Neumann Vintage U87 Clone : Build Thread.

[poctop]

However the implicit assumption here is that the U87 biasing string depends on the Zener keeping it's node at 24V!  It could be the original JFET's didn't pull much current.  However a modern 2N3819 can range anywhere from 2mA IDSS all the way up to 20mA!  That is a tremendous range for a single topology to cope with!

i tought the the use of the zener here was to keep the node just above the breakdown Knees at about 22V to keep the curent in the linear Portion To stabilize the Voltage without pulling too much of current for no reason.  :-\
Dan,

 

[poctop]

I just received some other boards if someone would like to give it a shot.
Dan,

 

[wthrelfall]

Dany

I removed the transformer connections, and the connections to the capsule, and also removed the fet and R11.  original bom components elsewhere.  Here are my voltages:

Phantom power is 48.6v

top of L1 - 48.6v
top of L2 - 45.9v
top of R18 - 48.6v
top of R19 - 45v
bottom of R17 - 45v
top of R17 - 23.18v
Zener 23.17v
FET drain pad 23.15v
FET source pad 0.089v
top of R12 23.15v
top of R13 42.4v
top of R14 23.14v
top of R8 41.8v


What is R9 & R10 doing?  I got 0v there..

 

[HellfireStudios]

Has anyone tried the 1.5-2.2 pF across the the Gate and Drain of the FET yet? This supposedly makes newer 2N3819's closer to the old ones (there was a reference to black magic in the post I read).
If a cap is implemented in this way, would biasing be carried out in the same manner? Any other circuit considerations? Thanks for all the info contributed so far.

Here's the original post: I don't know who to quote here, but you know who you are...

Hi Dan,
Did you measured amp specs via test input? There is Neumann U87i specification at page 8:
 http://www.neumann.com/download.php?download=copi0042.PDF
I had a problem to meet frequency response specs, as well as THD specs. 
After some research I found simple way to tune the circuit to original specs by adding 1.5-2.2pF capacitor between Gate and Drain of FET. Why so? I think that modern 2N3819 FETs have much lower reverse capacitance than original ones from 70s.
Hope it can help in U87 black magic.

 

[wthrelfall]

Ok, so I put the 25k pot on R11 and the 50K pot on R17.
With the fet drain set at 10.5v by R11 (1.67k), I adjust R17 to give me 23.17v at the Zener (turned right down to 65 ohm!!). 
I experiment with some other combinations, and I find that with R17 adjusted to a point where the zener hits 23.17'ish it then begins to be ineffective.  at that point the FET reads 17.66v with R11 @ 4.64k. 

But I did manage to measure my 3 fet's.  they were 11.4ma, 10ma & 9.7ma.  Not sure if these figures are higher than 'normal'?  anyway, I installed the 9.7ma fet, and put all my resistors back to the original values specified on the BOM.  This time I installed a 50k pot at R11, and following matador's guidelines, I was able to verify that while reducing the resistance from maximum to minimum, the fet behaved as expected (with the sudden rise near the lowest resistance).  With the drain reading around 10.5v, the pot was giving 5K of resistance, and I'm getting 18.3v at the zener.  I swapped the pot out for a 5k1 resistor and tested the mic, and yes it sounds alot better, actually pretty good.  Signal is low, needs tons of gain for a decent level.  I also did as Dany asked and put a 10k resistor at R11 and tested it.  sounds pretty much the same as the 5k1, but the zener is getting over 20v.

You are very close!  The next step is replace R17 with a 50K pot, and start reducing it until the Zener voltage reaches 24V.  You shouldn't see a lot of modulation of the drain voltage, however your output swing before distortion will be better (assume a fixed 10.5V drain value for the chosen bias!).

The 10.5V value is not sacrosanct:  it was chosen basically because it's exactly half of 21V, which on the original U87 was selected as the "VDD" value.  For the JFET's being used, you aren't reaching 21V, so 10.5V is no longer "optimum", based solely on signal swing capability.

All of those currents you measured were in the datasheet ballpark of roughly middle between 2mA and 20mA, so don't worry too much.

As for gain:  what transformer did you end up using?  The gain of the circuit is (roughly) -R12/R11, which for R12=47K and R11 =5K is about 10.  And the step-down transformer will throw part of that back away:  for example a 10:1 Cinemag CM-2480 will reduce the overall gain of the circuit down to unity.  The original Neumann transformer was (I believe 7:1) which means the overall gain is only 3dB.  So unless you are close miking stuff with hot transients the mike is going to be on the quiet side.

 

[tskguy]

Hi all,

I just ordered some of these boards..Looks like a fun project! I am going to use this as a test bed for some capsule design. Anyway I found this and thought it was interesting read.
http://recordinghacks.com/microphones/Neumann/U-87-Ai

 

[Matador]

Dany

I removed the transformer connections, and the connections to the capsule, and also removed the fet and R11.  original bom components elsewhere.  Here are my voltages:

Phantom power is 48.6v

top of L1 - 48.6v
top of L2 - 45.9v
top of R18 - 48.6v
top of R19 - 45v
bottom of R17 - 45v
top of R17 - 23.18v
Zener 23.17v
FET drain pad 23.15v
FET source pad 0.089v
top of R12 23.15v
top of R13 42.4v
top of R14 23.14v
top of R8 41.8v


What is R9 & R10 doing?  I got 0v there..

