bluebird
Well-known member
Well, I just ate a particularly fantastic turkey sandwich. :grin:
FETs, and a question for PRR (and others)
- Thread starter AP
- Start date
Help Support GroupDIY Audio Forum:
VacuumVoodoo
Well-known member
I thin it might be worthwhile to study Erno Borbely's articles on the subject matter:
http://www.borbelyaudio.com/ae599bor.pdf
http://www.borbelyaudio.com/ae699bor.pdf
http://www.borbelyaudio.com/ae599bor.pdf
http://www.borbelyaudio.com/ae699bor.pdf
thermionic
Well-known member
- Joined
- Jun 3, 2004
- Messages
- 1,671
A business associate was telling me about his friend's hi-fi amp recently. On the front-panel, large itallic letters proclaim "Passive Linear Volume Control"... So it's got a pot in it then, forgive me for not controlling my excitement... :roll:
Several oems I could name utilise BJTS and FETS in their designs; they tend to implement FETs where high input impedance is required such as DI inputs. A guess some designers might have a religious motive (or marketing) for specifying certain semiconductor types, whereas other designers implement whatever they feel is right for the job at hand with no particular dogma.
Cheers,
Justin
AP
Well-known member
Yes, I read those papers just a week ago, but mainly from the circuit topology standpoint. I just re-read them from a 'technology justification' standpoint, and yes - they do speak to that. So thanks very much for pointing that out.
I quoted Borbely in my initial post to this thread:-
He mentions the term "resolution" a number of times, without really saying what he means. If he simply means "clarity" (or somesuch), I would have been very pleased if he had explained why he thought that was, and what he thought were the mechanisms. Still, it is illuminating, all the same. He did seem fairly pleased with the distortion figures he achieved, although they didn't really seem too astonishing to me. The examples he shows though, are of configurations with local feedback only, and no use of overall negative feedback. The other article at www.borbelyaudio.com/borb503.pdf does show designs with overall feedback, and the figures are rather better. In fact, I believe his figures are doing his design an injustice - the lowest he ever measures is .0011%, and I suspect that is the limit of measurement of his HP test set!
So yes - very interesting. However, I would still be very interested in other views.
Regards
Alan
FredForssell
Well-known member
Hi guys. Sorry for not being too present in this discussion. I've just wrapped up a recording project that took a couple of months and my brain is mush. It's also summer here in North Idaho and I find it quite difficult to sit in my office when it is so nice outside. IMO, there are some good questions and well written replies in this thread (hell, in The Lab in general, which seperates it from most other forums), and I don't know how much I can contribute to this discussion.
Perhaps you count yourself as one, but you can be hardly put in the same bucket with the crowd that reveres the unidirectional RCA cable.
Well no, I don't count myself in with that crowd. But don't get me wrong, if I could hear a difference by reversing direction on an RCA cable, I find the best sounding direction (to me) and use it. I've tried. I don't hear a difference. YMMV. I do hear a difference in interconnects and other cabling and it is not within my technical grasp to explain objectively why this could be so. None the less, I use the cabling that I think sounds the best. I try to keep an open mind on this stuff, but my mind is a finite space and I don't try to keep up with all that is going on in the audio industry.
I've spent a lot of money of test equipment. I've spent a lot of money on listening set-ups. I use them both in the design and evaluation process of my work. My designs tend to start with concepts rooted in an intuitive approach to solving a problem or an operational situation. The solutions to those problems or situations tend to be rooted in the technical/engineering side of things. In my little world there needs to be balance between the two.
So I have a good listening set-up in my lab/office, another in my house, and I do recordings in a studio that has really nice stuff in it. I design equipment to use in those systems, in the studios, or in the remote recording rig that I use to do live concert recordings because it is what sounds or functions the best to me.
I also have a large tech library (35 years of collecting audio books) and I think this is a very important aspect of learning this stuff. I also do a lot of experimentation, listening to a ton of music (real and reproduced) every day, and have a lot of enegry and time to work on this stuff. That all helps too.
I design circuits on paper and build them on the bench. I make sure that they are working and behaving themselves with my test equipment. If all is well, I start listening and tweaking the design until I come up with what I feel sounds the best. I measure the circuit again to insure that it is still well behaved and operating well within the safe operating area of its components. I typically then run it though tests that simulate real-world extreme operating conditions to see that it performs propely under those circumstances. Adjust things if needed. Then I listen a bunch more, and finally, take it out into the world to see how it does. That's when it goes to other people to find out how a design works for them.
Fred, I don't believe you would choose to line up with the latter!
More on the point though, from a few of the things I have seen, I suspect you to be a FET fancier... If so, is there an objective case for FET superiority in audio applications?
