John,
there is no "set up" in my question, at least not intentionally. I clearly made the point in my first question that there is a difference between GFB and LFB and at least the current source should be considered LFB.
But this is not an emitter follower where the feedback mechanism could be debated to be a mathematical issue rather than an "intentionally" feedback loop. The CC source and regulated power supplies are indeed two feedback systems where lots of feedback intentionally is applied to achieve the desired function.
If you claim that say a preamp that is using these topologies is "without global feedback" - well that might be true. But it can not by any means be said to be "without feedback" as far as I see it. Please enlighten me!
/Magnus
there is no "set up" in my question, at least not intentionally. I clearly made the point in my first question that there is a difference between GFB and LFB and at least the current source should be considered LFB.
But this is not an emitter follower where the feedback mechanism could be debated to be a mathematical issue rather than an "intentionally" feedback loop. The CC source and regulated power supplies are indeed two feedback systems where lots of feedback intentionally is applied to achieve the desired function.
If you claim that say a preamp that is using these topologies is "without global feedback" - well that might be true. But it can not by any means be said to be "without feedback" as far as I see it. Please enlighten me!
/Magnus
Swedish Chef said:John,
If you claim that say a preamp that is using these topologies is "without global feedback" - well that might be true. But it can not by any means be said to be "without feedback" as far as I see it. Please enlighten me!
/Magnus
Please don't. More than 50 pages were uselessly written in similar semantic quarrel - just wasting time of us all.
To Hugh : I agree with you in all , some of amps are " tuneable " ( if you know how ), but before you must mainly to know " customer ", if you take it as business. Some peoples are rathed desoriented in terms, which is the best, marketing have in our world realy strong position and peoples are offten " blockhead " , you sure know it.
, some of amps are " tuneable " ( if you know how ), but before you must mainly to know " customer ", if you take it as business. Some peoples are rathed desoriented in terms, which is the best, marketing have in our world realy strong position and peoples are offten " blockhead " , you sure know it.
I like the Swedish chef photo
The feedback debate isn't black and white. All circuits have some feedback paths, either intentional or parasitic. They often have paths that are extremely sneaky: such as through power supply coupling and grounding.
I suppose the debate is more about whether and to what extent it is being deliberately employed for a purpose. At one extreme are those who try to avoid it as much as they can and on the other are those who try to use it as much as they can. Neither achieve their aim completely.
The feedback debate isn't black and white. All circuits have some feedback paths, either intentional or parasitic. They often have paths that are extremely sneaky: such as through power supply coupling and grounding.
I suppose the debate is more about whether and to what extent it is being deliberately employed for a purpose. At one extreme are those who try to avoid it as much as they can and on the other are those who try to use it as much as they can. Neither achieve their aim completely.
John, Upopa,
Thank you for your posts; I am gratified to have some consensus here; it's most encouraging.
Pavel, you make the point about marketing. People will only buy what they want; perception plays a huge part in this. Windows 95 was the highest selling operating system of all time, and it is no coincidence that $US500M was spent on its marketing. Most of us have some understanding of the limitations of this product, but then, many here of technical bent also know the limitations of Linux, which are of an entirely different kind. We could design the world's best amplifier, but the criteria of assessment are one thing, the market perception another, and the commercial success or failure another thing again.......
I am exasperated with the long, circuitous, and tendentious discussions about feedback. I think almost any topology can be made to sound good with care, experience, component selection, operating point, layout, etc. Doubtless someone like John could make germanium transistors and selenium rectifiers sound good! Heck, some even managed years back to make the 12AX7 sound good, and that's saying something.......
What shines through here (or perhaps it is a black hole?) is the manipulation of consumer preference. You have two choices when you design and field a product; you either make the product you think the market wants (notice use of 'wants'), or you create a 'need' with clever marketing, and then fill it with something which is perceived to fit the bill precisely. This latter course appears cynical but we see it all the time in the fashion, cosmetic and music industries. It certainly works, and it has design/manufacturing advantages because it enables genuine long term planning, which is good for business investment. Follow the evolution of auto technology and you see something like this. Almost all prestige autos, for example, are now using four valve cylinder heads with double camshafts whether they need it or not.
So, my conclusion is that you can build an amplifier with or without global or local feedback, balanced input, error correction or other feedforward techniques. You can use tubes or SS. You can choose point to point wiring or even teflon boards! But the use of these topologies, active components and layout techniques can never guarantee a good sound, or that the 'sound' will accord with the wishes of your target market, which incidentally is highly diverse. Therefore, anything goes, and the results are determined largely by the skill of your design team (or designer, as appropriate) and the sales success by your marketing team. In truth, and if the MS experience is anything to go by, the marketing (NOT the R&D) is the most important part, and this is why Stereophile is so incredibly important. Much as the technically inclined might scoff at this august publication, the high end rags influence a huge number of purchasers, particularly in the booming Asian market.
