Oktava MK-319 output level increase.

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

vmanj

Well-known member
Joined
Oct 3, 2017
Messages
297
I welcome everyone.

I have an Oktava MK-319.
I increased the microphone output level, as indicated in the image.
But I have doubts that this disrupts the bias of the transistor, or is everything all right ?
 

Attachments

  • OktavaMK319 HO.jpg
    OktavaMK319 HO.jpg
    50.5 KB · Views: 99
The DC bias is "only" set by the resistors; that 47uF capacitor increases the AC gain.

https://www.electronics-tutorials.ws/amplifier/emitter-resistance.html

By moving the connection point of that capacitor, in addition to increasing the overall gain, you will have moved the cut-off frequency of that resulting (high-pass) filter about an octave lower.
 
Thanks for the answer.

If I understand correctly, do I need to change (increase) the value of the capacitor C3 to lower the cutoff frequency ?
 
No, that already happened due to it being in parallel with both of the source resistors, not just one.

https://electronics.stackexchange.com/questions/140553/how-do-i-calculate-the-cutoff-frequency-of-a-low-pass-rc-circuit

Double the resistance => halve the frequency.

If you want to keep the same low-frequency cut-off point as before, you'll want to halve the capacitor value (down to 22uF).
 
Unfortunately making the suggested modifications would indeed change the frequency response.

The RC combination in question before modification is ((1/gm) +R7 ) C3,
where gm is the transconductance of the FET.

After modification we have (1/gm) C3
That is we have decreased the time constant thus moving the pole (or cutoff frequency) up.

Let's say gm is 0.001 s, then the cutoff frequency will have gone from 1.2 to 3.4 Hz

Say gm is 0.01, then it's 1.8 to 33.7 Hz (starting to get audible).

Note that since you have reduced degeneration (as they call it in the lingo) by 1.8 kOhm, you can work out gm approximately from the change in passband gain like:
(1/gm + 1.8k)/(1/gm) = GR
gm = (GR - 1) / 1.8k

Can you share what FET you are using?
Do you have an idea of the capacitance of the capsule? What size is it?

Edit: Couldn't find the exact details, but safe to say the numbers discussed are ballpark correct. It is safe to increase C3 to say 470 uF. You can maybe find a cap in the same form factor, this is a rather old design.

Still interested in knowing the FET. Low gate capacitance is a must in this topology so choice is limited. I did see mention of a 2sk170 mod which flies in the face of that notion. Maybe people enjoy the increased distortion from capsule loading. Modding is fun!

Edit edit: Forgot to specify:
GR is gain ratio ie. GR = (gain after mod) / (gain before mod)
GR > 1
 
I've replaced a couple of times FETs in this Oktavas with the 2sk170 (just because I had only them on my table, there was no another particular reason) and this operation significantly decreased noise. The mics had different Soviet/Russian FETs originally,  Octava's FET choice is always a Russian Roulette.

BTW I've read that the capsule was designed for a tube circuit and 60V,  hence it doesn't work properly in the 219/319 and needs at least a DC/DC converter for p48. A couple of tube versions of this mic were in production but I've never had an opportunity to check them out.
 
A 170 is not a good jfet to use in that circuit.

Also note that there should be numbers written on the transformer
Look close at the biasing it is a mix of voltage divider at the gate and source biasing. with resistors selected for the jfet
 
Yes, I know that it has to be rebiased.

Could you recommend something for this circuit instead? Yes, Toshiba's capacitance is a problem but it's better than old Russian FETs anyway.

It's interesting what they put in it now. As I see they started to care about their pro audio division a bit more nowadays and "just put this random part there" era is over.
 
The problem with excessive gate capacitance is not actually loss of signal (since a capacitive voltage divider is noise-less and gain is cheap it's no huge deal) but distortion generated from loading the capsule with "deadweight" capacitance.

You could try something like a 2sk209 or something else with modest capacitance. Or pick a different topology that doesn't seem to have such toob envy  ::)...
 
Thank you. I've checked my local supplier's prices, 2sk117 and 2sk246 are available and look promising as well.

I suspect that the primary FET for this design was the КП364, 6pF.  They used it in the 012, it's the 2sk170bl now. But there is a report of factory-installed 2sk163 in the 319 that has a huge capacitance.

Replacing the entire circuit is always what I want to do with this mics when I see them  :) And the ringing body. So basically it is easier to install its capsule into a modified Chinese donor.
 
VictorQ said:
Unfortunately making the suggested modifications would indeed change the frequency response.

The RC combination in question before modification is ((1/gm) +R7 ) C3,
where gm is the transconductance of the FET.

After modification we have (1/gm) C3
Actually most of your analysis is wrong, in particular the formulae that ignore the interaction between R7, R8 and gm. Indeed the cut-off frequency changes, but not in the way you describe. The original circuit has a rather flat LF response because the gain outside LF is not great, about 18dB. ther is not much difference between LF and HF because the capacitor acts on half of the source resistor, so the gain variation cannot be more than 6dB. In addition there is some selective feedback that boosts LF.
Now, after modification, indeed the mid and HF gain increases significantly up to about 30dB; now the difference between LF and Hf gain is significant.

