R/E/P Community
R/E/P => Klaus Heyne's Mic Lab => Topic started by: byacey on February 19, 2012, 03:56:16 AM
-
Hi Have a U87 that has been sitting for a number of years due to the capsule base plastic self-destructing. I've since replaced the base with the newer type with the magnetic reed switches. After re-assembling and testing, it turns out the 2N3819 FET had shorted for some reason. I had some 2N5457s on hand, so I put one of those in. After re-biasing it seems to sound fine. Has anyone any experience with using this particular FET in this application?
Thanks,
Bill Y.
-
I am afraid the answer is not as straight-forward as throwing you a simple part number or referring to spec sheet for a given FET. I have found that the two most relevant parameters of J-FETs, noise and headroom, cannot be gleaned from 1/F, max voltage or other information provided by the manufacturer.
If I were to rely on that information, many a FET would be preferable to the good old 2N3819 or other stand-bys which Neumann used in more microphone models for more years than any other FET.
I tried a few of these "superior" FETs (as per spec sheet) and found that their headroom was so grossly inadequate that I was scratching my head where that discrepancy may stem from.
I am not an electronics engineer and not even the kind of layman who would dedicate enough time necessary to be reaching deeply into this matter, but I can speculate that the enormous input impedance the gate sees in a single-FET microphone (in excess of ten thousand million Ohms) throws off the usual lab specs of the device, similarly to the impossibility to predict tube noise in this type of high impedance environment, no matter what the tube tester (or tube salesman) says.
The bottom line: you need to make simple noise and headroom tests, whereby ONLY the FET becomes the variable under test. Feed a clean sine wave into a high quality speaker, which you place 1 foot in front of your mic. Increase the level of the speaker output until you can clearly record (or hear) the sine wave at the output of the mic collapsing into third harmonic distortion (while the speaker output remains undistorted) Measure the mic's output at onset of the audible distortion, and repeat the test with a sine wave fed directly into the input of the gate. Then move on to the next FET and do the same again.
What complicates the test somewhat (though not as much as the specs seem to indicate) is that different J-Fets have different gain. For me, though, this is a secondary consideration when selecting a FET for a mic, as most J-FETs will only vary the mic's output by a few dB.
On the other hand, the amount of a FET's headroom is absolutely crucial fora microphone application, and there the surprises come in: The headroom of different J-FETs commonly used in condenser mics varied as much as 18dB in tests I made last year (and whose findings I discussed at length with Andreas Grosser who generously had supplied some of the test FETs for me over the years and keeps finding new candidates).
Another variable raised its ugly head in these tests: FET quality varied too much for my taste to rely on any FET from a given batch: For example, I tested several dozen Fairchild 2N3819 (recently available again!) and found that too many of them either did not power up, could not be properly biased, or had noise problems. Besides, none of these was equal in noise or headroom to the venerable Texas Instrument 2N3819. So selection, even within one batch, is an important step to finding a superior FET.
P.S.: I assume you know how to properly bias a FET in a condenser mic- the pre-requisite for this whole discussion? The goal is to bias for the golden cut- where lowest distortion and highest output meet.
I am curious to learn how others arrived at their favorite ("best") FET.
-
Klaus,
Thanks for the detailed reply. I initially just biased the drain for 10 volts.
Today I hooked up the scope and audio generator to the test point.
I fed a 1KHz signal into the test point and kept raising the level until the output waveform of the mic started clipping. I then juggled the selectable resistor value until I had symmetrical clipping of the waveform. This results in maximum headroom.
It turns out the selectable resistor worked out very close to 4.7K, achieving 13.4V on the drain of the FET.
-
Already one useful result of individual biasing, compared to going "by spec"!
-
...kept raising the level until the output waveform of the mic started clipping... juggled the...resistor value until I had symmetrical clipping... achieving 13.4V on the drain of the FET.
Watch out for drain DC changes with and without a clipping signal applied.
If the DC changes, the mic will cutoff at high levels.
This is, unfortunately, the 3rd variable (besides distortion and gain) to be taken into account.
P.S.: I assume you know how to properly bias a FET in a condenser mic- the pre-requisite for this whole discussion? The goal is to bias for the golden cut- where lowest distortion and highest output meet.
I am curious to learn how others arrived at their favorite ("best") FET.
I've build a mic frontend configuration which replaces the drain resistor of the FET with an adjustable low imp. supply.
It uses source-gate feedback too.
The advantages of this circuit are:
- it yields equal dynamics with a big range of different FET's, so you just have to select them for noise.
- the distortion in the normal level range can be trimmed to nearly zero.
- No high level cutoff appears.
- Source resitor doesn't need to be closely selected, it just needs to be designed for the basic type of FET (A, B or C, which specifies the the FET's drain-source current).
- gate resistor can be much smaller then usual (22M) without compromising low frequencies due to feedback used.
Downside:
-5 parts (current source, FET, trimmer, 2 capacitors) needed to replace usual drain resistor arrangement (2 capacitors, 2 resistors).
- Not useable as direct replacement in U87 (wouldn't do this anyway).
Regards
Kai