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

Where we discuss new analog design ideas for Pro Audio and modern spins on vintage ones.
Locked
emrr
Posts: 570
Joined: Sun Mar 02, 2008 10:27 pm
Location: NC, USA
Contact:

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

Post by emrr »

We have an observable!
Best,

Doug Williams
Electromagnetic Radiation Recorders
User avatar
JR.
Posts: 3700
Joined: Sat May 24, 2008 7:21 pm
Contact:

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

Post by JR. »

That lower 1/F noise looks good.
---
I've mentioned this before, I have actually encountered issues with at least one series of electrolytic cap that exhibited audibly noisy leakage current with phantom voltage applied. I would like to consider that one cap brand/series an outlier, that we black listed from purchasing approved vendor list (just the one bad series not the vendor's other caps as it appeared to be a specific design issue). I didn't personally deal with this. It was discovered by factory QA (it was that audible) and handled by one of my senior engineers, when I was over the mixer area at Peavey. So it seems to me any leakage current from this vector "could" help your DC coupled capacitor-free input measure even better (actually cap coupled inputs could measure slightly worse) in high resolution bench tests.

Another observation, a DC blocking cap in the gain leg (like most manufacturers use) will generally reduce gain at very low frequency where the I/F noise is rising. You don't say but I ASSume your 60 dB gain stage is flat to DC. I have seen Mackie on some models use marginal (IMO under-sized) caps in this gain node, to effectively roll off audible LF noise at very high gain settings (impacting above 20 Hz), despite specifying flat frequency response for the preamp (shame on them). :roll: The consumers who didn't know any better perceived the preamps as being lower noise, not reduced LF frequency response.

Another thing on my uber preamp wish list is the ability to tweak input stage current density to be a better match NF wise for the real source impedances encountered. Most mics will be higher than 100 ohm (with a few lower), while this may be a lot of work for very little benefit. With a wide band A/D hanging off the back end, i could almost automate noise measurements and current density tweaks, but this is way out there and not very practical. Probably more practical to tweak while listening to the noise floor, but even that is a little unrealistic as we would probably boost it to a much higher sound level than it would normally be encountered at for listening tests. If done cybernetically noise weighting would need to be used.

I am unclear about how DA would express in this application, but I am open for an exploration. Perhaps some overhang after transient input overloads (similar to your settling time findings).

Another metric that your DC coupled front end should improve upon is both LF and HF CMRR, significant issues for premium transformer-less mic preamps.

I believe we've discussed most of these issues before, just a bit of a refresh.

JR
Cancel the "cancel culture", do not support mob hatred.
User avatar
mediatechnology
Posts: 5444
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

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

Post by mediatechnology »

Thanks Doug and John
JR. wrote:That lower 1/F noise looks good.
Thanks. I'd been theorizing we would see an improvement. Although it might not have enough gain for a ribbon mic as-is, with a 30R source, the differences should be even larger. You mentioned switching out caps earlier and many designs do provide an input capacitor bypass for ribbons. Obviously not needed here.
JR. wrote:I've mentioned this before, I have actually encountered issues with at least one series of electrolytic cap that exhibited audibly noisy leakage current with phantom voltage applied... So it seems to me any leakage current from this vector "could" help your DC coupled capacitor-free input measure even better (actually cap coupled inputs could measure slightly worse) in high resolution bench tests.
I agree and remember you mentioning that. I've seen/heard it too though it was not brand-specific and it cleared up after electrolyte re-formation. This is one of the issues with turn-on time: The longer phantom is off, the longer you have to wait for it to be truly quiet. What I heard was rather annoying too - not white or pinkish noise but a "spitting" and whistling sound. Like popcorn and shot noise.
JR. wrote:Another observation, a DC blocking cap in the gain leg (like most manufacturers use) will generally reduce gain at very low frequency where the I/F noise is rising. You don't say but I ASSume your 60 dB gain stage is flat to DC.
Well, yes and no. Yes it's DC-coupled in the traditional sense because there's no Cgain capacitor. At 60 dB gain, the differential servo provides an equivalent Cgain with a 20 Hz response of -2 dB. At lower gains the corner falls. I may eventually tweak the servo C.
JR. wrote:I am unclear about how DA would express in this application, but I am open for an exploration. Perhaps some overhang after transient input overloads (similar to your settling time findings).
DA (dielectric absorption) means that you can never fully discharge the input coupling capacitors in a conventional preamp or - if you wait long enough to think you've discharged them - they'll re-charge. So it creates this constantly changing input offset. I fought this trying to develop a "click-less" tip/ring switch. Discharge the caps all the way (something on the order of 15 minutes), remove the discharge bleeder, and in a few minutes you'll see maybe 3-5 volts develop.