Cheers,
Hugh
Thank you for your posts; I am gratified to have some consensus here; it's most encouraging.
Pavel, you make the point about marketing. People will only buy what they want; perception plays a huge part in this. Windows 95 was the highest selling operating system of all time, and it is no coincidence that $US500M was spent on its marketing. Most of us have some understanding of the limitations of this product, but then, many here of technical bent also know the limitations of Linux, which are of an entirely different kind. We could design the world's best amplifier, but the criteria of assessment are one thing, the market perception another, and the commercial success or failure another thing again.......
I am exasperated with the long, circuitous, and tendentious discussions about feedback. I think almost any topology can be made to sound good with care, experience, component selection, operating point, layout, etc. Doubtless someone like John could make germanium transistors and selenium rectifiers sound good! Heck, some even managed years back to make the 12AX7 sound good, and that's saying something.......
What shines through here (or perhaps it is a black hole?) is the manipulation of consumer preference. You have two choices when you design and field a product; you either make the product you think the market wants (notice use of 'wants'), or you create a 'need' with clever marketing, and then fill it with something which is perceived to fit the bill precisely. This latter course appears cynical but we see it all the time in the fashion, cosmetic and music industries. It certainly works, and it has design/manufacturing advantages because it enables genuine long term planning, which is good for business investment. Follow the evolution of auto technology and you see something like this. Almost all prestige autos, for example, are now using four valve cylinder heads with double camshafts whether they need it or not.
So, my conclusion is that you can build an amplifier with or without global or local feedback, balanced input, error correction or other feedforward techniques. You can use tubes or SS. You can choose point to point wiring or even teflon boards! But the use of these topologies, active components and layout techniques can never guarantee a good sound, or that the 'sound' will accord with the wishes of your target market, which incidentally is highly diverse. Therefore, anything goes, and the results are determined largely by the skill of your design team (or designer, as appropriate) and the sales success by your marketing team. In truth, and if the MS experience is anything to go by, the marketing (NOT the R&D) is the most important part, and this is why Stereophile is so incredibly important. Much as the technically inclined might scoff at this august publication, the high end rags influence a huge number of purchasers, particularly in the booming Asian market.
Cheers,
Hugh
john curl said:
In order to establish PIM as the cause, it's necessary to correlate PIM to open-loop bandwidth. I'm still waiting on my AES CDs, so I'll reserve final judgement until I've read all of Matti's articles on this subject. However, everything I've read so far really points to PIM as just a different manifestation of the same root cause as TIM. That is, the non-constant input stage gm as the differential input voltage varies. In the case of TIM, this would be distortion in the tanh() characteristic of the bipolar input diff amp. It's been shown a number of times by several people that Otala's conclusions about the relationship of TIM to open-loop bandwidth were in error. I'll repeat this again for those that haven't seen it yet. Consider two hypothetical op-amps that are identical in all respects except DC open-loop gain and open-loop bandwidth, yet have the same gain-bandwidth product. For a fixed output voltage the differential input voltage of the op-amp with the lower open-loop bandwidth (and thus higher DC open-loop gain) will be less than or equal to that of the one with the wider open-loop bandwidth at all frequencies. Thus its TIM distortion will be less than or equal to that of the wide open-loop bandwidth design at all frequencies. PIM as I currently understand it is nothing more than the modulation of the gain-bandwidth product gm/Ccomp as the differential input voltage, and thus gm, varies. For a given design, if you minimize the variation in the differential input voltage, you minimize the variation in gm and thus minimize PIM. But as I said before, I'll reserve final judgement until I've completely understood Otala's original PIM articles.
Putting subjective considerations aside per the thread title... from a purely technical standpoint, I have not yet seen any valid technical justification whatsoever for a wide open-loop bandwidth at all in the over twenty-five years that I've been interested in the TIM problem. So the statement above is like saying that there might possibly be some valid automobile tire designs that are not square. In the "Superman and the Bizarros" world of high-end audio http://theages.superman.ws/Encyclopaedia/bizarro.php I suppose that's not so unusual.
AndyC, just because YOU can't find any justification for high open loop bandwidth, does not eliminate the fact that most real audio designs are limited by a real phase modulation due to a non-linear input stage. Look at REAL IC op amps, as they potentially have a serious problem in this regard.
It is true that a super linear input stage will reduce this problem significantly, but then I suspect that we would find another problem buried in the actual performance of these devices.
You equate subjectivity with experience, but it isn't necessarly so. Only a fool doesn't gather experience over time and trials, as to what works and what does not.
It is true that a super linear input stage will reduce this problem significantly, but then I suspect that we would find another problem buried in the actual performance of these devices.