Alan, yes I am a FET fancier... actually a JFET fancier. I started designing stuff with tubes back in 1970. I also messed around with BJT designs which I found far less rewarding sonically and a lot more demanding of me as a designer. That was my limitation as a designer. I had some good mentors in this process (a great designer named Geoff Cook, and another name Bascom H King) who were leagues ahead of me in knowledge, experience, and IQ. One of the really important things that I learned from them was about listening as part of the design process.
Somewhere around the mid 1970's a comany called Crystalonics came out with a really good low noise JFET called the C413 (which became the 2N6550). It was the first solid-state device that you could pretty much simply hook-up (like a triode) and have it pass totally acceptable signal through-out the audio passband. It was extremely quite, unlike most single tube circuits, and it sounded great all be itself. We used it for front-end diff-amps, and all by itself as a moving coil cartridge pre-preamp.
Later Siliconix came out with V-Channel MOSFET which to me, where the first really good sounding MOSFETs out there. We used the VMP-1 a lot as output devices in discrete opamps. I have also used depletion mode MOSFET in place of JFETs at various times, but I haven't liked the results as much as the JFETs. But you can get 500 volt depletion mode MOSFETs and you can't do that with JFETs.
One of the coolest things about JFETs to me (being a tube designer) is that you can get P devices. That really opens up the world of topologies to one as a designer, compared to working with tubes. So with modern JFET devices, I can take some of the design approaches and ideas that I use as a tube design and use them in JFET designs, but to have the added flexibility of P and N devices. The currently available (at least for now) JFET devices like the 2SK170/2SJ74 and the 2SK389/2SJ103 are large geometery devices with high Gm (22 mS) and low pinch off voltages. These work really well for audio circuits, IMO. Building a simple three-stage amplifer with these devices and running it open loop can result in an amplifier that is flat out to 10-20 KHz and that has very low distortion. I've never been able to do that with BJT. I do it with tubes all the time. Yes there is local degeneration with the JFET designs as there is with some BJT and with tubes. But without global negative feedback, the JFET and tube designs function acceptable within the audio passband. The BJT designs never have (at least not my BJT designs).
Because these designs tend to be more linear without global negative feedback they have a better "feel" to me as a designer. They are easier to work with, being more stable, less prone to becoming RF detectors, and do not suffer from the breakdown characteristics of BJT. JFETs are voltage controlled devices and do not require large amounts of drive current (although one needs to be aware of the gate capacitance, which requires low source Z and symetrical source/sinking of drive current for symetrical slew rate performance). I love not having to worry about using darlingtion connected driver stages, or dealing with the more complicated bias requirements, and SOA consideration of BJT.
However, the bottom line for me is that using JFETs I can design circuits that are simple, stable, sound good (to me at least), and still perform all the functions needed in the pro audio world. The fact that these same circuits perform well the audiophile world as well, says something to me.
When mixing a project it is always about how that mix will translate out there in the real world. We work hard at building mixing systems and rooms that allow us to build a mix that sounds good to us right then and there during the mix session. We then hope that the results will translate well outside of that room. For me, designing circuits is kind of like that... If I do it correctly, then it will sound good to me in my listening set-up, measure well on my test bench, and translate well out there in the real world. Designing with JFETs and tubes consistantly produces those results for me. Using BJTs, IC opamps, and (so far) MOSFETs does not provide the same results for me. That's probably a limitation of mine as a designer. As a designer, YMMV.
My recommendation is that you play around with this stuff in its most basic form and find approaches that feel and sound intuitively correct to you. I'm not sure you will ever find absolute answers to questions like "Are there advantanges to using FETs over BJTs in audio circuits." Nor will you find absolute answers to is Cabernet better than Merlo. It a personal thing. The answer is formed in one's own brain and that makes it subjective.
Sorry for the long post. The morning coffee kicked in.
Cheers.
Perhaps you count yourself as one, but you can be hardly put in the same bucket with the crowd that reveres the unidirectional RCA cable.
Well no, I don't count myself in with that crowd. But don't get me wrong, if I could hear a difference by reversing direction on an RCA cable, I find the best sounding direction (to me) and use it. I've tried. I don't hear a difference. YMMV. I do hear a difference in interconnects and other cabling and it is not within my technical grasp to explain objectively why this could be so. None the less, I use the cabling that I think sounds the best. I try to keep an open mind on this stuff, but my mind is a finite space and I don't try to keep up with all that is going on in the audio industry.
I've spent a lot of money of test equipment. I've spent a lot of money on listening set-ups. I use them both in the design and evaluation process of my work. My designs tend to start with concepts rooted in an intuitive approach to solving a problem or an operational situation. The solutions to those problems or situations tend to be rooted in the technical/engineering side of things. In my little world there needs to be balance between the two.
