Old: A Direct-Coupled Input-Capacitorless Active Mic Preamp

[JR.]

It seems we discussed this before. I've black balled some cap series because of noise only when phantom on.

IIRC we speculated that maybe they would form in and quiet down over time.

Does your I/'F noise get any better after letting the cap run energized for a while (overnight)?

JR

[mediatechnology]

It seems we discussed this before. I've black balled some cap series because of noise only when phantom on.

IIRC we speculated that maybe they would form in and quiet down over time.

Does your I/'F noise get any better after letting the cap run energized for a while (overnight)?

Well, it seems to, but it takes a long time to form.
I haven't turned it off long enough in this test setup to see how long it takes to re-form again.

I recall from some previous 1510 experiments that if it were off overnight it would take another 15-45 minutes or so to re-form again the next day.
So the formation isn't permanent.

[mediatechnology]

I decided to leave the preamp with coupling capacitors installed completely off overnight so that they could fully discharge without being formed by bias currents.

I started a timer at power and phantom turn-on while monitoring the output via FFT.
I concentrated on the 10-100 Hz band specifically looking at the 10 Hz level.
The phantom off level in this band is typically -105dB.

It took about 4-6 minutes for the level at 10 Hz to be at the -100 dB level only some of the time.
The spectra from 10-100 Hz could be seen to "percolate" to -80 or -90dB even after 4-6 minutes and followed an inverse 1/f type curve.

At 8-10 minutes 10 Hz was floating around -100 dB most of the time.
At 15 minutes it was usually below -100 dB at 10 Hz.
At 30 minutes the LF noise spectra from 10-100Hz was tranquil appearing like it does with phantom off with the decade from 10-100Hz hovering around -102 to -105 dB.

For the DC-coupled input-capacitorless preamp the stabilization time is determined by the microphone.
The condenser mics I have here are good-to-go and quiet in about 2-3 seconds.

[JR.]

Good data..., the more I think about it, I suspect that cap I black-balled back at Peavey was just not formed in, and the brief on-time during build up and QA test was not long enough for it to form in and get quiet. If I could have figured this out at the time, we wouldn't have had to rework an entire batch of mixers and PCBs, just let them run until quiet,

That adds some credence to the phools who insist on breaking in their electronics, or leaving them turned on..

I believe there is a self-annelaing phenomenon where zenered semiconductor junctions can heal and get quieter over time from forward current, but I can't say I've ever measured or confirmed that.

JR

[emrr]

Good batch of evidence.

The WAV files sound like nothing but ambient noise.

[mediatechnology]

The WAV files sound like nothing but ambient noise.

That's kinda what I thought but I'm not used to listening to the house at 75 dB gain.
I'm inspired to setup an MS or stereo pair of the AT-3525s out in the yard or down in the woods and just record ambience.
But, the Coyote I saw in the back yard this morning may inspire me not too.
The AT-3525s seem to have a fairly low self-noise.

I also rechecked the 1/f coupling cap noise after an hour of "rest" e.g. the capacitors pre-formed, the circuit powered, but phantom off via back-grounding the 6K81 resistors.
It gets quiet in about 2 minutes and by 5 minutes looks like it does with phantom off.

Based on the difference in servo correction with both phantom on and off it looks like the re-formed coupling capacitors add about 1.6 uA input offset current.
That seems reasonable.

I "get" what SSL did in the 82E149 to pre-bias/pre-form them. I may have already posted this in this thread but it's a long thread.
Pin 1 is "always on" 48V, the left side connects to the switched 6k81 pull-ups.
The positive ends of the caps point to the middle.
R13 and R14 lead to the input transistor bases and protection.


SSL82E149 Mic Input Phantom Pre-Bias

[mediatechnology]

Discovered an error in my previous screen captures: The 40 Hz 12 dB/oct HPF was on! Duh.

This doesn't change the previous relative measurements of capacitor forming and the "warm-up" time.
With the HPF Off however the rising 1/f component is far more obvious when coupling capacitors are installed.

This is a comparison of AC Coupled both with Phantom On and Off along with the DC-coupled version.
The 1/f component the capacitors introduce - due to them being in series with the 150R source - is obvious.
The DC-coupled version does not show the 1/f rise either with Phantom On or Off.

Note also that the AC-coupled preamp has 1K5 X2 bias resistors. The differential load on the right-hand side of the capacitors is 3K.
The DC-coupled preamp uses the Phantom pull-up resistors as the bias resistors. The total differential load is 5K.
Thus, the DC-coupled preamp should be at a slight disadvantage: But it isn't because the 150R source impedance is directly in parallel with the differential load. The THAT1570 "sees" 170 Ohms. (150 Rsource plus 10R/leg current limit is 170R.)
In the AC-coupled test circuit, the input capacitors reactance, with 23.5 uF equivalent capacitance adding 678 Ohms to the 170 Ohm resistive source at 10 Hz, presents an effective source impedance nearly 5 times higher.


AC-Coupled Preamp Phantom Off 47uFX2 150R 40Hz Off


AC-Coupled Preamp Phantom On 47uFX2 150R 40Hz Off After 5 Minute Capacitor Reforming


DC-Coupled Preamp Phantom On With Coupling Capacitors Removed

[ricardo]

There's another mechanism that is at least as important for LF noise. It's the noise current of the preamp through the evil caps even though they are 'perfect' or 'perfectly formed'.

This is a M-Audio DMP3 modified for within 0.5dB of Millenia Media HV3 noise at 200R. It's a dedicated Ribbon preamp so we bin the evil caps. Hiss is about 4dB less but look at the LF due to 10u+10u on each leg.

DMP3 has INA163 with noise current 0.8pA/rtHz. THAT 1510 etc have 2pa/rtHz so will be worse.

[edit] Du.uuh! I see Wayne already sorta mentions this!

[JR.]

Is that roll off below 4 Hz your bench or a HPF in the pre?

47uF and 150 ohm is 22Hz, so there will be a rising slope from say 22Hz?

Of course this is all not very audible but as long as we are chasing the angel's footsteps on the pinhead.

JR

[mediatechnology]

It's the noise current of the preamp through the evil caps even though they are 'perfect' or 'perfectly formed'.

Exactly. The noise current develops across the series combination of the source impedance and the two capacitors.
Without the capacitors you just have the noise current develop across the source and 10R current-limiting resistors.
Your plot seems to show the same thing and the 10 uF certainly exagerates it.

And then there's a Catch-22: If you chose to use a pair of film capacitors to avoid the leakage current noise and formation time of electrolytics (and DA) you're forced into lower values having a higher reactance for the noise current.

John:

There is a pole around 7 Hz in the AC-coupled "line receiver" that follows the preamp, a pole in the differential Deboo servo that varies with preamp gain, and a third one in the sound card.
The Soundcard, an EMU-0404 internal, has a LF peak around 6 Hz with a shorted input.
I'm not sure below 10 Hz we can resolve much more than slope and relative level.

I think the slope should begin rising about an octave higher than 22 Hz since it's not actually 47 uF but half that since it's 47uF/leg or 23.5 uF 150 Ohm.