help me design/select/optimise a mic circuit - LDC multipattern

I'd like to come up with a circuit for a multipattern LDC.

criteria:
* engaging natural detailed sound
* simple circuit with modest parts count
* flat frequency response, with possibility of adding some treble de-emphasis if needed.

so first question is what circuit topology?

options in my current order of preference:

1. fet with concertina balancing outputs, followed by a pair of transistors for current drive. I like the look of this topology as its balanced from the outset, but I have never seen it used in a proper condenser mic, only in electrets, does it have some terrible flaw?

2. fet source follower, followed if necessary by a transistor for lower output Z (like oktava MK-012)

3. fet voltage amp followed by transformer (like U87 or KM84)

can anyone offer arguments for and against each of these?

..for simple and transparant, look at e.g. Neumann KM-84. The important choice will be the capsule, not so much the electronics..

Jakob E.

Hi gyraf.

I'm thinking of replacing the board in my Rode NT2000, i suspect the capsule can probably do better than it does with the provided electronics.

I'd like the circuit to be of general application though.

The KM84 is option 3 above, certainly simple not sure how transparent that is  - it all depends on the quality of the transformer, would anyone like to recommend a transformer for that circuit?

why go with that transformer circuit topology rather than one of the others I mentioned?

a couple more things:

a DC converter: can anyone recommend a circuit? or even a chip? I presume I will want to polarise the capsule at 60v+

I don't want a pad or HPF, but I'd be interested in recommendations about the best way to do the multipattern switching.

..simplest to use an external polarization source, like we do in tube mics..

Jakob E.

gyraf said:

..simplest to use an external polarization source, like we do in tube mics..

Jakob E.

Click to expand...

that might look simple on a schematic, but its not going to be simple to build, with an extra wire to the mic, and an external power supply.

I think the design should run on standard 48v phantom power.

A DC converter is not that complex. And the basic amp design could/should be usable with phantom power polarisation or higher polarisation from  a DC converter or whatever. Its really a separate bolt on extra.

I'd be happy to discuss if you want, since I have gone through the exercise with a commercial mic design.

Just some semi random comments to start:

1. fet with concertina balancing outputs, followed by a pair of transistors for current drive.

Click to expand...

Probably the most used circuit these days. Yes it's balanced from a signal polarity standpoint, but one must keep in mind that the
output impedance is not. On the source side Zout approx = 1/gm, on the drain side  Zout=Rd. Not a big deal with proper care though.

Bipolar output transistors have limited swing particularly if phantom powered (Schoeps circuit) and driving a transformer input pre.

Clean at low levels. Significant  low order harmonic distortion at high levels.

2. fet source follower, followed if necessary by a transistor for lower output Z (like oktava MK-012)

Click to expand...

Not balanced! Typically needs a transformer. Or "impedance balancing".

As one goes up (or down?) the food chain I separate things this way:

Resistively loaded single stage (like followers)...clean at low levels, significant distortion at high levels (a feature?)

Active loaded... a little more complex, potentially much lower high level THD

Fet input op amps with huge open loop gain... basically dead clean

This only scratched the surface... Tons of other things to think about.

FWIW, my latest design uses resistively loaded source and emitter followers, followed by a custom designed ransformer.

And finally the biggest issue in the design process for me:
The limitations of phantom power.

Les
L M Watts Technology

thanks for the response Les!

I'm personally not that bothered about distortion at high levels because I'm mostly recording acoustic instruments, and in any case I'd be more concerned about the character of the distortion than its absolute level.

I've had good results with OPA2604 opamps in other applications, but they draw too much current for phantom powering so i'm pretty much ruling out chip opamps here.

My experience suggests that the most musical result and even the subjectively most natural result has little to do with distortion levels in any absolute sense. Some artifacts are much better tolerated than others.

with an unbalanced source follower, - yes its unbalanced, but the output can still be pseudo balanced for noise rejection, quite few mics are done that way, though it is a reason to prefer option 1.

I'm shying away from a transformer since getting one of suitable quality is likely to be expensive, and trying out many different ones even more so, and i've yet to hear one that was really transparent. I have loads of valve pres with transformers in if I want to add that colour.

with the concertina topology the output Z will still be the same after the current drive transistors will it not? so as long as the input Z of the balanced pair of transistors is high enough there should be no problem?

for sake of argument, if I were to start with that topology how should it be optimised? how should the fet be biased? it seems not quite as easy to bias with a trimpot in this topology. I have 2SK170 and J305 fets on hand.

what is a good transistor to use for the next stage? I have BC549C and 2SC2910 & probably some other things, what about a fet for the 2nd stage too? I'm presuming that there will be a coupling cap each side at this point, or is there a way to DC couple?

