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

[ricardo]

Just for Jurassic perspective, here's the Mk4 Soundfield PSU with evil voltage doublers. 2 diodes & 2 caps. I think the transformer was 20-0-20VAC and chosen for the doubler. C65, the main '+15V' rail cap on which the doubler sits is the critical one needing Lo ESR etc. I just used a much bigger value. Top transformer winding works overtime.

The 53V which feeds the mike is beefy cos it has 4x5534s with 10x LN gain. Single chip HV regulators had just come out but I didn't have time to try cos I was already planning my move back to speakers.

The waste heat was used to warm the capsule assembly.

Standing as far back as possible, the heater was probably my most important contribution to the Soundfield project. Who cares about that low noise, low distortion and flat acoustic response sh*t? Happiness is a warm condenser mike.

With 30+ year hindsight, I can clearly see the mistakes I made, even with the technology of the day. But the stone age PSU wasn't one of them.

[mediatechnology]

Very cool! Looks like D18/D19/C53/C64 do the same (Self) deal by bootstrapping a halfwave doubler perched on top of the +Vbulk across C65.

the heater was probably my most important contribution to the Soundfield project

Funny. We were always told to the best way to dry the spit out of a microphone capsule was on top of a hot console. The bus routing switches on a JH-500 control surface were warmest spot to lay the capsule.
Looks like self-heating was a more elegant solution.

[ricardo]

Very cool! Looks like D18/D19/C53/C64 do the same (Self) deal by bootstrapping a halfwave doubler perched on top of the +Vbulk across C65.

Du.uuh! It is indeed C65 and not 55. Proves I'm blind as well as deaf

I'll change the original post.

[mediatechnology]

I took a break from tweaking the power supply I built because I wanted to use it to test a new "recovery" amplifier for the 1510 output.

I came up with this: https://www.proaudiodesignforum.com/for ... ?f=6&t=557


AC-Coupled Line Input with T-Bias Network to Raise Input Common Mode Impedance

I have insisted on using film caps to recover the floating 1510 output and realized that I needed an additional stage to realize common mode rejection due to the relatively high cutoff frequency of the Common mode servo.

The "-" input ties to the floating Vcm output, the "+" input to the 1510 output.
Not shown in the above schematic are 100R series resistors and input-to-rail diode clamps.

Low-value caps are desireable here to minimize the stored charge coupling from the 1510 output to following non-floating stage.
Film caps also are in keeping with a "non-electrolytic" design.

Self points out that you really want the LF cutoff determined by something other than an electrolytic.
In this preamp there are two LF time constants:
(1) The differential servo whose cutoff varies with gain (<=20 Hz at 60 dB moving lower with lower gain) and
(2) The "recovery" stage.

The beauty of using this circuit over the single-ended stage I had was that I can now use a bipolar op amp rather than a BIFET and also realize some additional common mode rejection.

[JR.]

With 30+ year hindsight, I can clearly see the mistakes I made, even with the technology of the day. But the stone age PSU wasn't one of them.

If enough time passes we can look at our old designs with almost fresh eyes. It' nice when we aren't embarrassed.

JR

[mediatechnology]

ricardo -

Your 48V supply looks very clean. I noticed that you'd built in short-circuit protection since it appears to enter a cable.
I'm thinking about feeding the preamp regulated 66V from the PSU and then putting the 48V ripple eater on the preamp board.
Been mulling whether I should provide 48V short-circuit protection but thought it unnecessary since the emitter will go straight to the phantom switch and be current-limited by the 6K8 pull-ups.

Also revisted the ECM8000 unequal phantom current per leg thing from here: viewtopic.php?f=6&t=14&start=137
I re-checked the numbers and the servo, with roughly "6dB" of DC headroom still left, is able to easily correct the 15% current imbalance of the ECM8000.
It could easily be made to provide more correction by lowering resistors but I don't think it will be needed.

[ricardo]

Don't think you need SC protection on P48V. Although its not in the P48V spec. we used to sneer at desks which wouldn't give good P48V on one channel even with all 71 other mike inputs shorted to earth. Some early famous Name UK desks were notorious for wimpy P48V

It's only 14mA per channel.

48V ripple eater on the preamp sounds good from that standpoint.

I think ECM8000 imbalance is worst case.

[mediatechnology]

Thanks ricardo.

My thinking was I'd take 66V to the actual preamp board and put the 48V ripple eater there so there is very limited opportunity for it to see a dead short. Glad you agree that it would not be required.

I'm looking at what it would take to use a 1570/1583 in this application. I think I've come up with a "differential" Deboo servo. I need to test that on a separate board.

My girls are coming by for Thanksgiving today. I might be able to talk my daughter Robin into cutting a vocal with the prototype.

[mediatechnology]

The more I look at using the THAT1570 or THAT1583 in this project the more I like it.

The 1570 has the performance improvements THAT write about in the datasheets but in this circuit the 1510/1512 have one thing going against them that require a work-around.
The 1510 and 1512 realize common mode rejection.

The DC-coupled input-capacitorless preamp's flying rail generator provides three outputs: One of them is the Vcm "center tap" artificial ground.
Vcm is required for two things: The differential servo reference and the 1510's ground reference, pin 5.
The 1510 reference pin is the "ground" connection for the differential to single-ended stage.
When the rails are flown, pin 5 needs to fly with them to provide a "pseudo" ground and center the output between the flown rails.
The problem with pseudo ground is that it's well, pseudo.
In order to have full common mode rejection, the "recovery" stage on the right-hand side of the film capacitor(s) need to be differential and sense not only the 1510's output but it's pin 5 Vcm AC potential.
The output needs to "see" the difference between Vcm and Preamp Output.

A 1570/1583 has a few extra things going for it in this application:

A THAT1570 has no "ground" pin and does not realize common mode rejection.
Its differential Deboo servo has no ground connection.
It has a common mode gain of "1."
The 1570 does not require a low-gain Vos trim because it has no added offset of the 1510's output amplifier.
The 1570's low gain offset appears differentially and is thus servo'd.

It just makes more sense to use a 1570/1583 because it requires less "stuff."

Sure an Ac-coupled differential amp is required in either case, but a whole op amp is eliminated.
Plus, the Vcm servo no longer has to drive a direct load: It drives only the flying rails to bracket the supplies around the input.

Test circuit for a non-flying, non-phantom powered, 1570 preamp used to check the servo:


THAT1570 Servo Test Circuit with A Differential Floating Capacitor Deboo Servo

All resistors in the above are 1%.
Note that the differential Deboo is non-inverting: The servo correction outputs cross-couple to opposing inputs to provide the proper polarity.
The noise performance is virtually identical when Rcm = 0 Ohms or 100K.

[mediatechnology]

I rebuilt the DC-Coupled Input Capacitorless Preamp with a THAT1570.

The output is recovered by the AC-Coupled high impedance line amp followed by a double-balanced differential amp.
All I can say is wow. This thing rocks.

Here is spectrogram of the noise floor with the circuit built on an open protoboard. 0 dB = 0 dBu.


DC Coupled Input Capacitorless Preamp Noise Floor THAT1570 60dB Gain 154R Source Impedance

It should be noted that the 1/f noise below 10 Hz is present at about the same level with the inputs to the analyzer shorted.
The noise floor rises a tad, maybe 3 dB, with phantom on.

FWIW the Beringher ECM8000 still requires about +/-7V of servo correction to satisfy it's quiescent current imbalance.