Noise issues in a mic build in omni (Aurycle FET) Solved, thanks!

Help Support GroupDIY:

midwayfair

Well-known member
Joined
Apr 7, 2015
Messages
526
Location
Baltimore
I made some changes to the Aurycle FET mic circuit during a recent build:
http://aurycle.com/manual/a460-diy-schematic.pdf

I wanted to make a PCB with my changes (and some other mods) pre-made on it, but I'd like to understand what's happening before I do.

I made most of these one at a time, and reduced the noise a little bit each time. Some of them I didn't expect to make a difference, but they seem to have. Others may have had no effect.

My testing method isn't that great, but it does work a little: I set the gain to a certain level on my interface (45 dB for this test), record a clip, run an analyzer, and stack the comparison clips one behind the other on a loop. So I can tell if the noise difference is a decibel or two but I can't make fine measurements, and I don't replace the capsule with a capacitor. I also can't measure the raw self-noise figure because my analyzer doesn't go below -100 and because the pre adds its own noise.

Stuff in bold is things I KNOW reduced the noise.

0. Changed the 2SK170 for a 2SK30. The 2SK30 is lower gain. I was also under the impression that it's lower capacitance than the 170, though. This one might be a wash or might have been a mistake. I have no way of testing that will produce identical gains between the two transistors and I mostly made this change because I didn't really trust the transistor that came with the kit. Should either transistor be expected to contribute less noise?

1. Removed R1, R2, and C1. I came across a post (I think it was here) from a while ago that explained that these add noise and serve no other purpose (they were for something in the U87's testing circuit). This was a huge reduction in noise. Why does a connection to ground on a part of the capsule that is only connected to the polarization circuit (and not the FET amplifier) add noise, though?

2. Changed out C4 with a 100uF low-ESR electrolytic ... and ALSO bypassed R6 with a 100uF low-ESR. I got less noise this way than (a) bypassing the source completely with only one capacitor or (b) bypassing only one resistor. And by this I mean "less noise in the abstract" not just "better signal:noise." My guess is that the single capacitor can't fully filter this node. I'm also going to guess that a 2SK170 would have been able to make more use of the gain, possibly with ill effects on the headroom.

I also replaced R7 with a trimpot; I'm guessing that the trimpot wasn't great for noise performance, but when I pulled it out and replaced it with a 1K temporarily before adding the second capacitor, I was unable to measure a noise difference.

Another thought: Since the source is now fully bypassed, I'm not sure if there's much of a reason to bias the FET via the gate instead of just putting the gate to ground and trimming the source resistance. Gain is at max either way. Current consumption would go up but probably not by much. Would referencing the gate to ground be likely to have any effect on the noise performance compared to referencing it to a slightly positive voltage that's completely filtered? Not really something I know how to answer.

3. Removing the NFB cap (C5, the 470nF) had the opposite effect from what I expected: I got far more white noise when it was removed but didn't get a corresponding increase in gain.

4. I flipped the capsule connections and I'm taking the output from the backplate. I did this to add patterns (see the other thread). I can't remove C2 or R4 without a negative polarization voltage, but I've seen conflicting information about the leakage of certain capacitor types. I used a box cap for C2. Some mics use polystyrene there. Is one or the other a better choice if I must use an input cap?

5. The Aurycle has a pad created by connecting 150pF across the capsule. I didn't have a 150pF film cap, so I used silver mica for the pad. Is this a poor choice? I figured I'm looking at hundreds of megs of leakage resistance in a silver mica. I'm not sure if I should be worried about that dividing with the 1M or with the 1G. If it's dividing with the 1M, then I would say it's fine because the noise contribution from the leakage can't be a big deal. If it's with the 1G, though, then any noise contribution would matter. I'm okay with opening the mic back up and replacing it with a different cap if needed; the closest I have is a 220pF Panasonic ECQ-V.

