sreten said:
Hello -
The link that Steven posted to Hawksford's paper apparently omitted the last page. In it Hawksford addressed your question. Essentially he said that since the loop in the error correction circuit encompassed only one or two stages (transistors) instead of the whole amplifier, you could apply more feedback without running into stability problems. (Remember that the error correction circuit has gain in the feedback loop itself.) The short loop (typically just the output stage) has a wider bandwidth and less phase shift than does the longer loop typically used.
But again, this is all from a theoretical standpoint. In my experience the gains aren't really worth it in the real world. You can significantly reduce both the distortion and the output impedance of the output stage simply by increasing the number of output devices (as pointed out by another poster).
Hope this helps,
Charles Hansen
AKSA said:
Hello -
As pointed out in previous posts, there have been several amplifiers that use the Hawksford circuit (or some variation thereof). However, they do not sound as you imagine them to. The best known current example is often described as sounding "sterile" and "unemotional". In my experience, this is a characteristic of feedback circuits themselves, and not just the Hawksford circuit.
But the best thing to do (as always) is to build it yourself and draw your own conclusions.
Best regards,
Charles Hansen
Oh, talking about feedback in the real world...
In studio recorded tracks it's usual to be able to hear the mix from the monitorizing headpones of the singer/musician feed back through the microphone into the recording [though it gets 'masked' when all the tracks are finally mixed]
But the stronger feedback is found on live recordings or multiple-microphone-on-the-same-room recordings where everything feeds back through every microphone, including lots of reflected sound and the mix from the monitorizing speakers and the PA
Now, try to compare these evil-music-thrashing-feedback-phenomena present on nearly all recordings to those ridiculously low amplifier non-linearities
Having feedback problems? 😉
In studio recorded tracks it's usual to be able to hear the mix from the monitorizing headpones of the singer/musician feed back through the microphone into the recording [though it gets 'masked' when all the tracks are finally mixed]
But the stronger feedback is found on live recordings or multiple-microphone-on-the-same-room recordings where everything feeds back through every microphone, including lots of reflected sound and the mix from the monitorizing speakers and the PA
Now, try to compare these evil-music-thrashing-feedback-phenomena present on nearly all recordings to those ridiculously low amplifier non-linearities
Having feedback problems? 😉
Charles Hansen said:
I'm not arguing that they do, just that the explanations are poor.
Or in other words wrong. Its easy to hear differences.
What is better ubiquitously is harder.
After stripping and soldering umpteen combinations of CAT5
cable for single and biwiring I can confidently say bi-wiring is
tosh, a single run of better cable is better.
And for 8 ohm speakers 3 CAT5 runs braided with all pairs in
parallel to give a nominal impedance of 8.3 ohms is best.
So what, any idiot without even trying it can disagree.
Fair enough, but I dislike assumed technical justifications.
I just assume for similar speakers to mine results would
be similar.
And the sites that show braiding of pairs into one strand,
and then braiding the strands together are losing any
sense of what is a reasonable engineering exercise.
For example, I braided 3 CAT5 together. Looks nice. I'm sure
I couldn't tell the difference between braided, twisted or parallel
runs to the extent I didn't even try it - not that I didn't try every
combination of pairs for biwiring, and another cable for biwiring.
(With all the combinations that involved)
Added to that I also tried 15R parallel resistors for the bass/ mids.
With some of the cable combinations, and it worked with some.
But in the end the single run won out.
🙂 sreten.
Steven:
Nice circuit. i just found this thread. I have always been curious about the output stage error corrector circuits from Hawksford and others. Tandberg used it in a production power amp with the Hitachi lateral MOSFETs as output devices. Akai used it in a European market power amp also. I have schematics someone in my pile. Unfortunately, it's in storage while we move.
I was conversing with someone in the Netherlands in the very early 90's via email and paper mail (i.e. before the www) about these circuits. Was it you? Seems like contact was lost after a job change and relocation.
Nice circuit. i just found this thread. I have always been curious about the output stage error corrector circuits from Hawksford and others. Tandberg used it in a production power amp with the Hitachi lateral MOSFETs as output devices. Akai used it in a European market power amp also. I have schematics someone in my pile. Unfortunately, it's in storage while we move.
I was conversing with someone in the Netherlands in the very early 90's via email and paper mail (i.e. before the www) about these circuits. Was it you? Seems like contact was lost after a job change and relocation.
Steven said:
Long thread and very interesting!
No feedback? Depends on the application/specifications of course. Solid-state devices used for amplication are inherently non-linear over their entire operating region. Operation over small portions of their operating region, they can be fairly linear.
