Having looked at my scrap box parts now looking at proper soulutions. ADM7151 looks an idea.
Low noise 1.0 μV rms total integrated noise from 100 Hzto 100kHz. 1.6μV rms total integrated noise from 10 Hz to 100 kHz
Noise spectral density: 1.7nV√Hz from 10 kHz to 1 MHz
Power supply rejection ratio (PSRR) at 400mA load >90 dB from 1 kHz to 100 kHz, V=OUT= 5 V >60 dB at 1 MHz, V OUT= 5 V
I need about 1.6 amps so banking them would be ideal. Perhaps a 0R22 summing resistor to make it possible to each output. The regulator is like LM317 in the voltage setting by resistor divider times reference voltage ( 1+ R2/R1 ) x ref. . If so some statistical noise cancelling should be a bonus. I can add a pass transitor which defeats spending the extra money.
Looking at the noise density TL431 seems a non starter at 50 to 170 nV root/ Hz. This seems to relate to about 2 uV noise which is 20 dB better than bog standard regulators. LM7151 seems to imply 1.7 nV. However the overal noise density is quoted at 1.6 uV . Hardly seems worth the trouble for that alone. The winning question is 60 dB rejection at 1MHz. I wonder with the output capacitor of my simple design does that relate to a real advantage?
Low noise 1.0 μV rms total integrated noise from 100 Hzto 100kHz. 1.6μV rms total integrated noise from 10 Hz to 100 kHz
Noise spectral density: 1.7nV√Hz from 10 kHz to 1 MHz
Power supply rejection ratio (PSRR) at 400mA load >90 dB from 1 kHz to 100 kHz, V=OUT= 5 V >60 dB at 1 MHz, V OUT= 5 V
I need about 1.6 amps so banking them would be ideal. Perhaps a 0R22 summing resistor to make it possible to each output. The regulator is like LM317 in the voltage setting by resistor divider times reference voltage ( 1+ R2/R1 ) x ref. . If so some statistical noise cancelling should be a bonus. I can add a pass transitor which defeats spending the extra money.
Looking at the noise density TL431 seems a non starter at 50 to 170 nV root/ Hz. This seems to relate to about 2 uV noise which is 20 dB better than bog standard regulators. LM7151 seems to imply 1.7 nV. However the overal noise density is quoted at 1.6 uV . Hardly seems worth the trouble for that alone. The winning question is 60 dB rejection at 1MHz. I wonder with the output capacitor of my simple design does that relate to a real advantage?
In the John Curl Blowtorch thread, the issue of distrtion audibility came up again. As always, we're none the wiser.
I have seen several papers, stating that we can't hear THD of 0.3% to 0.003%.
We all know that tube gear produces THD figures of 3% and more, while IM figures are not unusually even worse. Yet, we have a good number of people who claim that sound is much more true to life than what others want from solid state systems.
How do we reconcile these directly opposing views?
And in general, we do NOT have this under control by default. We may have amps with just 0.008% THD and IM, but ultimately, our overall THD and IM will be defined not by the amp alone, but with our source distorion factored in, and woe, with our loudspeaker distortion as well.
Where does this leave us?
I have seen several papers, stating that we can't hear THD of 0.3% to 0.003%.
We all know that tube gear produces THD figures of 3% and more, while IM figures are not unusually even worse. Yet, we have a good number of people who claim that sound is much more true to life than what others want from solid state systems.
How do we reconcile these directly opposing views?
And in general, we do NOT have this under control by default. We may have amps with just 0.008% THD and IM, but ultimately, our overall THD and IM will be defined not by the amp alone, but with our source distorion factored in, and woe, with our loudspeaker distortion as well.
Where does this leave us?
Very easily, Dejan - systems have distortion, not components; and distortion is complex, not simple. The only reasonably capable device for measuring subjectively meaningful performance, at the moment, is the ear/brain ... and at the rate measuring techniques are evolving that will likely be the case for the next century,
...I agree completely, Pavel.
It's the logical background that intrigues me. Theoretically, the less distortion produced in the amp, the better, yet this doesn't seem to take into account how that low distortion was achieved.
