Member
Joined 2009
Paid Member
we do like to regurgitate old texts and information and whilst the bulk of it is valuable and in general avoids us reinventing the wheel there are examples where modern information should take precedence or where context is important to avoid misunderstandings. In the case of item a) this is one where context is important.
I have read that whilst small amounts of feedback may not be helpful, once you get enough of it the overall distortion of an amplifier will be reduced. In the case of at least one amplifier I have designed and built, the feedback factor is relatively high and the sound superb.
And I have read that a distortion profile that includes the 5th may not sound bad providing that there is a monotonic fall off in the magnitude of harmonics. However, this is mostly the purview of SE amplifiers, which the JLH (arguably for some) is not.
I've also noticed that many people using modern multi-driver speakers are getting good results with powerful feedback amplifiers with low measured distortion and relatively low output impedance that this affords.
That's the one. The PYE silent room is now DB Cambridge. I have used them to measure things. Just as RFI testing came in Philips sold it off to the staff just to get rid of it!
That book says 100 preproduction prototypes would be made of anything new. I totally doubt that. Not least because most things were variations of previous designs. I might be wrong. I doubt it.
In about 1974 I needed a job to fund college. 23 years later I had almost finished my education ( no ) and was still at that job. I was the service engineer. The boss of the company thought the service engineer sent things back to the original company. I hated that. Whilst I had the pieces of paper I had no real idea of what to do. I simply phoned the companies and could in 5 minutes get to things to test first. Creek Audio were very helpful to the extent of Mike asking what it sounded like to me. Mike always offered the latest idea as a repair.
Like engines the similarities become how I remember them.
About 1973 hi fi was at it's most popular and it took a very bad turn. Suddenly a group of revievers took it onto themselves to look for zero distortion. Radford HD series ( HD250 ) claimed to have got there. A horrid sounding design. Looking up it should have been very good. I am told it's the current limiters doing that. Bland and grey sounding.
Like it or not the market insisted ( and still does with they who do not compare ) on zero distortion. Euphonic sound was a putdown for makers who didn't go on that journey. The DIN standard for caseette decks allowed 5% THD at 333 Hz ( 333? ) and yet 0.1% would be Euphonic. Lies, damn lies and statistics I think ? A good speaker even today 2%.
When you repair things you find nearly every theory you have is wrong. However after 55 years of soldering a truth you never thought real comes to the surface, it's the ears of the designer. Take the Quad 303. It's second harmonic is circa -80 dB 10 watts. The third harmonic about - 93 dB and like the JLH the trend is Euphonic. In theory the 303 is 40 dB better than the true science demands ( 1% THD Euphonic or 0.1% if second cancelled out ). The Quad sounds like it has 1 % to the lovers of zero distortion. It doesn't sound like hi fi, it sounds like music. If very efficient speakers used it can even rock. I have had 115 dB 2 metres using iPhone app from a 303 and had to stop. It was Beatles as you never heard them. As the speakers I used are very good ( 1950s cinema clone ) it was the least compressed sound I ever heard and very dangerous. At 90 dB it was very low distortion. That would be at about 1/4 watt. Nothing heard at a hi fi show got anywhere close and PA too coloured.
Many say the 303 is a failed design and shows a lack of understanding. In this time we live in idiots like that are not rare. If you like hi fi can often be dominated by people with tin ears. Use your own ears. The boss of EMI records told he could read music and play it. New arrivals couldn't. They could splice tape better than him. His system, any turntable, Quad 405, B&W801. Any turntable was where I told him off. Paul McC has Garrard 401 Harmon Kardon Citation and Abbey Road JBLs. Well done Paul and I would be jealous. He has a Garrard RC80 Shure M3D, best bargain ever if you can do the mechanical work it needs. It has rumble which I can live with. 401 is very good on that. I phoned EMI to see if this guy really was who he said. He was, they were very amused about the splicing tape.
Like engines the similarities become how I remember them.
About 1973 hi fi was at it's most popular and it took a very bad turn. Suddenly a group of revievers took it onto themselves to look for zero distortion. Radford HD series ( HD250 ) claimed to have got there. A horrid sounding design. Looking up it should have been very good. I am told it's the current limiters doing that. Bland and grey sounding.
Like it or not the market insisted ( and still does with they who do not compare ) on zero distortion. Euphonic sound was a putdown for makers who didn't go on that journey. The DIN standard for caseette decks allowed 5% THD at 333 Hz ( 333? ) and yet 0.1% would be Euphonic. Lies, damn lies and statistics I think ? A good speaker even today 2%.
