Graeme Cohen "Double-Balanced" mic amp.

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

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

Jean Clochet

Well-known member
Joined
Jun 25, 2011
Messages
387
Location
Currently taking a break to smell roses and the co
Not sure if I should have really posted this in The Brewery?  If someone could move it if needs be, that'd be cool.  Thanks.

Anyway, I've been working on a new low noise mic amp and, while the thing I'm doing isn't related to the double-balanced topology, I found a page with quite a bit of info on Graeme Cohen's work, including some notes from Graeme along with a copy of the original white paper on the Double-Balanced concept.
Just in case anyone hasn't seen it before, here's the link:

http://leonaudio.biz/cohen.htm

If this stuff has already been shown here then I apologise,

Cheers,

J.C.
 
This is the quietest practical mike preamp in the known universe.  Millenia Media use this circuit as does probably Earthworks.  The Earthworks is the quietest practical mike preamp you can buy.

There are several (DIY or commercial) copies & kits including the SSL9K but none of them even approach its performance.

On the same website, is Great Guru Baxandall's last thoughts on LN design in Microphone Engineering Handbook.
 
ricardo said:
...There are several (DIY or commercial) copies & kits including the SSL9K but none of them even approach its performance.

Yep.  Although one of the well respected members and moderators here - Samuel Groner - has some VERY nice examples on his site here:

http://www.sg-acoustics.ch/analogue_audio/microphone_preamplifiers/index.html

I haven't built any of his circuits myself but I can't see any reason why they wouldn't give absolutely superlative performance!


ricardo said:
On the same website, is Great Guru Baxandall's last thoughts on LN design in Microphone Design Handbook.


Yes, that's also well worth the read.  Peter was a unique human - a massive wealth of knowledge that was always unselfishly shared with any who cared to ask.
I wonder what the £ figure total would have been if there was a penny given by everyone who made use of his original EQ topology?

Not to mention his research on distortion mechanisms wrt feedback, the "super-pair", the active volume control ...  on and on...

Anyway, Graeme Cohen's circuit is explained well on the posted site for anyone who was curious.

Thanks Ricardo,

J.C.
 
Chapter 8 of the Microphone Engineering Handbook, Baxandall on mic preamps and transformers, is the best for a one stop shop on mic preamp noise.
 
Yup, that got discussed back a few years ago.

When I first started participating here, I was confused when people mentioned the "Cohen" topology, since I had seen it around and was using it in consoles well before his AES paper publication, so from my perspective he was a later adopter. FWIW he doesn't claim to have invented it.

I talked about the benefit of the topology's open and closed loop gain tracking each other in a discussion of Paul Buff's Transamp in a 1980 console article I wrote, and even published a kit/design article in early '80s in popular electronics using a variant on that topology for a moving coil phono preamp front end. 

Yes it is good, and well known to those skilled in low noise design.  Of you look inside most modern mic preamp ICs you will see variants on that topology. While it was exotic 30-35 years ago, today not so much thanks to these popular mic pre ICs. 

JR
 
Good read, the lot of it.

In the cohen pdf there are pictures of the circuit before potting it in epoxy. Were opamps and transistors available "open" like that, so you had to solder the bond wires and could actually see the chip internals? Or is that some kind of a special brew?
 
Kingston said:
Good read, the lot of it.

In the cohen pdf there are pictures of the circuit before potting it in epoxy. Were opamps and transistors available "open" like that, so you had to solder the bond wires and could actually see the chip internals? Or is that some kind of a special brew?

IIRC he was working on developing an IC for phillips.

The LM394 is a relatively large transistor die to get the Rbb small for low noise.

Back in the day it was not unusual (but quite expensive) to make hybrids by throwing a few IC or discrete devices on a common substrate and wire them together. The same wire attachment system is used to connect package pins to large IC silicon chips.

