• WARNING: Tube/Valve amplifiers use potentially LETHAL HIGH VOLTAGES.
    Building, troubleshooting and testing of these amplifiers should only be
    performed by someone who is thoroughly familiar with
    the safety precautions around high voltages.

tubes & microphonic

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hi

on nr. 402 od Audioreview in Italy I wrote an article around microphonic aspect of the tubes. It ia an old argument well treated by Mullard in the past, there are some copy of dcuments on web. The tese are done with ECC88, three models; just to make some test. I attach just some picture.

In Foto_9 the basic test set; an amplifier, a Fostex 10 cm, full range, a test board with a 88 on a beautiful McMurdo socket anti-vibration, Foto_9. The circuit is a simple gain stage coupled with a capacitor. 1 watt at the speaker with mainly sweep from 20 to 20kHz; the input shorted and with signal from one of my two AP, in this case AP sys1.

Of course the isolation from directly mechanic vibration to the desk where the test set is installed is not perfect but I have applied some solution that made the circuit and the tube almost free from disturbs.

After some intial test I have noted that the main issues are above 2 kHz this mainly because the mechanical structure of the signal tube is little so the components tend to vibrate at mid frequency .

I published many photo, here some of them.

Test_A7 is a sweep limited from 2 khz to 10 kHz sent to Fostex. test with a E88CC Siemes (blu line) and a 6H23 russian, green line (with a getter fixed in two points) the signal is taken at the output with the input short The difference is evident with a lot of spurious signals fro 6H23 starting from 2 kHz to 5 kHz, then the shape is better until 10 kHz.

Teast_A6 a sweep to 88 Siem; in green the spurious, in red after the O-ring was installed in the middle of the envelope, the result is interesting. The blu line at the bottom is the residual signals without excitation.

Test B6: this is a a test with 88 Siem; white noise sent to Fostex and a FFT analisys , level comparable with the previous , input in short. The blu line is the resdual, the red line with a Teflon strip on glass of the tube, the green is a O-ring in the middle of the glass. At the end the O-ring gives the best results.

Test_B9 in this thes I sent a 10 mV 1 kHz signal to the 88 Siem; then sent the white noise by Fostex; I get the results from capacitor. The blu line is the results with the 88 free; the green is with O-ring inserted. the peak at 10kHz an internal resonance.

There are other pics on this tests publishd, very interesting.

The use of O-ring is always suggested because it (almost) kill the vibrations of the glass that are one of the most important cause of the issue.

Of course we are speaking of a low signals, from 550 uV to 2 mV, but in a chain where the gain is 40-60 dB, as a phono stage + line stage, it must be considered.

Walter
 

Attachments

  • Foto_9_Fostex.jpg
    Foto_9_Fostex.jpg
    880 KB · Views: 387
  • Foto_8_zoccolo.jpg
    Foto_8_zoccolo.jpg
    584.2 KB · Views: 383
  • Test A7.jpg
    Test A7.jpg
    512.8 KB · Views: 371
  • Test A6.jpg
    Test A6.jpg
    656.8 KB · Views: 367
  • Test_B6.jpg
    Test_B6.jpg
    757.8 KB · Views: 363
  • Test_B9.jpg
    Test_B9.jpg
    742.6 KB · Views: 126
waltube,

Thanks! That is an interesting test.

When testing using a swept frequency source, you have to sweep quite slowly. A fast sweep may go past a tube's high Q mechanical resonance and not even activate it.

I first learned about his kind of effect of frequency sweep speed versus filter width (fo/fwidth = Q) when I used swept spectrum analyzers. This issue is also present in FFTs (it takes enough record time to get narrow FFT bin widths); and how fast you frequency sweep across the tube's mechanical resonance, even if the FFT bin width is much wider than the tube's mechanical resonance width.

The white noise you used is a good way to help reduce the effect of fast swept testing; but even then, not all frequencies are presented at once, and are transient in nature.
 
Last edited:
tubes & microphonic

Steve Bench has done some testing and proposed some solutions using composite elastomer materials.

It’s dated but dependable information:

“Quantatively, the Visco' mounting provided approximately 16 dB isolation, and the tube dampers added another 4-6 dB isolation, totaling about 21 dB isolation. The combination of the mounting AND the foam in the phono section provides slightly over 30 (!) dB isolation. The result is you can pretty well pound on the chassis, and the only microphonics you hear are the 6SL7 (predominates), followed by the line amp 5842, peaking out at 4-6 kHz, with almost no contribution from the phono section, even though it is at the lowest signal level. Also, the really nasty very high frequency resonances in the 5842's seem to be tamed quite well.”

