Topic: KM84 FET Replacement: Type and Procedure (Read 18476 times)

squareneck · « **on:** January 17, 2013, 09:02:51 AM »

Hello!

I have a Neumann KM 84i that had no output, I found out that it was the FET that was broken. I bought a Fairchild 2N3819 and replaced the broken one wich was a SK107. Now the mic is working again but the output is low and quite noisy.

Do I have to change some other component to get the 2N3819 to work properly?

best regards/ Bo Savik

klaus · « **Reply #1 on:** January 17, 2013, 01:41:47 PM »

Assuming that the Fairchild FET you bought is low noise, you still need to individually bias any FET (as Neumann originally did in your mic when they built it.) That's why the existing bias resistor (RS) needs to come out, and proper biasing needs to be set up for the new FET.

Here is a simple way to get very close to the ideal biasing point:

Attach a wire to the gate of the FET (or take the head off, and squeeze the wire into the slit of the two gold-plated wires the head's contact needle attaches to), attach this wire to a signal generator putting out 1kHz @ ca. -25dB or lower (you will adjust the exact output level of the generator later), remove the current bias resistor, and solder a wire to that connection point on the circuit board. Attach the other end of that wire to a multi-turn trimpot (10-20kΩ), and attach a second wire to a convenient ground on the mic's circuit board whose other end will be soldered to the second contact of the trim pot.

Plug in the mic into proper 48V phantom (no need to squeeze the housing tube over the test wires), and observe the meter of your mixing board/pre-amp rise as you increase resistance on the trimpot. Turn the trimmer from zero Ω until you hear a faint (and very distorted) signal of the test tone, continue increasing the resistance, until you get the highest output with the least amount of distortion at 1kHz.

As you keep increasing resistance, you will notice two things happening:
1. the signal gets lower again, and 2. the signal starts to distort again.

Back off the trimmer to the least distorting point, unsolder the trimmer wires, measure the resistance you have dialed in, and find a fixed 1/8W or 1/4W resistor that comes closest to that value.

Done.

Some final notes of advice: use headphones for this procedure, at very low, barely audible listening levels, and don't have any other noises in the room that will interfere with your hearing. It makes the final set point that much easier to find.

Speaking of final set point: this process explained above is even more precise and yields lower distortion than how most manufacturers do it.

squareneck · « **Reply #2 on:** January 18, 2013, 03:01:16 AM »

Thank you so much Klaus for your fast and professional response! I will try this, I'm really looking forward too it.

Is there some way I can tell if the 2N3819 is noisy? I have four of them so if there is a simple way to test them maybe I should do that to get the best one into the KM 84i.

best/Bo

klaus · « **Reply #3 on:** January 18, 2013, 05:10:10 AM »

You will know once the FET is properly biased.

squareneck · « **Reply #4 on:** January 18, 2013, 05:25:02 AM »

Ok. Thank you!

Jim Williams · « **Reply #5 on:** January 21, 2013, 12:59:27 PM »

Listening tests have shown that any THD below about 3% is very difficult to identify, especially if you are attempting to find the lowest THD setting.

This is where the test equipment is used. Any decent THD analyzer can find the bias spots easily, without any guesswork. Do it by ear, then re-check with the test gear. You will find your settings are probably off to some degree. This should be done by any mic moder or repair person as certainty is what is being paid for.

klaus · « **Reply #6 on:** January 21, 2013, 04:43:20 PM »

I guess, I am one mic modder who does not use a distortion analyzer anymore. I found the exact same value of the correct source resistor through both methods enough times, that I leave the analyzer in the basement these days, and just do it by ear.

Besides: if I cannot hear distortion, who cares what I might see? Granted, measuring instruments can quantify what I hear, if I hear it., but that to me seems a secondary pursuit, after detecting anomalies by ear.

Kai · « **Reply #7 on:** January 24, 2013, 05:46:23 AM »

Quote from: klaus on January 21, 2013, 04:43:20 PM

if I cannot hear distortion, who cares what I might see?

If you like to actually hear distortions way below 1%, use the differential distortion method.

Feed in a two tone signal like 19kHz + 20kHz sine.
You probably won't hear these any more if you are male and older then 25.
But you can easily hear the distortion, because it's 1kHz.