If you open up an old metal case transistor or IC the chip will usually be sitting open like that with the bonding wires visible. When I was a young technician, one of my tasks was to grind the top off of failed TO-3 power transistors, so we could literally look at the device to determine failure mode. You could see a melted base wires from over current, or a small region of melted silicon from local heating after an over voltage punch through and short CE. 

JR
 
JohnRoberts said:
Yup, that got discussed back a few years ago.

Yes, I thought it probably had been but figured I'd chance it just in case  ;D



W
JohnRoberts said:
When I first started participating here, I was confused when people mentioned the "Cohen" topology, since I had seen it around and was using it in consoles well before his AES paper publication, so from my perspective he was a later adopter. FWIW he doesn't claim to have invented it.


That old "Shaker" of "Spears" said it quite eloquently:

"If there be nothing new, but that which is
Hath been before, how are our brains beguiled,
Which, labouring for invention, bear amiss
The second burden of a former child."


Another example that comes to mind of reinventing older wheels being Deane Jensen's patent pertaining to the use of inductors in // with the degeneration R's in the emitters of the input diff. pair.  Which, as we know, provides low Z at audio frequencies for the noise advantage while the resistors take over above audio for stability reasons.
There is clear prior use/art of this technique dating from quite some time before Deane's application - On page 36 of the PDF link below, we see an Analog Devices amp from 1966 with this clearly shown:

http://www.analog.com/library/analogDialogue/archives/39-05/Web_ChH_final.pdf


Anyway, thanks John.  Do you happen to know of any particular early examples prior to the "so-called" Cohen that you could point to online?  Just for idle academic curiosity's sake. 

MJK & Kingston:  Yes it's all a ripping good read  :)

J.C.

 
When LEDs are used like this as a fixed voltage source, is there any special property of RED leds that makes them more desirable to use than other color LEDs or is it just a matter of looking for the right voltage drop across the LED and red just happens to be a convenient 2V(ish)?
 
Lowest noise. There was an interesting thread somewhere on DIYAudio where someone measured all kinds of diodes for different currents.
 
Jean Clochet said:
Anyway, thanks John.  Do you happen to know of any particular early examples prior to the "so-called" Cohen that you could point to online?  Just for idle academic curiosity's sake. 

MJK & Kingston:  Yes it's all a ripping good read  :)

J.C.

I've posted my old phono preamps (there were more than one of them)
P10.gif
  note: this was the MM version, the MC version doesn't have a unique schematic but used discrete PNP devices and flipped the polarity around.  IIRC  pub 1981

http://www.circularscience.com/des_art.pdf  figures 10 and 13 discuss transamp in broad strokes, while it was not an electronics magazine, more of a general discussion.  IIRC pub 1980

Wayne's forum has a number of cites from his research into earlier examples of this topology.
http://www.proaudiodesignforum.com/forum/php/viewtopic.php?f=6&t=164

http://www.proaudiodesignforum.com/forum/php/viewtopic.php?f=6&t=107

I thought there was more discussion about early cites but can't find it now...

JR
 
Ethan said:
When LEDs are used like this as a fixed voltage source, is there any special property of RED leds that makes them more desirable to use than other color LEDs or is it just a matter of looking for the right voltage drop across the LED and red just happens to be a convenient 2V(ish)?

Red LED's just look cool, sod the technical specifications!  8)

Seriously, as Volker said, there is an optimum current to turn on the LED at which they generate less noise.  Same deal with other diodes such as a zener.  Although a zener would generally be quite a bit noisier regardless of current.
I don't remember all the details for each type of diode but 5mA is about where it's at if memory serves.  There is also an impedance difference and changes to the voltage drop with the changes in current.

