Charles Hansen said:
I still have that old correspondence from 2001. I was asking how the null pins were connected, as they are not shown in the simplified schematic on the data sheet.
I don't suppose you want to add your exchange with Barrie to the Jung, Cordell, and Curl letters?
😉
No, I wasn't serious.😀
Scott, we are perfectionists. Charles is even crazier than me, and LESS tolerant of global negative feedback. He inspires me to do better. You could learn from him as well.
We WANT you to succeed in audio design, as much as you have been successful in commercial design. Really, honest! Right now, I am conducting some tests on distortion changes by adding a current source to the - output. It isn't easy, but I will not allow the actual distortion to be hidden by total noise, because the ear doesn't hear that way. I am now signal averaging to get the noise out, then I can measure with an output of .3-1V without compromise. If you don't, you better give it a try. Look for 7th harmonic distortion. This is the worst harmonic in the musical scale. ANY 7th is a problem, no matter how low it appears with negative feedback applied.
We WANT you to succeed in audio design, as much as you have been successful in commercial design. Really, honest! Right now, I am conducting some tests on distortion changes by adding a current source to the - output. It isn't easy, but I will not allow the actual distortion to be hidden by total noise, because the ear doesn't hear that way. I am now signal averaging to get the noise out, then I can measure with an output of .3-1V without compromise. If you don't, you better give it a try. Look for 7th harmonic distortion. This is the worst harmonic in the musical scale. ANY 7th is a problem, no matter how low it appears with negative feedback applied.
john curl said:
...I will not allow the actual distortion to be hidden by total noise, because the ear doesn't hear that way.
Something that tube detractors refuse to comprehend. The ear is able to listen beneath the noise inherent in tube circuits in a manner that is counterintuitive unless you consider it as an adaptation to detecting the stealthy footsteps of an approaching predator under the sound of rustling grass. Once you view it as a survival trait, it makes it somewhat more understandable.
The question now becomes: What's the difference between tube noise and solid state noise? Why can't you listen "beneath" solid state noise as easily as tube noise? There's more tube noise--no one denies that--yet it's easier to listen through. Probably a spectral content thing, at least in part, but I'm not set up to measure that sort of thing easily and that would still leave begging the question of interpretation.
I approach it the same way I do bandwidth. For whatever reason, it seems that it takes more solid state bandwidth to approach the perceived bandwidth of a tube circuit. In terms of noise, I reduce the total solid state noise in an attempt to increase the listener's ability to hear into the signal. It's a brute force solution but it seems to work...partially. Even a nearly silent solid state system doesn't quite approach the "hear into it" effect of tubes. This irritates me to no end. Fortunately, noise is easier to manage in solid state, so at least that's on the plus side of the equation. I need all the help I can get.
For what it's worth I've always felt that solid state and tubes modulate the noise differently when playing music. It then becomes a dynamic problem, not a normal static bench-test thing. Yes, this begins to shade into THD+noise measurements, wherein you lump all the non-music things together after subtracting the original signal, but what I'm envisioning is different; not sine waves...something more demanding, like a square, or better yet, a spike. Look at noise after the signal. If you want an analogy for the effect, think of the noise ghost that follows the signal up and down when a limiter or active noise filter is in the circuit. The initial impulse drives the circuit's window fairly quickly, but there's an overhang as the circuit drags behind, not closing as quickly as the signal does. In gross form, people speak of pumping and similar terms, but that doesn't quite cover the effect I'm describing. There's an aspect to the quality of the noise, not the quantity. One question I have yet to resolve is whether the noise is spectrally related to the music, hence blending with it in the same manner that 2nd or 3rd harmonic distortion become part of the music...or whether it's so spectrally different that the ear is able to set it aside as unrelated, hence 'ignorable.' If you're listening for a predator, you can ignore the rumble of distant thunder, it's not relevant to the immediate threat.
(Unless you want to treat it as dramatic foreshadowing, but that's another matter entirely...)
Grey
Even more simply, THD uses WIDE bandwidth that creates a HIGH noise floor. The ONLY way to get below this floor is to signal average during FFT processing. Another way is very slow harmonic sweep with a non-FFT spectrum analyzer.
GRollins said:
For what it's worth I've always felt that solid state and tubes modulate the noise differently when playing music. adowing, but that's another matter entirely...)
