- Messages
- 3,703
A thread about building two Neumann U87 clones
- Thread starter Cirrus
- Start date
Cirrus
Roadie
- Messages
- 312
Now we're getting somewhere.
FETs selected and soldered on.
Tested for IDSS, which I touched on in the first post. It's the max. current drawn when the FET is biased as hot as possible - just shorted to ground. The FETs in question are 2n3819s, and their datasheet says it's nominally 10milliamps, but can be between 2 & 20 which is a big range. And it matters in the U87 because it doesn't have a lot of current to play with. The ones they used in the late '60s and '70s were apparently at the lower end of the IDSS scale, around 4 or 5 mA. I sourced 9, and the lowest I found was 8.35mA going up to 13mA. Damn. Then someone I know randomly posted that they'd found a whole treasure trove of different transistors, all sorted and labelled.
"Hey don't suppose there are any old 2n3819s in there...?"
He sent me ALL the 2n3819s he had, told me to pick my pair, and send the rest back. So tonight in went two lovely old Texas Instruments FETs, measuring 3.75 & 4.58mA. Yum!
Then lots of scrubbing with 99% isopropyl alcohol, because the back of the boards can't be too clean. Those 1G and 68m resistors present a problem - if you're asking the electrons to go through a really high value resistor, you can't also give them a path through finger oil and solder residue that will form a more attractive path for current. It's why the whole project's been done wearing gloves, and it's why I've used a toothbrush on the back of the PCBs.
Now that the front PCBs are screwed on these are beginning to feel like they'll be mics soon. Another good session of connecting the two boards and wiring up the XLR socket, and I'll be ready for the next leap into the unknown: Biasing the FETs with a test tone & oscilloscope.
FETs selected and soldered on.
Tested for IDSS, which I touched on in the first post. It's the max. current drawn when the FET is biased as hot as possible - just shorted to ground. The FETs in question are 2n3819s, and their datasheet says it's nominally 10milliamps, but can be between 2 & 20 which is a big range. And it matters in the U87 because it doesn't have a lot of current to play with. The ones they used in the late '60s and '70s were apparently at the lower end of the IDSS scale, around 4 or 5 mA. I sourced 9, and the lowest I found was 8.35mA going up to 13mA. Damn. Then someone I know randomly posted that they'd found a whole treasure trove of different transistors, all sorted and labelled.
"Hey don't suppose there are any old 2n3819s in there...?"
He sent me ALL the 2n3819s he had, told me to pick my pair, and send the rest back. So tonight in went two lovely old Texas Instruments FETs, measuring 3.75 & 4.58mA. Yum!
Then lots of scrubbing with 99% isopropyl alcohol, because the back of the boards can't be too clean. Those 1G and 68m resistors present a problem - if you're asking the electrons to go through a really high value resistor, you can't also give them a path through finger oil and solder residue that will form a more attractive path for current. It's why the whole project's been done wearing gloves, and it's why I've used a toothbrush on the back of the PCBs.
Now that the front PCBs are screwed on these are beginning to feel like they'll be mics soon. Another good session of connecting the two boards and wiring up the XLR socket, and I'll be ready for the next leap into the unknown: Biasing the FETs with a test tone & oscilloscope.
SillyOctpuss
Shredder
- Messages
- 1,913
That's exactly what I was going to post before seeing your reply. It's the most interesting diy thread I've seen in ages.
Cirrus
Roadie
- Messages
- 312
Both boards wired together and XLR lead connected. So now, if I plugged it in and turned on +48v... it should BECOME ALIVE.
But it wouldn't produce any signal, because neither the capsule or the transformer's not wired in yet. Why? Because the FET needs to be biased *before* the capsule + transformer gets wired up. Why? I simply don't know. I might have misunderstood something. I'm going to go get a glass of Whisky and start Googling old forum posts to better understand why.
Next step, then, is to make leads that go jack->2x crocodile clips, so I can inject signal into the FET input and pick it up on the other side.
But it wouldn't produce any signal, because neither the capsule or the transformer's not wired in yet. Why? Because the FET needs to be biased *before* the capsule + transformer gets wired up. Why? I simply don't know. I might have misunderstood something. I'm going to go get a glass of Whisky and start Googling old forum posts to better understand why.
Next step, then, is to make leads that go jack->2x crocodile clips, so I can inject signal into the FET input and pick it up on the other side.
Cirrus
Roadie
- Messages
- 312
Aaaaaand they're biased!
It seems my Audient iD22 thinks 48v phantom power should be 45 volts, but 3v under isn't going to do much to the bias point. The process: Plug an XLR lead in, and turn on phantom power. This bit's always a proper IT'S ALIVE moment for me - that first application of voltage to the circuit, wondering if anything's going to explode.
Nothing exploded.
Then, I got the jack to crocodile clip leads. One is plugged into an audio interface out, it's going to carry the 1kHz test signal from my software oscilloscope - I'm using "Soundcard Oscilloscope" and it did the job nicely. Free, too. There's a particular part of the circuit to inject the signal into the FET input. The second lead then connects directly to the FET output leg and goes back into an input of my audio interface to return the signal to the oscilloscope.
There's a little trim resistor, a blue box, that sets the bias voltage for the FET, and the process is simple: start with a low bias voltage. Apply cautious low level signal. Scope and listen to the result. It's a pure sine wave, so it should sound like one. Increase the test signal volume until it starts distorting. On the Oscilloscope that will show up as either the top or the bottom of the sine wave flattening out, and actually I found it really easy to hear the distortion harmonics start to creep in. Then turn the trimmer, and the flat tops will round out again. Increase test signal level again, watch and listen for distortion, trim bias and repeat.
