Looking, QST construction article about measuring noise of a resistor

A couple of FET buffers measuring the same resistor, amplified, digitized and cross-correlated. A scope can compute the cross-power spectrum. The key thing is that the voltage noise of each amplifier affects only its branch, so that their contributions are uncorrelated, and so average to effectively zero if you wait long enough.

The current noise of the buffers of course just adds to the resistor noise and shows up in both branches, so you really need to use FETs.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC / Hobbs ElectroOptics 
Optics, Electro-optics, Photonics, Analog Electronics 
Briarcliff Manor NY 10510 

http://electrooptical.net 
http://hobbs-eo.com

You can buy opamps with current noise measured in aA per root Hz. I wonder how they measure that.

--

John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

That's what I'm finding as I read about low noise microphones. But it seems like you need to spend several hundred if not thousand dollars to get one of those low noise large diaphragm microphones. The guys recording nature sounds seem to have interest in and info about low noise mics. Mikek

I wonder how parabolic his reflector really is? It's a lamp shade.

Mikek

PS, Parabolic reflector is what they use at football games, would a exponential horn be better? Even if unwieldy. Ok, did some reading on exponential horns, they work, but have their problems like a parabolic dish.

** The Australian built Rode NT1-A is one such and not too expensive.

But you will need a supply of phantom power for the mic.

.... Phil

They don't. They just publish sqrt(2 e I_bias). Leads to some fairly shocking underestimates for parts with leakage cancellation, not to mention drain circuit noise coupling via C_dg.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC / Hobbs ElectroOptics 
Optics, Electro-optics, Photonics, Analog Electronics 
Briarcliff Manor NY 10510 

http://electrooptical.net 
http://hobbs-eo.com

Lies, damned lies, and data sheets.

--

John Larkin         Highland Technology, Inc 

lunatic fringe electronics

There's a fun blog titled "Lies, Damned Lies, and Start-up PR".

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC / Hobbs ElectroOptics 
Optics, Electro-optics, Photonics, Analog Electronics 
Briarcliff Manor NY 10510 

http://electrooptical.net 
http://hobbs-eo.com

Yeah, like that. ;)

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC / Hobbs ElectroOptics 
Optics, Electro-optics, Photonics, Analog Electronics 
Briarcliff Manor NY 10510 

http://electrooptical.net 
http://hobbs-eo.com

But the noise voltages add as sqrt-of-sum-squared.

--

John Larkin         Highland Technology, Inc 

lunatic fringe electronics

"Build the JCan to Measure Resistor Noise" appears in the July 2007 issue. The author, Joe Geller, kindly makes it available for free at his website.

Thank you, 73,

--
Don Kuenz KB7RPU 
There was a young lady named Bright Whose speed was far faster than light; 
She set out one day In a relative way And returned on the previous night.

amdx wrote in news:q5228n$9en$ snipped-for-privacy@dont-email.me:

shade.

their

Get Mr. Klipsch to fashion one for you.

IIRC one of his early models had 36 feet of tuned enclosure cavity.

Except instead of an output transducer, he makes a acoustic gathering horn for you.

Remember those toothless old men with the little horn tube pressed into their ear?

Am 26.02.19 um 03:49 schrieb John Larkin:

Yes, they do. And for amplifiers, they disappear with sqrt(N).

< >

I get < 220 pV/rtHz from these 10 ADA4898 pairs. That's about 3 Ohms. The problem is that the noise current starts to hurt. With JFET opamps I would get 1 nV/rt Hz, just a BF862 :-( Why does no one make a serious FET op amp? Some Interfet IF-37 ?

(And Phil, thanks to remember me about that xcorrelation fail) There are also lots of analog switches, a SPI receiver in a Coolrunner for control and prehistoric HP optocpouplers.

S.E.D. is quite interesting as long as we do not talk about politics.

I also have made a FET version with 2 IF3602 pairs <

>

Features 190 pV/rtHz. It needs more current than promised in the data sheet, but the coolers are only there for thermal mass.

It has heroic bootstrapping of the cascode to remove the FET capacitances but stability is sensitive against everything, moon phase included, already in simulation. Part of the problem is that I insist in 1 MHz BW.

I did remove the bias loop. Input coupling C must be _much_ bigger for noise reasons than for f-3dB. And the bias loop must be much slower than input RC. One is constantly waiting.

A TL431 minus one BAV99 drop and 1/3 therefrom is about the right temp-compensated FET bias.

The right thing for rainy weekends.

Cheers, Gerhard

What's the low cut-off frequency you need? Is it a high Zin or should it be 50 Ohms?

Jeroen Belleman

I missed that you said all resistors of the "same value". My bad

I was surprised at Joe's using a BF244A JFET for the input stage. This VHF/UHF amplifier JFET isn't what I would call a low-noise JFET, but then I saw that he's measuring noise voltage on up to (and above?) 1 megohm resistors, and super-low input capacitance was one of his goals. 1meg and 3pF has a 50kHz response rolloff.

My answer to that situation would be to measure noise current, rather than voltage. Yes, the Johnson noise current from a 1M resistor is only 0.13pA/rt-Hz, or 13pA for a 10kHz bandwidth, but it's not hard measuring into a short with TIA amplifiers. I've got free RIS-617 PCB boards for anyone who wants to try.

Choose the version of bare PCB you'd like, customize component choices, and warm up your iron. email: snipped-for-privacy@yahoo.com

--
 Thanks, 
    - Win

Yeah that's my only thought. But whereas the voltage noise is a direct measure, for current noise you've got to subtract off the resistor noise. One would also be tempted to use a large value resistor, which at some point is going to reduce the overall bandwidth.

Yeah the cross correlation trick gets rid of uncorrelated voltage noise but now I've measured the sum of both current noises (from amps) and the resistor noise... so I'm not that much closer to measuring the current noise of one amp. (One can also worry about equal bandwidths in each channel.)

George H.

Wow! Awesome. Say in the second one (RIS-617B-1) The photodiode boot strap (Q1) goes through the cap multiplier low noise bias supply (Q2)? Does there have to be some gain there? I'm trying to sort out the various R and C values. (Gain needed to overcome C10/C15 current loss?)

Thanks George H.