Grey
Tube diff-pairs biased with really bad carbon comp resistors generate lots of signal modulated noise and you could easily convince anyone that they were hearing it. Noise modulation from non-stationary gm in mixers is also a problem. Yes these effects are real, I think there is plenty of room for some good theoretical work there just as people got into PM not all that long ago IIRC.
john curl said:
We WANT you to succeed in audio design, as much as you have been successful in commercial design. Really, honest!
I feel the love.

I doubt it, and I think that you are overly concerned about tube resistor noise. To be sure, there are bad carbon resistors, but usually they are weeded out and replaced by good ones.
john curl said:Doing the same analysis with the LT1028, I found that there was almost no extra current available for the output stage. The trend, it seems, is to starve the output stage. We are going backwards. We might as well update the UA709, with its class C output stage. 😕
LT1028 OP stage bias is only about 0.5mA ...
More opinions (questions?) from the balcony:GRollins said:
Something that tube detractors refuse to comprehend. The ear is able to listen beneath the noise inherent in tube circuits in a manner that is counterintuitive unless you consider it as an adaptation to detecting the stealthy footsteps of an approaching predator under the sound of rustling grass. Once you view it as a survival trait, it makes it somewhat more understandable.
The question now becomes: What's the difference between tube noise and solid state noise? Why can't you listen "beneath" solid state noise as easily as tube noise? There's more tube noise--no one denies that--yet it's easier to listen through. Probably a spectral content thing, at least in part, but I'm not set up to measure that sort of thing easily and that would still leave begging the question of interpretation.
Do tubes exhibit 1/f noise or pop-corn noise? I ask because I remember tube noise as mostly a high frequency hiss with little low frequency content or modulation. It would be far easier to hear coherent signals (such as music) through a steady unmodulated high frequency noise envelope than through a truly random noise source with a lot of low frequency content.
If the noise spectra are the same or even similar, disregard this post as blather.
My home system is all solid state, I like the seductive tube sound, but all the tube amps I've heard that I can afford, fall down with complex musical passages or with more than a few musical instruments playing at once.
hermanv said:
but all the tube amps I've heard that I can afford, fall down with complex musical passages or with more than a few musical instruments playing at once.
Absolutely correct. That is why I cannot vote for tube power amps, as I listen to classical music/philharmonic orchestra mostly. The euphonic sameliness of tube sound is driving me mad 😉
Tubes can sound great, but it takes big money to make it so, unless you DIY it. Even then, you better know what you are doing. Of course, they have sort of an IM 'haze' when you use just a stock Dyna or Heathkit.
Tubes can sound great, but it takes big money to make it so, unless you DIY it. Even then, you better know what you are doing.
Even if you diy it, the money is not inconsiderable!
hermanv said:
Do tubes exhibit 1/f noise or pop-corn noise?
They do - lots of. It's coming through the current contribution of the cathode surface trapping sites. A deep dive into the 1/f noise origin and physics here:
http://physics.bu.edu/~redner/542/1overf.pdf
syn08 said:
They do - lots of. It's coming through the current contribution of the cathode surface trapping sites. A deep dive into the 1/f noise origin and physics here:
http://physics.bu.edu/~redner/542/1overf.pdf
l love this stuff especially the age of the universe vs the shortest possible observable time, what was it 59 decades? Who was it Doug Sax or Keith O. Johnson that said digital will only be 'right' at 256 bits. Smart people say really stupid things sometimes.
I did once buy a pair of Quicksilver V4 tube mono-blocks unheard. These amps get quite good reviews. Mine had been modified with audiophile grade coupling capacitors (inside the feedback loop) and the power supply electrolytics bypassed with a couple of microfarad audiophile grade film capacitors. The chassis had been damped and connectors were upgraded.
I did not find the sound as clean as my old Conrad Johnson solid state MF2300A nor near as good as my newer Pass Labs X250 (thanks Nelson, nice job). Used the Quicksilvers and the X250 were the same price.
I received an email from the new V4 owner, he felt the modifications made noticeable improvements to a stock Quicksilver and asked for details so his friends could modify their amps to match.
The V4's were beguiling on acoustic instrument solos, especially any wooden bodied instrument. Even quartets sounded quite good but for impulse or large orchestral pieces, no joy. Somewhat muddled, certainly congested, unable to resolve individual instruments not fatiguing but not pleasing either.
I don't know where in the quality spectra of tube amps these fit, but they should have been a credible example of decent tube design. I guess I'm a solid state aficionado.
I did not find the sound as clean as my old Conrad Johnson solid state MF2300A nor near as good as my newer Pass Labs X250 (thanks Nelson, nice job). Used the Quicksilvers and the X250 were the same price.