Check the DC converter used in the Microtech Gefell micros:

http://www.twin-x.com/groupdiy/displayimage.php?pid=2731&fullsize=1

The autotransformer could be a obstacle though...

YW.

Thought i'd chime in, since I've been eat/sleep/breathing this stuff.

It's a mature technology, so the exercise is more a history review than modern analog design.

I like and use 2604 a lot, despite their distortion specs not being as advertised. But not in microphones.

My design is a small diaphragm condenser, specifically designed for acoustic string instruments like guitar. It should be a pretty good drum overhead too.
I'm going for a vintage colored  (high low order THD) sound, hence resistively loaded followers and transformer.

I'm shying away from a transformer since getting one of suitable quality is likely to be expensive,

Click to expand...

Yes. It's not so bad with a custom design in larger quantities though. And yes, you can "pseudo balance". That's what I meant by the term impedance balance in the previous post.

with the concertina topology the output Z will still be the same after the current drive transistors will it not? so as long as the input Z of the balanced pair of transistors is high enough there should be no problem?

Click to expand...

That's correct. Common mode output impedance will be about Zin/beta, very low. Not that such an impedance is best for CMRR. A transformer is of course very very high, although differential output impedance can be very low.

for sake of argument, if I were to start with that topology how should it be optimised? how should the fet be biased? it seems not quite as easy to bias with a trimpot in this topology. I have 2SK170 and J305 fets on hand.

I'd bias for equal max voltage swing on the source and drain sides. I use LSK170, similar to the old 2SK170. For single ended, I bias well below half supply though. These units have a very high Idss with some bins, so phantom current capability,noise, and voltage swing tradeoffs are made. For that reason, I do not direct couple to the output stage. It is biased differently, well above half supply. I do use BC549 for the output...it's fine. You could use a FET, but bipolar will have lower output impedance. I use a higher voltage 546
as a pass transistor in the voltage regulator though.

You can still bias the concertina with a 1 meg trimmer across the source resistor, with wiper to gate resistor. Be aware that this has a bootstrapping effect on that gate resistor. There are numerous example files in our yahoo micbuilder's list. I invite you to join.

One last thing that comes to mind is the noise budget...you might wonder about that.

Typically I find the largest noise sources (in decreasing order) are:
1. The transformed pressure noise of the capsule acoustic resistance
2.The 1/f noise of the gate ||capsule bias resistor
3.The johnson noise of the concertina drain resistor
4. The voltage EIN of the FET

Note the FET noise is way down on the list. So the very low noise of 2SK170 really gives no advantage. It's high transconductance might though.
Even a lowly TL072 FET op amp is not usually a big noise source with large diameter capsules. Not the biggest anyway.

Last thought: If I wanted lowest distortion and maximim dynamic range, I'd use op amps and an external supply for sure.

Les

leswatts said:

YW.

You can still bias the concertina with a 1 meg trimmer across the source resistor, with wiper to gate resistor. Be aware that this has a bootstrapping effect on that gate resistor. There are numerous example files in our yahoo micbuilder's list. I invite you to join.

Click to expand...

i'll have a look for that list.

do you mean wiper to the ground end of the 1G gate to ground resistor?

do you mean wiper to the ground end of the 1G gate to ground resistor?

Click to expand...

Exactly. So the gate resistor is no longer connected to ground. It's referenced to the divider between source and ground formed by the pot.

Les

is there a particular schematic I should look at as an example of this concertina topology?

I haven't seen this used in an externally polarised condenser, whats a good example?

In Yahoo micbuilders many examples of concertina phase splitter circuits are shown in the file zapnspark.
This includes both electret and conventionally polarized circuits as well as trimmer biasing.

Many many mics use this topology. Just not mine.

You'll have to join to see the files, but we'd be glad to have you. It's free of course.

Thie location of the schematics may change because just today we noted that our files section is full...no more memory.
We must move stuff around soon.

Les

I joined the group and looked at the schems.

Can this topology produce as good results as the others or is it a budget setup?

what is a high end mic using this topology and how does it differ from the basic shanghai mic style design?

those schems run the fet at only 10v or so, what's the optimum voltage?

Again, I'm not using that topology, but I'll have to say it's a very clever circuit. It was invented by Jo Wurtke at Schoeps, and is commonly referred to as the Schoeps circuit. It uses PNP bipolars for output, so the existing 6.81k phantom resistors can be used for degeneration in the emitter followers.


As far as the 10V....Junction FETs can have some unfortunate noise generation  if VDS exceeds about that value. Phantom power limitations are also a factor.

Yes, virtually every Shanghai mic uses it. For good reason. I don't consider it a compromise in any way. Not a budget setup at all. Some are considered implemented with cheap components though, and modders love to change them.

Does that mean it has the highest dynamic range and lowest distortion? Again, no.