Many years ago, in 1979, I borrowed a German made power amp, made by a company long gone called LAS (Linear Audio Systems). It sounded WAY better than I would have expected it to, so it intrigued me. I picked it apart.
Simply done, but with a tremendous focus on the end result. I posted it here twice, do't have it available right now, but I'll look it up. Some very novel features for that time, from which I learnt a lot.
It had an OL bandwidth of just 4 kHz, yet offered a slew rate of 100 V/us and a peak current of approximetaly 40 Amperes, using BD 249/250 C output devices (4 pairs per channel), which had 1.5 Ohms emitter resistors. I use their overvoltage/overcurrent protection to this day, damn well thought out. And it was the first audio amp I ever encountered which contained much thinking about settling times, not just rise times. Model was Mega 1, because its response extended to 1 MHz.
In brief, it was made in exactly the way I would never make it. Yet, it sounded truly excellent, easily the best amp sound I had heard until that time, including the likes of SAE and Accuphase amps.
It forced me to accept the fact that it's not all so simple at all, and that looks can cause wrong conclusions. In other words, it made me realize that schematics and figures are all well and nice, but until you hear it, you don't really know much about it.
A similar thing happened with some tube audio. It measured rather poorly, but it made some excellent sounds.
Go figure.
It's the logical background that intrigues me. Theoretically, the less distortion produced in the amp, the better, yet this doesn't seem to take into account how that low distortion was achieved.
Many years ago, in 1979, I borrowed a German made power amp, made by a company long gone called LAS (Linear Audio Systems). It sounded WAY better than I would have expected it to, so it intrigued me. I picked it apart.
Simply done, but with a tremendous focus on the end result. I posted it here twice, do't have it available right now, but I'll look it up. Some very novel features for that time, from which I learnt a lot.
It had an OL bandwidth of just 4 kHz, yet offered a slew rate of 100 V/us and a peak current of approximetaly 40 Amperes, using BD 249/250 C output devices (4 pairs per channel), which had 1.5 Ohms emitter resistors. I use their overvoltage/overcurrent protection to this day, damn well thought out. And it was the first audio amp I ever encountered which contained much thinking about settling times, not just rise times. Model was Mega 1, because its response extended to 1 MHz.
In brief, it was made in exactly the way I would never make it. Yet, it sounded truly excellent, easily the best amp sound I had heard until that time, including the likes of SAE and Accuphase amps.
It forced me to accept the fact that it's not all so simple at all, and that looks can cause wrong conclusions. In other words, it made me realize that schematics and figures are all well and nice, but until you hear it, you don't really know much about it.
A similar thing happened with some tube audio. It measured rather poorly, but it made some excellent sounds.
Go figure.
To me, the low distortion is just a necessary condition. Not necessarily the satisfactory condition. We may have link stages that measure 0.001% everywhere in audio band and still sound different.
We have tools like HF analysis, time domain analysis and we are able to find differences. We are able to set the design rules that would lead, with high probability, to a good sounding device. However, we still need ears and intensive listening tests to make a final decision. It is not frustrating, it is challenging and it is the reason why we have so much fun with audio.
The first thing to undestand is why a distortion exists. If it is simple bend distortion of an amplifying device it is a more benign distortion than crossover distortion. First requirement is an analyser to know the nature of the problem. If you see 1% THD at full power with regular harmonics that shows very little fifth harmonic it is safe to assume it will sound undistorted. If you look carefully at valve amps they seem to fall at every fence. A 3 kHz square-wave test should send them to the dustbin. Listening usually proves them to be oddly excellent. There will be a few problems that can be mistaken for distortion. Low damping factor and LF and HF reduction. If a SE valve amp drives something like a 12 Lta PA driver from 100 Hz to 12 kHz it might be ideal. A transistor bass and whatever class A tweeter amp ( OPA 604 with class A complimentary feedback pair dumpers, no loop feedback. Or ECC82 in PP with small transformer ).
