Quick question about a microphone pre-amp

[Howard]

So the resistance the opamp output sees should be 1/Rt = 1/(Rf+Ri) + 1/Ro, where Rf+Ri = ~1k, Ro = 10k (line-in). I get about 2.2mA at a 2V output. I think I should be safe, since the opamp can do much more than that.

I'm just going to ignore the N+R since I'm only using a WM-61a.

 

[Howard]

What reduces noise more? A capacitor between + and -, or a capacitor between both + and ground and - and ground? The latter has the caps in series (I think), so let's say the latter caps are twice as big as the single cap between + and -.

 

[mindless1]

Capacitors between both +/Gnd and -/Gnd.

 

[Howard]

I was hoping that wasn't the case. 🙁 Instead of upgrading the bulk cap, would it make more sense to increase the size of the local decoupling caps? I've read that 0.1uF works best around 3MHz, so should I increase the size to drop it closer to the audio band?

 

[mindless1]

You may not hear any difference, I wouldn't worry about it much.

 

[Howard]

Well, the ripple from my unregulated supply is pretty much nonexistent after the regulator and the PSRR of the opamp, but the noise from the regulator is still relatively high. The noise is 30uV/V (10Hz to 10kHz but I'll extrapolate to 20kHz), so 360uV from the regulator alone, and if I can add the 30uV from the noise-reduced TLE, that's 390uV altogther. At 20kHz, the PSRR is only 50dB, which means that the noise is only ~-80dB below a 1V input

Actually, just checked and it's -118dB below a 1V input... Still, the input will be much lower than that.

EDIT: Looks like the gain resistor noise is the most important, if the decoupling caps reduce supply noise significantly.

 

[mindless1]

The regulator might be a little quieter if you put a 10uF tantalum or 0.1 to 1uF ceramic cap between the middle of the regulator voltage divider resistors and ground (relative to the regulator so really the negative amp rail). With the tantalum it would probably be a good idea to add the protection diode as shown in LM317 datasheet.

 

[Howard]

Unfortunately I don't have enough holes to do that on my current board. I'll see if I can add one in a future build. It seems as though the impedance of good lytic caps is much lower than that of equivalent-size tantalum caps, though.

The TLE chip has ripple rejection on its output, too, with the noise-reducing capacitor installed. The graph is on page 16 of the datasheet. Does this mean that it also contributes to overall noise rejection or just on the negative rail?

 

[mindless1]

Considering the amount of work you're doing, I would go with a professionally made PCB next with a rebuild, and that brings me back to something I forgot to mention. When considering noise are you testing outside of a shielded case?

Impedance at which frequency, 'lytics often seem rated at only 100KHz, and does size mean physical or capacitance? You won't need a lot of capacitance for the regulator, and I'd be surprised if a electrolytic capacitor the same size as a 10uF or tantalum or 1uF ceramic is as low an impedance, though I must admit never trying to find a tiny low impedance electrolytic capacitor.

I don't know specifics about the TLE's noise-reducing capacitor, and unless I missed it TI's datasheet doesn't give enough info to know.

 

[Howard]

Well, I'm trying not to spend that much money.

I think it's just a normal 1uF ceramic. A monolithic should be OK.

EDIT: Damn it, why the heck do fixed regulators have better noise characteristics than adjustable ones?

 

[mindless1]

What? Adjustable have better specs (unless I overlooked something), maybe it's just the omission of the capacitor in the feedback loop you aren't seeing the effect of?

For what this is, overall I think it's close to time to stick a fork in it and call it done.

 

[Howard]

Originally posted by: mindless1
What? Adjustable have better specs (unless I overlooked something), maybe it's just the omission of the capacitor in the feedback loop you aren't seeing the effect of?

For what this is, overall I think it's close to time to stick a fork in it and call it done.

Haha yeah, I'm pretty much done myself. But the noise figures do seem to be superior, what with both types of regulators tested with caps on both ends.

http://www.onsemi.com/pub_link...lateral/MC78L00A-D.PDF

 

[bobsmith1492]

I wouldn't worry about noise specs on voltage regulators. Modern regulators all have extremely low noise (i.e. don't use an LM317 because they're ancient...). Plus, if you're simply powering op-amps, the op-amp will have tens of dBs of power supply rejection.

