Guess what, after some searching I found a site that is less careful with copyrights and coincidentally also Czech! They published complete copies of the Hawksford paper and the Cordell paper. So for as long as the papers stay there:
http://wes.feec.vutbr.cz/~gratz/Download/Hawksford.ZIP
http://wes.feec.vutbr.cz/~gratz/Download/Cordell.ZIP
Steven
http://wes.feec.vutbr.cz/~gratz/Download/Hawksford.ZIP
http://wes.feec.vutbr.cz/~gratz/Download/Cordell.ZIP
Steven
Thanks. And here are the files with Pavel Dudek's amplifiers, I have his permission:
http://web.telecom.cz/macura/dpa_386.pdf (mosfet)
http://web.telecom.cz/macura/dpa_222.pdf (bipolar)
Some minor improvements were done in the input circuit, not shown here.
I am starting new thread about these amps.
http://web.telecom.cz/macura/dpa_386.pdf (mosfet)
http://web.telecom.cz/macura/dpa_222.pdf (bipolar)
Some minor improvements were done in the input circuit, not shown here.
I am starting new thread about these amps.
andy_c said:
Andy,
I think I know the patent you are referring to. It's probably 5,892,398 from Candy. I have a copy, but have not yet studied it in detail. It's dated 1999!
Another similar patent is from Iwamatsu, dated 1984, number 4,476,442.
So, all are of later date than Hawksford's publication in the JAES.
If you experience trouble with http://patft.uspto.gov you can use the European patent site. E.g. the one from Iwamatsu:
http://l2.espacenet.com/espacenet/viewer?PN=US4476442&CY=gb&LG=en&DB=EPD
and the one from Candy:
http://l2.espacenet.com/espacenet/viewer?PN=US5892398&CY=gb&LG=en&DB=EPD
Don't forget to put US in front of the number for US patents if you are searching for a number and separatores are not needed (allowed?). A nice thing of this site is that you get the pages as pdf instead of tiff. Too bad it is only one page per pdf file so you end up with a lot of separate pages and need some Adobe program to stitch all pages together.
In 1983 we had no means for simulating everything, I got my first PC (XT) only a few years later. So, we just built a circuit and looked at stability with different loads. It took some tuning, but the amplifier is perfectly stable in a lot of loads, including capacitive ones, without output coil or Zobel network.
Steven
Hello -
As the designer of several high performance commercially successful zero feedback amplifiers, I thought I would offer a few comments.
1) Somebody suggested that one couldn't build a zero feedback amplifier with good measured characteristics. I suppose this depends on how one defines "good", but it isn't too difficult to build a high power design that has a bandwidth in excess of 200 kHz, a damping factor of 40 (Z0 = 0.2), and distortion at 100 watts of around 0.1%. In my opinion, there is no need to use feedback to try to improve these numbers.
2) Someone else asked about various definitions of feedback. The original poster confused things somewhat by calling a "non-feedback amplifier" an "NFB" amplifier. This is very confusing terminology because "NFB" is normally used (especially in Japan) to denote a "negative feedback amplifier", which is quite the opposite.
Please note that there are many more kinds of feedback than there are terms to describe them. This leads to a large degree of confusion. Most commercial designs that claim "no feedback" usually mean "no *global* feedback". In my opinion this is a terribly misleading use of terminology. Let me give a few examples of this:
a) The original Marantz 9 (tube) amplifier. This circuit was basically a traditional tube amplifier with feedback from the transformer secondary to the cathode of the input tube. However, for some reason they added one additional gain stage outside the feedback loop, in front of what would normally be the input tube. Technically, this is a "no *global* feedback" circuit, but if you believe that global feedback is bad, this design doesn't do *anything* to alleviate those problems.
b) The Boulder (solid state) power amps use a discrete op-amp module that is taken from Deane Jensen's design published in the JAES in the '70s. This is by necessity a high-feedback design. However, the Boulder amps use two of these modules in series. So while there is a tremendous amount of feedback (around several gain stages) they can claim "no global feedback".