When you repair things you find nearly every theory you have is wrong. However after 55 years of soldering a truth you never thought real comes to the surface, it's the ears of the designer. Take the Quad 303. It's second harmonic is circa -80 dB 10 watts. The third harmonic about - 93 dB and like the JLH the trend is Euphonic. In theory the 303 is 40 dB better than the true science demands ( 1% THD Euphonic or 0.1% if second cancelled out ). The Quad sounds like it has 1 % to the lovers of zero distortion. It doesn't sound like hi fi, it sounds like music. If very efficient speakers used it can even rock. I have had 115 dB 2 metres using iPhone app from a 303 and had to stop. It was Beatles as you never heard them. As the speakers I used are very good ( 1950s cinema clone ) it was the least compressed sound I ever heard and very dangerous. At 90 dB it was very low distortion. That would be at about 1/4 watt. Nothing heard at a hi fi show got anywhere close and PA too coloured.
Many say the 303 is a failed design and shows a lack of understanding. In this time we live in idiots like that are not rare. If you like hi fi can often be dominated by people with tin ears. Use your own ears. The boss of EMI records told he could read music and play it. New arrivals couldn't. They could splice tape better than him. His system, any turntable, Quad 405, B&W801. Any turntable was where I told him off. Paul McC has Garrard 401 Harmon Kardon Citation and Abbey Road JBLs. Well done Paul and I would be jealous. He has a Garrard RC80 Shure M3D, best bargain ever if you can do the mechanical work it needs. It has rumble which I can live with. 401 is very good on that. I phoned EMI to see if this guy really was who he said. He was, they were very amused about the splicing tape.
You've offered Crowhurst as a reference for feedback defects for some years now but as Bigun may refer, Crowhurst was writing in 1953 with reference to the Williamson amplifier - a milestone of when..1947? Let's look at what Putzeys had to say regarding SOTA electronics in recent times before exhuming our ancestors to look at the evidence as it stood near 70 years ago. Applying feedback then, was not as well understood and precise as it has become now, in the solid state era.
[Quote by SY, originally from Morgan Jones....
...when cheap gain became readily available, designers became very excited by the possibilities and implications of the feedback equation, and set out to exploit it by designing amplifiers that were thought to have high levels of feedback. In practice, these amplifiers did not have high levels of feedback at all frequencies and power levels, and it was the lack of feedback to linearise these fundamentally flawed circuits that caused their poor sound quality.[/Quote]
Further discussion in the lounge forum, here: Today's Version of The Williamson Amp. BTW, unless you are an AES member, I found access to Crowhurst's article in Radio Electronics Magazine, now has a US $33 fee applied for non-members to take a look
What a cheeky grab for revenue from a magazine article! Thankfully, there are other legible postings on the web.
I've read Putzeys (and mentioned "the F word" somewhere here abouts about a week ago)
The problem is that while an infinite amount of feedback might result in a complete reduction of forward distortion (i.e. that created by the "alpha" in the classic equation) solving the problem of item (a) it does nothing positive for the remains of the items on the list. Indeed it makes many of them significantly worse.
Bruno's thesis also requires infinite headroom and a perfect differential amplifier.
Crowhurst remains the first easily accessible* author to point out that all forms of distortion are not equal and, more specifically, that tweaking an amplifier to reduce distortion can provide a measurably worse amplifier. While we've moved from valves to transistors, the maths hasn't changed.
Nor has our basic auditory biology.
Now Gnobuddy has an interesting thesis that our musical preferences have changed as our background environmental noise has changed over the last 150 years. I certainly subscribe to the theory that very few people are exposed enough to unamplified voice and instruments to know what "real music" actually sounds like, with obvious consequences in terms of audio reproduction.
If you really must have a "modern" reference, and one specifically related to conventional ClassAB design, perhaps look up the musings of Hugh Dean, of Aspen Amplifiers (also AKSA or NAKSA) who was involved in the Melbourne AES.
He takes the view that all amplifiers are imperfect and that one needs to pick your trade offs to get the best result for your context.
*I believe that there was a paper or article from the late 30's but I haven't found it yet. Haven't tried very hard, either
I think your last reference is well known to exploit euphonic distortion as a means to an end i.e. building a better sounding amp. and probably one that has some valve/tube character to its sound quality. Anyone who obviously pays as much attention to having declining levels of harmonics as their order increases, is clearly onto something, as was Jean Hiraga.