The same applies to musical signals; sibilants e.g. are suffering from these kinds of distortions, because they aren't harmonic and their spectrum is far apart from that of the original signal.
The strongest test I know: use a triangle (perc. instrument).
You shouldn't hear any "wooden" knock together with the clear hit.

Explanation: differential distortions aren't separate from harmonic distortions, it's just another and sometime more obvious way to show the same nonlinearities of devices.

Regards
Kai

Jim Williams · « **Reply #8 on:** January 28, 2013, 12:10:10 PM »

The CCIF twin tone IMD tests are best for that. Shove a 15k sine with a 14k square wave and you will find anomolies. This test also has far higher resolution than any common THD+noise test which also incorporates the noise readings. The CCIF method eliminates the noise contributions so measurements can be seen at a far lower level. Typical Audio Precision resolution of the THD+noise tests are .0005%. CCIF goes down to 1 PPM.

If you can't hear these non-linearities that doesn't make them magically go away. Just because you can't hear the residual THD doing a bias set by ear also doesn't mean someone else with finer tuned hearing won't notice.

"Trust, but verify".

klaus · « **Reply #9 on:** January 28, 2013, 12:49:35 PM »

Quote from: Jim Williams on January 28, 2013, 12:10:10 PM

...If you can't hear these non-linearities that doesn't make them magically go away. Just because you can't hear the residual THD doing a bias set by ear also doesn't mean someone else with finer tuned hearing won't notice.

I have done enough of these over the years that I feel comfortable about passing on this simple FET biasing method without reservation. Let's talk again about 1PPM THD resolution once we have eliminated MP3 encoding and 20khz brickwall filtering.

Jim Williams · « **Reply #10 on:** January 29, 2013, 11:00:07 AM »

Some don't do MP3, there are some people doing high resolution recordings. After investing thousands in an old mic, I believe they would appreciate the extra fine tuning.

If things down stream are so destructive to the sonics of a well tuned microphone, you might as well save those thousands of $$$ and buy Chinese stuff if it doesn't make any difference.

klaus · « **Reply #11 on:** January 29, 2013, 01:47:15 PM »

Final note from me on this: to my ears, biasing FETs beyond the level of resolution that my method yields has not shown an audible improvement.

squareneck · « **Reply #12 on:** January 29, 2013, 03:48:34 PM »

Now I have done what you told me and found that 4.8 kOhm gave me the best result. I have now ordered a new resistor á 4.7 kOhm. The one that belonged to the old FET was 3.9 kOhm so I guess I will have more signal and less noise now :-) I'm very happy for your support and will report again when the new resistor is in place!

Kai · « **Reply #13 on:** January 30, 2013, 09:10:18 AM »

Quote from: Jim Williams on January 28, 2013, 12:10:10 PM

The CCIF twin tone IMD tests are best for that. Shove a 15k sine with a 14k square wave

I've mentioned the twin tone method mainly because an untrained ear can hear the distortion much clearer.
No analyser needs to be used.
For that purpose, the exact frequencies are of minor importance.

The CCIF method mentioned is even more demanding, as the square wave produces several different (sub-) harmonics. For listening it might be a bit too much.
You need a generator capable of producing a 14kHz square wave - this is not possible if, e.g. a CD is the source, because the 20kHz band limit converts the square into a sine.

The results of different distortion measurement methods, btw., aren't much different, if the harmonics lie within the usable bandwidth.
1% nonlinearity stays the same, however you measure it.

Regards
Kai

Jim Williams · « **Reply #14 on:** January 30, 2013, 12:09:26 PM »

With modern test gear like the Audio Precision 2600 series you can set the test up and see and measure all the pertanent perameters at the same time with one sweep.

You can see THD+noise, frequency response, IMD and phase response, you also get a noise vs frequency plot to sus out the quietest transistors.