So to me, a claim of "no global feedback" is essentially meaningless.
3) Delving further into the realm of different types of feedback circuits, the complementary feedback pair (CFP), as used in the input of the circuit posted earlier in this thread, obviously uses feedback (as implied by the name of the circuit. It is true that this loop is shorter than an overall loop around the entire amplifier, but it certainly is feedback.
4) The error correction circuit published by Hawksford is also a feedback circuit. The loop is typically just around the output stage, so it is shorter than a "global" feedback loop, but longer than the loop of a CFP. The twist is that there is gain in the *reverse* direction of this loop. This technique was first used (to the best of my knowledge) in the early Quad solid state amp that they called the "current dumping" circuit (I forget the model number). It was also used in the "Trans Nova" circuit that Jim Strickland did for Acoustat and later Hafler. However, the Hawksford paper formalized this approach, explained it, and made it easy to adapt to existing output stages.
(By the way, Bruce Candy of Halcro lied through his teeth when he (or some representative from his company) first claimed that his amp didn't use error correction. As anyone who has looked through the patent knows, this circuit is simply an adaptation of Hawksford's error correction paper.)
5) A bipolar emitter follower is inherently unstable when presented with a capacitive load. This is a difficult problem to solve, and tends to become worse when using Hawksford's technique. The usual solution is to employ a combination of an output inductor (to isolate the load capacitance) and series base resistors (to cancel the negative base impedance created by the capacitive load).
6) I agree with the poster that found Hawksford's error correction technique to not work all that well. When all is said and done, in my experience you are far better off to simply double the number of output devices.
7) The original poster asked about a "non feedback amplifier". So far this thread has focused on a nice design that happens to have several (short) feedback loops. Does anyone want to post a true zero feedback circuit?
Hope this helps,
Charles Hansen
Ayre Acoustics, Inc.
As the designer of several high performance commercially successful zero feedback amplifiers, I thought I would offer a few comments.
1) Somebody suggested that one couldn't build a zero feedback amplifier with good measured characteristics. I suppose this depends on how one defines "good", but it isn't too difficult to build a high power design that has a bandwidth in excess of 200 kHz, a damping factor of 40 (Z0 = 0.2), and distortion at 100 watts of around 0.1%. In my opinion, there is no need to use feedback to try to improve these numbers.
2) Someone else asked about various definitions of feedback. The original poster confused things somewhat by calling a "non-feedback amplifier" an "NFB" amplifier. This is very confusing terminology because "NFB" is normally used (especially in Japan) to denote a "negative feedback amplifier", which is quite the opposite.
Please note that there are many more kinds of feedback than there are terms to describe them. This leads to a large degree of confusion. Most commercial designs that claim "no feedback" usually mean "no *global* feedback". In my opinion this is a terribly misleading use of terminology. Let me give a few examples of this:
a) The original Marantz 9 (tube) amplifier. This circuit was basically a traditional tube amplifier with feedback from the transformer secondary to the cathode of the input tube. However, for some reason they added one additional gain stage outside the feedback loop, in front of what would normally be the input tube. Technically, this is a "no *global* feedback" circuit, but if you believe that global feedback is bad, this design doesn't do *anything* to alleviate those problems.
b) The Boulder (solid state) power amps use a discrete op-amp module that is taken from Deane Jensen's design published in the JAES in the '70s. This is by necessity a high-feedback design. However, the Boulder amps use two of these modules in series. So while there is a tremendous amount of feedback (around several gain stages) they can claim "no global feedback".
So to me, a claim of "no global feedback" is essentially meaningless.
3) Delving further into the realm of different types of feedback circuits, the complementary feedback pair (CFP), as used in the input of the circuit posted earlier in this thread, obviously uses feedback (as implied by the name of the circuit. It is true that this loop is shorter than an overall loop around the entire amplifier, but it certainly is feedback.
