Diode noise sources and Calibration Club

[rhb]

I just "finished" packaging my BG7TBL "nose source" and am looking at the DC to 500 MHz response on my Instek MSO-2204EA.  Both are limited.  It's a 200 MHz DSO and the noise source is anything but flat.

Around 2007-2008 Franco Rota published a couple of articles in "VHF Communications" on the subject of diode noise sources.  His 2nd article described the construction of  small batch of diode noise sources given out to attendees of the 13th EME conference in Florence.

www.vhfcomm.co.uk/Noise%20diode.pdf

https://www.rf-microwave.com/app/resources/uploads/diodes/VHFComm_NW303.pdf

I'd like to organize replicating the EME source.  That was built by milling out a piece of round aluminum bar for the enclosure.  However, the only thing of importance is the dimensions of the cavity containing the circuitry.  That is more easily produced by hydraulic forming with a rubber pad.  One makes a male die of the required shape, places a sheet of aluminum on top and then applies pressure using a soft elastomeric pad and a hydraulic press.  This is widely used for making airplane parts such as wing former ribs.  During WWII a very large press, (20' x 80' ??) was used to make hundreds of parts in a single operation using dies made of micarta. A big advantage of this is the surface finish is much better than milling if the die is smooth.

I have a Clausing mill and a 20 ton press.  So I can easily tool up and make enclosures at a cost of a few cents.  I'd like someone with experience playing with DC to daylight  to design a board and cavity dimensions based on Rota's  articles. @svenskelektronik has expressed an interest in promoting OSHW.  He operates a PCBA company, so he is in a position to manufacture the PCBs.

If enough people want one, I'd like to propose that a sample be sent to NIST for measurement and then circulated among the builders as a calibration reference.With a good design, well executed, we might even get lucky and  find a NIST staff member who would take on a "Swiss Navy" project.  In any case, spread over a number of participants, the cost per person should be quite nominal.

I also have a lot of experience writing DSP and other scientific software.  So if the diode noise source includes flash memory for calibration constants which can be read via USB, applying the corrections would be very easy.

[cellularmitosis]

Very interesting, I would be interested in a flat noise source.

What is the significance of the milled aluminum enclosure?  Is that important to the circuit performance, or would something like a Hammond diecast aluminum enclosure work as well?

I have recently acquired an LT1088, and it would be nice to have a calibrated noise source to calibrate or verify the LT1088.

Edit: Oh, now I see:

I selected a very small box in order to
avoid both the secondary propagation
modes and the microwave absorber; the
size that I used gives no trouble up to
10GHz.

If someone wants to increase the size of
the box (internal size) it will be necessary
to use a microwave absorber.

Edit 2: this thread looks useful: https://www.eevblog.com/forum/projects/ever-built-a-working-rf-noise-source/

[dkozel]

I've been working on a USB controlled power supply for noise sources. The next step would be to make a noise source to use with it. My target would be 1-10 GHz as a starting place.

The milled aluminum enclosure helps as a thermal mass, reducing temperature variation over the course of the measurement. When measuring LNAs it gets very difficult to remove the uncertainties to get precision down in the 0.1 dB level. Most commercial systems will only specify +-0.2 dB accuracy under general usage.

Keysight has a good summary in this app note.
http://literature.cdn.keysight.com/litweb/pdf/5952-3706E.pdf

There are some lower cost high ENR sources available:
http://g8fek.com/noise-gen.html

Paul Wade, W1GHZ, did some experiments with different diodes:
http://www.w1ghz.org/QEX/Noise_Measurement_and_Generation.pdf

Paul Marsh, M0EYT, built some very recently with a variety of ENRs
https://twitter.com/uhf_satcom/status/927299003481501701

I've always enjoyed reading this series of posts to see one way of pushing the limits of LNA measurement. Sergey Zhutyaev, RW3BP, ended up building an antenna to use the RF quiet sky as a very stable and cold noise source.
http://www.vhfdx.ru/apparatura/accurate_noise_figure_measurements_1296_mhz
http://www.vhfdx.ru/apparatura/rw3bp_1296mhz_lna_optimization
http://www.vhfdx.ru/apparatura/1296-mhz-small-eme-station-with-good-capability-part-5

[rhb]

 Thanks!  Lots of good reading material *and* a source of quality noise sources I can afford!

As you have researched the subject fairly extensively, do you think one could calibrate out errors?

To wit, Use a formed inner enclosure mounted inside a larger thermal mass with a thick silicone mat between the two.  Embed thermistors in the mat to monitor the temperature and either output a correction table over USB based on the temperature or adjust the output level if that is feasible.

FWIW The cover to a WD hard drive sitting on my desk looks as if it is likely done by hydraulic forming.  I don't see any of the fine scratch marks  that are typical of metal dies in the corners with a 14x Hastings triplet.

[dkozel]

In the end the flatness of the noise source is nearly irrelevant if you can get it calibrated by a good lab with enough frequency vs ENR points. You could calibrate them at home with a setup using resistors in boiling distilled water and liquid nitrogen as hot and cold sources, but you can see how that starts to get somewhat involved. I know one person who does this. The materials costs for the noise sources is very small, all the cost is in the calibration and stability.

Yes, if you can measure the true temperature of the noise diode you can at least reduce the errors somewhat. I've also considered the idea of ovenizing them, but I'm a long way away from being able to measure the drift in noise output reliably. Remember also that your device under test is also going to thermally drift, changing the gain at least.

How low of a noise figure are you trying to measure? If it's above 1 dB then most of the concerns are essentially academic because the error factors mostly reduce as the DUT gets noisier. Below that, connectors and other components start contributing very meaningfully to the noise figure. If it is below ~5 dB then a low ENR source is useful for reducing errors.

[rhb]

Actually, I'm not trying to measure anything yet.  I want to build some things and a good quality, well  calibrated noise source seemed a good start.

This is really a tools for making tools sort of exercise.  I've got a full metal shop and wood shop, though not organized enough yet to be very usable.

I spent about 3 years studying sparse L1 pursuits.  It's a general name for compressive sensing among other things.  I need to study the noise source calibration problem and see if I can apply a sparse pursuit to it.