You equate subjectivity with experience, but it isn't necessarly so. Only a fool doesn't gather experience over time and trials, as to what works and what does not.
quote:
Originally posted by PMA
I made a lot of listening tests and audience splitted into 2 halves - 1st prefered output stage outside global NFB, 2nd prefered output stage inside global NFB.
Statistically, nothing conclusive can be inferred from this result...
__________________
The artist formerly known as mikek
Statiscally, it can be inferred they could all hear a difference....
It might also be inferred the two NFB regimes present different nonlinear spectra to the speakers. And which the the listeners perceive as qualitative difference in the musical presentation.
The listeners sorted out NFB taste on the lines of preference for musical genres which are quite different in their demands on the amplifier. I think that's interesting, also.
FrankWW said:
The listeners sorted out NFB taste on the lines of preference for musical genres which are quite different in their demands on the amplifier. I think that's interesting, also.
Statistically this said much about the listeners taste in music, and hardly aything about the amps. per se...
I think I agree with you Andy.
I'm having some trouble with the precise definitions of high and low bandwidth which you are using. Are you saying, if I had two op-amps with identical unity gain OL frequency of, say, 1Mhz, and 6dB/oct roll off prior to this frequency, that the high bandwidth op-amp might have a flat OL gain of say, 10, from dc to 100kHz and the low-bandwidth op-amp would have, say, a flat OL gain of 1000 from dc to 1kHz?
Are you are saying that for the same output voltage the low bandwidth op-amp will see a smaller differential voltage that the high bandwidth one at frequencies below 100kHz and thus less non-linear distortion from the differential stage? If so, then all other things being equal, I agree. I also argue that the low bandwidth device is always superior because correction power is greater below 100kHz due to the higher OL gain.
I'm having some trouble with the precise definitions of high and low bandwidth which you are using. Are you saying, if I had two op-amps with identical unity gain OL frequency of, say, 1Mhz, and 6dB/oct roll off prior to this frequency, that the high bandwidth op-amp might have a flat OL gain of say, 10, from dc to 100kHz and the low-bandwidth op-amp would have, say, a flat OL gain of 1000 from dc to 1kHz?
Are you are saying that for the same output voltage the low bandwidth op-amp will see a smaller differential voltage that the high bandwidth one at frequencies below 100kHz and thus less non-linear distortion from the differential stage? If so, then all other things being equal, I agree. I also argue that the low bandwidth device is always superior because correction power is greater below 100kHz due to the higher OL gain.
I assess using my ears.
I strive to find electrical measures that correlate well with my ears and theorems that allow me to make progress in my designs.
It's a no-brainer that there is room in the world for all sorts. So what?
My experience is that no hifi sounds real to me. The difference between live and reproduced is marked. I'll even venture to claim that 99.9% of all hearing-enabled humans could tell the difference between a live saxophone and a duplicate via a hifi within a few seconds, every time. I can't be bothered with these marketing nuance things - make something that sounds artificial sound a little "sweeter" or a little "faster" or a little this or that. To me this is avoiding getting to grips with the bigger challenge. It is a business-driven approach rather than an excellence-driven approach.
I'm only interested in making hifi sound real.
I strive to find electrical measures that correlate well with my ears and theorems that allow me to make progress in my designs.
It's a no-brainer that there is room in the world for all sorts. So what?
My experience is that no hifi sounds real to me. The difference between live and reproduced is marked. I'll even venture to claim that 99.9% of all hearing-enabled humans could tell the difference between a live saxophone and a duplicate via a hifi within a few seconds, every time. I can't be bothered with these marketing nuance things - make something that sounds artificial sound a little "sweeter" or a little "faster" or a little this or that. To me this is avoiding getting to grips with the bigger challenge. It is a business-driven approach rather than an excellence-driven approach.
I'm only interested in making hifi sound real.
Good point. I think you are right to question the realism of recordings. Many are poor, a few are very good. In my experience the primary loss of realism usually occurs at the reproduction end - in our homes. Recording is somewhat less error prone due to the sophistication of professional recording equipment, the low powers involved and the great improvements in digital transcription in the last decade. But you are right that we are helpless to fix errors in the recording.
It is sort of odd what you've said if you think about it. It sort of suggests that either your judgement is out or that you don't get pleasure out of reality. My experience is that real live instruments, played with competence, sound inherrently pleasurable. Or at least they are stimulating. If a hifi sounds real it must be both a managerie of detail and highly stimulating. I don't hear this combination...yet.
