Rerurn to Romy the Cat's Site
In the Forum: Analog Playback
In the Thread: 7788 tube mic pre design
Post Subject: High Gm Tubes and "Sound Quality"Posted by Lucian on: 1/3/2008
Hello Romy,
Quite clearly, a high Gm device will have a higher voltage gain than a lower Gm (for pentodes, gain is Gm*Rload if single ended) and this is something we are after in a MC stage. Noise performance is crucial and below I am analyzing a bit of that...
Like any other amplifier, a tube has an input noise voltage (mostly defined by an RMS sum of the thermally generated noise, mechanically generated noise and internal noise voltages like partition noise for pentodes plus other sources of noise - 1/F noise, cathode interface noise, etc -) and a (relatively low) input current noise, whose origin is in the thermally and electrically generated leakages in the grid(s)... Let us just see whether the high Gm is warranting good noise performance...
As defined by the datasheet, E810F has a conductance of 42-58mA/V, an Ri=42K if pentode and a gain µ1µ2=57. Excellent choice for a high gain stage, it seems, specially if we do not ignore the equivalent noise resistange of 110 ohms at 45MHz... Does all this promise good noise performance in a WB application? Clearly, but only when used at maximum gain (i.e. as a pentode or cascoded with another one like it). You obviously can see that this performance is relative (to use an Atkinson's favorite, "good for a tube amplifier" :-)...
Let's make some measurements, then: an E810F measured for noise shows about 50-100nV/sqrtHz@20Hz, 10-20nV/sqrtHz@1KHz and about 10nV/sqrtHz for the rest of the higher frequencies noise voltage densities as measured with my HP3585 spectrum analyzer. These values are practically the same for a very high quality ECC83 (TFK used as a comparison), but generally 10-30dB worse for the garden variety ECC83/12AX7 bunch.
The ECC83 has a 1.6mA/V Gm vs. 220K load, while the Gm of E810F is 50mA/V vs 20K load... these values will convert to a (calculated) input noise voltage of 1.1µV for ECC83 vs. 1.4µV for E810F. When taking gain into account, the output noises are 55µV for ECC83 and 38µV for E810F, not much of a difference, but clearly better for E810F.
What my is point, considering the measurements above? The Gm of a tube is not so clearly illustrating the noise performance of a tube by default, but only in a direct comparison (1.6mA/V versus 50mA/V only yields a noise improvement of 2.5dB, while the direct Gm increase is 29dB).
Not to mention that the noise densities (measured) of a 2SK170BL JFET are 10nV/sqrtHz@20Hz, 0.4nV/sqrtHz@1KHz and the rest, yielding a WB input noise of 0.33µV with a Gm of 80mA/V :-)... Is the JFET better? Noise performance-wize, vastly, at only 2.2µV output noise (+24dB SN improvement as compared to E810F or about +21dB including the -non-trivial- noise current of the JFET).
Is the JFET better? Noise-wize, far better if the JFET is used with very low source impedances such as MC.
Does it also mean that the JFET will "sound" better? Now clearly this is a whole totally different sort of discussion and, as you have said, we must tackle this (very interesting) subject separatedly :-)...
Best regards,
Lucian
Rerurn to Romy the Cat's Site