I don't think it matters to Hugh or many others here for that matter, what audio design philosophy, techniques or criteria are applied to achieve that "ahhhhh, wonderful music, wonderful sound" as a listening impression. He's tried quite a few different types of design, topolgies etc. and studied the ideas of a lot of the successful designers here, over many years and he'd be quite aware that numbers mean little to the average buyer who just wants sublime sound quality - and purely by subjective appraisal too, if you don't mind.
I don't think it matters to Hugh or many others here for that matter, what audio design philosophy, techniques or criteria are applied to achieve that "ahhhhh, wonderful music, wonderful sound" as a listening impression. He's tried quite a few different types of design, topolgies etc. and studied the ideas of a lot of the successful designers here, over many years and he'd be quite aware that numbers mean little to the average buyer who just wants sublime sound quality - and purely by subjective appraisal too, if you don't mind.
I think we're having a heated agreement
The only item I'd pick out, and it's based on my readings of Hugh and the odd e-discussion a decade ago, is that he was very interested in how his amps measured. Specifically that they measured as well as they possibly could before adding NFB.
That is, he found his best sounding amps still had a little residual 2nd near full output (but less open loop gain) than closely-related circuits with less THD at 90% output which relied on maximising gain and NFB.
If I read you correctly (and I said "if") your inference is that we universally prefer a little H2, whereas my inferences are:
1. in reality (i.e. into a real load with real music) the high-loop-gain amp is producing more "nasties" which we can hear 60+db down (here I lean on Steve Temme's 1992 article for Bruel and Kjaer on detecting manufacturing defects in speakers)
2. when (not if) the high loop gain amp clips, it's going to take longer to recover.
Related to 2 above, any "junk" injected back into the amp from the speakers (be it box resonances, inter cone interference or demodulated RF) it's going to be a bigger problem for the high NFB amplifier.
I concur that a little H2 will sweeten a lone voice or instrumental solo.
But with large orchestral music or massed choirs you don't need much to create a mush of IMD.
So I've come to the thesis that it's not just our rooms and speakers we need to match with the amplifer: it's also our musical tastes and what flaws offend us vs the features that excite us. (As I'm sure I've said before to really enjoy Depeche Mode's "Personal Jesus" in my system I need to swap my JLH out for something with at least 6db more headroom and a lower F3. For the other 95% of my listening, I'll keep the JLH )
Real music can sound very distorted compared to the latest trends in hi fi. I am not talking amplified music only. Real music often suffers from euphoric colourations. Amplified music can make acoustic instruments sound the same only better. John Williams did this with the Rodrigo guitar concerto using A&R A60 and Spendor BC1. I redid it at Oxford town hall. I doubt anyone knew. If not the guitar would not be heard. Personally I would have placed the guitarist in the audience and asked the orchestra to keep time. I have doubts about the BC1 as it sounds wrong like real music. It can sound a bit polite.
I can't argue with your thesis and there's no heat in the debate, as it's an old one. I have followed AKSA's threads for 10 years now and read that from early days, he had a lot of trouble with dissent and ridicule from the engineering fraternity, who's views on distortion or rather the reasoned application of it, remain in strong opposition. If you follow his latest thread and many of XRK's he supports, you'll see how much priority he places on carefully grooming the harmonic content of his designs, albeit those predicted by LTSpice simulations. Just look for the sims....it's not hard to see their purpose.
I think that if one is in the business of selling audio product to a fussy group of critics as much as to the general public, you must identify and meet their specific criteria or you lose favour and business. Money is the great motivator to get the details right and it's no slave to theory or idealistic faith in what should be right rather than what buyers find to sound right in their experience.
So, JLH or a 10-20W ultra low distortion class A amplifier? Which would you or others prefer to use with your favourite recordings?
What thoglette posted about feedback is based on old designs where feedback was limited. Baxandall wrote about small amounts of feedback causing new harmonics being generated that were not present in the original signal. As the feedback increases these signals diminish.
Modern amplifiers using flat gain transistors and higher feedback levels do better.
In investigating those old designs, a standard Class AB amp from 1970's designs typically generated 0.2% distortion - which sounds (on paper) not bad - but in fact led to rather larger amounts of intermodulation distortion, which seemed to me to be far more serious in degrading the actual sound quality. When THD can be got to below about 0.01% the IMD become much less and the overall sound much better as a result.