4) The error correction circuit published by Hawksford is also a feedback circuit. The loop is typically just around the output stage, so it is shorter than a "global" feedback loop, but longer than the loop of a CFP. The twist is that there is gain in the *reverse* direction of this loop. This technique was first used (to the best of my knowledge) in the early Quad solid state amp that they called the "current dumping" circuit (I forget the model number). It was also used in the "Trans Nova" circuit that Jim Strickland did for Acoustat and later Hafler. However, the Hawksford paper formalized this approach, explained it, and made it easy to adapt to existing output stages.
(By the way, Bruce Candy of Halcro lied through his teeth when he (or some representative from his company) first claimed that his amp didn't use error correction. As anyone who has looked through the patent knows, this circuit is simply an adaptation of Hawksford's error correction paper.)
5) A bipolar emitter follower is inherently unstable when presented with a capacitive load. This is a difficult problem to solve, and tends to become worse when using Hawksford's technique. The usual solution is to employ a combination of an output inductor (to isolate the load capacitance) and series base resistors (to cancel the negative base impedance created by the capacitive load).
6) I agree with the poster that found Hawksford's error correction technique to not work all that well. When all is said and done, in my experience you are far better off to simply double the number of output devices.
7) The original poster asked about a "non feedback amplifier". So far this thread has focused on a nice design that happens to have several (short) feedback loops. Does anyone want to post a true zero feedback circuit?
Hope this helps,
Charles Hansen
Ayre Acoustics, Inc.
Hello -
I forgot to touch on one other topic in my previous lengthy post -- servo loops. Again this is a form of feedback. It uses gain in the reverse direction, just as found in the Hawksford error correction circuit. The feedback is nominally restricted to DC, but in actuality it extends upwards into the audio band. As John Curl pointed out, this can result in sonic degradation. By moving the low-frequency corner of the servo loop lower and lower, this degradation is reduced.
Another approach is to build a true DC amplifier, with no servo loops and no coupling caps. This gives excellent results, but requires a fair degree of care with both the circuit design and construction methods. For those interested in that approach, a very good starting point was Erno Borbeley's 3-part series many years ago in Audio Amateur magazine. He had a design called the "DC-100" that was DC coupled, and the article took you step-by-step through the design process.
Hope this helps,
Charles Hansen
Ayre Acoustics, Inc.
I forgot to touch on one other topic in my previous lengthy post -- servo loops. Again this is a form of feedback. It uses gain in the reverse direction, just as found in the Hawksford error correction circuit. The feedback is nominally restricted to DC, but in actuality it extends upwards into the audio band. As John Curl pointed out, this can result in sonic degradation. By moving the low-frequency corner of the servo loop lower and lower, this degradation is reduced.
Another approach is to build a true DC amplifier, with no servo loops and no coupling caps. This gives excellent results, but requires a fair degree of care with both the circuit design and construction methods. For those interested in that approach, a very good starting point was Erno Borbeley's 3-part series many years ago in Audio Amateur magazine. He had a design called the "DC-100" that was DC coupled, and the article took you step-by-step through the design process.
Hope this helps,
Charles Hansen
Ayre Acoustics, Inc.
Charles Hansen said:
Charles,
I agree with you that my circuit uses several small feedback loops and can only claim to have no overall feedback. It was designed as such and it is more or less accepted terminology to call it such.
Even Nelson Pass said in one of his posts that one of the Stasis amplifiers just got the feedback from the driver stage back to the input and keep the output transistors outside of the loop for a.o. reasons of marketing "no overall feedback".
Anyway, I think local feedback is almost unavoidable and required to get acceptable linearity. It still gives advantages to avoid multistage feedback.
You claim to have designed several high performance commercially successful zero feedback amplifiers. Are you using local feedback? Or are you the first to post a true zero feedback circuit?
