Dave,
I was fairly sure that I was not alone when Frank’s results triggered my (mechanical) engineer’s reflexes: I reached for my calculator, the library catalogue, and the internet, in about that order.
Step one was to convert Frank’s input voltage densities to output voltages. This confirmed that pentode output noise was, indeed, as much as 14dB higher than the output noise of the same valve strapped as a triode. OK. That hurdle cleared, I calculated partition noise from Frank’s measurements of noise in pentode mode compared to the noise of the same valve in BestPentode mode. As much as 3dB of partition noise for one valve and as little as 0.5dB for another … hmmm … even allowing for variability in the ratio of screen to anode current, there was more to this than met the eye.
More calculations followed to estimate thermal noise contributions from the rest of the circuit - relatively small. More estimates of excess noise from resistors and the MPSA44A - larger than thermal noise but not sufficient to explain the deviation from the classic theory. I may have calculated incorrectly (mechanical engineers can do that) but, since I couldn’t see anything amiss with Frank’s method or equipment (somebody PLEASE give me a Rohde and Schwarz precision audio analyser and I’ll test pentodes ‘til the cows come home!), I decided to re-visit valve noise theory to find out how it was derived and tested.
Step two: hit the books, as most of the references are far older than the internet. The
RCA Review has a succession of articles on noise in valves written by D.O. North, W. A. Harris and B. J. Thompson over the period January 1940 to July 1941 under the title ‘Fluctuations in Space-Charge Limited Currents at Moderately High Frequencies’. Harris and North also wrote on ‘Fluctuations Induced in Vacuum-Tube Grids at High Frequencies’ in the
Proceedings of the Institute of Radio Engineers in 1941. E. W. Herold reported on ‘An Analysis of the Signal-to-Noise Ratio of Ultra-High-Frequency Receivers’ in the
RCA Review of January 1942, and Aldert van der Ziel appeared in print with M. J. O. Strutt on ‘Signal-Noise Ratio at VHF’ in the September 1946 edition of
Wireless Engineer.
William J. Stolze (also from RCA) set out the basic results and formulas in the February 1947 edition of
Communications in an article on ‘Input Circuit Noise Calculations for F-M and Television Receivers’, much of which subsequently appeared in the
Radiotron Designer’s Handbook. William Harris neatly summarised the state of play in a compendium article in the September 1948 edition of the
RCA Review, collecting his own works along with North’s fundamental derivation of the space-charge effect on shot noise, re-stating the classic valve noise formulas derived in earlier papers and confirming their utility by comparison with experimental data for triodes and pentodes at radio frequencies.
The common thread in the historical material was the (understandable) attention concentrated on noise at radio frequencies. The experimental data was, with one exception, derived from tests made at 200kHz or more, and normally at more than 10Mhz. The sole publication on low frequency noise was a note from E. J. Harris and P. O. Bishop on ‘Low-Frequency Noise from Thermionic Valves Working under Amplifying Conditions’ in
Nature (No. 4103 of June 1948, page 971) in which they lament that:
Although there is a good deal of information concerning the noise produced by thermionic valves at radio frequencies, there do not appear to have been published data for the low-frequency noise intensity that they produce at under amplifying conditions. It has long been known that the noise intensity increases as the frequency to which the observational system is tuned is diminished.
Harris and Bishop, from the Biophysics Research Unit of University College, London, found that, at frequencies below 1kHz ‘flicker’ or ‘1/f’ noise became measurable above the shot noise and:
if expressed as an equivalent fluctuation EMF at the grid, the low-frequency noise was remarkably constant from valve to valve.
My hopes were up, only to have them dashed by Chapter 8 of Aldert van der Zeil’s 1954 book,
Noise, in which he recounts the rather inconsistent results obtained when measuring 1/f noise in tubes at low frequencies. On page 230, he concludes that:
Because flicker effect may vary strongly from tube to tube and from type to type, few suggestions about low-frequency circuits can be given. One should, of course use tubes with low flicker noise: tube types 6AG5, 6CB6, Philips EF40 are recommended; it is in general advisable to select tubes.
[
I hope we’ve all bought our stashes of these ...]
Aldert van der Ziel tentatively suggested in his 1959 contribution to the Smullin and Haus collection of papers
Noise in Electron Devices that 1/f noise might be cased by fluctuations in the voltage drop across the thin surface layer on the oxide coating of the cathode. This is repeated on page 490 of Gewartowski and Watson’s 1965 book
Principles of Electron Tubes … and that’s about it, folks.
What, no model? No coherent theory, no causal mechanism? No consistent set of measurements with a few outliers? In the words of Captain Jack Sparrow:
Well, that’s just maddeningly unhelpful.
And it must have really bugged van der Ziel, because step 3 (my internet trawl) turned up a paper he co-authored in 1984 when was at least 73 years of age, ‘Partition 1/f Noise in Vacuum Pentodes’ (
Physica C, 124B (1984) pages 299 to 304). The paper is one of a number written in conjunction with his PhD students, reporting on experiments conducted in an effort to confirm that Peter Handel’s 1/f quantum theory applies to electronic noise. Those experiments form much of the basis for van der Ziel’s 1986 book,
Noise in Electronic Devices and Circuits, published by Wiley.
Yippee! Are we there yet? Well, no. While Handel’s theory was championed by van der Ziel for electronic noise, its applicability is hotly contested and is currently denied by the broader scientific community. With that caveat, the 1984 paper tested 6AU6 pentodes and reported the interesting result that, for sharp cut-off pentodes, partition noise was not much suppressed by space-charge (on page 304). At this juncture, the engineer in me will accept any empirical crumb that falls from the scientific table ...
The upshot is that, unless there is a relevant body of work written in another language and yet to be uncovered from Philips or Mazda or some other manufacturer, in 2013 we are without an entirely satisfactory model for noise in valves at audio frequencies. We are reduced to testing them type-by-type, valve-by-valve and, for this alone, we owe Frank Blöhbaum a debt of thanks.
For now, I look forward to some more interesting reading (thank you, Ian), and the odd spot of solder-slinging. Pass me that Rohde and Schwarz analyser please, Jocasta darling.
Kim