I consider degenerative feedback a form of feedback and very useful for local feedback and sometimes appropriate as part of some global feedback. Old Fender and Gibson Tube Amps use various forms and amounts of this feedback, as do many modern designs. Even the small valued resistors used for current sharing when paralleling devices is degenerative feedback, I think. In this case, it is used for a different purpose.
I believe most loudspeakers produce more than 1 percent harmonic distortion. An audio amplifier can have some distortion and not sound bad depending on the type of distortion. This has been addressed in this thread. Having .1 percent harmonic distortion excluding harsh types such as crossover, hard clipping, etc., seems ok to me. Lots of feedback and especially large global feedback produces a transient/dynamic related distortion that I don't like. Everyone seems to be in agreement with that. For decent quality audio, my thinking tells me some feedback of some type is needed for linearization. However, by thinking outside the box, it is probably possible to produce a good design with no feedback. Maybe a design with multiple stages of voltage amplification in which each device operates in it's most linear region so that the combined operation of these parts produces good large signal linearity, is possible. This thread has certainly caused me to think in that direction more than I have in the past.
An amplifier design shown and talked about in this thread is said to have very low output impedance without global feedback. In my opinion, designs like this are desired for obvious reasons.
The search for the (near) ideal amplifier continues.
No feedback? Depends on the application/specifications of course. Solid-state devices used for amplication are inherently non-linear over their entire operating region. Operation over small portions of their operating region, they can be fairly linear.
I consider degenerative feedback a form of feedback and very useful for local feedback and sometimes appropriate as part of some global feedback. Old Fender and Gibson Tube Amps use various forms and amounts of this feedback, as do many modern designs. Even the small valued resistors used for current sharing when paralleling devices is degenerative feedback, I think. In this case, it is used for a different purpose.
I believe most loudspeakers produce more than 1 percent harmonic distortion. An audio amplifier can have some distortion and not sound bad depending on the type of distortion. This has been addressed in this thread. Having .1 percent harmonic distortion excluding harsh types such as crossover, hard clipping, etc., seems ok to me. Lots of feedback and especially large global feedback produces a transient/dynamic related distortion that I don't like. Everyone seems to be in agreement with that. For decent quality audio, my thinking tells me some feedback of some type is needed for linearization. However, by thinking outside the box, it is probably possible to produce a good design with no feedback. Maybe a design with multiple stages of voltage amplification in which each device operates in it's most linear region so that the combined operation of these parts produces good large signal linearity, is possible. This thread has certainly caused me to think in that direction more than I have in the past.
An amplifier design shown and talked about in this thread is said to have very low output impedance without global feedback. In my opinion, designs like this are desired for obvious reasons.
The search for the (near) ideal amplifier continues.
Eva said:
Not that I want to disagree, but this doesn't cause the sort
of feedback effects you get in a standard class aB amplifier.
Due to the the effects of standard single pole feedback for
aB amplifiers if the output stage is not properly biased the
the 19th harmonic of a 1khz tone can be nearly the same
as the 2nd, especially for a CFP pair, which has higher
harmonics than the EF pair.
(The CFP has lower THD at powers near maximum, not
particularly useful for music if its worse at lower powers)
🙂 sreten.
I think I'll take part in this mass confusion exercise 🙂 .
First I think we should not bother with an internal feedback of amplifying devices - whatever makes them to behave in a certain way is taken for granted and their characteristics are described in the device datasheet. Any particular bipolar transistor, valve, FET has got transfer parameters etc. and we can not do much about the inherent character of these things.
I would argue that a follower does have almost 100% negative feedback. After all that what all reference books would tell you 🙂 . Don't forget that the control input of (for example) a FET in a follower configuration is still its gate-source junction. Gate-Source voltage (Vgs) is controlling the current through the device. In a follower the full input voltage is essentially divided into output voltage and control voltage. In my view it is as good example of a negative feedback as any - all output voltage is compared with the input voltage and the difference between two is controlling the current.
There is another rather interesting way to differentiate a circuit with a negative feedback from one without. Many years ago Baxandall did show that when a negative feedback used, the harmonic distortion picture changes in a certain way. If the device produces (for instance) only 2-nd and 3-rd harmonics without the feedback, then with a negative feedback present the distortion order is increased, 4-th, 5-th, and higher harmonics do appear.
