mosfet limiter ?

[diggy fresh]

Hi all,

I got a few Mccurdy's ap-276 cards from a dumpster that i intend to re-use as sonic crayons.

They were intended as talkback amps, basicaly a simple preamp with a simple "agc" limiter.

Mccurdy usualy has clever designs, this looks like late 70s.

What i was intrigued about is the fact that the limiter uses a Mosfet instead of a jfet, i'm still green but i never saw that, and it looks interesting so i thought i'd share this with y'all. I am clueless as to how this works.


If anyone sees a way to add variable attack and release controls that would be very cool.


here's a link to the mosfet datasheet:
https://www.datasheet4u.com/mobile/1128417/MFE3003.html

 

[JohnRoberts]

I dislike schematics where I have to guess which lines are connected... especially for designs I do not completely understand. I have designed multiple shunt limiters using JFET but never with a MOSFET.

It is unclear if that design is using single gate mosfet or dual gate mosfet.

The distortion trim looks about right for a JFET active device (nominally injecting the drain voltage into the gate -6dB) but if it is actually a single gate device that distortion cancellation is shunted with 100uF making it ineffective, which suggests a dual gate MOSFET.

Speaking in broad strokes that gain stage is constantly running something like 200x (attenuation occurs before the gain stage so it is always adding noise).  The op amps look like old school (maybe LM301/748??), so not very quiet.

As an effect perhaps, but looks more like a white noise generator, than limiter...

JR

 

[diggy fresh]

Thanks John for taking a stab at it!


additional info:

input opamp is :  mc1741

limiter opamp is : mc1709

and the balanced out is done by a
"dual matched opamp"  mc1437
So yea, all ancient.


The original mosfet is mfe3003 but a service note states it was replaced by a 3n163 mosfet at some point if that gives any clue.  I barely know anything about mosfets but both datasheets don't mention anything about them being "dual gate".


Specs in the service manual states that. : "equivalent input noise (dynamic microphone) is better than -115dbm.


The 3p3t input switch seems to switch in/out a pad for mic/line and also alter the feedback/gain of the first stage, and also adapting the input impedance (i think).  The second position is adapted for carbon mics.


I also read in the manual that the freq response is +/-2db - 100hz -10khz. They probably limited it to minimize distortion and noise.

So yea probably i'll use it as an effect.
I originaly wanted to guts for parts because it uses a nice little Beyer "peanut" transformer but i might as well try it. i would feel a bit shamefull to just destroy them.

But why use a mosfet instead of a jfet?  I don't understand how the limiter work myself, does it work like a normal jfet comp?  or the mosfet does something like hard clipping?

 

[JohnRoberts]

Thanks John for taking a stab at it!


additional info:

input opamp is :  mc1741

limiter opamp is : mc1709

and the balanced out is done by a
"dual matched opamp"  mc1437
So yea, all ancient.


The original mosfet is mfe3003 but a service note states it was replaced by a 3n163 mosfet at some point if that gives any clue.  I barely know anything about mosfets but both datasheets don't mention anything about them being "dual gate".

data sheet for latter part shows 4 leads, with the case /substrate connected separate from the gate.

The distortion trim (AC signal) appears connected to this substrate, separately from the gate which appears to be a DC voltage with 100uF cap smoothing. 

Specs in the service manual states that. : "equivalent input noise (dynamic microphone) is better than -115dbm.


The 3p3t input switch seems to switch in/out a pad for mic/line and also alter the feedback/gain of the first stage, and also adapting the input impedance (i think).  The second position is adapted for carbon mics.


I also read in the manual that the freq response is +/-2db - 100hz -10khz. They probably limited it to minimize distortion and noise.

So yea probably i'll use it as an effect.
I originaly wanted to guts for parts because it uses a nice little Beyer "peanut" transformer but i might as well try it. i would feel a bit shamefull to just destroy them.

But why use a mosfet instead of a jfet?  I don't understand how the limiter work myself, does it work like a normal jfet comp?  or the mosfet does something like hard clipping?

Taking a WAG (wild as_ guess) perhaps there is a distortion lowering mechanism from injecting that -6dB AC voltage onto the substrate.

I never used a MOSFET this way, or heard of the distortion lowering trick used on MOSFETs (only saw it on JFETs).