“The quantative measurements were taken with signal gen feeding a power amp and speaker generating about 95ish SPL at the preamp face, and true RMS meter reading the preamp's output with inputs grounded and gain cranked max.”

Full text: http://diyaudioprojects.com/mirror/members.aol.com/sbench101/Microphonics/m_fonix.txt

Photos:

IMG_6652.JPG

IMG_6653.JPG
 
Last edited:
Regarding the sweep the setting of AP1 seems to be fine, in my opinion. I made a lot of "offline" test before to set the proper one to publish. It is a first step for future investigation but the way to follow is almost clear. But the start point is very interesting. In commerce there are lot of solution about O-ring also very expensive but we can find something efficently on RS or Digikey, with the ones for high temperature; they works fine. Also for socket , the McMurdo are one of the best but not in production.

Regarding white noise Mullard suggest absolutely this way because all the frequency are involved and I think is the best way to test the mic ( in addition with sweep)

After this test I believe that also the quality of the glass is very important!!!

The article is 6 pages and I sent just a few of pics.

Walter
 
The reason for a valve to become microphonic is the mica spacers no longer hold the elements of the valve in place and allow a tiny movement of the grids, this causes microphonic. The cure is not over priced rubber/neoprene 'O' rings, that do not do anything except make the misguided purchaser happy. The cure is a new valve that has not been abused by vibration.

How we used to do it .... YouTube
 
Last edited:
The heater is the cathode and that is not the problem. The problem is movement with the grid.
The cathode can be a problem. You'll notice in pretty much every tube that is built for low-microphonics that the cathode has some sort of extra holder. On the 7025 there is a spring of sorts that holds the cathode in place better to avoid microphonics. 5755 is the same way.

There are no directly heated valves that will be used for pre amplifying, so not an issue.
The 1S5 and 1U5 are a DHP made specifically for AF voltage amplification. I don't see them mentioned much in hi-fi but that's probably more of an availability problem than it not working. The 1U4 is also an excellent DHP preamp offering.
 
The use of O-ring is always suggested because it (almost) kill the vibrations of the glass that are one of the most important cause of the issue.
Yout test proceedure is most interesting. However, it is the vibrations of the elements inside the glass that cause the issue.Hi-Fi News did a technical report on tube dampers, but found very little difference with or without the rings.

The following site explains the cause of microphony in valves and suggests what to do if it occurs: Microphony in Valves - What Causes it? - ampvalves.co.uk
 
Well done Walter for setting all that up and providing some results.

The Mullard test results used a shaker table, with the valve in a socket that wasn't intentionally mechanically isolated in any way. The results show a response spectrum that heads in to the noise floor above 10kHz, and shows peaks above the noise floor of 20-30dB, for an EF86.

Given the Mullard tests are the only other tests like this that I know of, it would be great if you could try and do some benchmark spectrum results for a small sample of vintage EF86 (given that any particular sample could well be +/- 20dB from the next sample with respect to smoothed levels). Vintage EF86 samples may show up higher levels than the new samples that Mullard used, but I'd be expecting many of the spectral resonances should line up pretty well.

In post #1, can you clarify Test_B9 result plot, or are you able to replot the data, as there appear to be three results, but two are effectively using the same colour line, and you refer to a 10kHz resonance, but the resonant peak shown in the plot is at 14.5kHz.

Do you have a link to your article? Did you do some plots that overlay stepped power levels sent to the speaker? Did you put a can screen over a sample and compare outputs?

The EF86 appears to have quite a thick glass envelope - it is likely the thickness of glass that is a major difference with that of modern valve types.

The reason for a valve to become microphonic is the mica spacers no longer hold the elements of the valve in place and allow a tiny movement of the grids, this causes microphonic.
The cure is not over priced rubber/neoprene 'O' rings, that do not do anything except make the misguided purchaser happy. The cure is a new valve that has not been abused by vibration.
I don't agree with that view in general. Valves are microphonic from manufacture - they don't 'become' microphonic over time - that is what the Mullard tech report clearly spells out. They certainly do become more microphonic over time.

Not all of us have a bountiful supply of valves to pick and choose from, or the money to purchase more - and even then, it is pot luck if you can find a valve with really low microphonic levels.

Imho, a valve is likely to become more microphonic from thermal stress over time, rather than vibrational 'abuse'. Aspects of construction like the contact of mica disk edge 'fingers' against glass appears to be a key issue, and variation of that contact as a result of manufacture would appear to be a major cause of the sample to sample variation that Mullard report. And relaxation of that contact over time certainly appears to be a major reason for age related microphony, which would I suggest be more about thermal expansion/contraction cycles.