It is easy to see that from this criteria a follower most certainly has got a negative feedback - it's distortion pattern is much wider than for a common source amplifier, thought the level of distortions is much lower.
x-pro
First I think we should not bother with an internal feedback of amplifying devices - whatever makes them to behave in a certain way is taken for granted and their characteristics are described in the device datasheet. Any particular bipolar transistor, valve, FET has got transfer parameters etc. and we can not do much about the inherent character of these things.
I would argue that a follower does have almost 100% negative feedback. After all that what all reference books would tell you 🙂 . Don't forget that the control input of (for example) a FET in a follower configuration is still its gate-source junction. Gate-Source voltage (Vgs) is controlling the current through the device. In a follower the full input voltage is essentially divided into output voltage and control voltage. In my view it is as good example of a negative feedback as any - all output voltage is compared with the input voltage and the difference between two is controlling the current.
There is another rather interesting way to differentiate a circuit with a negative feedback from one without. Many years ago Baxandall did show that when a negative feedback used, the harmonic distortion picture changes in a certain way. If the device produces (for instance) only 2-nd and 3-rd harmonics without the feedback, then with a negative feedback present the distortion order is increased, 4-th, 5-th, and higher harmonics do appear.
It is easy to see that from this criteria a follower most certainly has got a negative feedback - it's distortion pattern is much wider than for a common source amplifier, thought the level of distortions is much lower.
x-pro
x-pro said:
I am not sure about that. For example, if a chip amp has gain / feedback resistors integrated into the chip (many low-end boombox-type chips do that), would you consider that amp feedback-free?
x-pro said:
I am surprised that it took people that long to see it, 🙂
light interlude
Sreten
'After stripping and soldering umpteen combinations of CAT5
cable for single and biwiring I can confidently say bi-wiring is
tosh, a single run of better cable is better.'
Way way off topic, but you could try putting the X-over right next to the amp and bi-wiring from there to the speakers. This is more logical and I find it makes a big difference. It's worth one more experiment..🙂
over and out and back to the solid state big guns...
Sreten
'After stripping and soldering umpteen combinations of CAT5
cable for single and biwiring I can confidently say bi-wiring is
tosh, a single run of better cable is better.'
Way way off topic, but you could try putting the X-over right next to the amp and bi-wiring from there to the speakers. This is more logical and I find it makes a big difference. It's worth one more experiment..🙂
over and out and back to the solid state big guns...
millwood said:
I was writing only about single active devices, not integrated circuits 🙂 .
x-pro
Charles Hansen said:
I don't see how this is any different than with the Darlington.
In the Darlington, the first transistor drives the second transistor and the output is fed back to the first transistor.
I mean, in a single device follower, the output is fed back to the input. If you add another device, whether of the same polarity as in the Darlington or of complimentary polarity as in the Sziklai, the output gets fed back to the input.
How do you figure? With even a single device follower, the output is fed back to the input. So how is this not also the case with the Darlington?
Yes.
The Darlington has a feedback loop as well, which must be closed before it can function.
The output of the Darlington pair gets fed back to the input and the output of the Sziklai pair gets fed back to the input.
Again, I don't see the difference. Both appear to be feedback pairs and both appear to have pretty much the same amount of feedback.
Sure. Most every transistor I'm aware of has a gain greater than one.
And the reason an even a single device emitter follower has a gain of unity is because there is effectively 100% feedback. The reason a Darlington pair has a gain of unity is because there is effectively 100% feedback. The reason a Sziklay pair has a gain of unity is because there is effectively 100% feedback.
What's a "unity gain transistor"? Again, most every transistor I'm aware of has some inherent non-zero forward gain. In a follower, that gain is effectively brought to unity by way of negative feedback.
In a Darlington, how can the first transistor's gain be brought to unity until its output has been fed back to its input? I don't see that it can. So what you have is the gain of the first transistor being multiplied by the gain of the second transistor, and then this is brought to unity when the loop is closed.
But the Darlington has high gain as well (basically the product of the two transistors) which opeartes at unity once the feedback loop is closed.
So again, I don't see any difference between the two.
You can do the same thing with a Darlington as well, can you not?
I don't know what you intend to mean when you say "operational amplifier."
se
Charles Hansen said:
Hello Steve,
Based on some past experiences, I'm not going to answer your questions. This is probably unfair on my part, as you seem to be asking in a genuine way. But the bottom line is that this particular question highlights your misunderstanding of the circuits we are discussing.
If someone else wants to edify SE on this topic, please feel free. Otherwise I'd recommend spending some time with a good textbook. One of my favorites (although the sections on logic circuits are getting outdated) is The Art of Electronics by Horowitz and Hill.