I suspect though that Mr Cohen may have specified RED due to requiring the lesser of voltage drops from the supply rail, most other types of LED's having greater drops.  Could be wrong on that though?  Maybe he just happened to have a junk draw full of REDs on hand?
If we wished, we could fashion an even quieter, very stable reference by other means but, as I think I may have mentioned earlier - LED's look cool!  :)
 
JohnRoberts said:
I've posted my old phono preamps (there were more than one of them)

Thanks John.  That's a nice design. 
Yes it's of the same ilk as the "Cohen". 
What I see at a glance (my eyes are not so good so I'll need to save it and blow it up in Photoshop) is that, whereas your circuit leaves each transistor/Op-Amp compound stage via R11  and R13 and then converted to a single-ended signal in the following diff amp, Cohen's is fed to two seperate diff-amps, of opposing polarity to each.  I guess that's how the "Double-Balanced" moniker arose?
Still, a nice design you have there, I hadn't seen it before.

Thanks for the other links, I'll look them through as see what I can find.

Cheers,

J.C.



 
If we go back far enough red LEDs were the only color LEDs and surely the cheapest after other colors were developed. The early red ones were expensive...

JR
 
JohnRoberts said:
Yes it is good, and well known to those skilled in low noise design.  Of you look inside most modern mic preamp ICs you will see variants on that topology. While it was exotic 30-35 years ago, today not so much thanks to these popular mic pre ICs.
In my book, a "proper" Cohen isn't about the topology.

It's about sub nV/rtHz noise performance.

Only Cohen and the Earthworks has this for a fully working & protected P48V mike preamp.  Millenia Media "only" achieves 1nV/rtHz.  As does Sam in Monte Generoso.  I was a bit contemptuous of Millenia Media until I did a little paper exercise on how to achieve this performance.  Now I take my hat off to both John LeGrou & Sam.

But to be pedantic, Cohen's circuit has many subtleties that differentiate it from "near" copies like the excellent THAT 1510 family.  Apart from actually achieving sub nV/rtHz performance, it's noise performance also deteriorates less at lesser gains.

The topology dates back to Demrow in the 60's IIRC but Cohen's implementation is the one that really breaks the 1nV/rtHz barrier for a real life P48V mike preamp.

Can someone explain the reason for the 2 x LED's in his circuit?
 
Hi Ricardo,

ricardo said:
In my book, a "proper" Cohen isn't about the topology.


To a certain extent I agree.  But the double balancing aspect of the Cohen does certainly improve CMMR.  Although, like other similar topology's, CMMR does worsen as the emitter to emitter resistance is increased  when gain is lowered.

ricardo said:
It's about sub nV/rtHz noise performance.

That's certainly a mighty fine specification but I actually don't know that's it's absolutely necessary to use the Cohen topology or even a super matched input pair to achieve those kinds of noise figures.  It's possible in my experience to manage this  sub nV/rtHz figure by other means. 
Other's mileage may differ of course and this is only my own experience.

But, we do need to look at using low internal base resistance (Rbe) input transistors and we also usually opt for the PNP variety as well. 
There were some pretty good 2SD devices from Toshiba available in the past that worked nicely but, sadly, they are now unobtainium.  The old standby these days is the 2N4403 (Rbe of 10?) of which we can even use 2, 3 or 4 of the buggers in //.  Doug Self, in one of his low noise phono designs found that using 3 X 2N4403's in // did not require any current sharing resistors in their emitters for adequate current sharing to occur.  Of course, if we did need these resistors, our noise would have been higher  *
Operating the input devices at an optimum current for the lowish source Z of a microphone is important too of course.

Regarding Rbe:  Linsley-hood demonstrated that some mighty fine noise specs can be had by using just a single pair of medium power input devices rather than the small signal TO-92 type.  An Rbe of 4 is not uncommon for some of these medium beefy globs of silicon.


ricardo said:
it's noise performance also deteriorates less at lesser gains.

Yes.  The SNR issue at low gains is always one of the problems you have to wrestle with .  I can't say I've measured this wrt the Cohen. 


ricardo said:
The topology dates back to Demrow in the 60's IIRC but Cohen's implementation is the one that really breaks the 1nV/rtHz barrier for a real life P48V mike preamp.

Hey thanks!  I didn't know about Demrow, I'll look for a reference  :)


ricardo said:
Can someone explain the reason for the 2 x LED's in his circuit?


