Noise Sources And Noise Figure Measurements

There was a lengthy but good thread on the Microwave mailing list about Noise Diodes and Noise Figure meters and procedures. While he probably won't terminate the thread, Jeff had something to say about it (obviously, it was a lot :-)


Hi Guys!

A great topic, a lot of knowledgeable players contributing here. I remember N4FS's article in Ham Radio on noise figure measurement. Also Joe Rieserts VHF column. Good learning when I was a kid.

OK, now for my 2 cents worth! I have made my living in this area so may have a little to contribute.

In fact a great many of the solid state noise sources are 35 dB ENR (hotter than the Sun, 1000000 K) (I actually look for and collect high ENR sources!) and were attenuated down to 25 dB (for Sanders Noise sets, used to inject at test coupler ports of ECM gear, a flight line quick test box), or 15.5 dB ENR (this value chosen as a direct replacement for the Gas Plasma Tube invented by Mumford in the 1950's and extensively studied because of the extremely flat noise vs. frequency response), or the 5 dB ENR sources, which have a 30 dB pad in them, to mimic the 5.2 dB ENR of the 5722 noise diode tube source in the HP 345 used with the HP340/342 Noise figure meters.

Both AIL and HP exploited the waveguide noise tube (with its critical guide angle) that Mumford puzzled out. It was the switching video spikes from the HP 349 coaxial noise source (100-4000 MHz!) and its AIL equivalent which I think was the AIL7011 that gave gas tubes the bad name for “killing” transistor LNA's in the '70's.

BUT gas tubes are a very good source and we used them for GaAsFET measurements by putting a 10 dB pad on the 15.5 source, making it 5.5 dB ENR (allowing us to use the low range scale on the HP342) and also providing a shunt path for the transformer coupled high voltage video spike that was propagated to the coax connector by the wideband pickup balun transformer used in the source. BTW waveguide gas tubes dont have the video spike, as waveguide is a HPF, wont propagate the video energy, so old waveguide noise sources are GREAT for the microwave bands, and cheap!

Almost all the early AIL and HP solid state noise sources were made by MSC (Microwave Semiconductor Company) in N.J. (Steve Kostro used to work there!), supplied as modules (little gold blocks with an SMA connector on them). When MSC went tits-up, killed by the MIMIC program and Jim D., there were a few other noise source makers such as Micronetics, who continued to make noise diodes and supply to source makers. I dont know where HP (Agilent) and Maury Microwave (new owner of the AIL/Eaton product line) get their diodes from now. I doubt they make their own.

I have had AIL 74's (vacuum tube version), 75's (solid state), and now have Eaton 2075, also had HP 340/342's have two 8970's now (kinda my favorite). Have as well as a hand full of the Sanders things. All of these rigs, by different makers, seem to work, but take a little skill and practice to get meaningful number. The older ones were not inferior, just generally harder to use (read, “more skill needed”).

I attended both the AIL and HP noise figure measurement seminars, waaay back when in the 1980's, and remember that both companies claimed no better than plus and minus .5 dB absolute accuracy on noise figure measurement! This was due to mismatch, calibration inherent instrument accuracies, LSB's on ADAC's etc, all adding up! If you wanted better, they would teach use how to do manual Y factor measurements with hot/cold sources to get better accuracy on absolute measurements. Relative measurements (like noise figure contests at the MUD etc.) are different. You want to see how good your amp is relative to all others. But you really cant come away claiming your amp is.25 dB noise figure from an 8970 measurement setup! All you can do is say, “My amp is better than yours, on the bench, in a 50 ohm system, that is well behaved, with no out of band impedance weirdnesses presented to to my amp”!

I agree with the concept that using the antenna-preamp combo in situ is the best way to evaluate NF, but a little tough for the production line…. Also, hot/cold VSWR of source and noise source match of the first stage are very important, but mostly were considered second order effects, and didn't stand in the way of tweaking for optimum relative performance. We used to use a double stub tuner and a line stretcher to rotate a input mismatch around the Smith chart to evaluate effect of source impedance change on noise figure.

Simple enough to do at home if you really care. AIL had some GREAT little app notes on this stuff, and even a noise figure slide rule (I have one somewhere….). K2RIW probably has this stuff somewhere as well, he worked there!

My work now is measuring noise using astronomical radio star sources for T sys and G/T work. Even use a little liquid nitrogen in a cold load for really precise Y factor measurements. I use a 35 dB ENR diode injected through a 30 dB coupler and level set attenuator to calibrate the 21M R-A dish at L band here at Morehead State, for “on the fly” noise cal marks on radio astronomy observations.

So there are a variety of cheap instruments and sources out there, even venerable gas tubes which are useful for modern measurements if you know how to tame them. Why build a source, when you can have professional one pretty cheap! Cant make it as good as you can buy (25.00 for a gas source, whats your time worth?)(Unless its a learning experience, answer in of itself, but I usually build something because I NEED it to do some other work).

73

Jeff Kruth

WA3ZKR