6. I increased C12 (4u7), C13 (4u7), and C14 (10uF) and used low-ESR alumimum. I didn't have tantalum caps to try just increasing the capacitance first, so I changed both the cap type and value at the same time. I'm guessing this was more the type of capacitor than the capacitance, though, because 150K > 2u2 is still below 1Hz, and 4k7 > 10uF isn't that much different than 4k7 > 6u8.  The C13 & C14 change seemed to have a bigger effect than the C12 change. Any thoughts here? Would going bigger with the filter caps help things, or is there a detrimental effect to increasing their value?

7. Why does this use a PNP follower instead of NPN? I thought "upside down" transistor arrangements (meaning, PNP running off a + supply) added noise, but is that only when flipping them in voltage amplification?

This circuit almost identical to the Mxl V67G, 2001, and several other mics, even some of the numbering is the same, so knowledge of those mics applies here.

Complete modified schematic -- all the numbering is the same.

FET%20mic%20EF%20plus%20xfo.png


https://dl.dropboxusercontent.com/u/9878279/Jon%20Patton%27s%20layouts/Circuit%20ideas/FET%20mic%20EF%20plus%20xfo.png
 

midwayfair

Well-known member
Joined
Apr 7, 2015
Messages
526
Location
Baltimore
Bumping this hoping to get some ideas and understanding.

In particular, the omni mode is fairly noisy (white noise) -- Figure 8 mode has better signal:noise despite its (much lower) sensitivity, and the Cardiod is still extremely quiet. The omni isn't unusable, but not really for anything except loud sources and certainly not for vocals. I do not have a convenient way to disconnect the switching so I can test if simply shorting the membranes together produces a quiet omni.

The only thing I can really see that's different between the front membrane and the back membrane in omni is that there's a 220pF across the 1M on the front membrane. I had thought that was purely there as part of the negative feedback circuit (the 470nF etc.). Should I have included that on the back membrane's positive bias resistor? Should I have used identical values on both capsules' polarizing resistors, or does only the polarization voltage matter?

I'd really like to solve this problem so I can built the other half of the pair. Thanks for your help.
 

midwayfair

Well-known member
Joined
Apr 7, 2015
Messages
526
Location
Baltimore
Khron said:

Hrm. I was going to type that it might be a different issue, but ...

Removing the 470nF cap results in a whole mess of extra noise. When my omni mode is connected, there's no capacitive NFB from the R3/C3/C5/R17 junction like there is for the front capsule.

So this means that maybe I do need a capacitor there. Unfortunately, running it through the calculator puts the required capacitance to form the same low pass (1M and 220pF) as what's on the front capsule (723Hz) when using a 330K is all the way up at 660pF, which is not a value I have lying around and will form a radically different cutoff when in series with the 470nF. So that means pulling out the 330K and replacing it with the same 1M and 220p, and then re-doing the lower half of the divider for F8 to 1.3M (that is, 1M plus the two 150K so that the divider is exactly half).

Confirmations that this would work and is necessary, or any other ideas?

Just realized that the file for my version of the schematic isn't working. I'll re-upload it when I get a chance.
 

Khron

Well-known member
Joined
Apr 8, 2010
Messages
1,931
Location
Finland
I don't mean to sound like a broken record, but all this effort put into struggling to coax the stock circuitry to do what you want it to, could've / would've gotten you the results you wanted with the U87A circuitry (ie. including a DC-DC converter) :) But that's just my opinion...
 

midwayfair

Well-known member
Joined
Apr 7, 2015
Messages
526
Location
Baltimore
Khron said:
So you've never DIY-etched your own PCB's? :D

I don't mess with chemicals. I do perfboard (not vero) or PCBs. I could probably perfboard it, the power supply side at least, but I've had problems in the past with the perfboard material contributing (apparently) to noise even when PTPing the high-Z stuff.
 

Khron

Well-known member
Joined
Apr 8, 2010
Messages
1,931
Location
Finland
It's nothing truly hazardous, really :D Unless you're... "uninspired" enough to actually drink the stuff ;D Other than that, they can worst case be a bit messy (if you've got Parkinson's or something) :)
 

midwayfair

Well-known member
Joined
Apr 7, 2015
Messages
526
Location
Baltimore
Knowing full well my quest might be quixotic ...

