Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[AusTex64]

I 'm curious how others test tubes for noise in mics, like:

How much gain on mic preamp? I use a Hardy M1 with 60dB of gain.

How long do you burn the tube? I'm burning 3 days before making judgement.

Do you use other mics to compare to while you're listening? I put up my '70's era U87 and AMI U47 clone for reference.

Bypass the capsule with a capacitor? I've not done this (yet).

How much tube noise (not microphonics, but "white self noise") is acceptable? In addition to the 60dB of mic pre gain, I'm listening on cranked up headphones.

Thank you,

Robert Mokry

[Jim Williams]

Microphone self noise is tested without a mic preamp in the signal chain. I use a Audio Precision test rig. It will measure noise down to about -140 dbu in the analog domain.

You feed the mic a very clean 48 volts to power the mic, (or the  power supply for a tube mic) and run a noise vs frequency sweep. 20~22k hz bandwidth signal to noise ratio measurements are also done. I avoid using "A weighting" filters to obtain an honest measurement. Substituting a quality film capacitor for the capsule eliminates any acoustic sound pickup interference.

If you use PC based software test gear, your noise measurements will be limited by the converters you use.

[klaus]

Those are all pertinent questions. In case you don't have access to Jim's sophisticated test equipment, there are other ways to get to roughly the same results.
I have several tube primers on the PSW "Stickies" Forum, where most of these issues are discussed, some in great detail:
http://repforums.prosoundweb.com/index.php/board,28.0.html

Here is a summary:

How much gain on mic preamp?

I would use an average for a professional condenser mic: around -25 to-30dBu

How long do you burn the tube? I'm burning 3 days before making judgement.

That subject is also discussed in great detail in my 'Stickies' forum. The longer the test, the closer to 100% predictability you will get. I test for at least 48hrs. Some expensive tubes I test for a week continuously, first making sure that the power supply is adjusted so that the tube under test sees the correct voltages, before I go for the long distance run.

Do you use other mics to compare to while you're listening? I put up my '70's era U87 and AMI U47 clone for reference.

You are doing the right thing. By using two mics at the opposite extremes of condenser mic gain, you get a nice average. Just make sure that you readjust the mics' gain so that the mic/tube under test is at the same gain level as the comparator. I use a white noise generator as comparator when testing more than one of the same type of tube: once the noise generator's output is set to exactly the same level as the noise floor of a known good tube in a known well-working mic, I have a reliable reference for all the other tubes under test that follow. As my preamp has stepped gain settings, I can listen back and forth between noise generator and tube, and pinpoint the exact noise difference between the two.

This also beats bypassing the capsule with a dummy capacitor: you will test the whole real-world set-up, rather than add another variable: the capsule. Under real-world conditions, testing the mic as a whole is preferable to isolating out the capsule, as long as you have a really quiet room.

Bypass the capsule with a capacitor? I've not done this (yet).

See above

How much tube noise (not microphonics, but "white self noise") is acceptable? In addition to the 60dB of mic pre gain, I'm listening on cranked up headphones.

Once you have gone through all the other steps outlined above, you will know pretty well whether a tube does not meet your expectations. In addition, I would not crank the monitor level: this is an unrealistic situation in the field. With headphones on, listen back to the noise in-between talking into the mic at reasonable monitor levels. You will know instinctively whether the noise floor of a mic/tube would interfere with a quiet passage of a musical recording. Remember, noise is cumulative: the noise of two tracks recorded with the same mic will increase twofold. Four tracks = four times the noise, and so on.

So, if you find that the background noise of a tube is 3 dB higher than that of your well-working tester mics, it may already be too much. I usually reject tubes whose noise floor is 2dB higher than the norm. But that is a professional situation: I don't even want to think of a client having the slightest doubts about his mic's noise floor.

Finally, if the noise is not steady state, but consists of periodic discharges, crackling, sputtering, etc. our ears are even more sensitive to such non-musical disturbances. I often find myself rejecting a tube which crackles at 1/2dB of noise volume or less, even if the tube's white noise is negligible (this scenario is frequently the  case with AC701 tubes).