In his circuit, they ensure that the Op-Amps he used (5534's?) don't get too close to the rail.  The LED's reference V keeps the amps in their linear territory.

All of this is just IMHO of course.  I'm still learning as I go  :)
And you probably know all and more of this stuff anyway  ;)

Cheers


* Edit:  Of course all of our impedances need to be as low as practical as well -  and the higher device standing current we use for low noise with low source Z helps with this.
 
One thing I meant to say:
A good friend and designer recently came up with an input stage topology that is a basically a compound triplet of transistors - no Op-Amps and several of the input device transistors of the compound in //  There were a few current sources in places where they gave no added noise issues and a few other bits and bobs thrown in of course.  But, as best as I can tell with my test gear, it measured .7nV/rtHz which isn't too shabby really.  At 10V output, THD was nice with 3rd being approx. 135dB down and 2nd a few dB lower (-143dB).  Couldn't see any higher harmonics. 

Without giving all the game away as it may end up in someone's product?, the compound utilised a bootstrapped cascoding which sidestepped the emitter to emitter R problem and it appears to have really superb CMMR regardless of gain setting.  This was with ordinary, unmatched 1% tolerance resistors.

I don't know what he plans to do with this little circuit but, if I can, I'll try and post a schem.


 
We're always interested in seeing something new and better.
-----
I prefer to think in terms of NF with low noise preamp design. 0 dB NF means the preamp adds no measurable noise, compared to the self noise of the source.

Better preamps have been in the 1-2dB NF range for decades, so no amount of liquid nitrogen, or design magic will make more than a 1 dB improvement. A 1 dB difference in noise floor is not likely to be a game changer, while some mixer/console companies, who shall remain nameless, have invested millions of dollars advertising how their preamps are "so much quieter" (cough BS cough).

I miss the old low Rbb parts (2sb737 with typical 2 ohms  base spreading resistance), but these were better than needed for 150-200 ohm sources. Some tweaks prefer the newer noisier but faster parts, I don't generally hear things like that.  ;D

Becoming overly fixated on nV may not give the optimal result. For example the 737, that I happen to have the data sheet open for, delivers around .4nV rt Hz at 10 mA which is a nice low voltage, but the noise current increases as you would expect, so NF to a 200 ohm source (at 10 Hz) is 5dB. For 200 ohms the 737 delivers better NF at 1 mA even though the noise voltage rises to the horrendous  .6 nV rt Hz (yes I'm being sarcastic).

In my perspective mic preamps are a problem that has been long solved, for decades.

JR

PS: In my phono preamp I was dealing with MC carts that can have source impedance in 10-20 ohm range, so optimal bias current for NF was probably closer to 10 mA than 1 mA  (IIRC I used 2-3 mA).

========

Note: Ricardo made a comment in passing that GC has lower noise than other approaches at low gain... While this is a rather broad and sweeping claim. Surely some approaches are worse than others.

The noise at low gain can be more important than immediately obvious for interfacing with wide dynamic range A/D convertors, especially when these convertors are running from only 5V or 3.3V rails.

Ignoring topology and discrete device specs for a moment, the resistors in the feedback network also contribute noise, so just like we want low Rbb in a transistor, we need to use low value resistors in the feedback network to keep their contribution low. While this is no accident some of the newer uber-opamps have very healthy drive capability, this allows us to drop down the resistor values in the feedback network to lower their self-noise contribution.

At the end of the day, the active electronics inside the mic will probably dominate the path performance. I know this is an area of Ricardo's interest and expertise, not mine.

JR
 
JohnRoberts said:
We're always interested in seeing something new and better.