With a servo correcting this DA "see-saw" at the inputs it's less of an issue. With a preamp not having a servo it will have a scratchy gain pot (or clicky switch) despite having two input coupling capacitors and a Cgain.
(Another source of DC offset causing clicks that can appear on top of DA - but not caused by it - is RF rectification.)
JR. wrote:Another metric that your DC coupled front end should improve upon is both LF and HF CMRR, significant issues for premium transformer-less mic preamps.
Absolutely.
The problem with capacitors are the increasing effect of mis-match as Cin and Rbias get smaller.
If you make Rbias small to lower 1/f noise then you've "amplified" capacitor mis-match errors for a given value C and it usually requires a +/-20% tolerance electrolytic.
If you make Rbias large, in order to use 10% film caps, then the film cap has to be small in value 5-10 uF (because its as big as a barn) and you've also made mis-match worse. Now, 1/f noise is raised because Cin's reactance is contributing more to the overall source impedance. See: viewtopic.php?f=6&t=14&start=153
Make Rbias small and Cin big and you've reduced mis-match to improve low frequency CMRR but have provided a huge and destructive stored charge.
So whatever you do to control 1/f noise by lowering Rbias, or raise Rbias to eliminate the electrolytic and replace it with film, CMRR takes a back seat.

What I need to determine is the role of the CM servo's response on 50 Hz CMRR.
I don't see it as an issue but it might be.
I may end up, on the output of the preamp, cap-coupling both an AC-coupled Vcm signal and preamp output into a bootstrapped diff amp. (THAT1200).
I don't think it may be necessary but it might.

One thing I need to build for testing is a Schoeps topology PNP emitter follower for a signal generator - phantom capable - interface.
Stuff to the right of T1 here: http://www.ka-electronics.com/DC_Preamp ... s_Topo.jpg
User avatar
JR.
Posts: 3700
Joined: Sat May 24, 2008 7:21 pm
Contact:

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

Post by JR. »

mediatechnology wrote:Thanks Doug and John
JR. wrote:I am unclear about how DA would express in this application, but I am open for an exploration. Perhaps some overhang after transient input overloads (similar to your settling time findings).
DA (dielectric absorption) means that you can never fully discharge the input coupling capacitors in a conventional preamp or - if you wait long enough to think you've discharged them - they'll re-charge. So it creates this constantly changing input offset. I fought this trying to develop a "click-less" tip/ring switch. Discharge the caps all the way (something on the order of 15 minutes), remove the discharge bleeder, and in a few minutes you'll see maybe 3-5 volts develop.

With a servo correcting this DA "see-saw" at the inputs it's less of an issue. With a preamp not having a servo it will have a scratchy gain pot (or clicky switch) despite having two input coupling capacitors and a Cgain.
(Another source of DC offset causing clicks that can appear on top of DA - but not caused by it - is RF rectification.)
I do not question your observations. The standard model for DA is a multiplicity of parallel distributed internal RxCs. The classic observation of a cap appearing to recharge by itself if a shunt is briefly applied and removed, is well explained by the standard model as internal charge continues to redistribute and equalize. That said, when a fixed constant termination is applied the passive response to a step input driving function should be monotonic. The internal RxCs can have longer time constants than the external time constant depending on the specific termination. Also the e^-t/rc discharge curve asymptotically approaches zero, never reaching 0V. :D

Note: the same relationship defines the voltage after phantom is turned on, so the caps will be slowly approaching their final voltage for some time. 95% there will be relatively few time constants, the last 5% could be many more with large electrolytic caps. Even without DA that last 1% will take a while.

The response of a servo to this input step (derivative of a step) could be more complex as it involves voltage gain and could even be underdamped, leading to ringing. I don't know how much of your observation is just somewhat longer settling time, and possible interaction with active circuitry.

DA is well known as problematic when termination impedance is not constant (like sample and holds), I have never really thought of it as a settling time problem but it will have a subtle effect in that regard. I stand corrected, I was looking for some gross effect, like from input clamping. DA has been widely blamed for sonic ills that I haven't quite been able to easily identify, but that isn't the only one I am still searching for. :oops:

JR
Cancel the "cancel culture", do not support mob hatred.
User avatar
mediatechnology
Posts: 5444
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

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

Post by mediatechnology »

That said, when a fixed constant termination is applied the passive response to a step input driving function should be monotonic.
The problem is "amplified" (pun intended) by the fact that there are two capacitors and the preamp output is the differential response of a "multiplicity of parallel distributed internal RxCs." So it isn't necessarily monotonic.
At the output of the preamp you will sometimes see an oscillatory swinging from rail-to-rail then decreasing as it settles.
It's a wobbly thing.
Also the e^-t/rc discharge curve asymptotically approaches zero, never reaching 0V
Yes, which is where the fun begins when you want to turn off phantom.
How fast is fast enough?
Particularly since you'll never get all of the charge out of it anyway.

If you say "I'll take a time constant of 5 seconds - that ought to be long enough" then you still have with 48V at T=0 a whopping 17V at 5 seconds.
That's always made the case for back-grounding the 6K81's during phantom off for me.
By back-grounding the phantom pull-ups the tau is 320 ms to reach 17V.
And 17V is still a lot.
If it were my ribbon mic I think I'd still wait a minute or so before patching.