Any form of dampening of the glass envelop of preamp tubes is typically effective for tubes showing microphony - that certainly doesn't need to be expensive - I've used simple o-rings, and even rtv silicon - and a photo in post 3 shows a foam buffer. A layer of heatshrink tubing would even be useful.

Another summary type article on microphony is: https://www.dalmura.com.au/static/Microphonics%20in%20valves.pdf
 
Last edited:
Any form of dampening of the glass envelop of preamp tubes is typically effective for tubes showing microphony. . . Another summary type article on microphony is: https://www.dalmura.com.au/static/Microphonics in valves.pdf
The only significant comment on glass vibration in the above article is "The technical reports cited in the references do not discuss glass vibration, but it will have resonances - as a cocktail glass can 'ring'."

Note that the article gives no technical justification for glass vibration being the cause of microphony, nor does it give any analysis as to the extent of the contribution of glass resonances. Therefore, evidence that glass vibration plays a significant part in microphony is not supplied within this article

The article more usefully mentions decoupling the valve from the amplifier chassis. This is a much more effective method of preventing vibration from passing into the tube than using an O-ring damper on the glass envelope.
 
Galu, I agree that when I wrote that article it was collating previously presented information.

I wasn't aiming to portray that a light-weight o-ring by itself would make a significant reduction in microphony level, but rather that it is likely to make a change. Test jigs such as Walters is one way of trying to quantify that change, and Walter's results appear to indicate that some aspect of attaching an o-ring type device shows up as a microphony level benefit.

The Cunningham reference (http://www.triodeel.com/micro1.html) indicates the effort directed at the mica to glass connection, which is easy to comprehend as it is visible. That reference also identifies cathode vibration as a very difficult problem to overcome, especially as it would sustain the highest thermal stress cycling of any valve internal part.
 
trobbins, it was not my intention to devalue the rest of your contribution by picking out your comment on 'dampening of the glass envelope'. It's just that I can't get my head round the idea that vibration of the glass envelope is a cause of microphony.

It is outside influences such as chassis or air vibrations which cause the entire valve, electrodes and glass envelope alike, to shake. I don't agree that the glass passes on vibrations to the electrodes - both are shaken simultaneously.

An O-ring may well damp the vibration of the glass envelope, but it will not damp the major cause of the microphony - the vibrating electrodes.

If the change in microphony level caused by a tube damper can indeed be experimentally quantified it will be an important contribution.

P.S. The print quality of the Cunninghan reference is not very clear, nor is my eyesight, but I haven't yet spotted the reference to 'the mica to glass connection' which you mention.
 
Cunningham discusses the many methods used to obtain valve frame 'clamping' to the valve glass deployed for the insulating mica disks in the middle pages and last page.

The physical path of the envelope glass from the terminal end to the upper envelope section where the insulating mica disc is squeezed forms a plausible vibrating structural element.

If the mass of any o-ring or whatever added widget is significant relative to the mass of the valve, then the total mass increase can be a reason for reduced acceleration of the valve mass by an outside influence, given the compliance of the mounting has a low natural resonance with the valve mass.

A better quality scan of that Cunningham paper: https://www.americanradiohistory.com/ARCHIVE-RCA/RCA-Review/RCA-Review-1954-Dec.pdf
 

PRR

Member
Joined 2003
Paid Member
... The 1S5 and 1U5 are a DHP made specifically for AF voltage amplification.....

Specifically: for beach radios. These tubes took potentially large signal from the detector and volume control and drove the output tube. More "line level", not pickup level.

The Euro equivalents, Philips sheets, have a microphonic rating. If 40mV at the 1st audio grid made no more than 50mW at the speaker, you should be OK for microphonics (i.e., it should not howl-round with 1st audio in same box as speaker) (a rather antiquated way to specify the actual problem in making radios). Re-scaled to say a 5 Watt level, you want >400mV signal to make 5 Watts, again more line than pickup.

Yes, with the speakers several feet away you can put more gain in line.

OTOH while we don't want outright howl-round, we generally want MUCH better: not a trace of stray tube-shake in our precious audio (more precious than most AM air signals).

They are cheap. Try them. But to work down to pickup level with "NO" acoustic microphonics you may have to try a dozen and pick the best. Tube radios were built to Price, not perfection.
 

Attachments

  • 1U5-Emerson-558.gif
    1U5-Emerson-558.gif
    21.7 KB · Views: 238
maybe another factor you need to control:

...Making the filament supply somewhat positive relative to the cathode seems to help for in most instances.

Why do valves go microphonic?


I also read somewhere on a guitar amp forum (they really have issues) that it's important with a brand new tube to fully burn-in, allow it to heat up for quite some time and avoid cycling the temperature to avoid problems with the mica inside.
 
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.