I was impressed myself with the performance of the circuit I mentioned above although I won't say it's better.  It was just a different approach
to solving the already solved "problem"  ;) of amplifying a microphone signal quietly.  Nothing about it is rocket science and beyond the capability of any number of folks on this forum .  It was very quiet but certainly no quieter than the best we have out there already.  I was impressed though with how well it worked wrt CMMR, even at low gain settings but...  maybe I just don't get out often enough?  :eek:
The basic topology, as I said, isn't mine so I wouldn't feel comfortable posting a schem, without asking permission first.

JohnRoberts said:
Better preamps have been in the 1-2dB NF range for decades, so no amount of liquid nitrogen, or design magic will make more than a 1 dB improvement. A 1 dB difference in noise floor is not likely to be a game changer.

I would agree completely with you. 

JohnRoberts said:
while some mixer/console companies, who shall remain nameless, have invested millions of dollars advertising how their preamps are "so much quieter" (cough BS cough).

Damn, now I'll be lying awake tonight wondering who you mean!  Care to drop a few discrete clues?  :) 

JohnRoberts said:
I miss the old low Rbb parts (2sb737 with typical 2 ohms  base spreading resistance), but these were better than needed for 150-200 ohm sources. Some tweaks prefer the newer noisier but faster parts, I don't generally hear things like that.  ;D


Thanks for that!  Yes, that's the transistor I was referring to - I mistakenly wrote 2sD and I knew there was a 7 in it somewhere but... ?
I don't think I could hear the difference either (ignoring noise for a minute) with using a faster part but...

JohnRoberts said:
Becoming overly fixated on nV may not give the optimal result. For example the 737, that I happen to have the data sheet open for, delivers around .4nV rt Hz at 10 mA which is a nice low voltage, but the noise current increases as you would expect, so NF to a 200 ohm source (at 10 Hz) is 5dB. For 200 ohms the 737 delivers better NF at 1 mA even though the noise voltage rises to the horrendous  .6 nV rt Hz (yes I'm being sarcastic).

I don't remember exactly off hand but, for one particular ribbon mic pre I optimised, I was somewhere around 3mA per. 
Back when we all played back our music from vinyl sources, I don't think I used anything other than a 737 (or sometimes a couple-or-three  in //) on the front end.  Standing current being matched to the cartridge du jour's impedance but also generally ending up in the round about 3mA region. 


JohnRoberts said:
Note: Ricardo made a comment in passing that GC has lower noise than other approaches at low gain... While this is a rather broad and sweeping claim. Surely some approaches are worse than others.

Those're my thoughts too.  Although in a lot of circumstances with high level sources from, say, guitar cabinets, requiring low pre amp gain, you hope/expect that the sn ratio issue will be mostly masked and irrelevant.

JohnRoberts said:
Ignoring topology and discrete device specs for a moment, the resistors in the feedback network also contribute noise, so just like we want low Rbb in a transistor, we need to use low value resistors in the feedback network to keep their contribution low. While this is no accident some of the newer uber-opamps have very healthy drive capability, this allows us to drop down the resistor values in the feedback network to lower their self-noise contribution.

Yep.  Ignoring for now the loading from the stuff downstream, I feel quite comfortable these days with a good Op-Amp being able to drive quite low FB impedances.  I've had good luck with the LM4562 myself as a not too exotic (read uber expensive) lower noise, pretty much a straight drop-in replacement to "old faithful" - the 5532/34.  It (4562) seems quite unfazed with loads down to around 500 ohms. 


JohnRoberts said:
At the end of the day, the active electronics inside the mic will probably dominate the path performance. I know this is an area of Ricardo's interest and expertise, not mine.

I have zero experience with designing mic head amp electronics of any kind.  Other than altering the topology a smidgen on a few valve types that needed valve replacements but that we couldn't warrant spending the £1000 going rate for the "correct" valve replacement.  I've always deferred to people such as Ricardo for the more up to date in depth stuff. 

One thing:  I only just realised that the first pair of Op-Amps in the "Cohen" operate in current mode with a resistive current source, hence are operating class "A"  DOH!!!    This for the folks that want/need/insist on such things.


Cheers guys.


 

Latest posts

Back
Top