Getting capsule polarization voltage from phantom power in a Schoeps-style circuit

[skyy7]

Unfortunately, I don't yet understand everything well enough to answer the question myself, but would it be possible to create around 30v to polarize the capsule using a circuit like this in a Schoeps-style circuit? I've seen it used in xfmr circuits but never in Schopes-style circuits. IIRC, the old version of the u87 polarized the capsule in a similar fashion.

What would the downsides be aside from an increase in S/N?

 

[Khron]

would it be possible to create around 30v to polarize the capsule using a circuit like this in a Schoeps-style circuit?

Yes.

What would the downsides be aside from an increase in S/N?

~3dB less sensitivity than at 60v.

Look up sE2000 or sE2200A schematics.

 

[skyy7]

Yes.



~3dB less sensitivity than at 60v.

Look up sE2000 or sE2200A schematics.

Oh, that's not too bad. Could I technically try to compensate by increasing the zener voltage and adjusting R4 and R5 accordingly?

 

[Khron]

Oh, that's not too bad. Could I technically try to compensate by increasing the zener voltage and adjusting R4 and R5 accordingly?

Not really - as long as R4 and R5 have the same value, the gain of thr circuit won't change.

 

[kingkorg]

What would the downsides be aside from an increase in S/N?

View attachment 120557

I guess you meant decrease in s/n. Signal goes down, noise stays the same, ratio is decreased.

 

[rogs]

You might try and increase the value of R7 to - say - 6k8 (or possibly even a little higher?) to reduce the standing current in the zener?

I use THIS CIRCUIT to utilise phantom power for capsule polarisation..... It does not require the DC supply to be regulated, and so only draws around 2mA in total from the phantom power supply.
Allows for around 40v of polarisation, depending slightly on the 'real world' characteristics of the phantom power supply being used.
No individual JFET bias adjustment required for minimum distortion either !

 

[abbey road d enfer]

~3dB less sensitivity than at 60v.

More like -6dB. Output is directly proportional to polarization voltage.

 

[abbey road d enfer]

would it be possible to create around 30v to polarize the capsule using a circuit like this in a Schoeps-style circuit? I've seen it used in xfmr circuits but never in Schopes-style circuits.

Why would you want to do that?

IIRC, the old version of the u87 polarized the capsule in a similar fashion.

Since it derived its polarization voltage from P48, they had to deal with about 40V instead of 60. There was no other motivation than simplicity. Apparently, the outcome was not entirely positive, since Neumann designers used DC/DC converters in many subsequent models.

 

[rogs]

More like -6dB. Output is directly proportional to polarization voltage.

I was able to derive 42v as a polarisation voltage, using my dbx286s preamp. ( Its phantom power supply outputs 52volts).
That reduced the sensitivity by 3dB against using a 60v polaristion voltage.
My Sound Devices preamp only gave me c.40v, so a further 0.5dB drop in sensitivity..... Not really a big deal!

I only tried using phantom power for polarisation out as an expermient. Works quite well.
Adding a voltage multiplier - as I did with my op-amp version HERE - is not an expensive or difficult task, and gives more flexibility when selecting a specific polarisation voltage.

 

[Cqwet Dbdfte]

Strip board! Have not seen those in a long time.
Any interest in a PCB version? Could be 4 layers or more if needed.
Could be farmed out for pic n' place assembly....

 

[Khron]

Could be 4 layers or more if needed.

What for? I think i managed to squeeze even the +/- version onto a single layer. Of course, having the other-side layer as a full ground plane would prevent risk of interference...

 

[rogs]

Strip board! Have not seen those in a long time.
Any interest in a PCB version? Could be 4 layers or more if needed.
Could be farmed out for pic n' place assembly....

It's cheap! -- and I find it useful for prototypes and 'one offs'...... also better for making simple corrections and revisions, with minimal cost..
Mic Scharf has already done a PCB layput for most of my OPIC mics .. see HERE for example.
Double sided - no ground plane. No freebie Gerber files though.
Khron has done the layout for my RF mic ... more complex, double sided with groundplane. Works very well.
(Khron has very kindly made those Gerber files freely available to all.)

 

[abbey road d enfer]

I was able to derive 42v as a polarisation voltage, using my dbx286s preamp. ( Its phantom power supply outputs 52volts).
That reduced the sensitivity by 3dB against using a 60v polaristion voltage.

42V vs. 60V is 0.7 or -3dB.
30V vs. 60V is 0.5, or almost exactly -6dB (6.02 actually).