Steven
good job of chopping away at some of the "zero feedback amplifier" hype, you just happened to leave out the fact that that followers are 100% negative feedback circuits - with a single gain element in the loop
triodes also have internal negative feedback from the plate to grid voltage
so i think this leaves grounded emitter/source/cathode(no triodes need apply) circuits - and series degeneration impedance also qualifies as local negative feedback
its going to be pretty hard to get those low damping numbers without having effeicency worse than 1/damping ratio
triodes also have internal negative feedback from the plate to grid voltage
so i think this leaves grounded emitter/source/cathode(no triodes need apply) circuits - and series degeneration impedance also qualifies as local negative feedback
its going to be pretty hard to get those low damping numbers without having effeicency worse than 1/damping ratio
Steven said:
Hi Steven,
Thanks for the patent reference. I will look there to see if I can find the Halcro schematics.
I wasn't trying to say that the circuit wasn't stable. In fact, I said
If I overcompensate it so it's stable with both, I get no improvement in 20 kHz harmonic distortion over the uncompensated design
I really should have said "over the non-error-corrected design", so my choice of words was not very good there.
Though I didn't look at it, I'm quite sure there would have been an improvement in low-frequency distortion with the circuit when I compensated the error-correction loop for good transient response with both resistive and reactive loads as mentioned above. However, that wasn't what I was looking for in my specific design. That design has plenty of loop gain at the lower frequencies, but it had only about 14 dB of feedback at 20 kHz though. So I was looking for the error correction to improve the distortion at 20 kHz by about 20 dB. I wasn't able to achieve this while meeting the "no overshoot or ringing with a 2 uF load" requirement (which is tough to achieve). The design does seem to be easier to tame with a small number of output devices. For the high-power case with many parallel output devices, it got pretty sensitive.
Hello -
I was pretty sure that we could get into a "discussion" (or would that be "argument") about semantics. That was not my intention. However, I will state my views in this one post. After this I will not respond to further posts regarding semantics or terminology, but I will be glad to discuss the circuits themselves.
Please remember that the place that I am starting from is that words help us to describe some sort of external reality (or at least our experience thereof). To the degree that words can aid us in this task, then they are useful. But if they do not serve us, then we need to develop a new vocabulary.
At the present time, we have only a few terms to describe negative feedback in an audio circuit:
- Overall (or global or loop) feedback
- Local feedback (or degeneration)
- Nested feedback loops
However, these terms are insufficient to describe the various types of audio circuits in use today. (Error correction and servo loops are fairly well described by the existing terminology.)
As I noted in my previous post, there are many, many circuits that use large amounts of feedback and yet can be classified as "no global feedback". Then you have people that deliberately mislead their customers by taking one part of the circuit outside the feedback loop and claiming "no global feedback".
(I have no idea if the previous poster was correct in his claim about an amplifier design that took the output stage outside of the feedback loop in order to claim "no global feedback", but if true it seems rather backwards to me. If you want to be able to make a claim for marketing purposes only (instead of designing on a clear fixed principle), then I would think that you would be better off to do as the circuit currently posted in this thread and confine the feedback to the output stage.)
At any rate, there are those that claim that an emitter follower has "100% negative feedback". Please allow me to explain why I disagree with this terminology.
In the first place, the term "feedback" clearly implies that the signal is being "fed" "bacK" from a later point in the circuit to an earlier point in the circuit. In all circuits that use feedback, except the emitter follower, this is true. In the emitter follower, the emitter is simultaneously both an input port and and output port. So the signal is *not* being "fed" "back" from a later point to an earlier point.
The second reason is that if we look at the case of a common emitter amplifier with an unbypassed emitter resistor, we will see that there is an action in the circuit identical to what happens in an emitter follower. If one were to call this action "feedback" you would run into a difficult semantic problem. Specifically, take the case of a common emitter amplifier with *no* emitter resistor. Due to the Re of the transistor itself, the gain is limited. So if one were to claim that there is feedback with an unbypassed emitter resistor, to be consistent one would still claim there is feedback even with no emitter resistor whatsoever. In this case one would have to say that *every* circuit in the world has feedback, and the term itself loses all meaning. Remember that words are supposed to be useful tools that help us define and understand the world.