So I have a good listening set-up in my lab/office, another in my house, and I do recordings in a studio that has really nice stuff in it. I design equipment to use in those systems, in the studios, or in the remote recording rig that I use to do live concert recordings because it is what sounds or functions the best to me.
I also have a large tech library (35 years of collecting audio books) and I think this is a very important aspect of learning this stuff. I also do a lot of experimentation, listening to a ton of music (real and reproduced) every day, and have a lot of enegry and time to work on this stuff. That all helps too.
I design circuits on paper and build them on the bench. I make sure that they are working and behaving themselves with my test equipment. If all is well, I start listening and tweaking the design until I come up with what I feel sounds the best. I measure the circuit again to insure that it is still well behaved and operating well within the safe operating area of its components. I typically then run it though tests that simulate real-world extreme operating conditions to see that it performs propely under those circumstances. Adjust things if needed. Then I listen a bunch more, and finally, take it out into the world to see how it does. That's when it goes to other people to find out how a design works for them.
Fred, I don't believe you would choose to line up with the latter!
More on the point though, from a few of the things I have seen, I suspect you to be a FET fancier... If so, is there an objective case for FET superiority in audio applications?
Alan, yes I am a FET fancier... actually a JFET fancier. I started designing stuff with tubes back in 1970. I also messed around with BJT designs which I found far less rewarding sonically and a lot more demanding of me as a designer. That was my limitation as a designer. I had some good mentors in this process (a great designer named Geoff Cook, and another name Bascom H King) who were leagues ahead of me in knowledge, experience, and IQ. One of the really important things that I learned from them was about listening as part of the design process.
Somewhere around the mid 1970's a comany called Crystalonics came out with a really good low noise JFET called the C413 (which became the 2N6550). It was the first solid-state device that you could pretty much simply hook-up (like a triode) and have it pass totally acceptable signal through-out the audio passband. It was extremely quite, unlike most single tube circuits, and it sounded great all be itself. We used it for front-end diff-amps, and all by itself as a moving coil cartridge pre-preamp.
Later Siliconix came out with V-Channel MOSFET which to me, where the first really good sounding MOSFETs out there. We used the VMP-1 a lot as output devices in discrete opamps. I have also used depletion mode MOSFET in place of JFETs at various times, but I haven't liked the results as much as the JFETs. But you can get 500 volt depletion mode MOSFETs and you can't do that with JFETs.
One of the coolest things about JFETs to me (being a tube designer) is that you can get P devices. That really opens up the world of topologies to one as a designer, compared to working with tubes. So with modern JFET devices, I can take some of the design approaches and ideas that I use as a tube design and use them in JFET designs, but to have the added flexibility of P and N devices. The currently available (at least for now) JFET devices like the 2SK170/2SJ74 and the 2SK389/2SJ103 are large geometery devices with high Gm (22 mS) and low pinch off voltages. These work really well for audio circuits, IMO. Building a simple three-stage amplifer with these devices and running it open loop can result in an amplifier that is flat out to 10-20 KHz and that has very low distortion. I've never been able to do that with BJT. I do it with tubes all the time. Yes there is local degeneration with the JFET designs as there is with some BJT and with tubes. But without global negative feedback, the JFET and tube designs function acceptable within the audio passband. The BJT designs never have (at least not my BJT designs).
Because these designs tend to be more linear without global negative feedback they have a better "feel" to me as a designer. They are easier to work with, being more stable, less prone to becoming RF detectors, and do not suffer from the breakdown characteristics of BJT. JFETs are voltage controlled devices and do not require large amounts of drive current (although one needs to be aware of the gate capacitance, which requires low source Z and symetrical source/sinking of drive current for symetrical slew rate performance). I love not having to worry about using darlingtion connected driver stages, or dealing with the more complicated bias requirements, and SOA consideration of BJT.
However, the bottom line for me is that using JFETs I can design circuits that are simple, stable, sound good (to me at least), and still perform all the functions needed in the pro audio world. The fact that these same circuits perform well the audiophile world as well, says something to me.
When mixing a project it is always about how that mix will translate out there in the real world. We work hard at building mixing systems and rooms that allow us to build a mix that sounds good to us right then and there during the mix session. We then hope that the results will translate well outside of that room. For me, designing circuits is kind of like that... If I do it correctly, then it will sound good to me in my listening set-up, measure well on my test bench, and translate well out there in the real world. Designing with JFETs and tubes consistantly produces those results for me. Using BJTs, IC opamps, and (so far) MOSFETs does not provide the same results for me. That's probably a limitation of mine as a designer. As a designer, YMMV.