First, I retitled the thread since my noise issues appears to be in omni mode only. I also fixed the schematic file so it should be visible now.

I tried replacing the 330K/620K resistors with 1M/1.3M (well, I used 1.2M). Figure 8 still worked fine, and omni was still more noisy (several extra raw dB of hiss). I next tried adding the 220pF across the new 1M, and while this did kill some noise, it also killed a ton of signal along with it! That makes sense now that I think about it, because it essentially doubled the amount of negative feedback. Turning the gain up on the preamp and the noise was right there still.

I temporarily removed the Figure8 even though that mode didn't exhibit a problem.

I confirmed that shorting the front and back capsules produces a quieter omni mode (hiss ~ identical to cardiod) with higher output than connecting the back diaphragm to the + voltage on its own. Don't really understand why that should be, though ...

Last thing I could think of to try was to use the front membrane as the + voltage for the back membrane and for figure 8. While this dropped the white noise a little in omni mode, I'm sure you can see that the problem there is that now Figure 8 drops the front capsule voltage!That's no good at all.

So I went back to the first change (same as the schem but with the 1M/1.2M). I did notice that with the larger resistors the back membrane is a little fatter sounding while the front has a little more treble. That's interesting at least.

Oh well. I'm still baffled and a little disappointed, especially since the Fig. 8 was a decent success. Back to playing and writing music while I save up for the next project ...
 

Matt Nolan

Well-known member
Joined
Apr 29, 2015
Messages
399
Location
Bath, England, UK
I think you need a cap to ground across that R21 100M. Low impedance for AC to ground is needed there else your omni signal is going through a lot of resistance.
 

Khron

Well-known member
Joined
Apr 8, 2010
Messages
1,931
Location
Finland
In case you haven't yet (entirely) given up on the mod idea, i believe the schematic from the CAD GXL3000 might give you some ideas.

As far as i can tell, despite having a DC-DC converter, it only has a single (positive) output voltage, and yet it has cardioid / omni / figure-8 patterns.

http://recordinghacks.com/images/mic_extras/cad/gxl3000-schematic.png

I don't really see why you couldn't pull off a similar sort of trick, only using the phantom-power-derived voltage :)
 

midwayfair

Well-known member
Joined
Apr 7, 2015
Messages
526
Location
Baltimore
Matt Nolan said:
I think you need a cap to ground across that R21 100M. Low impedance for AC to ground is needed there else your omni signal is going through a lot of resistance.

This did it! Thank you. I used a EDIT: 220uF. It's very quiet now. (The edit was that the 10uF I used at first took several minutes to discharge when going out of omni mode. 220 discharges in about 30 seconds.)

Interestingly, now that the noise is gone, the back now sounds slightly crisper in omni, but that might just be poor matching on the capsule's membranes. Either that or perhaps a filter cap might be useful on the front membrane as well.

Khron said:
In case you haven't yet (entirely) given up on the mod idea, i believe the schematic from the CAD GXL3000 might give you some ideas.

As far as i can tell, despite having a DC-DC converter, it only has a single (positive) output voltage, and yet it has cardioid / omni / figure-8 patterns.

http://recordinghacks.com/images/mic_extras/cad/gxl3000-schematic.png

I don't really see why you couldn't pull off a similar sort of trick, only using the phantom-power-derived voltage :)

I considered something similar because the switching would have been really simple, but it polarizes the capsule to 1/2 B+ and then the patterns use a high voltage or a half voltage as the reference, losing the ability to have the full available polarization voltage for at least one of the patterns. On phantom power, it requires using 24V (or realistically, closer to 15-20V) as the main polarization voltage and then using a higher voltage to get figure 8 and the same half voltage for omni. When you're bumping the voltage up to 60V internally, the sacrifice in sensitivity is smaller compared to cardiod on 48V. Though to be fair that's not a big voltage difference, but I just didn't want to make that particular sacrifice.

Still, that schematic is really useful, so thank you for passing it along.
 
Top