 

[Voyager10]

Here's a tweaked version which gives (in simulation) just over 42V bias from a 48V supply:



(I've just included the components relevant to the DC operating point here)

I don't know what bias current you want to run the FET at, but 1mA (modelled by source I1) should be plenty. The changes are:

1) Adding (on my diagram) R8/R9. This is now a 'Vbe multiplier' circuit giving about 10V between the emitters and collectors of Q1/Q2. This will allow a much larger output voltage swing before the transistors saturate.
2) Increase R1 (R7 on the original schematic) to 12K. This just allows the Vcoll voltage to be higher (about 32V in simulation).

So, in simulation, the capsule voltage (Vbias1/Vbias2) is 42.5V, Vcoll is 32.0V, Vfet is 12.6V.

While I'm here...
- The C2/C4 values (47nF) give a bass rolloff -3dB point of 34Hz (with the 100K resistors). No problem if you're happy with that, personally I'd make them at least double that.
- (Little-publicised fact about this type of output) the 'RFI suppression' caps C8 and C9 impose a slew rate limit on the positive-going edges of the output signals, determined by the capacitor value and the supply current. In this case it's 0.8mA each side, giving a decidedly poor 36V/ms. IMO these could be significantly smaller (e.g. 2.2nF).

 

[thor.zmt]

Unfortunately, I don't yet understand everything well enough to answer the question myself, but would it be possible to create around 30v to polarize the capsule using a circuit like this in a Schoeps-style circuit? I've seen it used in xfmr circuits but never in Schopes-style circuits. IIRC, the old version of the u87 polarized the capsule in a similar fashion.

What would the downsides be aside from an increase in S/N?

View attachment 120557

My first stop would be to replace the PNP Transistors with low power P-Channel Mosfets (BSS84 e.g.) and a bias Network in the MOhm. Otherwise 47nF coupling capacitance give ~40Hz -3dB roll-off, on top of the capsule's roll-off.

The LSK170/2SK170 is not the best choice as J-Fet, fairly high input capacitance. This will potentially form a voltage divider with the Capsule, further reducing output levels and degrading SNR. At least use a second suitable J-Fet as Cascode. Another LSK170C would probably work.

R14/C3 seem misplaced, is it is intended to form an EQ.

Feeding the frontend from 12V reduces available headroom. The original design could work with P24 and even P12 Phantom power. If we are sure of 48V only operation (which is the case if we derive bias from Phantom Power), why not optimise the overload margins?

Run a 24V Zenner diode directly in the drains of the Mosfet's (or Collectors of BJT's) and use a 1k/100uF "denoise" filter. With 1mA current for the input, we have 23V for the frontend.

Set the source to drain (collector to emitter) voltage of the output devices to around 17V. This leaves ~ 40V to bias the capsule.

The frontend get's a pair of 5.6k resistors and can swing around 4V balanced which with a 10mV/94dB @ 60V Capsule operated at 40V, giving 14mV/94dB and theoretically 143dB unclipped.
Lastly, while 48V is the common Phantom Power Standard, there is nothing that stops you from increasing Phantom Power to 68V making in effect a new "standard" of P68.

A P68 powered microphone would provide 60V bias and need a switch to add a 22V Zenner diode to the power supply circuit to balance all voltages.

This would get us to 20mV/94dB with a 140dB overload ceiling and 3dB better SNR.
Here a schematic I have used on the usual "China KTV serious looking garbage mikes" PCB's to convert them to something that works pretty well:


The main circuit is modified mainly by replacing SMD Parts and adding a few.

The 10nF and smaller value capacitors are C0G, the remainder are Panasonic SMD Film. The parts creating the bias are on the bottom side of the PCB. Sorry, the Mic's are still in the Wherehouse in another country with my studio, so no photos for now.

In my own Mic's I take the J-Fet's off the PCB and place them directly the capsule, with two twisted wires carrying the signal to the PCB.

The Microphone bias parts shown leftmost are also directly at the capsule. So there are 4 wires from the capsule, not 2 and the capsule is essentially active.

The 22k resistor on the Emitter of T5 needs to be matched with the J-Fet. The J-Fet actually operates at iDSS, which is variable between J-Fet's.

I normally get a tape with a fair few J-Fet's and make nice matched pairs/quads for pairs of quads of identical mic's all then with identical "current adjustment" resistors. Note these are 2SK660, which are one of the better choices for actual capacitor microphones.