[AusTex64]

Hello Klaus,

Thank you very much for your detailed reply. It is so cool to be able to communicate with guys like you, David Bock, Jim Williams, etc on your forum.

No, I don't have an Audio Precision system. :-) Gotta use my ears....

I did read the stickies (lots on good info in the AC701 and VF14 threads specifically). But your answers here really cleared up a lot of questions for me. I think I'm being too demanding on my tubes in relation to noise! 60dB of mic pre gain + headphones turned up all the way is not a real world test. Fortunately the tubes I'm testing (Telefunken PCC85) don't sputter or crackle, just white, steady state noise. Your comments have reassured me that what I have is really pretty quiet, all things considered. For example, I'm gettiing a comparable level of noise between my mic with PCC85 and the 70's U87, and I'm thinking that's probably pretty darn good.

Thanks again,

Robert Mokry

PS - I'm a closet hippie too.... he he

[hasbeen]

I posted a similar question here a while back. In my case I was testing the difference of noise generated between a Telefunken VF14 and VF14m. I used a similar method to yours by using headphones and cranking the gain on the mic pre.  The tubes are both easily quiet enough for use in a U47 but I wanted to know which was more quiet.

Lacking sophisticated test gear, my ears were able to choose the VF14M only by being able to listen at high volume through the headphones after recording several passages of various signals, including just the room.

I too decided I was being too critical. But still use the VF14M and keep the other as a spare.

[Kai]

Something not yet mentioned:
You need a sound source, e.g. a speaker with pink, or better midband filtered, noise to adjust the mic's under comparison to the same overall sensitivity, say recording level.
This can be done by listening too, or use any meter.
Else the mic with lower sensitivity will always seem to have the lower noise.
We want to know the signal to noise ratio, so we need a signal first for setup.

Then I wait until 3 o'clock in the morning (the city sleeps) and do the comparison in my room.
Very quiet mic's I lay into a sealed box filled with foam.

... use an average for a professional condenser mic: around 25-30dB...
...  I would not crank the monitor level: this is an unrealistic situation in the field. 

Most mic-pre's are less noisy (related to the input) on higher gain.
If you do use only 30dB it can happen you listen to the pre's noise, not to the mic's.
Higher gain, 60dB or so, has another advantage - you don't need to crank up the monitors to listen, avoiding to break them in case anything loud happens.
Listen to the noise at a realistic level, as it would appear in a mix.
This way your natural ear response does the correct filtering to judge the annoyance of the noise.

... noise is cumulative: the noise of two tracks recorded with the same mic will increase twofold. Four tracks = four times the noise, and so on.

Fortunately it's not that bad:
2 tracks give 3dB (50%), four tracks another 6db (double), 8 tracks 9dB, 16 tracks 12dB (four times) more noise, etc.
At the same time our mix gets louder with more tracks active, else it would be impossible to do good quality recordings with vintage mic's.
If I have very quiet instruments I do use very quiet mic's, cause the track with the worst noise usually dominates the final result.
BTW: a good 50's tube mic is on par with a good 70's transitorized mic, noise-wise, at least if you compare LDC Neumann's.

Regards
Kai

[klaus]

If you do use only 30dB it can happen you listen to the pre's noise, not to the mic's.

It should have said "-30dBu". Thanks for pointing out the mistake. I have corrected the citation.

Here is another way to measure tube mic noise, which I recently posted on another forum:

I have arrived at a "semi-objective" test protocol which, over the years, has served me well to weed out marginal tubes, and not only VF14:

I have several healthy, low-noise-select, VF14M at my disposal. I install one of them in a perfectly restored, optimally-functioning U47. That mic is inserted into an acoustic isolation chamber which blocks external noise down to -120dB. I maximize the monitor level of the headphone output of a special Nakamichi mic preamp which also has a white noise generator. I switch back- and forth between mic noise and white noise, until I have calibrated the pre-amp’s white noise output to exactly match the level of the mic's white noise. I then substitute the calibration tube in the mic with the tube to be tested. I let the mic run for a day or two, after repeatedly re-calibrating the supply voltage to Neumann spec. Then I compare the tube's white noise to the Nakamichi-generated white noise. As my pre-amp is equipped with dB-stepped attenuator, I can quickly assess the tube's noise floor relative to that of the Neumann-spec VF14, represented by the pre-amp-generated, calibrated white noise.