If you're worried about a -118dB noise floor, you're slightly crazy. 😉 You'll never see that little noise which means the regulator's noise is inconsequential.

 

[Howard]

Originally posted by: bobsmith1492
I wouldn't worry about noise specs on voltage regulators. Modern regulators all have extremely low noise (i.e. don't use an LM317 because they're ancient...). Plus, if you're simply powering op-amps, the op-amp will have tens of dBs of power supply rejection.

If you're worried about a -118dB noise floor, you're slightly crazy. 😉 You'll never see that little noise which means the regulator's noise is inconsequential.

I ran the preamp (in a case) without a mic hooked up and I could definitely hear some noise after turning it up. Not sure if it's the 84.77uV from the feedback resistors or whatever noise from my own sound card or some crap picked up by the dollar-store interconnect, but it's there.

What performs better than an LM317?

 

[blahblah99]

Not to burst your bubble or anything but if you want extremely high performance out of a microphone pre-amp you will need to do circuit analysis and CAREFUL pcb layout. I once designed a mic pre-amp with 48V phantom power with an EIN of -125dBu and about -116dB of dynamic range. It took a lot of number crunching, pcb layout tweaking, and circuit modifications.

Select an opamp that has a low nV/root(hz) noise spec, enough gain bandwidth for the gain you plan to use, good CMRR on the supply and inputs, and never pick a FET-input opamp for audio circuits.

The best theoretical noise performance you can get in any mic circuit is approximately equal to the nV/root(hz) (taken over the 20-20k range) plus resistor noise. This is referred to as Equivalent Input Noise. Take that and multiply by the gain of the circuit to get your output noise. Any measured noise that is above the calculated noise can be attributed to layout issues or noisy power supplies.

Then there's frequency response and THD. 🙂

For frequency response I usually pick circuit values that gives as flat of a response as possible from 20-22k without compromising oscillations.

THD is easy - don't drive the crap out of the circuit, and try to stay away from caps that have high voltage coefficients.

 

[bobsmith1492]

Originally posted by: Howard

Originally posted by: bobsmith1492
I wouldn't worry about noise specs on voltage regulators. Modern regulators all have extremely low noise (i.e. don't use an LM317 because they're ancient...). Plus, if you're simply powering op-amps, the op-amp will have tens of dBs of power supply rejection.

If you're worried about a -118dB noise floor, you're slightly crazy. 😉 You'll never see that little noise which means the regulator's noise is inconsequential.

I ran the preamp (in a case) without a mic hooked up and I could definitely hear some noise after turning it up. Not sure if it's the 84.77uV from the feedback resistors or whatever noise from my own sound card or some crap picked up by the dollar-store interconnect, but it's there.

What performs better than an LM317?

Can you draw a schematic of your circuit? I looked through the thread but couldn't find one. You should use some simple low-pass filtering to keep everything stable and happy. If you don't bandwidth limit the input to your op-amp it may oscillate.

For lower-noise regulators, you just have to shop around. So the LM317 has 30uV/V noise... this one has 10uV/V; that should help. It's also rated 10Hz-100KHz and doesn't require feedback resistors.

For capacitors, you'd probably want ceramic for decoupling regulators and your rail-splitter... for signal coupling you'll want film caps (as I think you're doing?) Ceramic caps are ok but you have to keep your signal well under the capacitor's voltage rating. They become non-linear in impedance beyond about +/-25-50% of their rating. Electrolytics are bad unless you know the polarity of your input or use a bipolar cap.

Circuit construction is by far the biggest contributor to noise, though. I could come up with some tips if you still have noise problems.

Originally posted by: blahblah99
Not to burst your bubble or anything but if you want extremely high performance out of a microphone pre-amp you will need to do circuit analysis and CAREFUL pcb layout. I once designed a mic pre-amp with 48V phantom power with an EIN of -125dBu and about -116dB of dynamic range. It took a lot of number crunching, pcb layout tweaking, and circuit modifications.

