Hi,
Don't forget ABX Ace Tony Faulkner who somehow always scores way higher than average, I am aware of a number of highly publicised tests (including on watermarking for DVD-A) where his score was simply thrown out as 'lucky coin" (a statistical device that allows you to exclude outliers, which in case of a test of "can anyone hear this" incidentally is not allowable) because it way, way above average and in some cases would have forced a rejection of the null hypothesis.
Having been then as press in one of the these tests in London, all I can say that the test conditions where dreadful, the system lousy and music not at all to my taste. The effort needed to not get up and walk out was such that you could have inverted one channel polarity and I would have literally heard nothing...
Again, something to note, just like "excessive false negative identifications compared to chance" over many individual tests seem to plague ABX, ABX Tests of Tony Faulkner seem plagues by "excessive false positive identifications compared to chance", however the ABX Mafia makes it all readily disappear while grinning all the while like that proverbial cat from Cheshire.
Ciao T
Don't forget ABX Ace Tony Faulkner who somehow always scores way higher than average, I am aware of a number of highly publicised tests (including on watermarking for DVD-A) where his score was simply thrown out as 'lucky coin" (a statistical device that allows you to exclude outliers, which in case of a test of "can anyone hear this" incidentally is not allowable) because it way, way above average and in some cases would have forced a rejection of the null hypothesis.
Having been then as press in one of the these tests in London, all I can say that the test conditions where dreadful, the system lousy and music not at all to my taste. The effort needed to not get up and walk out was such that you could have inverted one channel polarity and I would have literally heard nothing...
Again, something to note, just like "excessive false negative identifications compared to chance" over many individual tests seem to plague ABX, ABX Tests of Tony Faulkner seem plagues by "excessive false positive identifications compared to chance", however the ABX Mafia makes it all readily disappear while grinning all the while like that proverbial cat from Cheshire.
Ciao T
Scott, could you tell me, which kind of front end you did use for the measurement ? The calculation gives the input noise voltage 800 nV approx. referred to the audio bandwidth and the capacitance can be estimated to 5 pF max, should be the measurement correct. And further, do you have any explanation for the deviation of the behavior of the mica cap from this being predicted for the ideal cap (as well as the actual polysterene) ?
This was a composite charge amplifier with a 50pF source capacitance and 5pf || 5G Ohm feedback network. I used a 2SK222 which has 10mS at 1mA and only a few pF's Cgd. The audio frequency noise of mica capacitors as it turns out is well documented even back in 1977, DA is a loss mechanism so it makes heat and must make noise.
Understood. Maybe the DA could be interpreted as a kind of loss resistance in the equivalent circuit being generalized some way. Unfortunately, the mica is a highly anisotropic crystallic material and thus its dielectric constant is described by complex tensor (or matrix), so the complete description of the behavior of capacitor will be horribly complicated perhaps...
The plates only excite the cap in one direction along the crystal axes, so it's not that complicated. The equivalent circuit of the ESR as a resistor in series with the ideal capacitor gives a good first-order approximation to the noise. Second order, the ESR varies somewhat with frequency for most caps.
There's an interesting extension to this model that will be published shortly by a friend of mine. I'll supply details once it's out.
There's an interesting extension to this model that will be published shortly by a friend of mine. I'll supply details once it's out.
That is correct, but the noise when looked at with care is almost a perfect 1/f (10dB/dec). I actually have some much better measurements from several years ago with several decades of this behavior. So I conclude that Pease's R/C ladder approximation to DA is as complicated as one needs at audio frequencies.
Without directing specific comment toward the mathematically terse (some would say, mathematically clear) PhD research papers of anyone in particular, I would like to give particular credit to the writting approach of Dr. Malcolm O. Hawksford. Dr. Hawksford's math skills appear second to none, yet in reading his research papers one can tell that he intends them to also be accessible to an practical engineering audience, not just an academic one. I wish that more PhD papers were written with that dual audience objective in mind. I've no doubt that it would make the worthy research work of other PhDs less obscure.
Last edited:
All research papers and PhD theses are supposed to be written in a way which makes them to some extent accessible to an intelligent non-expert. That is why most editors and referees insist on a long introductory section which contains material which will already be known to people working in that field. After that you just have to work hard; all the information you need should either be included or referred to. To fully understand a research paper in a field with which you are unfamiliar could easily take months or years, even if well written. If it only takes hours or days then it may not be saying very much - even an expert may take a few days to digest a new paper in his own field.
For the whole of a paper to be easily accessible to a practitioner then either he needs to maintain his knowledge at academic level (which some do, but many do not), or the paper needs to be very long by including stuff which an academic can work out for himself. Such a paper would never get published, as there is too much pressure on pages in academic journals.
For the whole of a paper to be easily accessible to a practitioner then either he needs to maintain his knowledge at academic level (which some do, but many do not), or the paper needs to be very long by including stuff which an academic can work out for himself. Such a paper would never get published, as there is too much pressure on pages in academic journals.
I don't think it is all about dielectric itself, construction matters. Fir example, I used NOS capacitors, silver mica KCO, in microphones, between capsules with 40-100V and FET gates, no audible noise. But recently restoring vintage BC-348 receiver found mica capacitors noisy. My conclusion was, during decades they absorbed moisture that made them noisy.
Used for coupling and hence larger there would be less noise, I substituted 560pF and could hardly see it, also it's only a few dB in any case.
Though as other physics professors have pointed out he got things fairly wrong in his article on wires.
Yes, they were exactly 560 pF coupling caps. Another that were noisy, were 10 nF, mostly B+ filtering. First (KSO) were in brown mold, last (don't know vintage American brand) in black mold. Obviously different materials for cases, I think it is the main difference. Either gigroscopic, or some acids for curing of thermo reactive plastic.
Ken, I agree with you about Dr. Hawksford.
Hawksford is unique, because he can both hear differences in audio components AND he has a good education that is constantly updated by his being at a university, teaching the material he is interested in.
Like Dr. Vandenhul, he has been criticized for noting and attempting to measure differences in wires, or creating a mathematical framework where wire differences are possible.
I know the guy, had him over at my office over 20 years ago, and find him brilliant and stimulating. However, once I saw one of my topologies put out in print by him, without attribution, so I don't invite him to my office, anymore.
Same thing happened to me with Bob Stewart, from 40 years ago. And so it goes!
Hawksford is unique, because he can both hear differences in audio components AND he has a good education that is constantly updated by his being at a university, teaching the material he is interested in.
Like Dr. Vandenhul, he has been criticized for noting and attempting to measure differences in wires, or creating a mathematical framework where wire differences are possible.
I know the guy, had him over at my office over 20 years ago, and find him brilliant and stimulating. However, once I saw one of my topologies put out in print by him, without attribution, so I don't invite him to my office, anymore.
Same thing happened to me with Bob Stewart, from 40 years ago. And so it goes!
No one is perfect.
Well, exactly. If people are doing interesting research (as Prof. Hawksford has been), they're likely to come up with wrong stuff as well as good stuff. No risk, no gain.
- Status
- Not open for further replies.