To explore the third reason, we must look at why people want to avoid feedback in the first place. In general, people are trying to minimize or avoid the use of feedback because they have found that there is a detrimental effect on sound quality. We don't know all the reasons why this would be true, but for the sake of argument let's assume that it is because feedback can't correct an error until after it has occurred. This is because there is obviously a time delay involved in a feedback loop. The circuit cannot respond instantaneously (if it did it would have infinite bandwidth) and so there is a time delay between the input and the output. So the feedback loop is always responding after the fact. But in an emitter follower, there is no time delay. This is because the emitter terminal is *both* the output port and one of the input ports.
So the bottom line is that to call an emitter follower a feedback circuit is not useful in any way, in my opinion. In fact it only serves to confuse the issue. Sometimes people refer to an emitter follower (or a common emitter circuit with an unbypassed emitter resistor) as having "local feedback". While this helps to distinguish it from "loop feedback" or "global feedback", I think that ultimately it is misleading, at least in the world of high performance audio circuits.
In my opinion, the best way to describe an emitter follower (or CE with an unbypassed emitter resistor) is to say that it has "degeneration". Of course, this term, while it is accurate, will never catch on. No commercial designer wants to talk about "degeneration" in their circuit because it just sounds like a "bad" thing to the lay person.
So feel free to agree or disagree with my reasoning or my points. And feel free to post your feelings. But don't interpret my lack of response as any other than the fact that I have better things to do than discuss the *semantics* of circuit design.
So on to the circuits themselves! Steven asked if I had really developed a zero feedback amplifier. The circuits I use have a combination of emitter followers and common emitters with unbypassed emitter resistors. So I would describe these has having local degeneration at each stage. But there is absolutely no feedback *loop* of any sort. At no point is the signal "fed" "back" to an earlier stage. There are no "complementary feedback pairs" that comprise a two-stage feedback loop, there are no error correction circuits, there is no overall feedback, and there are no servo circuits (DC feedback).
I call this a zero feedback circuit because I think that is the most accurate way to describe it. Based on both my experience in circuit design, *and* my experience with the sonic effects of different types of feedback circuits, this type of circuit *should* be distinguished from other circuits that *do* include feedback loops.
In fact, I would say that using a feedback loop of *any* sort imparts a coloration to the sonic characteristics of the circuit. So to me, a circuit that is completely zero feedback except for using CFPs in the output stage is closer to a normal design with an overall feedback loop than it is to a zero feedback amplifier. In my experience, once you have a feedback loop, it doesn't matter how long or short it is, you will hear the same kind of sonic effect.
So if someone wants to argue whether or not local degeneration is really "feedback", go right ahead. But if you conclude (erroneously in my opinion) that local degeneration in a single stage is really "feedback", then please kindly coin a new term that will distinguish between local degeneration and feedback where the signal is "fed" "back" in a loop from one point in the circuit to an earlier point in an attempt to improve the linearity of the overall circuit. Otherwise we are left trying to discuss different types of circuit solutions without any useful words to describe them.
Best regards,
Charles Hansen
Ayre Acoustics, Inc.
I was pretty sure that we could get into a "discussion" (or would that be "argument") about semantics. That was not my intention. However, I will state my views in this one post. After this I will not respond to further posts regarding semantics or terminology, but I will be glad to discuss the circuits themselves.
Please remember that the place that I am starting from is that words help us to describe some sort of external reality (or at least our experience thereof). To the degree that words can aid us in this task, then they are useful. But if they do not serve us, then we need to develop a new vocabulary.
At the present time, we have only a few terms to describe negative feedback in an audio circuit:
- Overall (or global or loop) feedback
- Local feedback (or degeneration)
- Nested feedback loops
However, these terms are insufficient to describe the various types of audio circuits in use today. (Error correction and servo loops are fairly well described by the existing terminology.)