Eventually, I got to the best bias point; increasing the test signal made the top and bottom of the waveform on the screen flatten out at the same time. At this point I found finetuning by ear MUCH easier than watching the scope; essentially when the test signal eventually is increased enough that the FET is running at the very top of its clean headroom, you end up in a very narrow zone where turning the trim pot either way even a small amount brings on audible distortion. At that point, when there was only maybe a 1/8 turn between the distortion onset on either the top or bottom of the wave, you couldn't see the waveform change shape on the screen but you could hear the 2nd harmonic start to creep in. I just set it equidistant between those two points and called it done.
And the result? The two mics biased at 9.3v & 10.45v. The 9.3v one has like 1% more THD at the same high signal level, but in practice the slight headroom difference isn't going to matter. 10.5v is the "ideal" Neumann spec, but in practice it varies depending on the individual FET and when Neumann send out replacement FETs for servicing U87s, they test them first and also send a hand chosen resistor to bias it up right. I'm happy that my numbers are in the ballpark.
The next stage is the scary one - tomorrow I'll wire up the capsules and transformers, and see if they actually sound good.
It seems my Audient iD22 thinks 48v phantom power should be 45 volts, but 3v under isn't going to do much to the bias point. The process: Plug an XLR lead in, and turn on phantom power. This bit's always a proper IT'S ALIVE moment for me - that first application of voltage to the circuit, wondering if anything's going to explode.
Nothing exploded.
Then, I got the jack to crocodile clip leads. One is plugged into an audio interface out, it's going to carry the 1kHz test signal from my software oscilloscope - I'm using "Soundcard Oscilloscope" and it did the job nicely. Free, too. There's a particular part of the circuit to inject the signal into the FET input. The second lead then connects directly to the FET output leg and goes back into an input of my audio interface to return the signal to the oscilloscope.
There's a little trim resistor, a blue box, that sets the bias voltage for the FET, and the process is simple: start with a low bias voltage. Apply cautious low level signal. Scope and listen to the result. It's a pure sine wave, so it should sound like one. Increase the test signal volume until it starts distorting. On the Oscilloscope that will show up as either the top or the bottom of the sine wave flattening out, and actually I found it really easy to hear the distortion harmonics start to creep in. Then turn the trimmer, and the flat tops will round out again. Increase test signal level again, watch and listen for distortion, trim bias and repeat.
Eventually, I got to the best bias point; increasing the test signal made the top and bottom of the waveform on the screen flatten out at the same time. At this point I found finetuning by ear MUCH easier than watching the scope; essentially when the test signal eventually is increased enough that the FET is running at the very top of its clean headroom, you end up in a very narrow zone where turning the trim pot either way even a small amount brings on audible distortion. At that point, when there was only maybe a 1/8 turn between the distortion onset on either the top or bottom of the wave, you couldn't see the waveform change shape on the screen but you could hear the 2nd harmonic start to creep in. I just set it equidistant between those two points and called it done.
And the result? The two mics biased at 9.3v & 10.45v. The 9.3v one has like 1% more THD at the same high signal level, but in practice the slight headroom difference isn't going to matter. 10.5v is the "ideal" Neumann spec, but in practice it varies depending on the individual FET and when Neumann send out replacement FETs for servicing U87s, they test them first and also send a hand chosen resistor to bias it up right. I'm happy that my numbers are in the ballpark.
The next stage is the scary one - tomorrow I'll wire up the capsules and transformers, and see if they actually sound good.
fretworn
Rock Star
- Messages
- 4,158
Will take them to practice tonight, probably put them up as drum overheads, maybe also guitar close mic if there's time for screwing around
SillyOctpuss
Shredder
- Messages
- 1,913
Will take them to practice tonight, probably put them up as drum overheads, maybe also guitar close mic if there's time for screwing around
You definitely need to test them in the field. Independently obviously. You're too close to this project to be objective @Cirrus. You better send them to me and I'll use them to record a variety of sources over the next six months to a year. The only way to be sure is a thorough test.
No no don't thank me @Cirrus im just trying to help
Cirrus
Roadie
- Messages
- 312
Dropbox
Acoustic guitar clips if anyone's interested, the mics are always 35-40cm in front of the neck-body join. Different performances so it's not very scientific, but the EQ/ transient detail is so obviously different between the U87 clones, the AT4050s and the SM7b that I don't think it matters. I'm happy with these mics.
Sascha Franck
Rock Star
- Messages
- 6,079
Great stuff. And soooo out of reach for me. Enjoyed reading and listening, kickass job!
Cirrus
Roadie
- Messages
- 312
Did a bit of a vocal mic shootout with my singer tonight - U87, AT4033, SM58 & an Electrovoice PL80 (basically a nice old dynamic). Like an idiot I left my SM7B at home, but I used that on him for some demos last year and thought it was fine if a little dark for his voice. I'll try to find time to put some clips up tomorrow.
Also tried one of the '87s on close guitar cab duty and wasn't really sold on it. Liked it as a room mic, but I've got a Sennheiser mic someone put an MD409 capsule in which just always wins on close guitar at the moment.
Also tried one of the '87s on close guitar cab duty and wasn't really sold on it. Liked it as a room mic, but I've got a Sennheiser mic someone put an MD409 capsule in which just always wins on close guitar at the moment.