I received an email from the new V4 owner, he felt the modifications made noticeable improvements to a stock Quicksilver and asked for details so his friends could modify their amps to match.
The V4's were beguiling on acoustic instrument solos, especially any wooden bodied instrument. Even quartets sounded quite good but for impulse or large orchestral pieces, no joy. Somewhat muddled, certainly congested, unable to resolve individual instruments not fatiguing but not pleasing either.
I don't know where in the quality spectra of tube amps these fit, but they should have been a credible example of decent tube design. I guess I'm a solid state aficionado.
scott wurcer said:
Tube diff-pairs biased with really bad carbon comp resistors generate lots of signal modulated noise and you could easily convince anyone that they were hearing it.
Your point is valid taken on its own merits, but I was speaking of tube circuits built with modern metal film resistors. I have never been a big fan of carbon composition resistors regardless of application. One of the happiest days of my life was the day I got sufficient 2% metal films in stock that I felt safe throwing all my 5 & 10% carbon comp resistors in the trash can. The 20% parts were already long gone.
(Yes, Virginia, there really were such things as 20% tolerance resistors once upon a time. Hell, if played your cards right, you only had to stock three or four values...then let the tolerance be your "other values." If the tolerance didn't meet your needs, just wait a year or two and remeasure. Something would surely have drifted into range by then.)
Grey
scott wurcer said:I don't suppose you want to add your exchange with Barrie to the Jung, Cordell, and Curl letters?
😉
No, I wasn't serious.😀
Just to let people in on the joke, Scott had forwarded my questions concerning the AD844 to Barrie Gilbert (he had designed it). This led to a nice series of correspondences where I learned a few things:
a) Barrie Gilbert is a true genius. That guy *really* knows his stuff. You can get a taste for some of the topics he brought up in his series at the ADI website:
http://www.analog.com/library/analogdialogue/leif1.html
b) He kept asking me tough questions, I think mostly because he wanted to see if it was even worth his time to correspond with me.
c) When I told him that I was using the '844 in a zero-feedback configuration, he was not amused. Basically his viewpoint was that no matter how linear of a circuit one made, that it would be more linear still with the application of global feedback. He didn't believe in listening tests, and he didn't believe in "golden ears".
Scott had warned me that Dr. Gilbert was not especially fond of these audiophile concepts and had learned not to bring them up with him....
I wish I had had more time back then to continue our correspondence. Unfortunately I had to bail out as my wife and I had a child and the demands of starting a new business were quite intense. But yes, I still have those letters, and hope Dr. Gilbert is still doing well.
Isn't it actually before processing? That is, synced averaging in the time domain before the FFT is run (and the FFT-processing with overlapped segments and averaging in the freq. domain, to further improce SNR). Earl Geddes has discussed this in this forum, and I could perfectly recreate his results. Like measuring harmonics of a signal which is down 90dB (ref. 2Vrms FS) from full-scale, the harmonics located at -140dB, noise floor at below -160dB (exept for some clearly identifiable artifact), only 20dB (at most) above the 24bit theoretical limit. With the conventional methods all harmonics were buried in the -140dB noise floor, so I got a 20dB+ improvement in SNR (at the cost of a very long measurement time). With a semi-pro soundcard...john curl said:Even more simply, THD uses WIDE bandwidth that creates a HIGH noise floor. The ONLY way to get below this floor is to signal average during FFT processing.
- Klaus
GRollins said:
Your point is valid taken on its own merits, but I was speaking of tube circuits built with modern metal film resistors. I have never been a big fan of carbon composition resistors regardless of application. One of the happiest days of my life was the day I got sufficient 2% metal films in stock that I felt safe throwing all my 5 & 10% carbon comp resistors in the trash can. The 20% parts were already long gone.
(Yes, Virginia, there really were such things as 20% tolerance resistors once upon a time. Hell, if played your cards right, you only had to stock three or four values...then let the tolerance be your "other values." If the tolerance didn't meet your needs, just wait a year or two and remeasure. Something would surely have drifted into range by then.)
Grey
I was simply postulating an experiment to make the effect audible to almost everyone, then you back off to a sane situation. You could also inject 1/f noise into your long tailed pair input and try to find your threshhold of detectability. I can't repeat a circuit that Walt and I built in 1982, I can no longer find resistors bad enough.
- Status
- Not open for further replies.
- Home
- Amplifiers
- Solid State
- John Curl's Blowtorch preamplifier