Best regards,
Charles Hansen
Actually, Steve, I'd like to know what you're defining as the feedback loop in a Darlington pair.
x-pro said:
even there the distinction between "single active devices" and "integrated circuits" is fuzzy: would you consider darlington a single active device? how about a mosfet (it usually has a diode / zener built in)? or a matched pair of transistors? or those Hexfet (many many little fets)?
it is tough in my view.
Charles Hansen said:
I've a copy sitting right in front of me.
What specifically do you recommend I read?
se
Hello -
Some people keep insisting that an emitter follower has "100% feedback". Now there is a certain logic in saying this, but I've explained in previous posting why I think it isn't useful (and indeed even misleading) to use this terminology. Probably some of you are thinking that I'm some kind of crackpot out peeing into the wind. Maybe so, but I'm in good company.
Some of the best solid state designers in the business agree with my use of terminology. For example, please refer to the data sheet for the Burr-Brown BUF600. This is an "open loop follower" that basically consists of a complementary pair of emitter followers driving another complementary pair of Darlington emitter followers. In the data sheet it says:
"The BUF600 and BUF601 are 3-stage open-loop buffer amplifiers consisting of complementary emitter followers with a symmetrical class AB Darlington output stage.... The amplifiers use no feedback, so their low-frequency gain is slightly less than unity and somewhat dependent on loading." (emphasis mine)
or refer to the data sheet for the Maxim MAX4200 open loop buffer:
"The MAX4200–MAX4205 are ultra-high-speed, open loop buffers....Since these devices operate without negative feedback, there is no loop gain to transform the input impedance upward, as in closed-loop buffers." (emphasis mine)
The point is that we need to use words to communicate our ideas to one another. We therefore need to agree on the meanings of these words.
Of course we could all agree that every single circuit has some form of negative feedback since there is no way to build a device that has infinite gain. But this isn't very helpful, since there are no distinctions created by using the term "feedback" in this way.
So I repeat my previous request -- if you insist on saying that all circuits (including emitter followers) have negative feedback, please give us an alternative term to use that allows us to distinguish between (say) an emitter follower and (say) a complementary feedback pair.
Best regards,
Charles Hansen
Some people keep insisting that an emitter follower has "100% feedback". Now there is a certain logic in saying this, but I've explained in previous posting why I think it isn't useful (and indeed even misleading) to use this terminology. Probably some of you are thinking that I'm some kind of crackpot out peeing into the wind. Maybe so, but I'm in good company.
Some of the best solid state designers in the business agree with my use of terminology. For example, please refer to the data sheet for the Burr-Brown BUF600. This is an "open loop follower" that basically consists of a complementary pair of emitter followers driving another complementary pair of Darlington emitter followers. In the data sheet it says:
"The BUF600 and BUF601 are 3-stage open-loop buffer amplifiers consisting of complementary emitter followers with a symmetrical class AB Darlington output stage.... The amplifiers use no feedback, so their low-frequency gain is slightly less than unity and somewhat dependent on loading." (emphasis mine)
or refer to the data sheet for the Maxim MAX4200 open loop buffer:
"The MAX4200–MAX4205 are ultra-high-speed, open loop buffers....Since these devices operate without negative feedback, there is no loop gain to transform the input impedance upward, as in closed-loop buffers." (emphasis mine)
The point is that we need to use words to communicate our ideas to one another. We therefore need to agree on the meanings of these words.
Of course we could all agree that every single circuit has some form of negative feedback since there is no way to build a device that has infinite gain. But this isn't very helpful, since there are no distinctions created by using the term "feedback" in this way.
So I repeat my previous request -- if you insist on saying that all circuits (including emitter followers) have negative feedback, please give us an alternative term to use that allows us to distinguish between (say) an emitter follower and (say) a complementary feedback pair.
Best regards,
Charles Hansen
Steve Eddy said:
Hello Steve,
I don't have mine in front of me, so I'm going from memory. But if you have the 2nd edition, you probably need to read Chapter 3 about transistors.
Best regards,
Charles Hansen
SY said:
Well, what's the feedback loop in a single device emitter follower? It's the loop formed between the base and emitter, yes?
A Darlington just adds another base/emitter junction to the loop.
se
Charles Hansen said:
Chapter 3's on FETs, which of course are transistors but I assume you meant Chapter 2 which is on bipolars.
But I've read them both quite a few times. And call me thick, but I haven't read anything in either chapter which talks about "open loop" followers or "unity gain transistors" or that the Sziklai uses more feedback than the Darlington.
se
- Status
- Not open for further replies.