As I noted in my previous post, there are many, many circuits that use large amounts of feedback and yet can be classified as "no global feedback". Then you have people that deliberately mislead their customers by taking one part of the circuit outside the feedback loop and claiming "no global feedback".
(I have no idea if the previous poster was correct in his claim about an amplifier design that took the output stage outside of the feedback loop in order to claim "no global feedback", but if true it seems rather backwards to me. If you want to be able to make a claim for marketing purposes only (instead of designing on a clear fixed principle), then I would think that you would be better off to do as the circuit currently posted in this thread and confine the feedback to the output stage.)
At any rate, there are those that claim that an emitter follower has "100% negative feedback". Please allow me to explain why I disagree with this terminology.
In the first place, the term "feedback" clearly implies that the signal is being "fed" "bacK" from a later point in the circuit to an earlier point in the circuit. In all circuits that use feedback, except the emitter follower, this is true. In the emitter follower, the emitter is simultaneously both an input port and and output port. So the signal is *not* being "fed" "back" from a later point to an earlier point.
The second reason is that if we look at the case of a common emitter amplifier with an unbypassed emitter resistor, we will see that there is an action in the circuit identical to what happens in an emitter follower. If one were to call this action "feedback" you would run into a difficult semantic problem. Specifically, take the case of a common emitter amplifier with *no* emitter resistor. Due to the Re of the transistor itself, the gain is limited. So if one were to claim that there is feedback with an unbypassed emitter resistor, to be consistent one would still claim there is feedback even with no emitter resistor whatsoever. In this case one would have to say that *every* circuit in the world has feedback, and the term itself loses all meaning. Remember that words are supposed to be useful tools that help us define and understand the world.
To explore the third reason, we must look at why people want to avoid feedback in the first place. In general, people are trying to minimize or avoid the use of feedback because they have found that there is a detrimental effect on sound quality. We don't know all the reasons why this would be true, but for the sake of argument let's assume that it is because feedback can't correct an error until after it has occurred. This is because there is obviously a time delay involved in a feedback loop. The circuit cannot respond instantaneously (if it did it would have infinite bandwidth) and so there is a time delay between the input and the output. So the feedback loop is always responding after the fact. But in an emitter follower, there is no time delay. This is because the emitter terminal is *both* the output port and one of the input ports.
So the bottom line is that to call an emitter follower a feedback circuit is not useful in any way, in my opinion. In fact it only serves to confuse the issue. Sometimes people refer to an emitter follower (or a common emitter circuit with an unbypassed emitter resistor) as having "local feedback". While this helps to distinguish it from "loop feedback" or "global feedback", I think that ultimately it is misleading, at least in the world of high performance audio circuits.
In my opinion, the best way to describe an emitter follower (or CE with an unbypassed emitter resistor) is to say that it has "degeneration". Of course, this term, while it is accurate, will never catch on. No commercial designer wants to talk about "degeneration" in their circuit because it just sounds like a "bad" thing to the lay person.
So feel free to agree or disagree with my reasoning or my points. And feel free to post your feelings. But don't interpret my lack of response as any other than the fact that I have better things to do than discuss the *semantics* of circuit design.
So on to the circuits themselves! Steven asked if I had really developed a zero feedback amplifier. The circuits I use have a combination of emitter followers and common emitters with unbypassed emitter resistors. So I would describe these has having local degeneration at each stage. But there is absolutely no feedback *loop* of any sort. At no point is the signal "fed" "back" to an earlier stage. There are no "complementary feedback pairs" that comprise a two-stage feedback loop, there are no error correction circuits, there is no overall feedback, and there are no servo circuits (DC feedback).
I call this a zero feedback circuit because I think that is the most accurate way to describe it. Based on both my experience in circuit design, *and* my experience with the sonic effects of different types of feedback circuits, this type of circuit *should* be distinguished from other circuits that *do* include feedback loops.