My recommendation is that you play around with this stuff in its most basic form and find approaches that feel and sound intuitively correct to you. I'm not sure you will ever find absolute answers to questions like "Are there advantanges to using FETs over BJTs in audio circuits." Nor will you find absolute answers to is Cabernet better than Merlo. It a personal thing. The answer is formed in one's own brain and that makes it subjective.
Sorry for the long post. The morning coffee kicked in.
Cheers.
AP
Well-known member
No apologies required at all! Thanks for the detailed response, it is much appreciated.
Alan
Fred,
You are an authority on this so I can't argue on the suitability of components. Most likely you also have a hearing that is better than 99.9% of the people in the audio consumer market. This is a topic little discussed. Most people like to think that they have perfect hearing.
We test our hearing yearly and I know that mine is not as good as my recording partner's. When healthy it drops off at 18K. When my sinuses go to party it is 16K at best. I do pretty well at low frequencies. On the other, hand my pal can hear a cricket chirping ten miles away.
I wonder how many people could tell the difference between a JFET992 and a JH990 when set up as a line amp with a gain of five or so?
Those that can hear the difference it is easy to form an opinion. The rest of us have to rely on a more quantitive analysis to improvements.
Tamas
You are an authority on this so I can't argue on the suitability of components. Most likely you also have a hearing that is better than 99.9% of the people in the audio consumer market. This is a topic little discussed. Most people like to think that they have perfect hearing.
We test our hearing yearly and I know that mine is not as good as my recording partner's. When healthy it drops off at 18K. When my sinuses go to party it is 16K at best. I do pretty well at low frequencies. On the other, hand my pal can hear a cricket chirping ten miles away.
I wonder how many people could tell the difference between a JFET992 and a JH990 when set up as a line amp with a gain of five or so?
Those that can hear the difference it is easy to form an opinion. The rest of us have to rely on a more quantitive analysis to improvements.
Tamas
> did you say only within the last day or so, that *you* couldn't hear resistors....?
On current amplifiers, resistor flaws are swamped by amplifier flaws.
In a super-high feedback amplifier, the amplifier flaws should vanish in the feedback. The feedback resistor flaws will dominate.
With high-quality resistors at "sane" voltages, I doubt the flaws will be audible. (But people do pay more for exotic resistors in passive attenuators and in no-feedback amplifiers.) I doubt I would hear it. Many years ago I saw a write-up on a 3,000 volt direct-coupled electrostatic loudspeaker driver, that should have given 0.01%THD, but voltage-sensitivity in the feedback resistors (exposed to kilovolts of DC and audio swing) added several tenths THD. He used a string of super-good resistors to reduce that effect.
> There will be Acoustic instruments, Microphones, Preamps, Midi Keyboards, ADCs and DACs, Power amps and Loudspeakers, and ...PCs (and not too much else).
The mike preamp should go away. Mike preamp design is (admit it) not doing anything really new. Those ADC designers are riding Moore's Law. Within a few years we will have over 24 bits real resolution (not these so-called 24-bit ADCs with 3 or 4 bits of garbage on the bottom), 144dB S/N, noise level near -130dBm and overload far above any phantom-power mike.
The analog power amp goes away. Working switch-mode rotates your dynamic range problems into frequency-range problems, so at present it is mostly good for subwoofers and some less-great apps like laptop audio and big car audio. But there are tricks, and some designers have made really good totally-digital power amps. They take a digital bit-stream input, juggle some big switches, and dump good analog audio into the speaker.
30 years ago Bell Labs had a digital earphone. Imagine an electrostatic speaker. Divide the surface into areas of 1, 2, 4, 8, 16,... square units. Drive each one with the coresponding audio bit. The sum of the displacement is the audio signal plus the switching spikes. In their app the diaphragm worked in a telephone handset, in a large chamber with a fairly small opening to the ear-- an acoustic low-pass filter.
> a page scanned from a book by John Linsley-Hood:
Thanks. I had (and lost) a more recent essay giving similar curves for a Type 50 tube. I think for the more common multi-stage amplifier, the situation is worse. (It may be best to lump all your distortion in one stage, rather than trying to strain each stage equally.)
Also note that this is unweighted. 0.1% 3rd is probably inaudible, but 0.001% 6th may not be. Although it may be below the noise level of most systems, pure tones are not masked by noise of "equal level". (Also I think many real amps produce more than 0.001% of the higher harmonics, while still giving low "total distortion" numbers.)
> I've come across that concept before (weighting the harmonic products), but I cannot recall where.
Try Big Red Book, Radiotron Designer's Manual, 4th edition.
> hoping to hear a robust (basically technical) case put forward for FETs being particularly suitable for audio applications.