Yikes, -116dB is not much range. 😉 I agree, though, cheap parts on a perfboard without careful layout consideration will be rough:

- Star ground reference point
- High-quality decoupling caps
- Careful decoupling placement (close to components)
- Careful power and signal routing
- Good shielding

 

[Howard]

Originally posted by: bobsmith1492

Originally posted by: Howard

Originally posted by: bobsmith1492
I wouldn't worry about noise specs on voltage regulators. Modern regulators all have extremely low noise (i.e. don't use an LM317 because they're ancient...). Plus, if you're simply powering op-amps, the op-amp will have tens of dBs of power supply rejection.

If you're worried about a -118dB noise floor, you're slightly crazy. 😉 You'll never see that little noise which means the regulator's noise is inconsequential.

I ran the preamp (in a case) without a mic hooked up and I could definitely hear some noise after turning it up. Not sure if it's the 84.77uV from the feedback resistors or whatever noise from my own sound card or some crap picked up by the dollar-store interconnect, but it's there.

What performs better than an LM317?

Can you draw a schematic of your circuit?

It's pretty much the same as Linkwitz's schematic with the addition of an LM317L and a TLE2426 (with 1.0uF noise-reducing cap between pin 8 and neg rail) in front of it to provide regulated split rails. Non-standard parts are an electro coupling cap (the bias on the cartridge should keep the cap from seeing any reversed voltages), X7R 0.1uF caps from opamp supply pins to ground, and a bulk 470uF electro between + and -.

You should use some simple low-pass filtering to keep everything stable and happy. If you don't bandwidth limit the input to your op-amp it may oscillate.

You mean I should put some inductance on the input(s)? I'm pretty sure the opamp isn't oscillating.

For lower-noise regulators, you just have to shop around. So the LM317 has 30uV/V noise... this one has 10uV/V; that should help. It's also rated 10Hz-100KHz and doesn't require feedback resistors.

Yeah, the UA78L12C has even less than that, so I'll probably use that for my next build - if I ever build another AC-powered one.

For capacitors, you'd probably want ceramic for decoupling regulators and your rail-splitter... for signal coupling you'll want film caps (as I think you're doing?)

Yep, using ceramic for decoupling. Again, the input coupling/DC blocking cap is an electro because it's a lot smaller. The cartridge is powered with -6V up to -9V, and the output from the cartridge will never swing that high, so the cap should always see a lower voltage on the cartridge side - I think.

Originally posted by: blahblah99
Not to burst your bubble or anything but if you want extremely high performance out of a microphone pre-amp you will need to do circuit analysis and CAREFUL pcb layout. I once designed a mic pre-amp with 48V phantom power with an EIN of -125dBu and about -116dB of dynamic range. It took a lot of number crunching, pcb layout tweaking, and circuit modifications.

Yikes, -116dB is not much range. 😉 I agree, though, cheap parts on a perfboard without careful layout consideration will be rough:

- Star ground reference point
- High-quality decoupling caps
- Careful decoupling placement (close to components)
- Careful power and signal routing
- Good shielding

I'll probably build one off an ESP board (or with the INA217) if I ever need to hook up professional mics. I've got an AC-powered preamp for use at the computer and for speaker testing, and I'm going to be building another battery preamp for portable head recordings (two WM-61A). Hopefully the mic noise should swamp the preamp noise.

 

[blahblah99]

Originally posted by: bobsmith1492

Originally posted by: Howard

Originally posted by: bobsmith1492
I wouldn't worry about noise specs on voltage regulators. Modern regulators all have extremely low noise (i.e. don't use an LM317 because they're ancient...). Plus, if you're simply powering op-amps, the op-amp will have tens of dBs of power supply rejection.

If you're worried about a -118dB noise floor, you're slightly crazy. 😉 You'll never see that little noise which means the regulator's noise is inconsequential.

I ran the preamp (in a case) without a mic hooked up and I could definitely hear some noise after turning it up. Not sure if it's the 84.77uV from the feedback resistors or whatever noise from my own sound card or some crap picked up by the dollar-store interconnect, but it's there.