In fact, I would say that using a feedback loop of *any* sort imparts a coloration to the sonic characteristics of the circuit. So to me, a circuit that is completely zero feedback except for using CFPs in the output stage is closer to a normal design with an overall feedback loop than it is to a zero feedback amplifier. In my experience, once you have a feedback loop, it doesn't matter how long or short it is, you will hear the same kind of sonic effect.
So if someone wants to argue whether or not local degeneration is really "feedback", go right ahead. But if you conclude (erroneously in my opinion) that local degeneration in a single stage is really "feedback", then please kindly coin a new term that will distinguish between local degeneration and feedback where the signal is "fed" "back" in a loop from one point in the circuit to an earlier point in an attempt to improve the linearity of the overall circuit. Otherwise we are left trying to discuss different types of circuit solutions without any useful words to describe them.
Best regards,
Charles Hansen
Ayre Acoustics, Inc.
PMA said:
Charles refers to simple emitter followers as "open loop followers." If this is the case there isn't any negative feedback as that requires a closed loop.
Which brings me to another issue I'd like to try and get a definitive answer to.
Charles claims that the CFP (Sziklai) employs substantially more negative feedback than a Darlington pair, saying that a Darlington pair is comprised of an "open loop follower" feeding another "open loop follower" resulting in substantially less negative feedback than a CFP.
However everything I've read of the two says that each has essentially the same gain multiplication and since they're both followers and both operate with essentially 100% feedback, that means that they would both have essentially the same amount of negative feedback.
In other words, the Darlington appears to be every bit as much a "feedback pair" as the Sziklai.
Yes?
se
Hello -
One person asked why Hawksford's error correction circuits hadn't been used more in commercial designs. In addition to the most obvious example of the Halcro, and the less obvious examples of the Quad and Acoustat I noted in an earlier post, there is one other example of which I am aware. That is the Tandberg.
One of the posters gave the Tandberg as an example of a zero feedback design. I'm not 100% sure that we are talking about the same model, but about 15 years ago, Tandberg offered a range of amplifiers that used Hawksford's error correction circuit with MOSFET output devices. I don't recall what the front end circuitry was like, but the error correction itself is a form of feedback. So there may or may not have been no *global* feedback, but certainly not a zero feedback design.
I know that there are several other commercially other amplifiers that use the Hawksford circuit, but unfortunately I can't recall what they are now.... sorry.
Hope this helps,
Charles Hansen
One person asked why Hawksford's error correction circuits hadn't been used more in commercial designs. In addition to the most obvious example of the Halcro, and the less obvious examples of the Quad and Acoustat I noted in an earlier post, there is one other example of which I am aware. That is the Tandberg.
One of the posters gave the Tandberg as an example of a zero feedback design. I'm not 100% sure that we are talking about the same model, but about 15 years ago, Tandberg offered a range of amplifiers that used Hawksford's error correction circuit with MOSFET output devices. I don't recall what the front end circuitry was like, but the error correction itself is a form of feedback. So there may or may not have been no *global* feedback, but certainly not a zero feedback design.
I know that there are several other commercially other amplifiers that use the Hawksford circuit, but unfortunately I can't recall what they are now.... sorry.
Hope this helps,
Charles Hansen
millwood said:
Its impossible to design any kind of amplifier, never
mind H-P, without some form of local feedback due to
inherent device parameters.
I find it totally ridiculous that a device such as feedback,
used ubiquitously throughout the recording and mastering
chain can suddenly have such a deleterious effect when
applied to the replay amplifier.
Its totally wrong, amplifiers have their problems but feedback
per se is not one of them, anyone who thinks so is gullible.
(Though I will say the way feedback is used in the majority of
amplifiers is misguided and there are better feedback topologies)
edit :
I'll just add error correction circuitry has the same bandwidth
limitations as feedback circuitry, so what's the difference.
sreten.Steve Eddy said:
Hello Steve,
I say "no" for the following reasons:
1) With the Sziklai pair (complementary feedback pair), the signal is literally "fed" "back" from the second transistor to the first transistor. In contrast, in the Darlington pair the two transistors are completely independent and there is no signal being "fed" "back" at any point in the circuit to any other point in the circuit.