Modern amplifiers using flat gain transistors and higher feedback levels do better.
In investigating those old designs, a standard Class AB amp from 1970's designs typically generated 0.2% distortion - which sounds (on paper) not bad - but in fact led to rather larger amounts of intermodulation distortion, which seemed to me to be far more serious in degrading the actual sound quality. When THD can be got to below about 0.01% the IMD become much less and the overall sound much better as a result.
Member
Joined 2009
Paid Member
this reasoning leads to the oft repeated advice to use no feedback or lots of it but don't dawdle in the middle lane. However, some folk have made amps with low feedback on purpose, Hugh had this objective in mind with a couple of his amps. There's really no simple rule for the best sounding amp anymore than there is a rule for the best written poem.
I have the issue of Wireless World mentioned by John Ellis (December 1978) where the Baxandall article appeared.
My reading of this is Baxandall was laying out a future case aiming to reduce the harmonics to a somewhat ideal 120 dB below the fundamental where higher order harmonics (generated consequent to use of low levels of negative feedback) would become negligible.
Linsley-Hood did not subscribe to low levels of feedback but to use enough without compromising transient response.
My reading of this is Baxandall was laying out a future case aiming to reduce the harmonics to a somewhat ideal 120 dB below the fundamental where higher order harmonics (generated consequent to use of low levels of negative feedback) would become negligible.
Linsley-Hood did not subscribe to low levels of feedback but to use enough without compromising transient response.
Sometimes a very large amount of negative feedback sounds fantastic. Many op amps using 100% AC and DC feedback sound much like a piece of wire with current gain. At a gain of 10 upwards they sound slightly euphoric but good. To my ears gains of 2 to 10 need more care. If designing a phono stage that might help you. The open loop gain is massive which needs thinking about. Should be bad but isn't. Simple reason is much thought went into them.
SE valve amplifiers like no obvious feedback. Even halving the gain can make them sound a bit nasel. This can be overcome if prepared to do lab work.
FET designs like the Hitachi MOSFET loves feedback and sounds very like real music.
It is argued a conventional class AB could better use 20 dB feedback or reduction by 10 which is modest and plenty of local feedback. Advocates of the science despise that. I suspect it's right even if rare. I t means bias is critical to get low high order distortion.
SE valve amplifiers like no obvious feedback. Even halving the gain can make them sound a bit nasel. This can be overcome if prepared to do lab work.
FET designs like the Hitachi MOSFET loves feedback and sounds very like real music.
It is argued a conventional class AB could better use 20 dB feedback or reduction by 10 which is modest and plenty of local feedback. Advocates of the science despise that. I suspect it's right even if rare. I t means bias is critical to get low high order distortion.
Negative feedback does far more than reduce distortion. It alters input and output impedance. This can be mistaken for distortion. When this is understood the learning really starts. Even positive feedback is reducing distortion. I saw a very cheap amplifier by Velleman using input positive feedback. Wow that's brave. It also uses it as the JLH does. Positive feedback might be as important as negative feedback. To for no great difficulty raise the amplifier input impedance makes driving the amplifier easier whilst keeping noise low. Suddenly an input only suitable for opamps can be driven by the beautiful anode follower valve circuits. I am not a fan of the automatic use of buffer stages. Even the cable is important and favours TV coax of less than one metre. Use conventional cable for bad sound. These are the real distortions.
Focusing on negative feedback is like thinking cosmetic surgery improves all parameters. I said to a friend I did some work for that I am hopefully more scientist than Audiophile. They know not to do something and I know why. This was about a misconception he had that would be very dangerous. If you knew the whole story it's also about becoming more modern and maintenance of standards.
Focusing on negative feedback is like thinking cosmetic surgery improves all parameters. I said to a friend I did some work for that I am hopefully more scientist than Audiophile. They know not to do something and I know why. This was about a misconception he had that would be very dangerous. If you knew the whole story it's also about becoming more modern and maintenance of standards.
Member
Joined 2009
Paid Member
If you think about it coax should be better than anything. Some turntable wires were more like coax. There is a belief that wires should have about four times the capacitance as coax for a better shielding. In fact very often unshielded cables given that it works can sound better. Twisted unshielded are worth trying. Specific TV cable is more engineeed for 700 MHz. It uses the skin and avoids the iron core. That is used to suit vertical aerial use.There is a plating on the iron alloy. Audio specific types use ofc.
