Noise source calibration: Looking for low budget DIY ideas

[rhb]

 I picked up a <$30 Chinese noise source on ebay which turns out to be pretty good.  More info on that here:

https://www.eevblog.com/forum/rf-microwave/another-ebay-noise-source/msg3014188/#msg3014188

So, this raises the question.  If you don't have a well equipped lab and are trying to build a basic lab on a small budget, how do you calibrate it?  Keysight advises making comparisons to a noise source calibrated by the UK NPL.  That's obviously not a low budget solution.  And the national labs didn't start out with a calibrated noise source.

While I'm not even traceable to NPL or NIST, I have a pretty good set of gear including an HP 386B as a reference.  I've got a nanoVNA which is needed to measure the SWR, and I've ordered a couple of RTL-SDR dongles just to see how well they will work as a substitute for a spectrum analyzer.  I've also got an SDRplay RSP2.   I *should* be able to do a decent job of comparisons of cheap kit to HP kit.

I've posed a puzzle to solve and am looking for suggestions.  Particularly, early papers on the problem.  My first order guess at a solution is a high Q tuned circuit and one of the Analog Devices log detector boards.  Doing this over 2 decades with a single tuned circuit seems improbable, so it gets a bit sticky. But if the tuned circuits have sufficient overlap it seems to me it should be doable.

I bought  an old set of the standard kit (8970B & 346B).  But it was still not cheap.  So how do we make it cheap?

Have Fun!
Reg

[MadTux]

Why do you need calibration for noise source?
I'd guess a cheap chinese noise level is dependent on temperature, input voltage..., so it wouldn't make much sense to calibrate it, if it drifts off anyway.

I mostly used my cheap china noise to display filter response on spectrum analyzer, before I got my HP 8444A tracking generator. No need for calibration there as well, just use trace subtraction function on spectrum analyzer.
A noise source is also great to track signals when repairing amplifiers.
Apart from that, I usually use clean signals from synthesizer or so, instead of white noise.

[rhb]

Go price a Keysight noise source and ask yourself why anyone would pay that much money.

[Conrad Hoffman]

IMO, you could DIY a thermal method and compare the result with an accurate DC source.

[David Hess]

I have calibrated noise sources up through RF frequencies by making spot noise measurements using a series of bandpass filters and an RMS voltmeter.  The filters are calibrated in a separate step to measure their loss and shape factor.

[rhb]

Obviously one needs the filters.    Building those without access to better gear becomes challenging.

I think I need to do some research into old NBS/NIST and NPL papers.

[David Hess]

Obviously one needs the filters.    Building those without access to better gear becomes challenging.

I think I need to do some research into old NBS/NIST and NPL papers.

The filters just require calibration with an RF signal source and voltmeter.  It is time consuming but not difficult.

[rhb]

Yes, but that presumes access to a good RF signal generator.  An F***Tech will get you to 60 MHz, but above that there are lots of issues with harmonics from cheap square wave sources like the ADF4351.  Though I remember when 30 MHz seemed like the moon.

My major interest is low cost T&M kit.  I'd been working on ideas for a cheap VNA when the nanoVNA appeared and made that moot.  At this point I'm focused on evaluating cheap ebay stuff and figuring out what else is needed to make it a viable piece of gear.

The ebay noise source that prompted this thread turns out to be quite good from 200 kHz to 1 GHz.  With serious HP & Tek gear for testing cheap kit,  I'm enjoying buying and testing things someone with less money would have as their only option. 

I just bought an HP 4145B  w/ 16058A test fixture on ebay a few minutes ago.  So my BSIDE ESR02 Pro and Peak Atlas DCA75 are about to go head to head with the big boys.  And testing stuff is a bunch of small discrete projects, none of which is difficult.  Though things like this are not simple if you have very limited resource, e.g. an EE student in the 3rd world from a middle class family of merchants and professional people. Or me as a 30 something PhD candidate.

Have Fun!
Reg