The construction of a valve is not a completely rigid structure. Any relative movement of the various electrodes not only affect sinter-electrode capacitances but also variations in current and mutualconductance. Such relative movements can be caused by vibration, such as sudden knocks or even ambient sounds.

Mullard presented the results of an investigation into micrphonics in valves which it published in two articles that appeared in the November 1962 issue of their publication "Mullard Technical Communications". A summary of their findings follows.

Valve Acceleration @50mW [12K]

The first test attempted to simulate the typical acceleration experienced by a valve due to ambient sound. This used 3 specially constructed accelerometers which were subjected to various sound frequencies. The sound was generated from a loudspeaker generating 50mW of power (note, milliwatts). The response varies considerably with frequency, however it is suprising to see peaks as high as 0.1g.


The next graph is illustrates the typical microphonic behaviour of an EF86 subjected to 0.05g acceleration (0dB is equivalent to a grid voltage of 6.67uV)(no, I dont know why they chose that voltage !).

The above graph is based on measurements taken on 75 valves from a large production batch. This next graph is concerned with the spread of microphonic sensitivity within the batch. To understand the graph, I've marked two points on the graph.


  1. At 50mg acceleration, 90% of valves generated a microphonic equivalent grid voltage of below 320uV
  2. >At the same level of acceleration, only 15% valves generated a microphonic equivalent grid voltage of below 100uV

This indicates the massive spread that can be expected between valves of the same type and manufacturer. But in order to select a "best" valve from a batch you will need to test a batch of perhaps as many as 10 valves.

Further tests involved the use of a vibrating bench to subject a valve to specific frequencies. Special valves with cut-outs allowed viewing of the various electrodes such that the contributions of each part of the valve structure could be uniquely identified.

[25K] Opposite is a view of the suppressor grid of a pentode, showning one particular grid resonating at 2.1KHz. Slight variation in the mountings of each grid wire will give rise to a spread of reasonances around this frequency.

[13K] The next picture is of the frame grid of a valve reasonating at around 37KHz. Again, slight variation between each wire will cause a spread of reasonances about this point.

Cathode [19K] The next view is of the top end of a cathode. The two cylinders that can be seen either side of the cathode are the supports for the grid, so it can be appreciated that the slightest movement of the cathode will be a not-insignificant change in grid-to-cathode distance and hence directly effect the anode current. In this instance the cathode reasonated at 600Hz.

Getter [11k] Although the getter of a valve does not play a direct part in the electrical performance of a valve, any reasonances of the getter may serve to vibrate other parts of the valve structure. In the picture opposite, the getter can be seen reasonating at 300Hz

Heater [15K] The next picture illustrates one half of a heater reasonating at 570c/s with the other reasonating at 600c/s.

Mounting the getter at two points significantly improves the rigidity of the structure, and this is reflected in the results of a comparitive test shown below :-

Getter response [13K]

This next graph illustrates the effect of using a symetrical anode :-

Anode shape response [11K]

The next plot shows the differences caused by securing both halves of an anode assembly within a mica spacer instead of just one of the halves. This is nowhere near as marked a change as has been recorded in the previous two graphs, so I wonder just how significant this difference is in the light of the fact that a 4:1 valve-to-valve variation is not unreasonable.

Anode mounting response [11K]

Finally, Mullard presented some mathematics estimating the effects of electrode movement upon the current flow and capacitance of a valve. Perhaps the most useful result was an estimate of the grid voltage at which small movement of the grid had no effect upon anode current. This voltage was given as :-

Element Diagram [1K] Vg = - Va
(1- 1
)   where 'u' = mu


It is not unreasonable to expect as much as a 4:1 variation in microphonic sensitivity even for valves from the same manufacturer and the same batch. Thus, for a valve intended for use as the early stages stge of an amplifier it is certainly beneficial to select a valve by testing a batch.

When designing an amplifier, there are in addition to the usual electrical characteristics some structural criteria to also consider when selecting a particular valve type. The case for a double mounted getter and symetrical anode structure have clearly been proved. Quality of construction will also play a part, such as the crimping of the two halves of the anode structure and the tightness of fit of the various electrodes within the mica spacers that are typically used.

There also appears to be a valve operating point at which, at least as far as the grid is concerned, minimises sensitivity to microphonics. However this may well be at a point that conflicts with the optimum valve working point from the point of view of electrical noise and distortion.

Testing the sensitivity of a valve using a speaker driven at 50mW clearly showns that there is an effect on a valve. In those tests the speaker would have been reasonabley close to the speaker so as to simulate the effect that would be seen in a typical domestic radio. With a Hi-Fi amp the speakers will be significantly further away, but will also be outputting significantly more power. This must result in an influence upon the resulting sound, but unfortunately my ears are not sensitive enough to quantify this (all the main reasonances seem >300Hz, which is above the highest frequency used by Motorhead ;-) If any of you audiophiles out there can help correlate sound quality with valve construction then let me know and I'll post your results here.

Download Part 1 of the Original Mullard Pulication [1.6MB]

Download Part 2 of the Original Mullard Publication [410K]

Page copyright
J.Evans 2000-2009
Last updated
16th November 2009