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 said:
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... ?

There were actually two similar low noise parts from the same company that formed a complementary pair.  The pnp was 2sb737 and the npn was 2sd786. As I recall the pnp was a rch (old machine shop term) lower noise than the npn. When I went to work for Peavey they already had the npn in the system so I just used them, rather than bring in a new slightly better part (on paper) that nobody would hear the difference from.

Don't lose any sleep over which company I was accusing of hyperbolic marketing, most do it because it works, more now than ever as technology advances reduce the real differences between many product platforms.

JR
 
Jean Clochet said:
But, as best as I can tell with my test gear, it measured .7nV/rtHz which isn't too shabby really.
As only Cohen & the Earthworks have proven sub nV/rtHz performance for a properly protected P48V preamp, your friend joins a very short list.  The best of the rest, Millenia & Sam at 1nV/rtHz, are also a very short list.
I don't know what he plans to do with this little circuit but, if I can, I'll try and post a schem.
So yes please, pretty please!

I have to say, I'd use THAT1512 (optimised for slightly less gain than 1510*) if I was building a preamp tomorrow.  The only reason for the complexity of Cohen would be an original STC4038 ribbon with nominal 30R output.  But even that has a workround eg PRR's

http://www.prodigy-pro.com/diy/index.php?topic=5743.80 or my own efforts on MicBuilders.

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.
Baxandall & I both looked at many medium power devices in da old days.  The problem with these is usually poor hfe which means more current noise.  But much nastier is they often had huge 1/f noise so you had to do a lot of selection.

When I found Hitachi 2sa1084/2sc2546, measured rbb 5R, I stopped looking.  Later the Rohm 2sb737/2sd786, rbb 2R, that JR likes came out but I've never used them.

There's also Tosh 2sc3329 rbb 2R, which has escaped from the Unobtainium Store  ;D  These three types are the lowest noise devices I've come across.  IIRC, better than LM394 etc but those have the advantage of 2 matched devices in one package.

My best effort is 0.28nV/rtHz with a huge total of 2 devices and 3mA but that's for Moving Coil cartridges and no way could that be a P48V mike preamp.

I don't think JLH understood noise.  Some of his "low-noise" designs were pretty dire.

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.
Actually once you get below 20dBA spl, the acoustic resistance noise starts to be important and you'll want to listen to the noise too.

Hence the above about STC 4038 nominal 30R.  The Coles version is nominal 300R and -65dBV/Pa.  THAT1510 can ju..ust do 1dB NF at 300R but not 30R.  I think THAT1510 or even 1512 is beyond reproach for any condensor even old ones with big step downs.  I may be wrong PRR

* Having 2 different versions is an example of JR's stuff on feedback resistors.  But Cohen also doesn't have evil current sources at sensitive points to muck up low gain noise.
 
> was a rch (old machine shop term)

An engineer tole me it was an old engineering term, slightly larger than the mythical BCH.

> we need to use low value resistors in the feedback network

That is where the Cohen published plan differs greatly from others.

There's two resistors from differential output to input emitters. In other amps these are many K ohms. Cohen used 300 ohms.

> some mixer/console companies...advertising how their preamps are "so much quieter" (cough BS cough).

Mackie sold me some "VLZ" or somesuch. Where others used 10K or so, these used 2K or so. The ads touted lower noise at lower gain. Don't need a weatherman to see that this is so. I think THAT Corp offers two parts identical except these resistors (one for legacy designs and one improved).

> with high level sources ...requiring low pre amp gain, you hope/expect that the sn ratio issue will be mostly masked

Not always. Especially with hot mikes on wide dynamic range (not e-gitar) live performance in quiet room. The input noise figure does rise annoyingly. Even back in days of tape, I could hear it, and struggled to understand how an input could be quite acceptable wide-open yet annoying at 20dB less gain.

The practical DIS-advantage of very low resistors is that your high-gain cross-emitter resistor gets down to an ohm or two. That gets into pot/switch contact resistance, lead inductance, and discourages a DC-block cap here. That's why THAT offers the higher resistance part, to suit available or in-stock pots.

> the active electronics inside the mic will probably dominate

AKG 414 has so much self-hiss that a TL072 after will hardly hurt. (That's absolute hiss; the signal output is very high and the S/N is quite fine.) While this may be generally true, some "simple" condensers use very steep step-down transformers and the output hiss is similar to a dynamic.
 
It's about sub nV/rtHz noise performance.

It's worth noting that typical dynamic microphones have output impedances of several hundred Ohms. Getting low current noise is at least as important as low voltage noise.

I think THAT Corp offers two parts identical except these resistors.

They altered the gain of the second stage to -6 dB (instead of 0 dB). The feedback resistors of the first stage are identical.

Samuel
 
PRR said:
> was a rch (old machine shop term)

An engineer tole me it was an old engineering term, slightly larger than the mythical BCH.