They are hard(er?) to go wrong with. I think Fred understates his ability as a designer, and I only know FETs superficially, but there seems to be MUCH more to think about in good BJT design than with FETs. Our experience and time is never infinite, and time spent pondering BJT quirks must reduce time spent, as Fred says, "tweaking the design until I come up with what I feel sounds the best..... Then I listen a bunch more..."
> Most people like to think that they have perfect hearing. ...I know that mine is not as good as my recording partner's.
Some study of deafness suggests that mildly deaf people are MORE sensitive to distortion than "normal" people. All of us males over 30 (or any age if we hang around drummers) have some hearing loss. So you may not hear the 19KHz ringing that your chum complains about, but you may be hearing more midrange hash.
On current amplifiers, resistor flaws are swamped by amplifier flaws.
In a super-high feedback amplifier, the amplifier flaws should vanish in the feedback. The feedback resistor flaws will dominate.
With high-quality resistors at "sane" voltages, I doubt the flaws will be audible. (But people do pay more for exotic resistors in passive attenuators and in no-feedback amplifiers.) I doubt I would hear it. Many years ago I saw a write-up on a 3,000 volt direct-coupled electrostatic loudspeaker driver, that should have given 0.01%THD, but voltage-sensitivity in the feedback resistors (exposed to kilovolts of DC and audio swing) added several tenths THD. He used a string of super-good resistors to reduce that effect.
> There will be Acoustic instruments, Microphones, Preamps, Midi Keyboards, ADCs and DACs, Power amps and Loudspeakers, and ...PCs (and not too much else).
The mike preamp should go away. Mike preamp design is (admit it) not doing anything really new. Those ADC designers are riding Moore's Law. Within a few years we will have over 24 bits real resolution (not these so-called 24-bit ADCs with 3 or 4 bits of garbage on the bottom), 144dB S/N, noise level near -130dBm and overload far above any phantom-power mike.
The analog power amp goes away. Working switch-mode rotates your dynamic range problems into frequency-range problems, so at present it is mostly good for subwoofers and some less-great apps like laptop audio and big car audio. But there are tricks, and some designers have made really good totally-digital power amps. They take a digital bit-stream input, juggle some big switches, and dump good analog audio into the speaker.
30 years ago Bell Labs had a digital earphone. Imagine an electrostatic speaker. Divide the surface into areas of 1, 2, 4, 8, 16,... square units. Drive each one with the coresponding audio bit. The sum of the displacement is the audio signal plus the switching spikes. In their app the diaphragm worked in a telephone handset, in a large chamber with a fairly small opening to the ear-- an acoustic low-pass filter.
> a page scanned from a book by John Linsley-Hood:
Thanks. I had (and lost) a more recent essay giving similar curves for a Type 50 tube. I think for the more common multi-stage amplifier, the situation is worse. (It may be best to lump all your distortion in one stage, rather than trying to strain each stage equally.)
Also note that this is unweighted. 0.1% 3rd is probably inaudible, but 0.001% 6th may not be. Although it may be below the noise level of most systems, pure tones are not masked by noise of "equal level". (Also I think many real amps produce more than 0.001% of the higher harmonics, while still giving low "total distortion" numbers.)
> I've come across that concept before (weighting the harmonic products), but I cannot recall where.
Try Big Red Book, Radiotron Designer's Manual, 4th edition.
> hoping to hear a robust (basically technical) case put forward for FETs being particularly suitable for audio applications.
They are hard(er?) to go wrong with. I think Fred understates his ability as a designer, and I only know FETs superficially, but there seems to be MUCH more to think about in good BJT design than with FETs. Our experience and time is never infinite, and time spent pondering BJT quirks must reduce time spent, as Fred says, "tweaking the design until I come up with what I feel sounds the best..... Then I listen a bunch more..."
> Most people like to think that they have perfect hearing. ...I know that mine is not as good as my recording partner's.
Some study of deafness suggests that mildly deaf people are MORE sensitive to distortion than "normal" people. All of us males over 30 (or any age if we hang around drummers) have some hearing loss. So you may not hear the 19KHz ringing that your chum complains about, but you may be hearing more midrange hash.
bluebird
Well-known member
Why are circuits with feedback considered to a lesser design? Some of the best preamps I've worked with, tube and transistor, have used feedback in thier designs. It seems all opamps (API and the like) use feedback to control volume.
Doesn't the Neve designs (BA 283?) have butt loads of feedback resistors all through out the circuit?
Doesn't the Neve designs (BA 283?) have butt loads of feedback resistors all through out the circuit?
AP
Well-known member
I did hear that negative feedback prefers tuna over turkey...
Alan
FredForssell
Well-known member
They aren't, to me anyway. But here is a way to think about negative feedback and how it used...
Think of negative feedback as a coat of paint on a house. Well applied to a well designed and built house, a coat of paint is a great and important final touch. Well applied to poorly designed and built house, it doesn't have the same effect. It can cover up the flaws, but they are still there.
