How quiet can a pure phono stage be? (or, Why not a pure tube phono stage for moving-coil cartridges?)
Tubes are just too noisy to be used with MC cartridges, and anyone who claims otherwise is blowing smoke.
All amplifying components (transistors, tubes, etc.) create a certain amount of noise. For triodes, this self-generated noise takes two forms: "Johnson" noise, which is a hissing "white" noise sound, and 1/f (pronounced "one over f") noise, which is "pink" noise characterized by greater amplitudes in the lower frequencies than white noise. Tubes are considerably noisier than transistors and other devices which employ transistors, such as integrated circuits.
The 6922 type tube is about the quietest tube that can be found (the 417A is quieter by about 3dB, a small margin). It self-generates about 6 microvolts rms (6µVrms) of noise across a 20kHz bandwidth. The noise is a mixture of Johnson noise and 1/f noise which starts to predominate below about 1kHz. While it's possible, even easy, to find 6922 type tubes that are a lot more noisy due to contamination or other flaws (which may not appear until the tube has been in use for many hours), it's theoretically impossible for a 6922 vacuum tube -- or pretty much any tube -- to have quieter background noise than what the 6922 offers.
When compared to a 5mV MM cartridge signal, 6uV of noise is a signal-to-noise (S/N) ratio of 83dB. Although this is a respectable number, the riaa playback network (which boosts bass and rolls off treble) emphasizes the tube's low-frequency 1/f noise so that by the time the signal and noise comes out of the phono stage, the S/N ratio has been reduced to 74dB. But considering that the S/N of a vinyl groove is usually not much more than 65 to 75dB anyway, the noise contribution of a quiet tube is seldom intrusive.
But moving coil cartridges put out considerably less than MM cartridges, so the noise of the tube becomes more of an issue. An MC cartridge with an output level of 0.5mV has one-tenth, 20dB less, of the standard MM cartridge signal. With a signal this small the noise becomes that much more audible. With a 20dB drop in signal level, the noise is 20dB louder by comparison, worsening the S/N dramatically to a miserable 54dB. To give you a reference, a 54dB S/N is comparable to the tape hiss from a 1970's cassette deck. Now the tube noise is really intrusive.
One technique which can be used to reduce tube noise is to run several tubes in parallel. I used this technique in my circa 1980 Counterpoint SA-2 MC pre-preamplifier, and there are products on the market today which use this idea. It works like this: when two triodes are wired in parallel, the noise goes down by 3dB while the signal level remains unchanged, providing a 3dB improvement in S/N. In the SA-2 I ran two tubes (or four triodes since the 6922 is a dual-triod tube) in parallel per channel. This afforded a 6dB reduction in noise over a single triode, giving the S/N ratio of a 70's cassette deck using the Dolby B sound reduction process.
But if we try to use this technique to lower the noise of the phono stage by 20dB to achieve a respectable 74db S/N when used with MC cartridges we soon see that this technique offers only limited benefit and has two very severe liabilities.
Since each doubling of triodes provides a 3dB better S/N, we need to double the number of triodes 6.7 times in order to achieve the desired 20dB improvement. That's 104 triodes in parallel-- 52 dual triode tubes. Per channel!
Obviously this is an expensive approach. Good-sounding NOS 6DJ8/6922 type tubes, like Siemens, Telefunken, or Amperex are very costly. It would require more than $10,000 for the tubes to populate the sockets in our mondo phono stage. Secondly, with NOS tubes you are faced with a diminishing supply, increasing cost, and the increased difficulty of getting carefully matched, quiet tubes in the future. It is hard enough now to get matched pairs of the best tubes, think what it would be like to do that for 104 tubes.
The other big time problem is one of reliability: tubes are always just itching to get majorly noisy, or microphonic, or fade away, and every time you double the number of tubes, the chance that one will fail also doubles. Some audiophiles know just how frustrating it is to find one or two quiet tubes which will stay quiet for a few months to a few years. Imagine trying to keep a stable of 104 tubes quiet for any length of time!
So the reason we don't use tubes in moving coil cartridge stages is because it requires 100 tubes to reduce the noise to the 1/10th value needed for a good signal-to-noise, and the product ends up with 1/100th the reliability.
Johnson noise: http://en.wikipedia.org/wiki/Johnson-Nyquist_noise
1/f Noise: http://en.wikipedia.org/wiki/1/f_noise