[R-390] Measuring sensitivity

[Roy Morgan]

At 12:30 PM 3/12/2007, Mark Huss wrote:
>...I have seen traceable measurements down around 0.2 uV (probably 2khz 
>bandwidth or less, it was not specified). Of course, the problem here is 
>that most of the readings you see on the Internet are not traceable (to 
>NIST standards), because calibration is expensive.

Mark and all,

May I comment on the "NIST Traceable" idea that comes up from time to time?

Note: I do work at NIST, but not in RF or Electrical calibration. I simply 
have a bit of information and some ideas on the topic of calibrations.

If someone tells you that an instrument has been calibrated "traceable to 
NIST" here is what they likely means:  The calibrations have been made in a 
laboratory which in turn has calibrated its instruments by means of 
materials or instruments that have been in turn calibrated against 
something that was actually AT NIST for calibration.

Let's see how this works: Let's say you want to know that your RF 
Millivoltmeter is accurate.

The NIST website is
http://www.nist.gov/
Click on
NIST Laboratories: provide measurements and standards for U.S. industry.
         ...
         Electronics and electrical engineering
and get to:  http://www.eeel.nist.gov/

The Eectronics and Electrical Engineering Laboratory provides the 
fundamental basis for all electrical measurements in the United States.
Then find Calibrations:  http://ts.nist.gov/MeasurementServices/Calibrations/
Then find the AC and DC current and voltage calibrations information at:
http://ts.nist.gov/MeasurementServices/Calibrations/Voltage.cfm

Among the many many calibrations discussed, one is called
"Special 25–Point Test of Digital Multimeters (DMMs), by Prearrangement 
(53202S–53203S)  This is a special reduced cost, 25–point test covering all 
five functions (ac and dc voltage and current, and dc resistance) of most 
precision DMMs. ..."
Under this test, we learn that the calibration point of 0.1 volt at 1 mc 
has a minimum uncertainty of  1000 parts per million.  That is 0.001 
volts.  This calibration costs $1600.

Reading further, we can see that the uncertainty involved in higher 
precision measurements (done with thermal conversion methods) is about one 
third of that.  Calibrations of this sort may cost $2000 or more.

So a commercial calibration lab can send its very best millivoltmeter or 
thermal conversion cell to NIST with a couple grand and be reasonably sure 
that it is correct within a very small amount.  Then the lab standards 
person compares the millivolt meter to be used for routine calibrations to 
the NIST calibrated standard to see if IT is right, and the accuracy drops 
an order of magnitude. Then when you add in the uncertainties associated 
with calibrating some one else's meter in the setup to be used, the 
accuracty may drop another order of magnitude.

You can see that if someone reports the sensitivity of an R-390A to be 
0.354 micro volts, that person is being quite cavalier with the numbers.

Mark goes on to point out that there are uncertainties in our home test 
setups, and also "... there are at least three methods to measure S+N/N 
ratio. All valid. All giving different results everything being equal."

It seems to me that we should pay attention to the setup we have, ask 
reasonable questions about how to get reasonable measurements, and pay 
attention to such things as leakage and what methods we use.

THEN, we might be able to say that we measured a radios sensitivity as "0.5 
microvolt, but it might be from 0.3 to 0.9" or some such.

There's a lot I don't know about all this.  For example, Mark says: "Then 
we would finish off by adjusting the IF Gain using Tangential Sensitivity 
readings."

I have no idea what that is, and would like to hear more about it.

Roy


- Roy Morgan, K1LKY since 1959 - Keep 'em Glowing
13033 Downey Mill Road, Lovettsville, VA 20180
Phone 540-822-5911   Cell 301-928-7794
Work: Voice: 301-975-3254,  Fax: 301-975-6097
roy.morgan at nist.gov --