Take a 1950s valve pre power set up and hear the sound you have always wanted. It will have very often the most miserable cable you ever saw. Do some cosmetic surgery and if you are honest you just killed it. The valve most likely an ef86. Leak Varislope and Stereo 20 come to mind.
Take a 1950s valve pre power set up and hear the sound you have always wanted. It will have very often the most miserable cable you ever saw. Do some cosmetic surgery and if you are honest you just killed it. The valve most likely an ef86. Leak Varislope and Stereo 20 come to mind.
I remember who first told me about coax in audio. It was John Dawson of Arcam. It was used for a DIN to RCA sockets. RG58. To quote Belden , all copper, 25 pF per ft and 90% shielding. Maximum load 40 lbs. This is a 50 ohm cable. 75 ohms has slightly less capacitance. Let's say 65 pF per metre whereas standard can be 220 pF per metre. The 40 lbs is for when supporting it's own weight I believe. Some masts can be more than 100 metres high. Steel wire, copper plus silver is better for non audio.
Low levels of negative feedback will reduce the low order harmonics and pump up the higher ones you don't want. In Baxandall's test to reduce the harmonics to the non-feedback level the sixth harmonic needed a feedback level of about 35 dB. That is about the level used in the Simple Class A amplifier.
In a conventional Class AB circuit 20dB of negative feedback will give a poor result since that will be insufficient to deal with the high order harmonics in the small signal stages with seriously more of these being generated in the power output stages that follow.
Scientist Dave Kimber ( if I get his name right ) DF96 died recently. Sadly he can not join us to explain. He was very insistent that higher order distortion is not created in feedback amplifiers. It is simply something that can not be corrected by feedback because the. Nyquist stability point has taken over. I did some very careful measurements and found he was right. The distortion always was there except it was harder to see. Douglas Self puts it another way and says if the Nyquist point can be extended performance might be enhanced. Two pole VAS compensation his answer. It looks dreadful in some simulations. It measures very well. If the amplifier can accept it the poles can be linked to the output stage. This supposes that VAS compensation is used. The JLH sort of does in it's VAS transistor capacitance. This makes changing it risky. The JLH shouldn't have Nyquist stability issues but does. In my opinion this is due to dividing the output current. Again playing with the second harmonic distortion may induce the very same problems being thought about.
Dave maybe was being pedantic. Effectively the residual distortion is the only thing left and negative feedback has squashed everything. This is a warning and says is it really such a good Idea to keep applying it when a remarkable performance is already available.
Let's look at that. If we raise the JLH gain to double we still might meet the JLH design brief of 0.1 THD. What we might have is a more dynamic sound with more micro detail in the music. You also would have an amplifier that can work directly from CD player levels or phono stages. Reducing complexity brings micro detail.
Dave worked with Professor Higgs. He thought Higgs was brave enough to publish something the others had discussed. Although I didn't ask I got the impression it was one of a myriad possibilities.
Dave maybe was being pedantic. Effectively the residual distortion is the only thing left and negative feedback has squashed everything. This is a warning and says is it really such a good Idea to keep applying it when a remarkable performance is already available.
Let's look at that. If we raise the JLH gain to double we still might meet the JLH design brief of 0.1 THD. What we might have is a more dynamic sound with more micro detail in the music. You also would have an amplifier that can work directly from CD player levels or phono stages. Reducing complexity brings micro detail.
Dave worked with Professor Higgs. He thought Higgs was brave enough to publish something the others had discussed. Although I didn't ask I got the impression it was one of a myriad possibilities.
The second harmonic is the result of the asymmetry of the cascade.
The third harmonic is caused by the nonlinearity of the amplification devices (beta-decay, input and output characteristics).
Higher-order uncorrelated harmonics are caused by switching push-pull or bridge amplifiers.
Higher harmonics are also given by a rigid amplitude limitation.
Negative feedback can add overshoot distortion during an oscillatory transient.
The load is affected.
The third harmonic is caused by the nonlinearity of the amplification devices (beta-decay, input and output characteristics).
Higher-order uncorrelated harmonics are caused by switching push-pull or bridge amplifiers.
Higher harmonics are also given by a rigid amplitude limitation.
Negative feedback can add overshoot distortion during an oscillatory transient.
The load is affected.
Last edited: