OPEN SOURCE DIY Mic Project - ORS 87 - Stripped Down u87

[Khron]

Think I've got KSA board properly populated so far. Can the jfet be soldered in and clipped?

C9 should actually be removed - i've yet to real-world-test anything myself, but you can see what happens as-is. If no audio comes out, remove C9 and re-test.

You'll want to install the 1G resistor as well, not just the JFET

Regarding transformer connections - P+ & P- are the transformer *P*rimary connection points - see what i did there?

If you have a single secondary winding, that goes between S1+ and S2- (by the same logic as the primary).
If you have a center-tapped secondary, center tap goes to S1-/S2+ (they're connected together), and you can skip the two 2k2 resistors. If you have two separate secondaries, same story.

 

[AdamFrisen]

The C3, 10pF in U87 original schematics (same as the C1 in U87 ORS original) has an impact if omitted as it has been in this project. I am not worried about -14 dB gain reduction. That I can compensate by increasing mic preamp gain.

But the frequency response is being affected too as you can see in the graphs with no C1, with 10pF and with 3.3pF. Also of course the phase.
The shape of the phase curve is different from when not using any C1 and when C1 is 10pF and 3.3pF.

If one wants to get close sonically to U87 original, I assume you should add that C1 capacitor.

I learned how to use QSPICE a little bit. Thanks to this project and also the gruesome (or proud?) history of SPICE.

 

[Khron]

But the frequency response is being affected too as you can see in the graphs with no C1, with 10pF and with 3.3pF.

Any chance you could scale the Y-axes so they're the same?

Or just copy the same circuit two more times, with the needed alterations of C1 value, and plot all three responses in one graph? Just for some perspective...

 

[AdamFrisen]

Any chance you could scale the Y-axes so they're the same?

Or just copy the same circuit two more times, with the needed alterations of C1 value, and plot all three responses in one graph? Just for some perspective...

I'll try to learn how to do it tomorrow .

QSPICE seems to be beginner-friendly (I think).

I think I would like this mod for using Arienne K47 flat for vocal. Remove C6 (220pF) and C5 to 15nF.

Keeping C9 probably saves some characteristics of U87 and also keep the HF flat.



Then I got this in simulation:

 

[mrchapp]

I'm happy to report my mic works too!

TLDR is at the bottom.

I made a lot of mistakes along the way.

As @Wordsushi said, there were different ways to approach this build, and my choice was to go the cheapest way. So I gathered all components from Aliexpress, with zero certainty on them being the right choice, and hoped for the best.

Initially I made a PCB design because designs at the time were for a different body (HL-95, while I had a BM-800 lying around) and it was cheap. Because I wanted to test that design first, there was a 2.54 mm grid on the PCB, just like a perfboard. After (badly) routing everything, printing the PCB on paper and trying it on top of a perfboard, I realized my first mistake here: routing was done on TWO layers!, which only complicated matters. I tossed the perfboard idea and went straight to PCB manufacturing. So much for the 2.54 mm grid.

Once the PCB arrived, all components were soldered as best as I could (IOW, not great). Because my donor was a BM-800, space was really, really tight (36x63 mm). With the height of the WIMA capacitor and the T-13 transformer, it was impossible to make it fit on the upper side of the BM-800 frame, thus it had to be installed on the underside of it. This presented a problem: the frame itself needed some space on the inside of the PCB, where some capacitors were placed. Lessons learned: center the WIMA capacitor, add a lot more margin to the PCB, and drop the 2.54 mm grid altogether.

I must have mishandled the capsule a lot because I later discovered some fingerprints on it. Much later, while desoldering, I must have pushed the diaphragm accidentally against the backplate as there were some marks (dents) in it. Oh, well. At least it wasn't Arienne's.

First time I tried the build there was absolutely no sound at all. I chalked it up on my poor choice of components (like the 1G resistor, I thought) the handling of the capsule, and the probably burnt transistor as I soldered it. I let that sit for a couple of weeks.

When you guys talked again about the BOM I looked again at my build. With the oscilloscope and REW I found that a 1 KHz sine wave was *very* distorted on the output. No matter what I had on the trimpot, it was really distorted... except sometimes it wouldn't work at all.

I was really excited at that point because it showed signs of life. After fiddling around, the problem became apparent: there was a bad connection in the C11 capacitor, so pushing it a little made it work!, at least with REW signal generator. When somebody mentioned that the JFET _could_ work either way (the Chinese transistor I got had no spec sheet), I tried that with a new transistor, to no avail. Left it again.

I fixed that and then tested with an audio interface. The output was BAD, like B.A.D. -- BAD, really noisy. The baseline for noise was -28.5 dB according to my DAW. Again, I thought about the poor choice of components. Left it again for some days.

Yesterday I decided to bring the transistor back to its place and make new measurements with REW, and decided to solder it with the gate leg lifted (no PCB), straight to the 1G resistor and the capsule connection. It worked!! And it worked amazingly! I am beyond impressed.

TL;DR: Lots of mistakes, but this is a very forgiving build, even with low quality Chinese components.

Here's a list of components (with links to Aliexpress and approximate costs). In total, for just one build (and hundreds of spare components) it's something like US$131.71, more or less.

Thank you, all, who shared your knowledge, your pictures, your ideas, and your suggestions. You guys rock!

 

[AdamFrisen]

I found how to overlay the 3 graphs with no C1, C1=3.3pF and C1 =10pF.

 

[Khron]

With the height of the WIMA capacitor and the T-13 transformer, it was impossible to make it fit on the upper side of the BM-800 frame

Why the HUMONGOUS capacitors, though? Ok, "bigger is better" in some instances, but I'll never understand the obsession with 400V capacitors in phantom-powered mics...

PS. Flipping the transformer "connector" by 180deg would net you much simpler routing (for the next revision) And you might want to consider connecting the mounting holes to the ground plane(s).

But was the ORS87-Plus not good enough(?) for fitting into a BM800?

 

[Khron]

Keeping C9 probably saves some characteristics of U87 and also keep the HF flat.

Keeping C9 makes C15 do its low-cut job. The HF is flat by default, once you removed the C6.

Designators as per your Qspice schematic, of course.

 

[AdamFrisen]

Keeping C9 makes C15 do its low-cut job. The HF is flat by default, once you removed the C6.

Designators as per your Qspice schematic, of course.

Without C9, the

Keeping C9 makes C15 do its low-cut job. The HF is flat by default, once you removed the C6.

Designators as per your Qspice schematic, of course.

Yes, I am just nitpicking and I cannot hear much above 13-15Khz anyway. Well I would hear it as perceived increase in sound level if there is enough frequency energy at higher range than I can "hear" (if I remember the transfer function equation correctly).

I wanted to keep the LF anyway when I use K47 flat capsule. I find the flat circuit without C9 picks up too much base in my speech without it.

Without C9, the gain seems to increase even more than when not having C1 at all (from about 14dB to 15.8dB).
I want to listen with the lower gain as well following the original U87.

I could have measured totally wrong too. It is the first time I am using any simulation tool.
But I am having fun.

I received the K87 capsule from 797 Audio. It was very well packed. Two thick layers of internal soft padding (each layer taped individually) and on top of that 2 padded envelopes. Compared to how the other capsules were delivered, 797 Audio wins in packaging.

Now, if I can just get the response of the 797 Audio capsule to get -2dB - 2.5dB starting from 3KHz to 20KHz, then I should be good to go for comparing.

When I tried the flat circuit with 3.3pF at the gate and drain, I measured the THD goes down to 10% compared to without it. For me that is another reason to add the C1.

 

[Khron]

I could have measured totally wrong too. It is the first time I am using any simulation tool.

If anything, you might want to see how the current way compares to applying the test signal across the 560 ohm resistor, "as it was intended" (instead of "emulating" the capsule).

 

[AdamFrisen]

Using the measuring input (messeingang), the shapes of the curves looks similar as in emulating. I see little difference in gain and phase.



Schematics below was used for the simulation without C1:

 

[anvl]

Dear all. Are you sure we are inserting this 10pF cap in the correct way? Regarding this original schematic this cap should be connected to the rear diafragm. Am I wrong?

 

[Khron]

View attachment 134721
Dear all. Are you sure we are inserting this 10pF cap in the correct way? Regarding this original schematic this cap should be connected to the rear diafragm. Am I wrong?

That would be true, if we were using the exact copy of the entire circuit, multipattern & isolated-backplate and all. That's not the case, since this project is the cardioid-only "reduction" of this circuit.

As both tests and simulations have shown, that capacitor does indeed work the same way, regardless of the exact connection nodes not being completely identical.

https://audioimprov.com/AudioImprov/Mics/Entries/2015/12/20_Modding_a_BM-800_Mic.html

 

[anvl]

Sorry, I don't understand why this 10pF is connected to the JFET gate in our cardiod-only situation.

 

[Khron]

Sorry, I don't understand why this 10pF is connected to the JFET gate in our cardiod-only situation.

You're concerned that it doesn't function in the same way as it does in the vintage U87i, or...?

 

[anvl]

In my opinion, this cap connected to the JFET gate probably helps for a better linear response as a negative feedback but it means a noticeable modification of original vintage schematic, and so for the U87 sound/character that we were trying to follow. It depends on what are you looking for.

 

[Khron]

but it means a noticeable modification of original vintage schematic, and so for the U87 sound/character that we were trying to follow.

And you're basing that claim on... what, exactly?

 

[anvl]

No claims. I only wanted to understand

 

[pasarski]

In my opinion, this cap connected to the JFET gate probably helps for a better linear response as a negative feedback but it means a noticeable modification of original vintage schematic, and so for the U87 sound/character that we were trying to follow. It depends on what are you looking for.

It's maybe not obvious in a first glance that the 10p cap is connected from drain to gate in U87's cardiod mode through the 470p (safety?) cap.

Edit: not a safety cap, there's already the 470p C4 for that...

 

[cello4u]

C9 should actually be removed - i've yet to real-world-test anything myself, but you can see what happens as-is. If no audio comes out, remove C9 and re-test.

You'll want to install the 1G resistor as well, not just the JFET

Regarding transformer connections - P+ & P- are the transformer *P*rimary connection points - see what i did there?

If you have a single secondary winding, that goes between S1+ and S2- (by the same logic as the primary).
If you have a center-tapped secondary, center tap goes to S1-/S2+ (they're connected together), and you can skip the two 2k2 resistors. If you have two separate secondaries, same story.

Don't see what you did there. Went with Primary +/- to S1+/- and Secondary to S2?

and finally? FD to plate and P1+ to the center capsule?