After all of that, I finally undertake an "in the field" test: from many years' experience I know at roughly what level of noise a tube will interfere with quiet passages of a recording session in a modern-day, state-of-the-art recording studio. I can extrapolate from that experience what noise would become objectionable by simply talking into the mic at a vocalist's distance from the capsule.  There is not much human tolerance for noise from a VF14 or other tube-equipped condenser mic: even a lay person will hear 4dB elevated noise from a tube instantly, (a trained ear can distinguish as little as 1/4 dB noise elevation, especially in direct comparison to another noise source).

[Jim Williams]

That's a lot of work. I don't trust my ears without verification.

A modern PC based analyzer can plot a noise vs frequency sweep in about 10 seconds.

The benefit of that is you see the spectral noise, it will vary vs frequency. Shot noise, high frequency hiss, power supply leakage are all shown in color.

Part of reducing noise is to discover what is creating it and at what frequency. Wideband spot noise measurements like 20~22k hz s/n ratio tests will not seperate these sources of noise like the noise vs frequency plots.

[boz6906]

Most DAW software will let one easily see the noise spectrum and make credible observations about the source.

In a recent test I compared the noise spectrum of a Neumann V476B preamp to that of a RME UFX mic pre.

I set both gains at a typical value (+40dB) and recorded 20 or 30 seconds with no input (you could also terminate the inputs with a 1k2 resisor).

The V476B showed a level of about -77dB, the RME -79dB, leading one to think the RME was quieter,

Here's the important step, normalize the recorded tracks and observe the noise spectra; in this case the V476B's noise was all below 100Hz, the RME's was spread over the entire range.

The V476B's noise above 100Hz was 15-20dB lower than the RME making the RME apparently much noiser in the audible bandwidth, despite the 'lower' measured noise.

[klaus]

What you mentioned in your last paragraph is the crucial aspect, why I still prefer audible testing:
Our ears are quite discriminating in the weighting of noise, and there is no clear-cut comparison between bands and objectionable noise.

That is why I prefer using white noise: it seems to fall into the range and distribution of greatest annoyance. It coincidentally is also the type of noise where VF14 usually indicate their state of health.

[David Satz]

boz6906, unfortunately the noise of a mike preamp with its inputs left "open" doesn't tell you anything about that preamp's noise performance when its input is shunted by the actual impedance of a microphone. Even the relative noise performance of two preamps can't be judged that way.

If you were to repeat your experiment, this time with a resistor across the inputs, having a value of somewhere from 150 to 200 Ohms so as to mimic the output impedance of a "vintage" studio condenser microphone with an output transformer--I think a lot of people here would be interested in the results.

It would also be interesting to see the results with a lower impedance typical of modern, transformerless condenser microphones, which typically have output impedances in the 35 to 50 Ohm range.

--best regards

[boz6906]

Yep, you are correct, a load on the input might yield a more real-world result, my goal was just a quick comparison of new vs. old German preamp design.

I'm also interested in the noise spectra of linear vs. switched supplies.

The RME is switched, the V476 rack came with a +24Vdc Chinese switching supply, I'm interested to see the level of power supply harmonics in the output.

[Jim Williams]

Mic preamp noise, which BTW should be left out of a mic noise test, will vary with load.

Some designs will show marked improvements with lower source impedances. Measure at the standard EU 200 ohms, than the US standard of 150 ohms, use a resistor across pin's 2+3 to test those loads.

I also test at 50 ohms to cover the modern SS circuits that are popular. Transformer input mic preamp designs will limit EIN to about -127 dbu with 1/5 ratio transformers. That is due to 2.2 db of transformer insertion losses. Dump the transformer for low noise transistors and that improves to -129.6 EIN at 150 ohms. Measure some designs at 50 ohms and you can get -133 db EIN, quite an improvement over any transformer coupled design. Still, you are looking at a 20~22k bandwidth s/n ratio of about 73 db.