Thanks to Gordon Mercer for the analogy (he said it better than I, but that was the 1970s and I can't remember much from that time period).
AP
Well-known member
It is rather difficult to design analog circuitry without negative feedback. So the people who are wont to pick on things to blame, usually point to overall negative feedback as the bad guy.
Certainly, the reckless use of overall negative feedback can bring on a whole load of problems - throw that in with a bit of slew-rate limiting and you are into all sorts of TID and nasty clipping habits, bursts of instability etc.. And that sort of thing just eats horns and tweeters... (and of course, sounds awful).
But anything, used poorly, can be more harm than good. Used wisely, it is one of the sharpest tools we have.
Alan
CJ
Well-known member
I can see that being a good designer requires two distinct talents, that as a listener, and that as a electronics engineer. It is rare to have both of these qualities in one person. It sure is handy as you don't have to have somebody hanging round to tell you how something sounds.
Great discussion!
:guinness:
Great discussion!
:guinness:
> It seems all opamps (API and the like) use feedback to control volume.
Remember what an Operational Amplifier originally was.
All amplifiers are imperfect. Their gain changes when the voltage changes, as the parts age, or when you change tubes. And if you build two the same, they won't be the same.
As the electronics industry matured, it became clear that gain is cheaper than precision.
In applications like radar controllers, analog computers, and telephone systems, you need an amplifier that has an EXACT gain, that won't change when you look at it funny.
The original work on feedback was to get telephone repeaters to amplify just the right amount. Too little and you can't hear, too much and the long telephone lines echo or howl.
An extension of that concept used the fact that in a feedback amplifier, the amplification is mostly set by the passive feedback parts. These do not have to be resistors. Using a capacitor on front or back gives an electronic equivalent of the mathematical operation of Integration or Differentiation.
Simple problem in Integration: You get in an airplane. Start the engine, take off, climb over a mountain, under a cloud, over a fog bank, then drop toward an airport, and shut off the engine. If you "land" more than a few feet above or below the ground, you crash. In real life this is obvious and expensive, so it is nice to have a trainer. The student pilot sits in the trainer and moves the controls to follow an assigned flight pattern. A computer follows the result of the control motion and computes the flight altitude. If the landing height is not the same as the landing runway height, a little bell tinkles to say "you crashed!". Since the trainer is just a seat in a barrel on the ground, the student can try again until he gets the hang of it.
To compute flight height, you wire the controls to variable resistors. Pulling back on the stick makes a positive voltage, pushing forward makes a negative voltage. You have to multiply by engine power, aircraft weight, etc. Then keeping a running average of the control voltage tells the current flight height. If the voltage at the end of the flight is not the same as the start (or the difference in altitude of two airports), you crash.
Integration is a very nasty thing to compute by hand. By the 1930s there were mechanical computers that used a lot of ball-and-disk integrators. By the 1940s, electronic analog computers replaced them. If you can express a mathematical operation as the ratio of two (possibly very complicated) impedances, and have an amplifier with a gain that is not known exactly but known to be much higher than the ratio being computed, you can do the mathematical operation with two impedances and an amplifier.
In audio, the math operation we use most is "multiply by a fixed constant". We also like "multiply by a constant set by a front panel control". This is just an op-amp plus two resistors.
And the classic op-amp has SO much gain that it is useless without feedback to set a reasonable gain. Like concrete is useless without a form to pour it into to define its shape.
> Some of the best preamps I've worked with, tube and transistor, have used feedback in their designs.
Of course. It is hard, nearly impossible, to build a highly accurate complete audio system without ANY feedback. I even include the "no feedback" tube systems used before the 1930s: these were mostly Triodes which do have negative feedback hidden inside. I'll even grant tetrode radios, because the overall gain was always controlled somehow: either a filament pot or AVC. For most of the last 70 years, feedback has been an essential tool in the designer's handbag. Like any tool, it has been abused: feedback around a swo's ear will not make a silk purse. However it often does make a nicer sow's ear. Or reduces the need for brute-force power filtering.
The wonderful benefits of negative feedback tend to out-shine the problems it causes. Of course the first problem is stability: feeding your output back into your input is bound to get you in trouble, and it took Black and Bode a while to clarify the conditions where NFB can work reliably. This business about complex wideband signals (music) eating their own distortion isn't so easy to grasp.
Remember what an Operational Amplifier originally was.
All amplifiers are imperfect. Their gain changes when the voltage changes, as the parts age, or when you change tubes. And if you build two the same, they won't be the same.
As the electronics industry matured, it became clear that gain is cheaper than precision.
In applications like radar controllers, analog computers, and telephone systems, you need an amplifier that has an EXACT gain, that won't change when you look at it funny.