I supect the most awful distortion is not enough current, true if valve or transitor. It is often very hard to show why. I am convinced that is the best way to thinking as all other things point that way. Other distortions are obvioius and only exist because people refuse to buy measuring equipement. Valve amps teach us much of transistor distortion is visualised by setting certain rules in place. Valve amps prove these rules are meaningless and yet are cures fror transistor amps. If you like fever is proof of a virus without being able to detect a virus. Valve amps less prone to viruses but will get bacteria as an analogy. I have seen transistor distortion. It is a crossover glitch not at the centre line. Mostly the amp sounds bad yet 95 % of tests say it's good. Simple distorion bacteria, nasty stufff as virus.
One reason I think valves work well is simplicity is a virtue and a need of the designs. My feeling with transisors is why use one resistor when 30 components would do a better job. This is somehow linked to commersial amps where a low grade PSU is likely. > 90% of what I see is designed this way. That is the people have been schooled by looking at common practice. That's great if designing a chip for an amp. Only a fool would dismiss the oppertunity to get it right as 30 extra semicondutors is no problem.
All who have not built a valve amp should. The PSU will be big trouble. Not least if like me you refuse to spend money. SE with EL 34 and no loop feedback is my advice. Double EL 84 and op amp ? Don't bother with other things as to be frank they won't impress and will need serrious money spent. If you don't get 1 % THD full power and 0.2% 1 watt you have failed. Also no loop feedback is something to want. Hum must be as low as a transisor amp and hiss. If not you have failed. EL 84 is possibly the best device of all.
Jean Hiraga thought it is not the distortion that matters. It is the slope between harmonics. He could not make his mind up if a linear or an exponential slope. His instinct was the latter as music works this way. Thus an amp with - 70 dB only 5 th harmonic will sound bad and a 1% THD valve amp to Hiraga rules sounds fine. Hiraga belives the distortion of such amps is like a TV picture. The grey is mistaken for black until grossly wrong ( < 2 % ? ). Conversely a tint of blue in a black and white movie is wrong. Old TV was bad in this way. Because I knew when a TV tube was going more than most it made it almost impossible to watch. I love black and white and hate blue shadows.
If someone says valve amps only sound good because they distort I will say nonsense. That's a bad valve amp. They do sound interesting but like the black and white they have a colour cast. The quality you might get is you almost hear things before you do. This is a transitor amp quality also. The odd thing is valves seem to do it better. Although very hard to prove loop feedback does seem to mildly change the grain of the photo. Dejan although your amps are very hard to brake down into the plumbing, the gas supply and electricity if you like I do support two ideas which I like. Local feedback in sensible amounts ( valves do if triode like it or not ). Low loop feedback and enough drive current. Where I doubt it matters is VAS linearity although I see no harm in it ( by making it more linear you reduce the gain. As loop gain it restors, it might just be better in the VAS than in the loop ? ). The extra standing current allows the clean up of the crossover distortion to be less critical as the gM doubling is deliberte. That is the amp is mildy class G in how it looks ( A + AB ).
Nige, as you know, I agree.
What prompted me to start this thread is that there seems to be no agreement regarding the absolute thershold of audibility, which has caused many engineers to get into the "zero race", i.e. how many zeros you have before you come to a number other than zero.
Regarding crossover distortion, I agree it's a particulary offensive form of distortion, but it's not too hard to detect. If it's happening, sooner or later up the bandwidth it will become THD. If your amp is showing relatively easy increase rates of THD as you go up the scale, you will reach a point when they suddenly take off out of any proportion. I would suggest this is where crssover distortion comes into play. Not a sure fire method, but I find that is so in say 90% of all cases.
And you can check it out fairly easily. If you add some bias current, and that sudden THD jump moves upwards, or is significantly reduced, you have the beast by its throat.
I believe peak THD values of 0.05% are inaudible, but we have seen people here claim they can hear 0.0005% differences - which I sincreley doubt.
What prompted me to start this thread is that there seems to be no agreement regarding the absolute thershold of audibility, which has caused many engineers to get into the "zero race", i.e. how many zeros you have before you come to a number other than zero.
Regarding crossover distortion, I agree it's a particulary offensive form of distortion, but it's not too hard to detect. If it's happening, sooner or later up the bandwidth it will become THD. If your amp is showing relatively easy increase rates of THD as you go up the scale, you will reach a point when they suddenly take off out of any proportion. I would suggest this is where crssover distortion comes into play. Not a sure fire method, but I find that is so in say 90% of all cases.
