It has been a few years....

The noise density in the JFET has a strong 1/f characteristic. This means it
increases by 6 db each time you halve the measurement frequency. For
silicon, I recall the corner frequency being in the ~ 100 Hz.range. The unit
"nV/root-hertz" is the noise density, not the noise. You need to measure the
voltage within a finite bandwidth. From that, you can derive
the noise power, then the noise density. When measuring below the corner
frequency you have to account for this 1/f slope.

Thermal noise in a 50 ohm system at room temperature is about 0.9
nV/root-hertz. In a 1000 ohm system it is about 4 nV/root-hertz. This will
set the lower measurement limit, you test equipment will probably limit you
to worse than this. Using adequate measurement bandwidth and correct circuit
resistance (within the confines of the JFET parameter range) will provide
sufficient noise voltage to actually measure, but it will take some gain
(via a very low noise measurement amplifier) to produce something you can
hook to a measuring device.

There are lots of subtle way to make noise measurement errors. I recommend
that you use the test setup that is specified by the manufacturer to make
the measurement, otherwise your results will likely not correlate with their
numbers.

-- Tom, N5EG


wrote in message
...
I need to measure the "equivalent noise voltage", in units of
nV/square root( hz) of a JFETs drain-source at a given Vds & Ids and
F=10 hz. How do I do that? What's throwing me is the F=10 Hz. What do
I do with that? I thought to take the measurement, I would bias the
FET to the required Vds and Ids and measure the voltage across the
drain - source with an RMS meter. Then divide that by the sqrt of the
bandwidth of the RMS meter. I've Googled, but didn't find anything to
straighten me out. Any help?

Thanks alot for your help, I really appreciate it. Would you look at
this FFT analyzer: http://www.srsys.com/products/SR760770.htm and do
you think it will "see" 20nV/sqrt(hz) or will I need an amp? What do
you use as a low noise amp? This FFT analyzer has noise floor of 10
nV/sqrt(hz) and dynamic range of 90 db at -60 dbV full scale range.
Again. I really appriciate any help.

(Tom Bruhns) wrote in message om...
It's probably a lot more common to express the noise referred back to
the input, but if it's output noise you want, that's fine too. You'd
want to either use a drain load that's much higher resistance than the
output (drain) resistance of the FET at that bias--or else account for
the shorting effect of the load.

Often the noise is so low you would be served well by amplifying by
100 or 1000 so your "meter" has enough to measure accurately.

If I were doing it, my "meter" would be an FFT-based spectrum
analyzer, which would be able to read the amplitude directly in
V/rtHz; then I'd just divide by the gain, if I'd used an amplifier.
(I have a very low noise amplifier I use for just such measurements,
so the amplifier doesn't contribute a lot to the measurement, but even
then, it may be necessary to subtract out the amplifier's noise
contribution.) If you really need to know it at 10Hz, you probably
should not use a bandwidth much more than a couple of Hz. An
FFT-based analyzer (or possibly other spectrum analyzer, or even a
wave meter/analyzer) makes that fairly easy. Not sure this will help
in your particular situation.

Cheers,
Tom

wrote in message . ..
I need to measure the "equivalent noise voltage", in units of
nV/square root( hz) of a JFETs drain-source at a given Vds & Ids and
F=10 hz. How do I do that? What's throwing me is the F=10 Hz. What do
I do with that? I thought to take the measurement, I would bias the
FET to the required Vds and Ids and measure the voltage across the
drain - source with an RMS meter. Then divide that by the sqrt of the
bandwidth of the RMS meter. I've Googled, but didn't find anything to
straighten me out. Any help?

Thanks alot for your help, I really appreciate it. Would you look at
this FFT analyzer: http://www.srsys.com/products/SR760770.htm and do
you think it will "see" 20nV/sqrt(hz) or will I need an amp? What do
you use as a low noise amp? This FFT analyzer has noise floor of 10
nV/sqrt(hz) and dynamic range of 90 db at -60 dbV full scale range.
Again. I really appriciate any help.

(Tom Bruhns) wrote in message om...
It's probably a lot more common to express the noise referred back to
the input, but if it's output noise you want, that's fine too. You'd
want to either use a drain load that's much higher resistance than the
output (drain) resistance of the FET at that bias--or else account for
the shorting effect of the load.

Often the noise is so low you would be served well by amplifying by
100 or 1000 so your "meter" has enough to measure accurately.

If I were doing it, my "meter" would be an FFT-based spectrum
analyzer, which would be able to read the amplitude directly in
V/rtHz; then I'd just divide by the gain, if I'd used an amplifier.
(I have a very low noise amplifier I use for just such measurements,
so the amplifier doesn't contribute a lot to the measurement, but even
then, it may be necessary to subtract out the amplifier's noise
contribution.) If you really need to know it at 10Hz, you probably
should not use a bandwidth much more than a couple of Hz. An
FFT-based analyzer (or possibly other spectrum analyzer, or even a
wave meter/analyzer) makes that fairly easy. Not sure this will help
in your particular situation.

Cheers,
Tom

wrote in message . ..
I need to measure the "equivalent noise voltage", in units of
nV/square root( hz) of a JFETs drain-source at a given Vds & Ids and
F=10 hz. How do I do that? What's throwing me is the F=10 Hz. What do
I do with that? I thought to take the measurement, I would bias the
FET to the required Vds and Ids and measure the voltage across the
drain - source with an RMS meter. Then divide that by the sqrt of the
bandwidth of the RMS meter. I've Googled, but didn't find anything to
straighten me out. Any help?