John Curl's Blowtorch preamplifier part II

Status
Not open for further replies.
Jump to Latest
gerhard

gerhard

Member
Joined 2005
hoffmann-hochfrequenz.de
missing pic
 

Attachments

  • DSC_7603_SCALED.JPG
    DSC_7603_SCALED.JPG
    139 KB · Views: 248
scott wurcer

scott wurcer

R.I.P
Joined 2004
Waly said:
I don't get it, paralleling 20 op amps makes me cringe. 2 x 10 BF862 in parallel, as an input diff stage, running at Idss, will do the same equivalent noise voltage, even after the 3dB penalty, and without any noise current. Then a precision op amp of your choice and close the loop for precise gain. Some compensation for the extra gain is required, nothing fancy. Cheaper, smaller, elegant, straightforward. Higher 1/f corner noise, though.
Click to expand...

Yes Ovidos did 32 BF862's (you need to damp the gates in groups with some L since they are 700MHz figure of merit). There are non-audio uses for low noise instrumentation, like for instance measuring the Shumann resonance with a wire loop.

The IRF3600 should do .25nV by itself, but it is essentially a massively paralleled JFET. These techniques are essentially brute force and don't by themselves show any particular sophistication. That is the AD743/45 is FET input and 2.9nV so I could parallel a 1000 of them to get .1nV or so. No expense to me but not useful to anyone else.
 
Last edited:
wayne

wayne

Member
Joined 2002
Paid Member
scott wurcer said:
Yes Ovidos did 32 BF862's (you need to damp the gates in groups with some L since they are 700MHz figure of merit). There are non-audio uses for low noise instrumentation, like for instance measuring the Shumann resonance with a wire loop.

The IRF3600 should do .25nV by itself, but it is essentially a massively paralleled JFET. These techniques are essentially brute force and don't by themselves show any particular sophistication.
Click to expand...

That makes it perfect for me and my stash of those plus they are big enough to solder.
 
gerhard

gerhard

Member
Joined 2005
hoffmann-hochfrequenz.de
Waly said:
I don't get it, paralleling 20 op amps makes me cringe. 2 x 10 BF862 in parallel, as an input diff stage, running at Idss, will do the same equivalent noise voltage, even after the 3dB penalty, and without any noise current. Then a precision op amp of your choice and close the loop for precise gain. Some compensation for the extra gain is required, nothing fancy. Cheaper, smaller, elegant, straightforward. Higher 1/f corner noise, though.
Click to expand...

Yes, and the BF862s give you drain resistors to match, 20*12 = 240 mA total Idss,
with enough voltage drop for the cascode, optionally ccs from the negative supply,
four times the number of devices for being differential at the same noise level,
constantly moving offsets because of missing thermal coupling and, and.
It's not that I have not done that. (2*8 2SK369...)
And the moving offsets hurt, I'm after low frequencies.

Some more examples are at
< https://picasaweb.google.com/103357048842463945642/LowNoisePreamplifiers?authuser=0&feat=directlink >

The effort to make 10 ADA4898 instead of 1 is here: (pic)
Hint: it's in the repeat(pre, 1, 10). The layout of the single
ADA4898 ist automagically repeated 10 times. (Altium Designer)

The preliminary writeup on this amplifier is in
< http://www.hoffmann-hochfrequenz.de/downloads/lono.pdf >

First I wanted only 100 KHz for the usual Stanford (or whatever)
FFT analyzer, but now that the 89441A is here, 1 MHz was the
minimum, so the overcompensation caps and the fixed lowpass Cs
had to go. Next incarnation will have only 10 dB/Opamp for flatness
(but the first stages) and R18 is really 340 Ohm.

regards, Gerhard
 

Attachments

  • Bildschirmfoto - 03.05.2014 - 22:21:22.png
    Bildschirmfoto - 03.05.2014 - 22:21:22.png
    59.1 KB · Views: 266
Last edited:
Bonsai

Bonsai

www.hifisonix.com
Joined 2003
hifisonix.com
Paid Member
gerhard said:
Yes, I'm using here 20 opamps in parallel ( 10 pairs ADA4898) for noise measurements.
Delivers 220 pV/sqrt Hz from a few Hz to 1MHz.

Blue line: input shorted,
brown line = noise of 60 Ohm = 1nV/sqrt Hz
green = noise of 4 Sanyo Eneloop NiMH AA in series

2nd pic is a 2/3AA Lithium Cell.

Probably the Eneloops are even better. It's time to update
my second preamp so that I can use the cross correlation
capability of the analyzer, that should bring 25 dB more. :)

I wanted to try NiCd, but I cannot buy them here any more.

Gerhard
Click to expand...

Can you share your circuit Gerhard?
 
Bonsai

Bonsai

www.hifisonix.com
Joined 2003
hifisonix.com
Paid Member
scott wurcer said:
Yes Ovidos did 32 BF862's (you need to damp the gates in groups with some L since they are 700MHz figure of merit). There are non-audio uses for low noise instrumentation, like for instance measuring the Shumann resonance with a wire loop.

The IRF3600 should do .25nV by itself, but it is essentially a massively paralleled JFET. These techniques are essentially brute force and don't by themselves show any particular sophistication. That is the AD743/45 is FET input and 2.9nV so I could parallel a 1000 of them to get .1nV or so. No expense to me but not useful to anyone else.
Click to expand...

I am looking at this approach as well. I've simmed with 8 JFETs (BF862 - no expense to me :D) but of course the actual noise figures coming out of LTspice can be believed with a pinch of salt. However, it does allow one to get a feel for what paralleling does. SMD ferrites in the gates to prevent oscillation.

Does 0.8nV/rt Hz sound about right?

For source I'm taking a worst case Shure 600mH+ 200 Ohms.
 
Status
Not open for further replies.
Top Bottom