And you can check it out fairly easily. If you add some bias current, and that sudden THD jump moves upwards, or is significantly reduced, you have the beast by its throat.
I believe peak THD values of 0.05% are inaudible, but we have seen people here claim they can hear 0.0005% differences - which I sincreley doubt.
I know what is being said and is not nonsense at all. I have seen a strange logic in designs. I asked what happens to a shunt regulator at 30 MHz? Everyone said it stops doing anything. To which I said what's the point? Shunt regulators look instinctively right. In truth they are mostly interesting. Take the cheapest high speed TV devices. No way would any of the high flung PSU's in DIY Audio do anything for these. And yet they work? How come? The decoupling is mostly the power source. If not so we would have made a lie. Maybe when calling them decoupling we closed our minds?
I was told that Crimson Electronics used a chain of resitors and zener diodes in their amps. They worked out step by step the needs for headroom. For pennies they built a vastly better design by doing this one thing. That's genius. They used a new step when they could. I was told to avoid zener diodes in 1972 because they are noisey. " You can filter a zener Nigel, but why bother when regualtors are vastly better Nigel ". Nigel trusted this advice. It was B-ll-cks.
Can I better this? Until I know more NE 5534 22 pF comp and fast big NPN power transitor ( 5534 is 4 nV root/Hz noise and will drive the transistor without needing a Darlington ). I suspect to use the higher voltage type of LT 6655 and least 5534 gain? I exspect the 5534 to be trouble. Any common device I have overlooked? TL431 is my bench mark and looks to be 2 uV.
LTC6655 - 0.25ppm Noise, Low Drift Precision References - Linear Technology
I was told that Crimson Electronics used a chain of resitors and zener diodes in their amps. They worked out step by step the needs for headroom. For pennies they built a vastly better design by doing this one thing. That's genius. They used a new step when they could. I was told to avoid zener diodes in 1972 because they are noisey. " You can filter a zener Nigel, but why bother when regualtors are vastly better Nigel ". Nigel trusted this advice. It was B-ll-cks.
Can I better this? Until I know more NE 5534 22 pF comp and fast big NPN power transitor ( 5534 is 4 nV root/Hz noise and will drive the transistor without needing a Darlington ). I suspect to use the higher voltage type of LT 6655 and least 5534 gain? I exspect the 5534 to be trouble. Any common device I have overlooked? TL431 is my bench mark and looks to be 2 uV.
LTC6655 - 0.25ppm Noise, Low Drift Precision References - Linear Technology
Last edited:
Nige, I still think my "virtual battery" stabilizer is overall the best way to go. There are better regulators, but they are much nore elaborate for what I feel is too small a difference in sound quality. Others may feel differently.
The reason why I include the predriver as well is twofold. One thing is that it provides a very stable current drive, that's obvious. But the second reason is that it comes on after the bias determining circuit (simulated zener), thus also indirectly stabilizing that circuit as well.
The reason why I include the predriver as well is twofold. One thing is that it provides a very stable current drive, that's obvious. But the second reason is that it comes on after the bias determining circuit (simulated zener), thus also indirectly stabilizing that circuit as well.
Hi Nigel !
My personal opinion is that one of the key ( but not the major one) of usually good sound of tube amps is the usually used Point To Point hard wiring layout technique , why this ancient layout technique is not used in transistor amps but only Printed Circuits Boards with more and more packing density of active& passive elements was always the question for me .
My personal opinion is that one of the key ( but not the major one) of usually good sound of tube amps is the usually used Point To Point hard wiring layout technique , why this ancient layout technique is not used in transistor amps but only Printed Circuits Boards with more and more packing density of active& passive elements was always the question for me .
DVV there may not be an absolute threshold but rather a range based on the listener experience and native language. People hear sound differently in places that have different languages . A learned skill. Different but not bad. Albert E. said make it as simple as possible but no simpler. Here is where good tube amp win and lose in where the line is . My take is design a functional unit that is stable and low distortion for each section use feedback as a band-aid not a splint . Power supplies are the heart of the unit . Just my thoughts.
- Status
- Not open for further replies.