diyAudio (
-   Tubes / Valves (
-   -   Mullard 12AX7 part uniformity (

trobbins 12th April 2012 09:21 AM

Mullard 12AX7 part uniformity
Valves never cease to amaze me, for such a partly-handmade analog device made decades ago and long before Japanese five 9's quality processes.

I just tested 47 used 12AX7 from the same Feb 1957 batch (mC1 B7B) and within the approx 1-2% tolerance and repeatability of test conditions they all were effectively the same with respect to voltage gain and DC bias conditions (320VDC supply; 100k anode resistor; 1k8 cathode unbypassed). For DC bias, the spread in anode voltage was from 212V to 222V with only 8 triode sections just going beyond.

I can't verify their operating hours, but it is likely to be substantial as they were in a church organ (manufactured late 57 or 58), and all in the same type of circuit (tone generators) - of the 60 valves, thirteen were ring-ins, which may indicate the level of past failures.

Unfortunately the organ didn't come with a golden NOS sample to use for comparison :(

DF96 12th April 2012 10:22 AM

Your spread in anode current is about 10%, which is at the better end of typical for a competent manufacturer. 20% is easily possible. Of course, it may be that they were selected if uniformity was a requirement for the circuit.

trobbins 12th April 2012 02:38 PM

The spread in the bulk of triode sections is within +/- 2.5%, with outliers within +/-4%, but that formally expands a bit with test process and circuit tolerances.

I doubt they were selected in any way - it was an English production line organ, and the circuitry is simple see-saw dividers plus tone oscillators. I expect the production was pretty mature by 1957 for the 12A-7 range after 8 or so years.

Do you have any feel for what parameters would be expected to change with use with respect to DC anode current changing in a fixed setup, and whether that change is as great as typical manufactured part variation?

DF96 12th April 2012 06:30 PM

I would guess that gm would roughly scale with anode current. Maybe the mC1 design had particularly tight parameters?

trobbins 13th April 2012 04:26 AM

I found some time look at a few references.

Metson provides a summary of what appears to be the typical form of degradation (as opposed to failure):
"The decay of mutual conductance with time is probably the most serious of the electrical forms of failure. It is due to two separate and distinct causes—oxidation of excess barium in the cathode leading to emission failure, and growth of a high-resistance interface between matrix and core. The first cause is clearly the destruction
rate of barium by residual gas exceeding the production rate by the activators over a sufficient period of time. Symptoms of failure are a fall in mutual conductance and an increase in valve noise due to inadequate emission. The second cause is the formation of a layer of barium orthosilicate between matrix and core during valve manufacture - the chemical reaction will be clear from eqn. (2). When the valve enters service the layer is highly active owing to the presence of excess barium atoms and its resistance is normally less than an ohm. As the layer loses excess barium by gas action or diffusion to other parts of the matrix, its resistance rises and may reach 50 or even 150 ohms for a two-watt core. The negative-feedback action of this resistance results in substantial degradation of mutual conductance. These two actions are at present thought to account for most cases of mutual-conductance failure."

Fig 4 of following link shows an example of the gradual reduction of transconductance with time.

An older 1926 paper on life testing indicates in Fig.1(c) that testing at 0V gate bias may be a health indicator of emission degradation. Perhaps a suitable test would then be to use a pushbutton short across the cathode resistor and test for anode voltage. The resulting variation from anticipated 12AX7 datasheet 'average transfer characteristic' may be an indicator.

From my limited experience with valve tester methods - one test is to reduce heater voltage and check how gain changes. Maybe I'll set up to repeat tests on a sample of the 47 long-plates and check anode voltage (aka anode current) variation for filament voltage at 6.3V and 5.7V (original tests had nominal 6.6V heater level). I haven't come across any representative curves - but it appears to be that linear variation would be expected as normal, but an increasing drop in gain as heater voltage reduces would indicate more degradation of cathode.

trobbins 13th April 2012 10:05 AM

The change of bias voltage required to maintain constant operational conditions, for changing heater voltage when cathode current is not limited by emission capability was discussed by Valley and Wallman in their 1948 Vacuum Tube Amplifiers section 11.6.

A fairly constant slope in bias voltage vs heater voltage occurs when emission capability is not a concern. So I'm guessing there would be a noticeable change in slope when emission does limit maintaining cathode current.

trobbins 13th April 2012 11:48 AM

From PRR's comments in link, it looks like a 12AX7 would have to be very very emission degraded to show up changes in normal circuit operating performance (circa 1-2mA cathode current) when lower heater voltage was reduced to even well below -10%.

Reduced Filament Voltage

All times are GMT. The time now is 12:48 PM.

vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2017 DragonByte Technologies Ltd.
Copyright 1999-2017 diyAudio

Content Relevant URLs by vBSEO 3.3.2