The original work on feedback was to get telephone repeaters to amplify just the right amount. Too little and you can't hear, too much and the long telephone lines echo or howl.
An extension of that concept used the fact that in a feedback amplifier, the amplification is mostly set by the passive feedback parts. These do not have to be resistors. Using a capacitor on front or back gives an electronic equivalent of the mathematical operation of Integration or Differentiation.
Simple problem in Integration: You get in an airplane. Start the engine, take off, climb over a mountain, under a cloud, over a fog bank, then drop toward an airport, and shut off the engine. If you "land" more than a few feet above or below the ground, you crash. In real life this is obvious and expensive, so it is nice to have a trainer. The student pilot sits in the trainer and moves the controls to follow an assigned flight pattern. A computer follows the result of the control motion and computes the flight altitude. If the landing height is not the same as the landing runway height, a little bell tinkles to say "you crashed!". Since the trainer is just a seat in a barrel on the ground, the student can try again until he gets the hang of it.
To compute flight height, you wire the controls to variable resistors. Pulling back on the stick makes a positive voltage, pushing forward makes a negative voltage. You have to multiply by engine power, aircraft weight, etc. Then keeping a running average of the control voltage tells the current flight height. If the voltage at the end of the flight is not the same as the start (or the difference in altitude of two airports), you crash.
Integration is a very nasty thing to compute by hand. By the 1930s there were mechanical computers that used a lot of ball-and-disk integrators. By the 1940s, electronic analog computers replaced them. If you can express a mathematical operation as the ratio of two (possibly very complicated) impedances, and have an amplifier with a gain that is not known exactly but known to be much higher than the ratio being computed, you can do the mathematical operation with two impedances and an amplifier.
In audio, the math operation we use most is "multiply by a fixed constant". We also like "multiply by a constant set by a front panel control". This is just an op-amp plus two resistors.
And the classic op-amp has SO much gain that it is useless without feedback to set a reasonable gain. Like concrete is useless without a form to pour it into to define its shape.
> Some of the best preamps I've worked with, tube and transistor, have used feedback in their designs.
Of course. It is hard, nearly impossible, to build a highly accurate complete audio system without ANY feedback. I even include the "no feedback" tube systems used before the 1930s: these were mostly Triodes which do have negative feedback hidden inside. I'll even grant tetrode radios, because the overall gain was always controlled somehow: either a filament pot or AVC. For most of the last 70 years, feedback has been an essential tool in the designer's handbag. Like any tool, it has been abused: feedback around a swo's ear will not make a silk purse. However it often does make a nicer sow's ear. Or reduces the need for brute-force power filtering.
The wonderful benefits of negative feedback tend to out-shine the problems it causes. Of course the first problem is stability: feeding your output back into your input is bound to get you in trouble, and it took Black and Bode a while to clarify the conditions where NFB can work reliably. This business about complex wideband signals (music) eating their own distortion isn't so easy to grasp.
bluebird
Well-known member
Very funny Alan :green:
So a circuit like the Hamptone has feedback? Is a grounded signal considered feedback?
Tube circuits like in a Fender Champ, where is the feedback happening?
AP
Well-known member
Yup - every single active device there has local -ve feedback.
The FET has an emitter^H^H^H^H (ha!) source degeneration resistor of 50R. As the input voltage increases, the FET will turn on harder. The increased current through the device will result in an increased voltage drop across the 50R, raising the voltage at the source, which will tend to turn it off again, which will decrease the current through the device. Use of source degeneration is negative feedback, and has all the same qualities - main ones are, decreases gain, increases linearity, stabilises operating conditions. Roughly speaking, if the source resistor is equal to the drain load resistor, -ve feedback would be 100%, and gain would be x -1.
The Darlington is essentially a source-follower (emitter follower), and has 100% negative feedback exactly as described above, except we are taking the output from the emitter, and are not interested in what the collector current is doing. Gain is (almost exactly precisely - nearly) x1.
Even the poor current source transistor that I have so vehemently maligned has -ve feedback (but not much). Because of the voltage across it, the 47k resistor will try to pass a current into the base of the ZTX transistor. This will cause it to turn on, and pass emitter current through the 100R. The increasing current through that will cause an increasing voltage drop across it, the voltage at the emitter (and hence the base too) will rise, reducing the voltage drop across the 47k, which reduces the available base current, which will tend to reduce the emitter current, etc etc. - Negative feedback.
I don't have that schematic (and I don't normally "do" valves). But I bet there will be some, err, 'thingy' degeneration resistors, and ummm, 'wotsit' followers there....
Alan
Transistors have Re that varies with current so there is always some form of local feedback.
