMD. Considerable difference in listenning tests with people not knowing the difference in technical solution.
My aim ist to find why, not to struggle for objective or subjective, which is only the way to hell 😉
PMA said:
I find that amazing. Maybe you didn't go far enough out in frequency in your thd / imd measurement? or the preamp doesn't generate much distortion at higher harmonics without the HF filter.
Did go farer than shown. Of course, there is a difference in step response and amplitude/phase characteristics - but far above "audio" range😉 .
To prevent any misunderstanding: all I want to say is that scientists have try hard to bring new measurement methods correlating with human perception. And they have to sit down from clouds to real life. During more than 30 years when I have been interested or engaged in audio there were many changes in scientific explanations, and certainly will be in the future.
To prevent any misunderstanding: all I want to say is that scientists have try hard to bring new measurement methods correlating with human perception. And they have to sit down from clouds to real life. During more than 30 years when I have been interested or engaged in audio there were many changes in scientific explanations, and certainly will be in the future.
I think that you all don't understant "much feedback" the same way.
If you think of feedback in percent, of course, an emitter follower and a sziklay have the same amount of NFB, 100%
But if you speak in terms of dB (here, I don't know the numbers), but an EF has much less feedback than a CFP, of an amp with global NFB, since the open loop gain of the 2 laters is much higher than the closed loop gain
If you think of feedback in percent, of course, an emitter follower and a sziklay have the same amount of NFB, 100%
But if you speak in terms of dB (here, I don't know the numbers), but an EF has much less feedback than a CFP, of an amp with global NFB, since the open loop gain of the 2 laters is much higher than the closed loop gain
does it matter what signal is being "feed to" and "feed back"? for example, the follower wouldn't have any feedback in terms of current input and current output. But there is clearly feedback in voltage input and voltage output.
I still have yet to figure out how a darlington has feedback (other than being used as a follower).
I still have yet to figure out how a darlington has feedback (other than being used as a follower).
True.
All, I'm bemused by the feedback discussion. I mean, feedback exists when a change in output condition results in a change in input condition. Isn't it that simple?
In any case you are missing the point. Whether there is feedback or not is not important in itself. You must get past the terminology and think about what the electrons are doing.
Eva said:
Yes!! ...if the output of a feedback amp is a exact replica of the input signal then the loudspeaker can only reproduce a accurate replica of the input signal...for the loudspeaker to give a diferent "sound" it need to be feed by a diferent signal...and this diferent signal is the result of the nolinearitys and intermodulation distortion produced by a amp without negatif feedback...
traderbam said:
That is clearly one definition. another one, which I often use, is that a negative feedback exists when the impact of an input signal reduces the system's deviation from its steady state.
This is particularly useful in understanding thermal stability (where either Vbe drop can be viewed as a "signal").
In my view it would be reasonable to clarify one point about negative feedback. It is clear that some kind of feedback is always present in all amplifiers, including internal feedback in the devices themselves, all kinds of parasitic feedbacks etc.
May be we should add "substantial" to "negative feedback" ? If the negative feedback effect is less than, say, 1dB = 10% reduction in gain it could be considered "unsubstantial" 🙂
x-pro
May be we should add "substantial" to "negative feedback" ? If the negative feedback effect is less than, say, 1dB = 10% reduction in gain it could be considered "unsubstantial" 🙂
x-pro
Hi x-pro, All,
IGBT: A MOSFET/BJT CFP as one device (or isn't it?)
It's rather different than discrete MOSFET and BJT wired together.
Regards,
Peter Jacobi
x-pro said:
IGBT: A MOSFET/BJT CFP as one device (or isn't it?)
It's rather different than discrete MOSFET and BJT wired together.
Regards,
Peter Jacobi
- Status
- Not open for further replies.