When I was a young lad working in a machine shop, I was taught that the "red" CH was the finest uint of measurement possible.  8)

JR

PS: OTOH I think that "blonde" CH is a myth that doesn't exist in the wild.

PPS: my apologies to the PC police.
 
Nice to read the comments from PRR, Samuel, Ricardo & John.  Some good stuff to think about.  8)

Ricardo, some back-peddling from me now:
After doing another noise measurent on this "non" Cohen circuit with that input stage designed by my friend (as best as it can be tested anyway with the jerry-rigged kit I have on hand!), I don't get quite as decent a noise spec. as I stated before.  In a prior post, I boasted a .7nV rt Hz spec but, this time I come up with somewhere round about 1.2nV rt Hz.  Not exactly terrible, but it isn't as good as Samuel's efforts and the rest of the best.
Mustn't have accounted for something or other last time around, or messed up my sums a bit!    :-[

Haven't put it in Spice yet so who knows if I can better it?

I think it'll be OK to post a schematic btw but I'll need to redraw it and transcribe the scribbles etc. first.


Cheers,

J.C.









 
Jean Clochet said:
In a prior post, I boasted a .7nV rt Hz spec but, this time I come up with somewhere round about 1.2nV rt Hz.  Not exactly terrible, but it isn't as good as Samuel's efforts and the rest of the best.
Jean, I'm sure all of us have had similar experiences  ;D

My best effort in a mike preamps was 1.7dB NF using a Lundahl and 5534, early 80's for Calrec.  But getting to something which measured like that was a lot of effort.

It's worth noting that typical dynamic microphones have output impedances of several hundred Ohms. Getting low current noise is at least as important as low voltage noise.
As I said earlier, I would only consider a Cohen if I had a pair of 30R nominal STC 4038.

However, if anyone wants to achieve true Cohen performance or compete with Earthworks ...

The better single chip preamps, THAT1510, SSM2019 all have around 2pA/rtHz and 1nV/rtHz

The exception is INA163 with 0.8pA/rtHz.  4 of these in parallel equal the Earthworks ( :eek:) on paper.  I twisted Earthworks' arm to get their noise current spec of 1.5pA/rtHz.
 
ricardo said:
Jean Clochet said:
In a prior post, I boasted a .7nV rt Hz spec but...
...I'm sure all of us have had similar experiences  ;D
Ricardo,
my noise now seems to be a bit better again ( .95nV rt Hz)  - I adjusted the current upwards for the input devices.

Edit: Am trying to post a reply with info on the circuit I'm working on but the forum won't let me!    It's a long post so maybe that's it? 

Will try again later.


Edit:  Long post from me in new thread instead...  Better to not clutter up discussion about the fine "Cohen" circuit with my doodlings anyway. 
 
> picture I'm attaching

Post the words, Edit, attach the files... then at least you know what the forum is objecting to.

> trying to post a reply with info ....but the forum won't let me!    It's a long post

If you are composing in MS-Word or other too-fancy typewriter, pasting into the forum compose box can fail for many obscure reasons. I know that newspaper online text with "smart quotes" often chokes, I must replace all left-right quote-pairs with plain "-marks. Support for non-ASCII (non-US) characters is often hit-and-miss, especially when pasting from a legacy wordprocessor like MS-Word.
 
Further to the discussion regarding transistors with low noise, I ordered and tried some ZTX790 as drop-ins to the 2N4403.  I couldn't find any info regarding the ZTX's base spreading resistance (I had only found the recommendation from a link on here), but I do measure slightly lower noise when comparred to a 2N4403 although it is a more expensive part - £1 per.

Also, I was looking for further info regarding why the theorectical noise advantage of 3dB per device doubling doesn't usually hold up and remembered reading this paper before which is pretty decent:
http://users.ece.gatech.edu/mleach/papers/Parallel.pdf

If anyone's interested, there are a few other articles by Leach which are worth a look IMHO and the page for downloads is:
http://users.ece.gatech.edu/mleach/papers.html

Cheers.
 
An earlier discussion of parallel devices to reduce noise can be found in "Low noise Design" by Motchenbacher & Fitchen  c. 1973

While Leach's entire website (RIP)  is worth reading.

JR
 
JohnRoberts said:
An earlier discussion of parallel devices to reduce noise can be found in "Low noise Design" by Motchenbacher & Fitchen  c. 1973

Ah! That's it, thanks.  I couldn't remember who it was and Marshall didn't list it in his references. 

JohnRoberts said:
While Leach's entire website (RIP)  is worth reading.

There's a good paper on low noise design in general, a coupla Fet limiters, lots of speaker design stuff which I know nothing about...  All well worth the read indeed.   
Yes, I never knew or spoke with him but RIP.  Seems like a he was a good human. 
 
Back
Top