All of that looks correct.  You are pulling 0.32mA through R17 because the Zener is sinking current to keep the node "top of R17" at it's breakdown voltage.  Once the Zener is regulating, R17 does nothing as the system acts like a constant current source (the Zener will change it's internal impedance to keep the voltage across it at 23.14V, making R17 "do nothing" as it's adjusted). 

The idea is that you turn R17 down until the Zener just starts regulating.  Once this happens, you re-adjust R11 until you get 10.5V at the drain, then check the Zener voltage to make sure it hasn't changed, and if it dropped, you lower R17 a little bit more, and iterate through the process.  Something like this:

1) Raise R11 and R17 to maximum.
2) Stick in your FET
3) If the Zener node is not 23.14V, lower R17 until it just hits 23.14V
4) Lower R11 until you see 10.5V
5) If the Zener node is not 23.14V, go back to step 3

After several of these sequences you should be able to get both conditions met, at least for that FET, and that Zener.  You can then remove the pots and substitute the closest values that you have if you want to.

R9 and R10 are there to provide an AC ground point for the capsule backplate.  C5 will not allow DC current flow through these two resistors, so your meter will not register a voltage there.

 

[wthrelfall]

Yes, that process is basically what I've been doing, over and over again and now my boards are losing pads due to the number of parts being changed.  I was feeling like I was getting 'good' numbers, and successfully got the sine input and tweaked it for 'sound' as well as for volts.  so feeling confident, i plugged in the capsule and tested the mic...  Now it's noisier and still ridiculously quiet.  I need at least 60 - 70db of gain to get a usable signal but then it's so noisy it's unusable.

Dany

I removed the transformer connections, and the connections to the capsule, and also removed the fet and R11.  original bom components elsewhere.  Here are my voltages:

Phantom power is 48.6v

top of L1 - 48.6v
top of L2 - 45.9v
top of R18 - 48.6v
top of R19 - 45v
bottom of R17 - 45v
top of R17 - 23.18v
Zener 23.17v
FET drain pad 23.15v
FET source pad 0.089v
top of R12 23.15v
top of R13 42.4v
top of R14 23.14v
top of R8 41.8v


What is R9 & R10 doing?  I got 0v there..

All of that looks correct.  You are pulling 0.32mA through R17 because the Zener is sinking current to keep the node "top of R17" at it's breakdown voltage.  Once the Zener is regulating, R17 does nothing as the system acts like a constant current source (the Zener will change it's internal impedance to keep the voltage across it at 23.14V, making R17 "do nothing" as it's adjusted). 

The idea is that you turn R17 down until the Zener just starts regulating.  Once this happens, you re-adjust R11 until you get 10.5V at the drain, then check the Zener voltage to make sure it hasn't changed, and if it dropped, you lower R17 a little bit more, and iterate through the process.  Something like this:

1) Raise R11 and R17 to maximum.
2) Stick in your FET
3) If the Zener node is not 23.14V, lower R17 until it just hits 23.14V
4) Lower R11 until you see 10.5V
5) If the Zener node is not 23.14V, go back to step 3

After several of these sequences you should be able to get both conditions met, at least for that FET, and that Zener.  You can then remove the pots and substitute the closest values that you have if you want to.

R9 and R10 are there to provide an AC ground point for the capsule backplate.  C5 will not allow DC current flow through these two resistors, so your meter will not register a voltage there.

 

[Matador]

Make sure you scrub the area around the capsule connections with isopropyl alcohol to get rid of solder flux:  the board can't be "too clean" in this regard.  The more you solder around this area (like where the gate of the JFET connects), the more flux splatter you have.  The high impedance nodes see any leakage paths and can cause noise like you are describing.

The U87 circuit is definitely not high gain in any case. . My Aurycle mikes need 50-60dB when used several feet back from quieter sources.

 

[Matador]

Thinking about this some more:  there is a duality with this circuit due to the capsule bias.  We can maximize gain by finding higher gm transistors and hand selecting higher IDSS from our stash.  We could then lower the drain resistor and probably squeeze another 25dB out of the circuit.  I can see why the 2SK170 is a popular mike JFET as it has 10 times the transconductance of the 2N3819. But those concerned about needing too much gain (and not recording highly dynamic sources that need lots of input headroom) might want to try this JFET in this circuit.

But the more current we pull from the phantom supply, the lower our capsule bias, which increases noise and reduces sensitivity, which ultimately steals back gain.