What performs better than an LM317?

Can you draw a schematic of your circuit? I looked through the thread but couldn't find one. You should use some simple low-pass filtering to keep everything stable and happy. If you don't bandwidth limit the input to your op-amp it may oscillate.

For lower-noise regulators, you just have to shop around. So the LM317 has 30uV/V noise... this one has 10uV/V; that should help. It's also rated 10Hz-100KHz and doesn't require feedback resistors.

For capacitors, you'd probably want ceramic for decoupling regulators and your rail-splitter... for signal coupling you'll want film caps (as I think you're doing?) Ceramic caps are ok but you have to keep your signal well under the capacitor's voltage rating. They become non-linear in impedance beyond about +/-25-50% of their rating. Electrolytics are bad unless you know the polarity of your input or use a bipolar cap.

Circuit construction is by far the biggest contributor to noise, though. I could come up with some tips if you still have noise problems.

Originally posted by: blahblah99
Not to burst your bubble or anything but if you want extremely high performance out of a microphone pre-amp you will need to do circuit analysis and CAREFUL pcb layout. I once designed a mic pre-amp with 48V phantom power with an EIN of -125dBu and about -116dB of dynamic range. It took a lot of number crunching, pcb layout tweaking, and circuit modifications.

Yikes, -116dB is not much range. 😉 I agree, though, cheap parts on a perfboard without careful layout consideration will be rough:

- Star ground reference point
- High-quality decoupling caps
- Careful decoupling placement (close to components)
- Careful power and signal routing
- Good shielding

I had it at -125 DR, but had to give up some DR for EIN. Little tradeoff, since the marketing folks seem hell bent on better EIN rather than DR for mic input.

The INA163 is a good choice, the linear LT1128 is slightly quieter with better specs, but GBW of only 20MHZ, which means you can only get about 60dB of gain with 20khz bw.

 

[Howard]

Good morning, I suppose.

http://www.diyaudio.com/forums....php?s=&postid=1773501

 

[Howard]

Originally posted by: Howard

Originally posted by: bobsmith1492
I wouldn't worry about noise specs on voltage regulators. Modern regulators all have extremely low noise (i.e. don't use an LM317 because they're ancient...). Plus, if you're simply powering op-amps, the op-amp will have tens of dBs of power supply rejection.

If you're worried about a -118dB noise floor, you're slightly crazy. 😉 You'll never see that little noise which means the regulator's noise is inconsequential.

I ran the preamp (in a case) without a mic hooked up and I could definitely hear some noise after turning it up. Not sure if it's the 84.77uV from the feedback resistors or whatever noise from my own sound card or some crap picked up by the dollar-store interconnect, but it's there.

I recorded without the preamp hooked up (no interconnect connected, even) and there's still quite a bit of noise. God damned onboard audio.

 

[mindless1]

Hmm, what's round two then? USB sound card as recording input, it's supply lines through USB severed and an LDO regulator placed inbetween system and card, maybe a LC filter too for fun?

 

[Howard]

Yeah. I think an E-MU 0404 might be in my future. What do you do with an LC filter? An RC's a low-pass, isn't it?

 

[mindless1]

Noise in a PC tends to be high frequency (when heard through the integrated audio), and if starting at 5V for USB, plus the drop across an LDO regulator, you don't want a lot of loss across a regulation stage too (I'm not even sure what the minimum voltage a typical USB *sound card* will tolerate), so an LC is called for.

 

[Howard]

Originally posted by: mindless1
Noise in a PC tends to be high frequency (when heard through the integrated audio), and if starting at 5V for USB, plus the drop across an LDO regulator, you don't want a lot of loss across a regulation stage too (I'm not even sure what the minimum voltage a typical USB *sound card* will tolerate), so an LC is called for.

Meh. 4 AA NiMHs with a cap would do the trick, I think. But we're going a bit overkill on it, aren't we? I mean, it IS designed for USB power.

 

[mindless1]

Maybe overkill, but the same power rail noise is present on USB as on your integrated audio, it too being designed to run from that power.