Actually, the -3dB point changes from an impossibly low DC to 3 Hz.
BTW Gm does not change because the op point is the same.
 

Attachments

  • oktava LF resp.jpg
    oktava LF resp.jpg
    35.9 KB · Views: 54
VictorQ said:
The problem with excessive gate capacitance is not actually loss of signal (since a capacitive voltage divider is noise-less and gain is cheap
Isn't it a bit overoptimisic? Any level loss at the input results in loss of S/N ratio, since the FET's noise is constant.

  but distortion generated from loading the capsule with "deadweight" capacitance.
Agreed, but distortion resulting from modulated Cgs is even more significant.
With the source fully decoupled, as per mod, modulation of Cgs is directly in parallels with the capsule. With the original configuration, Cgs is bootstrapped by the moving source.

Regarding "deadweight", note that the -10dB switch and the HPF add their share. Who cares? Many mics use a "charge amp" topology, that loads the capsule with just about a shiort-circuit...
 
Ok so by interactions you mean this(?):

((1/gm + R7) || R8 ) C3

to

((1/gm) || (R7 + R8)) C3


... right?

That means (assuming gm = 1 ms) lower cutoff goes from around 3 to around 4 Hz. All fine.
I don't see where your "impossibly low" figure comes from.. What are you using for gm?

I doubt that the small boost at LF is intentional. Seems like C7 should've just been "large" and 1 uF was already there ( note the excessive 160 V rating).

Regarding cheap gain: Yes the most important parameter to determine the self-noise of a mic is the sensitivity. Pretty much all competent head amps have EIN < 1 uV A-weighted (not this one though). It is indeed overly optimistic to not fret about wasting gain.

But who cares about microphone distortion indeed? Some people seem to like it even. You can't claim that the Neumann U89 for instance (which uses a charge amp) sounds severely distorted. It definitely does impart a 'color', though, and that's not ideal for everything. Sometimes you need a sharper knife. Sometimes you want to record a tambourine without having to remove sub-sonic crud after. But I digress.

I agree that Cgs modulation is worrisome as well. I strongly prefer to operate the input device in constant current and constant voltage. Bootstrap everything that can be. Like the C414 (which I find to be more versatile for sure than aforementioned U89 for one).

Anyway this circuit is not really worth spending too much time on. Oktava has other designs that are better. If the goal is a quiet mic without much color I think the first step is to start over.



Edit: BTW thanks for responding, Abbey! Getting some actual SPICE sims going tells me that the behaviour of this circuit is more complex than I thought.
 
VictorQ said:
... right?
Yes.

I don't see where your "impossibly low" figure comes from..
Just observe the frequency response. The VLF response is almost equal to the midrange response, and extends so down to DC. There is no -3dB point, so the response is actually DC. The capsule and coupling capacitors determine the actual overall -3dB point, not the source decoupling cap.

  What are you using for gm? 
I used the 2N3819 model in LTspice, that has Gm more or less equal to 2mS.

[/quote] I doubt that the small boost at LF is intentional.[/quote] I doubt it too. Nevertheless, it is there and contributes to whatever sonic signature this mic has.

But who cares about microphone distortion indeed? Some people seem to like it even. You can't claim that the Neumann U89 for instance (which uses a charge amp) sounds severely distorted. It definitely does impart a 'color', though, and that's not ideal for everything. Sometimes you need a sharper knife. Sometimes you want to record a tambourine without having to remove sub-sonic crud after. But I digress.
You raised the subject, rightfully. It's a can o' worms. Who am I to criticize Neumann on the basis of a mere physics formula? Anyway, by all accounts, the other distortion mechanisms (mechanical as well as electrical) seem to be dominant.

Anyway this circuit is not really worth spending too much time on. Oktava has other designs that are better. If the goal is a quiet mic without much color I think the first step is to start over.
That's the case with most inexpensive condenser mics; when one single FET can do the job, why complicate things for performance most buyers won't appreciate...
 
Greetings to all.
I removed the R3 - 200k resistor from the circuit.
I liked the sound better than with the resistor.
What does the R3 resistor affect in this circuit?
 

Attachments

  • OktavaMK319 HO 02.jpg
    OktavaMK319 HO 02.jpg
    50.5 KB · Views: 44
Last edited:
Greetings to all.
I removed the R3 - 200k resistor from the circuit.
I liked the sound better than with the resistor.
What does the R3 resistor affect in this circuit?
No, it's not NFB. At least, not AC NFB, but actually DC NFB.
It changes significantly the FET bias.
It actually helps stabilizing bias.
You should measure the DC voltage at the source and drain of the FET with and without it.
 
Last edited:
Why do you want to increase the output level?

Next look up mixed bias the microphone uses source and voltage divider biasing like I posted before in this thread

This allows a wider spread of JFET specs to be used

Vishay had an AN102 app on the web you might want to look for to understand the circuit. A google search will find it.

IMO the stock circuit is well optimized for the amount of parts and the parts selected for the circuit.

If you want more from this circuit you might want to build a different circuit
 

Latest posts

Back
Top