Still guessing about the actual nodes connected or skips not indicated in the poorly detailed schematic (not to mention old hand drawn schematics often contain mistakes).

AFAIK JFETs were lower Rds(on) so better parts for this kind of application.

Perhaps search for very old MOSFET application notes describing such a distortion cancelling trick... or not. . 

JR

 

[diggy fresh]

Thanks for your insight John!

Yes the schematic is kinda drawn confusing with all the limes crossing and grounds everywhere..  Just for the sake of curiosity i will try to redraw it and compare it with the board.


Another interesting detail, the transformer is a "studio grade" Beyer "peanut".

br 351 110 005
1:10
200ohm/20kohm
50-15khz

Seems overkill for a talkback amp.


I will redraw and update this thread when i'm done.

 

[diggy fresh]

As promised, here's a cleaned up schematic.

I'm curious why and how they used a 4 pins mosfet, for educational purpose. Anybody has an idea?

It seems like the negative portion of the signal at the output of ic2 is controlling the gate of the P-channel Mosfet. (?)  And then shunting the input signal?

Is c11 at the gate of the mosfet responsible for the release time?

And what about attack time? Is is dependent on the mosfet?

I can't wait to see how it sounds, this looks like a very simple circuit to diy if it can be improved. Cool little lo-fi effect maybe..
How about bringing the Distortion trimmer to a pot for a dist control?    Would the mosfet produce 2nd harmonics in this context?

Thanks!

 

[PRR]

> diy if it can be improved

Why do you think it needs "improving"??

It may be "OK" as it is. It is fundamentally so simple that if it is NOT satisfactory, it will take a major re-think to add the feature-items you wish for.

Most of the people who once knew dual-gate or 4-pin MOSFETs have quit the biz and gone fishing. I glanced once, and except for lo-cost RF preamps and mixers, felt there must be some better way.

 

[abbey road d enfer]

Is c11 at the gate of the mosfet responsible for the release time?

Discharge time-constant of C11 into R28 and portion of R33.

And what about attack time?

Charge time-constant of C11, mainly governed by current capability of IC2.

  Is is dependent on the mosfet? 

No.

Would the mosfet produce 2nd harmonics in this context?

Predominantly 3rd.

 

[diggy fresh]

Thanks guys!

I'll try it out in the next few days to see if it's worth using as an effect.

 

[VictorQ]

According to Johns & Martin (Analog Integrated Circuit Design) the body effect transposes the effective threshold voltage by a voltage (marginally) proportional to the square root (transposed to a rather flat part of the square root curve so as to not be TOO nonlinear) of the source-body voltage.

In the triode region (where this FET is being operated) the drain-source conductance is of course proportional to the negative of the drain-source voltage as well the negative of the threshold voltage. This dependence of the drain-source voltage is what generates third order harmonic distortion. The symmetrical dependence on the threshold voltage points the way to this distortion nulling mechanism: Through the body effect a signal dependent shift in threshold voltage in induced which aims to cancel out the dependence on drain-source voltage of the triode conductance.

The amount of signal required to null optimally depends on a bunch of process parameters like oxide capacitance and doping rates so is likely just dialed in by measuring.

Note that the usual -6 dB added to the gate-source voltage would also work here (and probably be easier to set and not require trimming). I'd guess that the substrate connection was a convenient way to trim the FET to work here directly using signal level as control voltage. It does make for an elegantly simple side chain. Breaking the distortion trimming around to there allowed them to simplify even more. Just my guess. Very likely no-one lost much sleep over this circuit.

Edit: Just to expand on the last point: this is a non-critical circuit in a non-critical part of a non-critical machine. The reason for doing things like this could easily be "because I was bored and wanted to see if I could make it work that way".

Edit2: Corrected "second order" to "third order"

 

[PRR]

> According to Johns & Martin... ... ... ... a convenient way to trim the FET....

What I understand of that looks right.

> ....this is a non-critical circuit in a non-critical part of a non-critical machine. The reason for doing things like this could easily be "because I was bored and wanted to see if I could make it work that way".

Quite true. Also, in those days, a console company ordering thousands of transistors and chips, their suppliers would be eager to supply samples of odd transistors. Even with no letterhead I had a box full of samples. I'd guess in a real operation the Designer's cubby could be overflowing with odd parts. Maybe just enough MOSFETs to do a single minor uncritical part of every console.

