Measuring noise with O'scope

[Optikon]

I am trying to come up with a setup to measure circuit noise with equipment on hand.

I wish to measure noise in a bandwidth of 1 MHz.
I expect the noise to be from about 1uV to 100uV peak.
I expect the noise to be white.

Has anyone done this with a high bandwidth scope successfully?
Obviously there needs to be a low noise amplifier and filter in the setup.
So I have considered the methods used in an appilcation note from linear tech (AN70). The two problems that bother me about the method are:

1) It calls out using some ancient equipment that I stand no chance of ever finding and even if I did, it's not worth my while to buy or otherwise get.

2) It describes building several complicated circuits and shielding boxes for use with the equipment and for noise isolation. This I can understand but MAN!! is it involved... got to be an easier way..

Anyone made these kind of measurments before with a setup that can take less than two days to build?? Looking for alternative methods to AN70..

Thanks!

 

[crust]

You can do it by taking several Fourier transforms of your time series data (if you have a digital scope) and then do the math to compute the PSD/Mag of the noise.

 

[Optikon]

You can do it by taking several Fourier transforms of your time series data (if you have a digital scope) and then do the math to compute the PSD/Mag of the noise.

I dont fully understand how this method would work and I'm suspicious that it wont provide accurate enough information. Thanks for the suggestion.

Anyone else make noise measurements on circuits?

 

[Russlk]

Here is the problem: You will need a gain of 10,000 or so in order to see the noise on the scope. With 1 mHz bandwidth the preamp has to have 10 gHz gain-bandwidth.
A spectrum analyzer might be able to do the job because it divides the spectrum up into small segments.

 

[Optikon]

Here is the problem: You will need a gain of 10,000 or so in order to see the noise on the scope. With 1 mHz bandwidth the preamp has to have 10 gHz gain-bandwidth.
A spectrum analyzer might be able to do the job because it divides the spectrum up into small segments.

I do need about 40dB of gain with high fidelity. Since what I am after is a noise bandwidth of 1MHz. actual 3dB bandwidth will be more like (1/SQRT(2)*NBW) about 707KHz. So I would expect say, 10MHz Bandwidth to faithfully reproduce 1.0MHz components. 10MHz at a gain of 10,000 is a gain-bandwidth product of 10GHz but is this not feasible with a multiple stage setup? IF I made one of these, could I:
1) connect to scope with input shorted and measure / record noise
2) then appliy circuit input and record noise
3) RSS calculate out the noise from amp+scope to yield ckt noise?

Also what about making a broadband true RMS meter to directly measure RMS noise? If I did this, I would only need to accomodate the 10MHz bandwidth and be measuring noise power directly? Is this right?

Anyone have any experience with this method?

Thanks for inputs!

 

[Roff]

I got most of this from literature available on the Internet, so take it with a grain of salt. I have no experience measuring noise but I played a noise engineer on TV. (If you're not old like me, that joke will go over your head.) I believe that the best op amps are just below 1nv/√Hz, but you have to also check the specs for 1/f noise (some are better than others), and their current noise limits them to low source resistances. Chopper-stabilized amps have no 1/f noise, but their wideband noise performance is relatively poor.
Even at 1nv/√Hz, a 1MHz bandwidth will give you 1uv of wideband input-referenced noise, so it looks to me like an amplifier is going to be a challenge. As you suggested, you might be able to RSS the amplifier noise out with a caibration scheme, but that means your amplifier noise really needs to be "white" (I think).
However, you gotta ask yourself, "How do they measure 1nv/√Hz?

 

[Optikon]

I got most of this from literature available on the Internet, so take it with a grain of salt. I have no experience measuring noise but I played a noise engineer on TV. (If you're not old like me, that joke will go over your head.) I believe that the best op amps are just below 1nv/vHz, but you have to also check the specs for 1/f noise (some are better than others), and their current noise limits them to low source resistances. Chopper-stabilized amps have no 1/f noise, but their wideband noise performance is relatively poor.
Even at 1nv/vHz, a 1MHz bandwidth will give you 1uv of wideband input-referenced noise, so it looks to me like an amplifier is going to be a challenge. As you suggested, you might be able to RSS the amplifier noise out with a caibration scheme, but that means your amplifier noise really needs to be "white" (I think).
However, you gotta ask yourself, "How do they measure 1nv/vHz?

Well.. I think you're 'bout 2x older but I've heard the "I'm not a real doctor joke.. a few times" - I don't know about the original instance! :lol:
A system noise cal to determine the noise floor before ckt measurement, I believe is the idea behind avoiding amplifier noise mixing with noise of interest. Also, I'm pretty sure that for most broadband op-amps noise is white beyond 1/f knee up until multi-GHz. I am very curious about how they measure very small noise over a wide bandwidth for the OA specs..

For that matter, if they have it tough, just imagine the tektronix guys having to measure input noise on the newer multichannel GHz scopes.

AN70 is a good read for what seems like practical(!?) noise measurements but it's an aweful lot of work to properly build the amps. Not to mention getting ahold of the ancient test equipment. I'll take a closer look at what they use and see if they end up with such phenominal overall gain-bandwidth.

An interesting side note regarding the 1/f noise and it's shape..
I once read that 1/f noise occurs naturally in a variety of situations and one has to wonder how low in frequency one must go until the 1/f curve is not followed. As one approaches DC, 1/f _should_ approach infinity or else follow a different relationship. One of the lowest frequency measurements ever made that followed 1/f was in the ancient middle-east where the people recorded the rise & fall of the Nile river for hundreds of years. With all of that sample data and such a low frequency process, it is amazing to discover that it still followed the 1/f curve almost perfectly.

never looked into it any further but thought it was interesting..

Thanks for the comments!

 

[Russlk]

"but I played a noise engineer on TV."

I am older than you, Ron, and I don't get it!

Russ

 

[Roff]

"but I played a noise engineer on TV."

I am older than you, Ron, and I don't get it!

Russ

I couldn't remember the specifics, but a little Googling led me to this page about the **broken link removed** I was playing off of. I never watched "All My Children", but I remember the commercial for Vicks 44.