Dave Oldridge wrote:
Owen Duffy wrote in news:Xns999955EE72868nonenowhere@
61.9.191.5:


Dave Oldridge wrote in
. 159:


Near as I could measure it, the NF of the receiver after my mod was
1.2db. I had to resort to boiling and freezing water and a tiny dummy
load to measure it at all.

snip


This

still demands high resolution measurement of noise power.


Yes, anything less than 4 digits is just about useless.

That would be necessary but not sufficient.

I suspect that other aspects of the measurement introduce greater
uncertainty than the voltmeter. For instance, do you know the
reflection coefficient of the load to 4 digits? (that would be knowing
Z to about 0.1 ohm, at the RF frequency of interest), and is it stable
with temperature to that level?

For instance, a high quality load from Maury Microwave (a 2610F ) is
specified to have a VSWR of 1.005 from DC-1GHz, which is a reflection
coefficient of 0.0025. But that's only at 25C.

An Agilent metrology grade cal kit with N connectors specifies
rho0.00398 for the lowband loads, but only within 1 degree of the
specified temperature.
See, for example:
http://cp.literature.agilent.com/lit...5054-90049.pdf

A good thinfilm resistor might have a tempco of 5 ppm, with metal film
being around 50 ppm, and thick film more in the 200 ppm area. For a 100
degree change, that's a 500 ppm (for the thin film) or a reflection
coefficient change of 0.00025. Clearly one doesn't want to use any old
resistor for the calibration load here.


Measuring RF power to an accuracy of 1% is challenging. Your system is
measuring a change in noise power of 100K out of 300K, roughly, so
you've got a 30% change in noise power into the system.

The Y-factor method essentially plots two points (one at 273K another at
373K, if you're using ice and boiling water), and then calculates the
intercept at 0K. Since zeroK is about 3 times farther away than the
measurement's width, errors in the measurement are roughly tripled at
the intercept, and then doubled because you're using two measurements,
so an error of 1% in the power measurement leads to about 5% error in
the NF (if you're around 100K) (and this also applies if you have
consistent errors.. say both power measurements are 1% high, the NF will
come back as 105K instead of 100K).

A 5% measurement uncertainty for power (0.2dB) gets you about 25-30%
uncertainty in NF.

The best way to improve the accuracy is to push the low temperature
lower (e.g. with dry ice (195K) or LN2 (77K)), but that, of course,
aggravates the change in reflection coefficient of your load with
temperature.