Or, just eliminate the caps.
User avatar
JR.
Posts: 3700
Joined: Sat May 24, 2008 7:21 pm
Contact:

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

Post by JR. »

mediatechnology wrote:
That said, when a fixed constant termination is applied the passive response to a step input driving function should be monotonic.
The problem is "amplified" (pun intended) by the fact that there are two capacitors and the preamp output is the differential response of a "multiplicity of parallel distributed internal RxCs." So it isn't necessarily monotonic.
At the output of the preamp you will sometimes see an oscillatory swinging from rail-to-rail then decreasing as it settles.
It's a wobbly thing.
I didn't raise the two cap tolerance thing, but even then it seems like a similar delta or bias should dominate in just one direction or the other. Active servo action could contribute to overshoot and hunting as it over corrects back and forth.
Also the e^-t/rc discharge curve asymptotically approaches zero, never reaching 0V
Yes, which is where the fun begins when you want to turn off phantom.
How fast is fast enough?
Particularly since you'll never get all of the charge out of it anyway.
Yup 1% is still 1/2V which is huge in a mic preamp input.
If you say "I'll take a time constant of 5 seconds - that ought to be long enough" then you still have with 48V at T=0 a whopping 17V at 5 seconds.
That's always made the case for back-grounding the 6K81's during phantom off for me.
By back-grounding the phantom pull-ups the tau is 320 ms to reach 17V.
And 17V is still a lot.
If it were my ribbon mic I think I'd still wait a minute or so before patching.

Or, just eliminate the caps.
The best cap is no cap... still..

why stop there? with dsp even filters can be made without using caps. 8-)

JR
Cancel the "cancel culture", do not support mob hatred.
User avatar
mediatechnology
Posts: 5444
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

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

Post by mediatechnology »

I didn't raise the two cap tolerance thing, but even then it seems like a similar delta or bias should dominate in just one direction or the other. Active servo action could contribute to overshoot and hunting as it over corrects back and forth.
It could also be the dynamic differences in leakage current as each cap re-forms.

IIRC I saw the effect in preamps having no servo.
For that matter there is the Cgain having to recover during phantom on/off transients.
There are also conditions where it could become reverse-polarized.

So yes the best cap is none at all.
ricardo
Posts: 262
Joined: Sun Apr 10, 2011 7:24 am

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

Post by ricardo »

There's a simpler explanation of the LF noise with smaller caps. THAT1510 has In = 2pA/rtHz.

With 2x22u on the input, this generates In / (2 pi f C) volts = 2e-12 / (2 pi 20Hz 11u) = 1.6nV / rtHz. Starts dominating the 1nV/rtHz En of THAT1510 etc. lower down.

Using INA163, 217 or the discontinued SSM2017 will see this occur at lower frequencies cos only 0.8pA/rtHz In

In a 'conventional' THAT1510/12 circuit, bigger i/p caps have to be balanced against the risk as Wayne says. I think 2x22u is the sensible choice.

Like JR, I've doodled many capacitorless mike i/p circuits over the years but have always shot them down cos real estate & number of bits. On the performance side, IMHO, Wayne's is the nearest to 'wanna build'. This is still far from stirring me from my beach bum lethargy cos, as you know, I begrudge even 2 extra resistors. :D

Somewhat off topic, any chance of THAT integrating the 4x1n4004s & 2x1n914, as in Phantom Menace 2, into future generations of 1510 / 12 ?
User avatar
mediatechnology
Posts: 5444
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

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

Post by mediatechnology »

On the performance side, IMHO, Wayne's is the nearest to 'wanna build'.
Thanks ricardo.
...Starts dominating the 1nV/rtHz En of THAT1510 etc. lower down.
Yep, the smaller the C the larger the reactance in series with the source x current noise = more noise.
[Would THAT be adding] 2x1n914, as in Phantom Menace 2, into future generations of 1510 / 12 ?
I recall at one time there being a discussion about making the internal rev-Vbe diodes more "manly."
Geometrically speaking they are fairly small.
FWIW and OT a casual glance at the external diodes might make one think junction capacitance would be an issue but it's not since they are bootstrapped.
The AC potential at both ends are essentially the same.

Dunno if there's existing die area big enough for 4X 1N400X along with the existing ESD diodes.
I'm using a DB-104 in the preamp which is a nice commonly available bridge.
The DB-104 in a 4 pin DIP-6 does make things a little tidier.

I'm getting a new roof today. Once the hammers stop and the dust settles I'll get back on this and post the input schematic.
I also made a couple of minor mods to the CM servo.
User avatar
JR.
Posts: 3700
Joined: Sat May 24, 2008 7:21 pm
Contact:

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

Post by JR. »

While I should let go of this, I am still trying to understand completely what is going on. Electrode forming is a chemical (plating) process with internal self-limiting based on condition of the electrode so will be a much more variable mechanism than DA. I am not clear about the un-forming mechanism (rate?) that should impact how much of this we encounter in normal use.

yes, caps suck... electrolytic caps suck in obscure ways.

JR
Cancel the "cancel culture", do not support mob hatred.
Locked