Pentode noise - some numbers for the 6BR7

A few weeks ago there was a discussion regarding the use of pentodes in
pre-amps. The following information is from an application report on the
6BR7 low-noise pentode by Standard Telephones and Cables Pty Ltd Sydney
Australia and dated January 1955.

Begin quote:

Valve Shot noise (Hiss): A certain amount of random noise will be
generated in the valve by the random arrival of electrons at the plate,
and this is further increased by the partition of the cathode current
between plate and screen, as the random collection of electrons by the
screen must have its effect as an increased variation in the number of
electrons arriving at the anode. This is inherent in the valve and cannot
be entirely eliminated. A major contribution to valve noise, however, is
noise produced by leakage between electrodes over the mica insulators and
in the base. Also a poorly activated cathode can produce a noise voltage
swamping the normal valve noise.

The shot noise in the 6BR7 has been reduced to a minimum by careful
design, and the noise due to leakage controlled by careful assembly and
inspection. A certain amount of leakage noise is unavoidable in a
mass-produced valve, and this reveals itself as a small variation in noise
level from valve to valve.

The curve No. 308.228 shows the maximum hiss voltage plotted against the
percentage of valves, using a grid resistor of 100 k Ohm and a bandwidth
of 10 kc/s. The curve is very steep and indicates that a hiss level on the
grid always lower than 7uV is to be expected.

The curve No. 308.230 shows the same parameters with the valve triode
connected. Here partition noise has been eliminated. As would be expected
the higher noise levels are unaltered because they are due mainly to
leakage which is not greatly affected by whether the valve is triode or
pentode connected. The lower noise level is reduced, and in fact falls
below the thermal agitation noise generated by the 100 k Ohm grid
resistor, which in a 10 kc/s bandwidth at 20 C is about 4uV.

Conclusions: While it is not claimed that the 6BR7 has exceptionally low
noise properties, it can be stated that it is a very great improvement
over average normal valves. Individual samples taken at random from a
large batch can be expected to give consistently good performance. As with
normal receiving type valves, by selection, it is possible to find
individual samples of outstanding performance, but whereas with normal
types the average product is many times inferior to the selected samples
in the case of the 6BR7 the average product is little different from the
best samples selected.

In spite of this, care must be taken when designing prototype equipment
for high gain, low noise applications, to ensure that the design is not
finalised from the experience acquired from a single sample of the 6BR7.

End quote:

Since I can't reproduce the graphs here, the table below shows the
measured noise levels for a given percentage of valves extracted from the
graphs. The bandwidth is 10 kHz. (use a fixed pitch font)

Percentage Noise (uV) Noise (uV)
of valves triode pentode
connected connected
100% < 7.10 < 7.10
90% < 5.70 < 5.70
80% < 5.20 < 5.36
70% < 4.86 < 5.20
60% < 4.60 < 5.12
50% < 4.40 < 5.03
40% < 4.20 < 4.95
30% * < 4.87
20% * < 4.80
10% * < 4.55

* The triode graph stops just below 40% as the grid resistor noise

Operating conditions
Plate Supply 300V
Grid resistor 100 k Ohms
Plate resistor 220 k Ohms
Cathode bypass 50 uF
Cathode resistor 14 k Ohms (Triode) 1500 Ohms (Pentode) Screen resistor
1.5 M Ohms (Pentode)

Published gain for these operating conditions is (following grid resistor
1 M Ohm) 14 times triode connected and 124 times pentode connected.

Observations: The noisiest valve pentode connected is < 5 dB noisier than
the best valve triode connected. You might be unlucky enough to find a
valve that is just as noisy triode connected as pentode connected. 50% of
valves pentode connected are only 1.5 dB noisier than the best valves
triode connected. A pre-amp with 200 mV input sensitivity would have a
worst case signal to noise ratio of 89 dB.

Regarding the 6U8 (for Patrick and flipper), in the May 1960 edition of
the AWV Radiotronics magazine there is a reprint of an RCA article titled
"The 7199 in High-Fidelity audio equipment". The "in this issue" page
describes this article as "An interesting discussion of the use of
triode-pentodes in AF amplifiers, particularly the 7199, hifi version of
the 6U8."