For example, I was at the London hifi show a couple of weeks ago and heard a $70k Krell system and a similarly priced Audio Research/Wilson/Theta system. Conditions were reasonably good - large room, about 30 people, a lot of attention to damping the rooms as you'd expect. Specially-picked recordings. Both these systems sounded what many observers would call "real" in that there was great, spacious sound stage and pinpointed instruments. However, to me it wasn't that realistic. It was clear and clean but not especially pleasurable. Life was missing. Plenty of power and shock value but lacked a certian dynamic expression and vitality that makes real instruments and voices complelling to listen to. I closed my eyes in my local jazz pub last week and imagined I was at the Krell demo again, but it was beyond my imagination: had this been the Krell demo I'd be out of pocket $70k today.
Real = pleasurable!
'Real' is very difficult, if not impossible for hi fi playback. All we can do is get the 'info' from the source to our ears.
If I listen to typical IC op amps, I hear a certain processed sound that appears to remove some of the 'info' from the listening experience. For example, I can't tell the difference between DVD and SACD reproduction on my Sony SACD-DVD player. I attribute this to the IC line amp, which I hope to change out with a discrete design one of these days.
Is it TIM? Is it PIM? Is it crossover distortion, or thermal feedback? I don't know for sure, just that IC's tend to remove subtle information from the audio source.
When it comes to 'Real' and EXCITING: I ALWAYS find that a quality phono playback will give this. Why? I don't know for sure.
I personally find CD playback almost always booring. Why, I don't know for sure, but SACD-DVD playback seems more interesting, but not as good as phono playback.
How many of you even listen to a quality phono system anymore?
Yes, I have questions. Professional questions that I must attempt to answer when I design new audio equipment. However, I use my listening experience and feedback from associates, to answer these questions.
Open loop bandwidth and the use of negative feedback is one of the most important questions that I must address. I still don't have any answer to this. I am now studying Barrie Gilbert's 1998 article that gives some real numbers associated with dynamic phase shift. It is complex and incomplete as far as details are concerned, after all it was only an overview of op amp problems, often overlooked by others. It looks promising to me that PIM generated by a TYPICAL input stage in IC op amps, causes problems with the listening experience.
If I listen to typical IC op amps, I hear a certain processed sound that appears to remove some of the 'info' from the listening experience. For example, I can't tell the difference between DVD and SACD reproduction on my Sony SACD-DVD player. I attribute this to the IC line amp, which I hope to change out with a discrete design one of these days.
Is it TIM? Is it PIM? Is it crossover distortion, or thermal feedback? I don't know for sure, just that IC's tend to remove subtle information from the audio source.
When it comes to 'Real' and EXCITING: I ALWAYS find that a quality phono playback will give this. Why? I don't know for sure.
I personally find CD playback almost always booring. Why, I don't know for sure, but SACD-DVD playback seems more interesting, but not as good as phono playback.
How many of you even listen to a quality phono system anymore?
Yes, I have questions. Professional questions that I must attempt to answer when I design new audio equipment. However, I use my listening experience and feedback from associates, to answer these questions.
Open loop bandwidth and the use of negative feedback is one of the most important questions that I must address. I still don't have any answer to this. I am now studying Barrie Gilbert's 1998 article that gives some real numbers associated with dynamic phase shift. It is complex and incomplete as far as details are concerned, after all it was only an overview of op amp problems, often overlooked by others. It looks promising to me that PIM generated by a TYPICAL input stage in IC op amps, causes problems with the listening experience.
traderbam said:
Yes, that's exactly what I'm saying. Just to clarify, I'll put it into formulas. Mathematically speaking, say we have two amplifiers that are almost identical, with equal gain-bandwidth products. The first one will have an open-loop gain of:
AOL1(s) = A1 * omega1 / (s + omega1)
and the second will have an open-loop gain of:
AOL2(s) = A2 * omega2 / (s + omega2)
Let's assume omega2 > omega1 and A2 < A1. Equal gain-bandwidth products requires that:
A1 * omega1 = A2 * omega2
so the numerators of both transfer functions are the same. When the frequency is high enough, the constant term in the denominator of both can be neglected. If
A1 * omega1 = A2 * omega2 = GBP (rad/sec)
then for high enough frequency, both open-loop gains can be written as
AOL(s) = GBP / s
So there's no benefit either way if the frequency is high enough. However, at low frequencies the gain AOL1(s) is greater than AOL2(s) (because A1 is greater than A2). So for a fixed output voltage at lower frequencies, its differential input voltage is less, therefore less TIM and/or PIM. So if everything else is equal (a tall order in practice I know), then you're never worse off with AOL1(s), and you're definitely better off with AOL1(s) at lower frequencies.
Also, I wasn't arguing subjective issues here, such as which one might be preferable to some listeners. It's just that there doesn't seem to be any hard technical justification for the wide open-loop bandwidth approach from what I've seen so far. Of course there's the "all other things being equal" bit, which in practice complicates the issue because this is rarely the case.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.