 

[VictorQ]

For anyone interested in the gory details, here are the relevant equations reproduced for your analyzing pleasure!

Edit: I obviously messed up trying to make the formula n/p agnostic... Please imagine hole mobility µ_p instead of electron mobility µ_n and you have the equations for a p-channel device. Not that it makes a difference for the analysis.

 

[JohnRoberts]

> According to Johns & Martin... ... ... ... a convenient way to trim the FET....

What I understand of that looks right.

> ....this is a non-critical circuit in a non-critical part of a non-critical machine. The reason for doing things like this could easily be "because I was bored and wanted to see if I could make it work that way".

Console designers, even for small companies learn to avoid obscure, exotic parts because lack of that one part could prevent shipping a multi-thousand dollar (EU) product (not to mention future service calls/repairs).

I agree there is some attraction to novelty after the step and repeat design repetition of console work.

For want of a nail the shoe was lost, for want of a shoe the horse was lost, for want of a horse the battle was lost.

Quite true. Also, in those days, a console company ordering thousands of transistors and chips, their suppliers would be eager to supply samples of odd transistors. Even with no letterhead I had a box full of samples. I'd guess in a real operation the Designer's cubby could be overflowing with odd parts. Maybe just enough MOSFETs to do a single minor uncritical part of every console.

I noticed a huge drop off in rep/free sample interest after I left Peavey...  :

Even after a designed a micro-based meter for a friends console company the microprocessor maker, gave them the free stuff and rep visits, not me...

JR

 

[PRR]

> avoid obscure, exotic parts because lack of that one part

Yes. But stuff happens. I've been getting samples. Sometimes I think it is a 6-pack, but they break the pack and send just one. Another time it was a pack of 8.... they shipped SIX 8-packs. A couple of times I have received a second sample, told them of the mistake, and they said keep it.

So say the designer wished for 10 transistors, and the salesperson and shipping clerk fumbled that to 10 boxes, 1,000 parts, and were indifferent about the mistake. With 1,000 parts in-hand and underfoot, a console maker doing 50 consoles a year with one talkback module each, and knowing that few designs last a decade, has a "lifetime supply" already in-house. I'd run with that.

 

[JohnRoberts]

> avoid obscure, exotic parts because lack of that one part

Yes. But stuff happens. I've been getting samples. Sometimes I think it is a 6-pack, but they break the pack and send just one. Another time it was a pack of 8.... they shipped SIX 8-packs. A couple of times I have received a second sample, told them of the mistake, and they said keep it.

So say the designer wished for 10 transistors, and the salesperson and shipping clerk fumbled that to 10 boxes, 1,000 parts, and were indifferent about the mistake. With 1,000 parts in-hand and underfoot, a console maker doing 50 consoles a year with one talkback module each, and knowing that few designs last a decade, has a "lifetime supply" already in-house. I'd run with that.

I wouldn't and I didn't.

I even had issues with a common function generator chip... (8038 ? or something like that.. that disappeared unexpectedly.) One per master section, but no chip, no master section, no master section no console, no console no money..

Good luck with your console company.

JR

 

[diggy fresh]

thank you all for having a look at this odd thing.

Especialy VictorQ, very interesting!

These cards were actualy part of a large modular intercom and talkback system for broadcast use.  around 1975.

Mccurdy was mostly in that market in Canada for state networks. Cbs etc.

here's the full manual for curious peeps:

https://www.google.com/url?sa=t&source=web&rct=j&url=http://www.steampoweredradio.com/pdf/mccurdy/manuals/McCurdy%2520CS9100%252010%2520by%252010%2520Intercom%2520System%2520Manual.pdf&ved=2ahUKEwi-r-7Lg8nkAhUlT98KHfScDyYQFjAFegQICBAB&usg=AOvVaw3LkoS3jDfwVibiHF89V13M


Probably a very expensive system with all cmos crosspoint matrix switching with around 20 2n5460 jfet per cards..


I understand for most of you this is just junk but i am from a younger generation that never was expose to these things.  And i'm a type of guy that like to creatively use junk for effects. Also cause i'm poor haha..

I will have time to test it tonight to see how it sounds.

I'll try to experiment with the time constant for a release control.
Very interesting how the sidechain is so simple.