2) The Darlington pair has (approximately) unity gain. The Sziklai pair (CFP) also has (approximately) unity gain, *once the feedback loop is closed*. Before the feedback loop is closed, the first transistor has a gain defined as Rc / (Re + re) which is always greater than one, and the second transistor also has a gain greater than one although it is undefined since the load resistor may vary.
So the Darlington is simply two unity-gain transistors in series, while the Sziklai (CFP) is a circuit with a high open loop gain that operates at unity gain once the feedback loop is closed. In fact you can close the feedback loop of a CFP with a resistor instead of a wire if you want an output stage with gain. So in this regard the CFP is really operating more like an operational amplifier.
Hope this helps,
Charles Hansen
sreten said:
Hello -
I can't argue with your logic. I agree that it doesn't make sense that having a zero feedback amplifier at the end of a chain that includes dozens of feedback amplifiers should make any sonic difference whatsoever. And by the same token, using an "audiophile" power cord for the last few feet after thousands of feet of ordinary wire shouldn't make a difference either.
But in my experience, both things that *shouldn't* make a difference actually do. Your mileage may vary.
Best regards,
Charles Hansen
True. If you have no global feedback, you must still use local feedback - particularly in the output stage to ensure it can drive a speaker. The speaker is the principle difficulty. The best I've seen here is a bipolar/mosfet CFP, in single end. I've measured 40 milliohm Zout with just these two devices in a 28W SE amplifier driven from a tube voltage amplifier.
This has been one of the most interesting threads I've ever seen here. I was aware of Hawksford's work, but knew nothing of Charle's or Pavel Dudek's products. I have been wrestling with non-switching output stages for about five years, and after many experiments recently finished the design of a single ended push pull, but although it is quite sophisticated I doubt it would compare with the elegance, efficiency and low parts count of this AB approach.
The idea of a Hawksford output stage with vanishingly low Zout, no global feedback, no switching, is beguiling. I cannot comprehend why the world has not beaten a path to his door, and why consumers the world over have not fallen in love with the sound. I would expect it to be as detailed as Class A but with even better dynamics.
Transistors are markedly less linear in common emitter than tubes in common cathode, and I can't help thinking a tube driving a Hawksford output stage would be almost optimal.
Andy_C, your quest to get such an amp stable and undistorted on sine wave into 2uF at 20KHz is ambitious; almost no conventional output stage can do this. If you backed off a little, could you not get something still hugely impressive? Are you designing specifically for electrostatics?
Steven, thank you for your response to my questions.
Thanks to all involved, and for posting schematics. My congratulations go out to Steven, a prophet in his own time, and Pavel Dudek.
Cheers,
Hugh
This has been one of the most interesting threads I've ever seen here. I was aware of Hawksford's work, but knew nothing of Charle's or Pavel Dudek's products. I have been wrestling with non-switching output stages for about five years, and after many experiments recently finished the design of a single ended push pull, but although it is quite sophisticated I doubt it would compare with the elegance, efficiency and low parts count of this AB approach.
The idea of a Hawksford output stage with vanishingly low Zout, no global feedback, no switching, is beguiling. I cannot comprehend why the world has not beaten a path to his door, and why consumers the world over have not fallen in love with the sound. I would expect it to be as detailed as Class A but with even better dynamics.
Transistors are markedly less linear in common emitter than tubes in common cathode, and I can't help thinking a tube driving a Hawksford output stage would be almost optimal.
Andy_C, your quest to get such an amp stable and undistorted on sine wave into 2uF at 20KHz is ambitious; almost no conventional output stage can do this. If you backed off a little, could you not get something still hugely impressive? Are you designing specifically for electrostatics?
Steven, thank you for your response to my questions.
Thanks to all involved, and for posting schematics. My congratulations go out to Steven, a prophet in his own time, and Pavel Dudek.
Cheers,
Hugh
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.