Do that with a set of jfets and you will see they are all over the place in spec. This is the most precise and predictable way to find the best parts and set bias for the highest performance, if you care to do that.

radardoug · « **Reply #15 on:** January 30, 2013, 03:07:46 PM »

Any thoughts from the panel on a replacement for the 2N3819 seeing they are getting hard to get?

klaus · « **Reply #16 on:** January 30, 2013, 07:45:15 PM »

They are readily available for little money from major electronics online shops under "Fairchild 2N3819".

radardoug · « **Reply #17 on:** January 31, 2013, 02:58:03 AM »

I did a quick check before I posted, none at Digikey, Farnell, Mouser.

usattler · « **Reply #18 on:** January 31, 2013, 09:12:04 AM »

Random leads to suppliers with current inventory of 2N3819:

http://www.newark.com/jsp/displayProduct.jsp?sku=33C8258&CMP=KNC-G-SKU-OTH-NTEELECTRONICS&mckv=swEKFcUeF|pcrid|17260993980|plid|

http://www.mouser.com/Search/Refine.aspx?Keyword=2N3819

http://www.jameco.com/1/1/390-2n3819-jfet-transistor-j-fet-n-channel-small-signal-92.html

klaus · « **Reply #19 on:** January 31, 2013, 09:38:10 AM »

To my knowledge, Fairchild is the only manufacturer which restarted production on discrete FETs. Hence my recommendation to google Fairchild + 2N3819

Jim Williams · « **Reply #20 on:** February 01, 2013, 12:07:38 PM »

I wouldn't waste time on Fairchild jfets. As I mentioned, their specs are all over the map so selecting them is a PITA, IF you want the highest performance.

A better source is:
www.linearsystems.com

This is a company in Fremont, CA that specializes in high performance jfets for high end audio. They have their version of the 2N3819 jfet, the LS846. This part is higher performance with a low 3 nv noise spec and only 15 pa gate leakage. I suggest that a $1000+ EU microphone is deserving of a better jfet, they only cost a buck or two.

Linear Systems also makes their version of the 2SK170, another common jfet used in condenser microphones. This is the LSK170, only 1 nv noise and a great sub in an U-87, if you want to lower the noise floor.

klaus · « **Reply #21 on:** February 01, 2013, 05:28:54 PM »

Thanks for the tip. I've been happy with the batch of Fairchilds I have been installing ( a couple of duds, here and there, but overall happy). But will give these a try as well.

squareneck · « **Reply #22 on:** February 05, 2013, 06:35:29 AM »

Hi again!
I have now soldered in a new resistor (R3). But still the output is very weak, I tried with another 2N3819 but there was no difference so I decided to measure some voltage (according to the schematics) and compare to my other KM 84 which is fine and I found out:

Working one: R10(46V) = 45,8V. GR1(24V). =23.8V. C5(21.5V)=20.6V. = Drain(10V)=6V
Bad one: R10(46V) = 45,7V. GR1(24V). =22.5V. C5(21.5V)=18.7V. Drain(10V):=1.7V

So I thought I must have done something wrong when I was trying to find out the value of R3. What I did now was once again placing a variable resistor on R3 and then adjusting until I had 10V at the drain. This gave me 8.2kOhm instead of 4.7 kOhm. So I will order a resistor close to 8.2 kOhm and then I'll see if that solves the problem. Do you think I'm on the right track?

Best/Bo

ps. I've also replaced C3, C4, C5 and C8 since I heard they could be old and leaky. ds.

klaus · « **Reply #23 on:** February 05, 2013, 02:45:16 PM »

Quote from: squareneck on February 05, 2013, 06:35:29 AM

Do you think I'm on the right track?

I don't think so. After you have undertaken many attempts which included four different FETs, proper FET biasing etc. You have so far not gotten full output of the mic.

I would start looking somewhere else for the reason why you are down so much: Have you made sure that the pads are bypassed, and the mic's output transformer is strapped to 200Ω? Does the capsule measure roughly 32-36pf capacitance, as measured between contact needle and outer capsule shell?

Jim Williams · « **Reply #24 on:** February 06, 2013, 11:15:08 AM »

Get out a scope and feed a 50 mv, 1k sine wave into the fet gate. Observe each stage for proper signal levels. You will find where it stops.

squareneck · « **Reply #25 on:** February 12, 2013, 02:30:12 AM »

Finally I've got the mic to work properly again :-)
I must have done something wrong the first time I did Klaus method for finding out the value of R3, cause when I repeated the procedure this time I found out that 8.2 kOhm did it for me and that also corresponds well to what I found out by measuring voltage which gave me 8.2 kOmh. So when I now soldered in the new R3 (8.2 kOhm) the mic sounds perfect again.

Thanks a thousands for all your help I've learned a lot on this project and looking forward to my next one!

R/E/P Community

Author Topic: KM84 FET Replacement: Type and Procedure (Read 18476 times)