I think a grounded cathode triode gain stage(or cathode bias with a big bypass cap)is as close as you can get to no feedback. There is still miller effect.
Is there anything like Re in a triode?
I think a grounded cathode triode gain stage(or cathode bias with a big bypass cap)is as close as you can get to no feedback. There is still miller effect.
Is there anything like Re in a triode?
FredForssell
Well-known member
It depends on the vintage of Champ Amp.
There was a version of the Champ that used a 6SJ7 front-end tube that did not use overall negative feedback.
I think most (all?) of the versions that used 12AX7 tubes used a resistor connected from the secondary of the output transformer to the cathode of the second stage. That's the negative feedback connection.
> Is there anything like Re in a triode?
Triodes have feedback from plate to cathode. The electric field around the grid is not the input voltage, it is the input voltage plus a fraction of the plate voltage. The fraction is set by the geometry of the plate, grid, and cathode. (You can think of it as the relative distances of plate and grid from cathode; in practice it is more about grid wire spacing than distance.) This ratio is Amplification Factor or Mu.
Since we often take output from the plate, this is output-input feedback.
If you work a triode with constant current, the voltage gain is Mu and extrememly linear.
On top of that, the cathode impedance is partly actual gain (transconductance) and partly plain lossy resistance. There is a thin layer where electrons rip away from the cathode according to the electric field they feel (grid and plate voltages), and thicker layers where electrons drag themselves through cathode material and empty space with little regard for the grid and plate voltages.
> So a circuit like the Hamptone has feedback?
The FET unbypassed source resistor is about 6dB-10dB NFB on top of the naturally resistor-y transconductance of an FET. The darlington emitter follower is heavy feedback, at least while in its linear range.
> in a Fender Champ, where is the feedback happening?
I don't know the Champ well. I did once see a small Fender with nearly no feedback except the Mu of the triodes. But in the bigger Fenders: the input stage generally works with no explicit feedback, a triode with moderate load and significant Mu feedback. The second stage often has an unbypassed cathode resistor, often with a small cap to raise gain (and reduce feedback) in treble. The output stage of the big push-pull Fenders always has a resistor from the speaker winding to somewhere in the cathode of the driver stage and applies around 6dB NFB around the whole power amp. This reduces change of gain with change of tubes, takes a bit of the curse of crossover distortion out, flattens amplifier frequency response. It also gives significant damping to the speaker, which would be undamped with naked 6L6 pentodes pushing it. The usual Fender open-back cabinet uses speaker resonance to cancel some baffle loss, but if totally undamped the speaker would slap itself silly at resonance. Small NFB and damping lets it bump a little without getting crazy. And finally: 6dB NFB means you need -half- the power supply filtering for the same hum level, a major economic and portability advantage.
Triodes have feedback from plate to cathode. The electric field around the grid is not the input voltage, it is the input voltage plus a fraction of the plate voltage. The fraction is set by the geometry of the plate, grid, and cathode. (You can think of it as the relative distances of plate and grid from cathode; in practice it is more about grid wire spacing than distance.) This ratio is Amplification Factor or Mu.
Since we often take output from the plate, this is output-input feedback.
If you work a triode with constant current, the voltage gain is Mu and extrememly linear.
On top of that, the cathode impedance is partly actual gain (transconductance) and partly plain lossy resistance. There is a thin layer where electrons rip away from the cathode according to the electric field they feel (grid and plate voltages), and thicker layers where electrons drag themselves through cathode material and empty space with little regard for the grid and plate voltages.
> So a circuit like the Hamptone has feedback?
The FET unbypassed source resistor is about 6dB-10dB NFB on top of the naturally resistor-y transconductance of an FET. The darlington emitter follower is heavy feedback, at least while in its linear range.
> in a Fender Champ, where is the feedback happening?
I don't know the Champ well. I did once see a small Fender with nearly no feedback except the Mu of the triodes. But in the bigger Fenders: the input stage generally works with no explicit feedback, a triode with moderate load and significant Mu feedback. The second stage often has an unbypassed cathode resistor, often with a small cap to raise gain (and reduce feedback) in treble. The output stage of the big push-pull Fenders always has a resistor from the speaker winding to somewhere in the cathode of the driver stage and applies around 6dB NFB around the whole power amp. This reduces change of gain with change of tubes, takes a bit of the curse of crossover distortion out, flattens amplifier frequency response. It also gives significant damping to the speaker, which would be undamped with naked 6L6 pentodes pushing it. The usual Fender open-back cabinet uses speaker resonance to cancel some baffle loss, but if totally undamped the speaker would slap itself silly at resonance. Small NFB and damping lets it bump a little without getting crazy. And finally: 6dB NFB means you need -half- the power supply filtering for the same hum level, a major economic and portability advantage.