Why am I not using it? My product is emulating a vintage sound. It actually has much more distortion. Distortion that people consider desireable "color".
Similar to a Neumann KM-84 or Sony C-38. When you combine the  cleanliness of a modern transformerless circuit with the accuracy of a modern Digital recording system the resulting sound can be too antiseptic for many. The hyped high end of many modern mics doesn't help either.



If you want a different sound, use op amps. Or active loads, cascodes, or White followers. That's the beauty of a knowledgable hobbyist making DIY circuits. You can have exactly what you want.

I strive to make what other people want most. Or what I think they want. If I make a mistake, I don't eat!

Les

greenmanhumming said:

I'd like to come up with a circuit for a multipattern LDC.

criteria:
* engaging natural detailed sound
* simple circuit with modest parts count
* flat frequency response, with possibility of adding some treble de-emphasis if needed.

so first question is what circuit topology?

options in my current order of preference:

1. fet with concertina balancing outputs, followed by a pair of transistors for current drive. I like the look of this topology as its balanced from the outset, but I have never seen it used in a proper condenser mic, only in electrets, does it have some terrible flaw?

2. fet source follower, followed if necessary by a transistor for lower output Z (like oktava MK-012)

3. fet voltage amp followed by transformer (like U87 or KM84)

can anyone offer arguments for and against each of these?

Click to expand...

Why re-create the wheel? 

http://recordist.com/ampex/schematics/neumann/u87sch.jpg
http://coreyeng.tripod.com/neumann.htm 
http://www.schematics-free.com/modules/PDdownloads/viewcat.php?cid=13

New, genuine Neumann capsules appear on Ebay fairly frequently (e.g, http://cgi.ebay.com/NEUMANN-K67-K870-CAPSULE-MICS-U67-M269-U87Ai-SM69-/110578374156?pt=LH_DefaultDomain_0&hash=item19befc160c).  Get a capsule, build the circuit, insert in your Rode stick, and enjoy.

DY

damnyankee, thanks, I should have been more clear....

I'm not seeking to recreate a neumann.  I probably shouldn't have used the word design in my original title, maybe "select" and "optimise" are better words.

Available parts have changed a lot since the classic neumanns were designed, there are options now they didn't have.

I'm looking for a good basic circuit topology I can use in a variety of mics,that can be implemented in a basic way in a tight space like a pencil mic, and expanded on where there is more space.

I'm also looking for discussion of the advantages or disadvantages of the different topologies, and parts recommendations for best performance in each. If there's a thread somewhere that covers this ground well please direct me to it, because I can't find it!

leswatts said:

Again, I'm not using that topology, but I'll have to say it's a very clever circuit. It was invented by Jo Wurtke at Schoeps, and is commonly referred to as the Schoeps circuit. It uses PNP bipolars for output, so the existing 6.81k phantom resistors can be used for degeneration in the emitter followers.
Les

Click to expand...

what happens if the phantom resistors are a different value?

whats the best PNP to use there? I have been looking for the Jim Williams recommended 2SA1084, but I can't find it for sale in or accessible to the UK.

It would work over a range I suppose. 6.81K is hardly ideal for a degeneration resistor, but it's there. Throw it in SPICE and have a look at different values.

I use 1K. I have to drive a transformer... a heavy load at low frequencies when the self inductance reactance poops out.
I'm using as much phantom current as I dare.

As far as transistors, 2n5087 is popular in the Scheops circuits.

Took a glance at 2SA1084 and it indeed looks excellent, especially in the high beta bin. At high collector currents (1ma or more)
it would work well with low impedance sources where current noise is not an issue. It's more of something I would consider for an active ribbon mic, or an old stype moving coil phono cartridge head amp.

To be fair the sub 1nV/sqrt Hz voltage noise is far below other noise sources, so would offer little sonic benefit. Same goes for the
5087 at about 2 nV/sqrt Hz.

It's a compromise in the Schoeps circuit because the source side of the fet can have an output impedance of as little as 50 ohms, but the drain side might be several K ohms. So a little voltage noise/current noise tradoffs. I haven't really analyzed an optimum, mostly because i'm not using that topology. Of course one could try to set up a FET where 1/gm=Rd. (gm is a linear function of Vgs, and varies hugely from one device to the next, even of the same part number) Doubt if that would offer any real advantage though.

The real way to drastically reduce noise is to reduce the capsule acoustic resistance and filter the resultant response peaks with the electronics.
As high as practicable gate and capsule bias resistor helps too. Suprisingly noise goes DOWN as those resistances go UP. Things can get shaky above a few gig though. Then there are charge amps and RF...whole other subject.

I guess i'm saying a lot of the semiconductor noise stuff is nitpicky compared to the HUGE noise from the capsule transformed acoustic resistance.

I'm off today evaluating Nickel plating of mic bodies...wish me luck!

Les