 

[HellfireStudios]

My Carvin CM-87S only has switches for pad and low-cut. I was wondering if there were any suggestions for mounting a switch inside the mic with this PCB.
On the mic's original PCB, there was a place to put a DPDT switch for the low-cut.  My original plan was to add a switch on the PCB, and replace the low-cut on the exterior with a pattern selector switch. Could this be adapted to work with the new PCB?
I would hate to have to add a relay system just for pattern switching, unless the geniuses on this forum can figure out a way to add a relay without replacing the 3-pin XLR with what I'm guessing is at least 5-pin.
If this is the only plausible solution, are there any suggestions on relays that will fit and work reliably?
Sorry about all the questions. My head is swimming with the 15 or so electronic projects I'm trying to finalize so I can order the goods. As the old saying goes, "Design twice, order once."

 

[wthrelfall]

Matador,
That seems like a good idea.  Could the aurycle transformer be adding considerable noise or be giving a lower gain?  I'm going to get a cm-2480 and a Peluso capsule so I think this should make some marginal improvements.  I'm also going to build up the Aurycle board using your ideas and BOM from your other thread, see how that sounds..  Have you used and heard what the stock Aurycle components sound like?  big differences using your BOM?

cheers.

Make sure you scrub the area around the capsule connections with isopropyl alcohol to get rid of solder flux:  the board can't be "too clean" in this regard.  The more you solder around this area (like where the gate of the JFET connects), the more flux splatter you have.  The high impedance nodes see any leakage paths and can cause noise like you are describing.

The U87 circuit is definitely not high gain in any case. . My Aurycle mikes need 50-60dB when used several feet back from quieter sources.

 

[Matador]

Matador,
That seems like a good idea.  Could the aurycle transformer be adding considerable noise or be giving a lower gain?  I'm going to get a cm-2480 and a Peluso capsule so I think this should make some marginal improvements.  I'm also going to build up the Aurycle board using your ideas and BOM from your other thread, see how that sounds..  Have you used and heard what the stock Aurycle components sound like?  big differences using your BOM?

Wait...you are using the stock Aurycle transformer with the poctop's boards?

 

[Matador]

Nevermind, you already said you were earlier in the thread.

We need to confirm that the transformer supplied with the Aurycle kit is indeed a 10:1.  The exact same mike from MXL (2001?) comes with a 2:1 transformer which is common for circuits that contain the PNP follower stage (as the 2:1 is cheaper).

If Aurycle is giving out 2:1 transformers, this would explain why the mike output is messed up, as the U87 common drain stage would be immensely loaded down.  If the mike was plugged into a pre with a 600ohm input impedance, this would look to the JFET stage like a 2K4 load, which means that 95% of the signal would be thrown away.

 

[wthrelfall]

Any way I can check the transformer ratio?

Nevermind, you already said you were earlier in the thread.

We need to confirm that the transformer supplied with the Aurycle kit is indeed a 10:1.  The exact same mike from MXL (2001?) comes with a 2:1 transformer which is common for circuits that contain the PNP follower stage (as the 2:1 is cheaper).

If Aurycle is giving out 2:1 transformers, this would explain why the mike output is messed up, as the U87 common drain stage would be immensely loaded down.  If the mike was plugged into a pre with a 600ohm input impedance, this would look to the JFET stage like a 2K4 load, which means that 95% of the signal would be thrown away.

 

[wthrelfall]

Cleaned up the boards, still noisy.  Here's a screen grab of the noise profile of the mic with the capsule disconnected, with 54db of gain.

 

[wthrelfall]

Here's another, with transformer disconnected, same gain.  (not sure if this actually confirms if the tx is noisy, is the circuit now incomplete?)

 

[Matador]

Here's another, with transformer disconnected, same gain.  (not sure if this actually confirms if the tx is noisy, is the circuit now incomplete?)

What point in the circuit are you measuring from in both cases?  From the output of the drain?  After the coupling cap?  From the XLR output?

How well matched are the phantom 2k21 resistors?  Were they the exact same value?

 

[wthrelfall]

measuring from the xlr output.
can't recall the exact measurements, but this was my voltage readings taken earlier
top of L1 - 48.6v
top of L2 - 45.9v
top of R18 - 48.6v
top of R19 - 45v

Here's another, with transformer disconnected, same gain.  (not sure if this actually confirms if the tx is noisy, is the circuit now incomplete?)

What point in the circuit are you measuring from in both cases?  From the output of the drain?  After the coupling cap?  From the XLR output?

How well matched are the phantom 2k21 resistors?  Were they the exact same value?

 

[Matador]

measuring from the xlr output.
can't recall the exact measurements, but this was my voltage readings taken earlier
top of L1 - 48.6v
top of L2 - 45.9v
top of R18 - 48.6v
top of R19 - 45v

That doesn't look right:  you should be pulling phantom current equally through R18 and R19, which means the common nodes on the "top" of these resistors should be the same.  This is why R18 and R19 are specified to be closely matched (to within 0.1 ohms or better).

The voltages you show above mean that there is a 2.7V drop across the transformer, which means there is DC current flowing through the secondary, which is going to be 50Hz hum city.

There's only two things that can explain this:  either of R18 or R19 is not the correct 2K21 ohm value, or the secondary may have a short to ground. 

It seems that most of your woes center around this Aurycle iron.   When you disconnect the transformer, do your "top of R18/R19" voltages re-equalize?