John Curl's Blowtorch preamplifier part II

[VladimirK]

Originally Posted by PMA

EMI (coupled by many paths) as a source of sound difference has been my main target for many many years. Anyway, this is just the issue that can be measured very precisely.

In such kind of measurements, IMHO, "a devil" is in interpretation of measurement details.
To illustrate what I mean, consider an interconnect cable. Presize measurements of a single pulse propagation through the cable, could reveal some front slope, some settling decay, and post-pulse settling due to energy internally stored by the cable and released after the pulse has finished.
What part of the resulting signal curve should br studied "via microscope" as the most important one?
Anyway, some info about your measurements and conclusions will be interesting.

[Pano]

Agree. Although I may be able to measure the EMI coupled into a circuit, how do you relate that to sound quality? What's the metric or technique?

[DF96]

Constant carrier RF could cause bias shift, which could create changes in gain or distortion. I guess you would have to plot field strength against gain/distortion.

Modulated RF could either directly create audio voltages at the output, or intermodulation. Again a plot would be interesting.

The snag is that this would have to be done over a wide range of frequencies, and for both radiated and conducted interference. Alternatively, concentrate on likely domestic interference sources: nearby AM transmitters, cell phones, wi-fi, CFLs, cheap Chinese SMPS.

Of course, I am making an assumption here: that if it changes the sound quality then it has to do this by changing the output voltage (or possibly output impedance).

[PMA]

BJT inputs are most sensitive to EMI. Not only due to exponential transfer function, but they may act as rectifiers.

JFETs and MOSFETs are much less sensitive, and tubes are least sensitive.

How to measure EMI? With HF spectrum analyzer, e.g. We are at diy forum, but we do not discuss DIY approach in this thread.

The less EMI coupled into electronics the better. Audio enclosures are especially poor in this respect.

[Chris Hornbeck]

Originally Posted by PMA

BJT inputs are most sensitive to EMI. Not only due to exponential transfer function, but they may act as rectifiers.

JFETs and MOSFETs are much less sensitive, and tubes are least sensitive.

This parallels completely the "ease" with which audio amateurs are able to make good sounding electronics. It's difficult, but not impossible(!), to make "bad" sounding vacuum valve based electronics.

Hence, my contention that electronics' performance at -60VU is both important and difficult to get right.

Thanks,
Chris

[john curl]

Chris, I completely agree with your -60Vu comment. The information that separates quality analog from digital seems to reside there, for example. I KNOW that analog tape recordings have lots of third harmonic distortion, yet it is not so bad as the numbers would imply, AND that the low levels are virtually distortionless, by comparison. That does not mean that analog tape recording is perfect, it obviously is not, but it can be 'magic' in the way it can convey the essence of a human voice, etc.

[Chris Hornbeck]

Originally Posted by john curl

-60Vu The information that separates quality analog from digital seems to reside there, for example. I KNOW that analog tape recordings have lots of third harmonic distortion, yet it is not so bad as the numbers would imply, AND that the low levels are virtually distortionless, by comparison. That does not mean that analog tape recording is perfect, it obviously is not, but it can be 'magic' in the way it can convey the essence of a human voice, etc.

This brings to mind VladimirK's comment earlier that what we might ought to be looking for are some *loss* mechanisms, rather than solely the *additive* classic distortion components. Are *all* information losses the results of an overlay of extra frequency components, harmonic, inharmonic, correlated or not, etc. ? Or are there significant information losses from *lossy* mechanisms, maybe from crossover distortion-like flat spots around zero crossing?

We understand some mechanisms pretty well - push-pull amplifiers for example - but others we just try to design out, like transformers' BH curves and the voodoo of electrolytic capacitors. What others do we ignore because they're messy?

Much thanks, as always,
Chris

[VladimirK]

On the intuitive level, I guess that human's audio perception operates both on analog and digital priciples. First (digital principle) one recognizes that definite event has happened (even at very low amplitude level, signal curve has happened to have bump, for instance). Second (analog principle) one analizes and makes some conclusions about shape of this bump.
I suppose, that due to a whole bunch of various mechanisms, including widely understood "noise" and transient responce properties under dynamically varying signals conditions, some sub uV level bumps, especially at high listenable frequencies, can be smeared out or hidden by those distortion mechanism, i.e. at the output signal we can not recognize some low-level bumps that existed at the input. This can be a basis of the loss mechanism.

Some time ago, I posted some translation from a paper in russian, saying about "thickening" of the transfer function of an amp, generalising all noise-like and transient imperfections of an amp. In that paper, it was rigorously proved, in terms of information theory, that such smeared transfer function puts fundamental limit on the transmission capacity of an information channel. The low-level high-frequency signal components can quite probably be cut out by the limited transmission capability of the channel.

PMA, speaking about EMI effects, I mean almost effects of electromagnetic interactions between various components of schematics itself, under big transient signals conditions, when essentials EM fields can be produced. Such fields can be localized and partly eliminated by using shunt power supplies. These effects add to the whole family of hardly measurable effects, like those from wire insulation, from capacitors, etc.

[john curl]

Vlad, your input on load line 'thickening' has been very valuable. I sent what you put up here to other designers, one of whom once worked with Enrico Fermi, in the long past.(interesting stories). These problems are subtle, but work around-able, by using 'what works' foremost, rather than insisting what measures good MUST sound good. We come to understanding with experience as well as listening for ourselves.

[VladimirK]

I would like to make clear one point about this transfer function thickening. We do not speak about measurement of TF at static or pseudo-static condition (apply 1V to the input, measure how many V at the output, and some smearing will be possible due to standard noise mechanisms).
We should think about measurements at transient conditions (apply the 1V to the input almost instantly, 1nS for example, and while settling is not completed, we apply 1uV to the input). The responce to 1uV in such conditions will not be the same as without 1V applied. Etc., were are many possible mechanisms of instability and "vibrations" of transfer function under real operation conditions of schematics.