Thor

 

[abbey road d enfer]

- (Little-publicised fact about this type of output) the 'RFI suppression' caps C8 and C9 impose a slew rate limit on the positive-going edges of the output signals, determined by the capacitor value and the supply current. In this case it's 0.8mA each side, giving a decidedly poor 36V/ms. IMO these could be significantly smaller (e.g. 2.2nF).

I couldn't run your simulation (unknown schematic syntax**) but my simulation (admittedly with different PNP's) show an operating current of about 3.7mA, which results in significantly higher slewrate, but with a voltage drop at the collectors that result in only about 22V for biasing the capsule.
Anyway, 36V/ms allows outputting about 0dBu at 20kHz, which is not too bad.

** Are you on Mac? Your file is in columns, when it should just be with spaces.

 

[abbey road d enfer]

- The C2/C4 values (47nF) give a bass rolloff -3dB point of 34Hz (with the 100K resistors). No problem if you're happy with that, personally I'd make them at least double that.

Although I don't make mics, I used the Schoeps output topology in my MicBooster. In this position I use 220nF. It has the advantage of maintaining the low output Z down to the lowest frequencies.

 

[rogs]

It is a testament to the simplicity and elegance of the Schoeps CMC5 schematic that it is still so highly regarded, some 50 years after it appeared.
There are probably hundreds of threads online discussing the various aspects that can be 'tweaked' to 'improve' it .

The fact remains though it utitlises components that were easily available 50 years ago. Some - like suitable discrete 'thru hole' JFETs for example - are becoming increasingly hard to find these days.
Following hobbyist Jules Ryckebusch's suggestion to try OPA164* series opamps in place of a separate JFET input a couple of years back, it will be interesting to see whether that kind of topology starts to replace 'Schoeps' style circuits in the DIY world, going forward?

I've been using the OPA1641 for a while now, and find it a very useful alternative to Schoeps style.....
• No need to regulate the DC, and thus limit headroom.
• No need to include any FET bias adjustment
• Very low distortion
• Rail to rail output swing
• Acceptably low noise.
• No noise creating resistors in the signal path
I have used the attached schematic as a simple way to interface with a conventional 34mm LDC condenser capsule, using phantom power for the capsule polarisation supply and find it seems to work at least as well as the Schoeps circuit......

Although the audio output is single sided, the line is passively impedance balanced, to help maintain a reasonable CMRR.

So far, there seem to be relatively few hobby folk who have tried this type of circuit, instead of using the ubiquitous 'Schoeps'...
As suitable JFETS get more and more difficult to find, maybe more DIY hobbyists will venture over to this 'dark side' alternative?

 

[Voyager10]

I couldn't run your simulation (unknown schematic syntax**) but my simulation (admittedly with different PNP's) show an operating current of about 3.7mA, which results in significantly higher slewrate, but with a voltage drop at the collectors that result in only about 22V for biasing the capsule.
Anyway, 36V/ms allows outputting about 0dBu at 20kHz, which is not too bad.

** Are you on Mac? Your file is in columns, when it should just be with spaces.

Oh, I'm sorry. I'm using LTspice for MacOS, v17.1.4. (I had to rename the .asc file to .asc.txt to get it to upload, but if I re-download it and rename it, it works for me).

It is a testament to the simplicity and elegance of the Schoeps CMC5 schematic that it is still so highly regarded, some 50 years after it appeared.
There are probably hundreds of threads online discussing the various aspects that can be 'tweaked' to 'improve' it .

Absolutely - I'm sure it's survived so well because it works even with all manner of component substitutions. Using two off-the-shelf inductors (relying on mutual coupling between them) rather than a custom-wound RF coil is a stroke of genius.

Another point: it was designed to work at 12V and 24V phantom power voltages, just by bridging a single resistor (see Schoeps for annotated schematic). Particularly at 12V, there is a very tight tradeoff between headroom in the FET stage and headroom for the output transistors - the R6/R7 resistors which supply the base bias current are 'tweakable' for this reason. (The 'Vbe multiplier' arrangement I suggested wouldn't work well here).

Also, I'm sure the 22nF capacitors were chosen with as a careful compromise between not hitting slew-rate limiting in practice, and minimising the amount of RF which bleeds through to the output (inevitable in a cramped circuit layout). I've tried taking those caps off MXL 440 and 990 boards (which use this same circuit) and you get a very audible "AM radio" whine on the output. (Copper foil tape helps a lot here).

 

[abbey road d enfer]

(I had to rename the .asc file to .asc.txt to get it to upload, but if I re-download it and rename it, it works for me).

That's what I did, but doesn't work. It seems it rearranges text as a spreadsheet.
Do you use LTspice' Mac version or under emulated Windows?