System Noise Figure
One important parameter which indicates how well a spectrometer is
working is the noise figure of the preamp and total system. An
inexpensive method of determining the system noise performance is the
"hot/cold resistor" test. Done with care, this can provide valuable
insight into spectrometer performance.
The first step is constructing the noise source. Take a piece of RG58
coax cable and carefully strip the cover back about 2 inches. Now
slide the shield back as far as possible (about 1 inch), cut the exposed
center conductor to a 1/2 inch in length and strip off the insulation on
a 1/4 inch of the conductor. Solder a 1% 50 Ohm metal film resistor to
the center lead. Cover with shrink tubing and shrink the cover tight
leaving the unconnected end of the resistor exposed. Slide the shield
down over the resistor and solder the shield and the unconnected end of
the resistor together to make a shielded noise source. Take care to
make all leads as short as possible. In most environments, it is very
important to use a shielded noise source to help reduce outside RF
noise. Interference renders this measurement useless.
Carbon resistors increase resistance as much as 20% when cooled in
liquid nitrogen. For this reason, two sources need to be prepared if
carbon resistors are to be used. One which measures 50 Ohm at 20 degrees
C and one which measures 50 Ohm in liquid nitrogen. Determine the
values with an Ohm meter. The metal film resistors change only a few
percent when cooled by liquid nitrogen and are therefore preferred.
Connect the shielded 50 Ohm noise source to the input of the preamp in
place of the probe. If the entire system noise figure is to be
determined, leave all devices for transmitter coupling in place. Set
the spectrometer to operate on the frequency of interest with a large
sweep width (+/- 10,000 Hz) and a gain level set to fill the digitizer
more than half way when acquiring a single scan. Collect a scan and
determine the RMS of the noise after Fourier Transform with no
line-broadening. This can be done with computerized routines or by
plotting the noise and drawing a line on top and bottom of the noise
such that about 10% of the maximum noise excursions fall outside the
lines. Any obvious spikes in the noise spectrum can be ignored, but
their presence indicates interference in the NMR system.
Repeat the above process with the 50 Ohm liquid nitrogen noise source in
place of the probe while the source is being cooled by liquid nitrogen.
Process the data such that the same normalization constants are used and
determine the RMS value of the noise as above.
These two RMS noise values can be used to calculate the instrument noise
figure as indicated by the equation below:
NF(dB) = -1.279-10log10[1-(RMSc**2/RMSw**2)]
where RMSc is the value of the RMS noise in liquid Nitrogen and RMSw is
the value at room temperature (20oC)
A table relating noise figure to the ratio of the cold RMS noise value
to the warm RMS noise value (RMSc/RMSw) as calculated by the above
equation is shown in the table below:
Ratio NF(dB) Ratio NF(dB)
0.99 15.73 0.90 5.93
0.98 12.74 0.85 4.29
0.97 11.01 0.80 3.16
0.96 9.78 0.75 2.31
0.95 8.83 0.70 1.65
0.94 8.06 0.65 1.11
0.93 7.41 0.60 0.66
0.92 6.86 0.55 0.29
0.91 6.37
This test is easy to do, but requires careful experimental technique.
If a noise level difference cannot be observed between a hot and cold
resistor, make sure the system can observe an NMR signal from a standard
sample. If you are unable to obtain a reproducible system noise
figure, problems in the system noise profile could be presenting
limitations on spectrometer performance. A good procedure is to strip
the spectrometer receiving system to the bare minimum. If possible take
out any unnecessary lock filters (turn off the lock) and decoupler
filters from the receiving system. Remove the transmitter coupler if
possible. With this minimum system determine the system noise figure.
Anything more than 3.0 dB needs to be improved. Now add back the
removed components one by one. Repeat the test after the addition of
each component to determine how they affect the system noise figure.
Today's spectrometers typically give overall system noise figure of less
than 2.0 dB.
All versions of NUTS including the shareware and demo versions have the
ability to calculate the NMR systems noise figure. A command (NF or
NoiseFigure in the non-2-lettered command mode) can be used to calculate
an NMR instrument's noise figure. First, replace the NMR spectrometer's
probe with a 50 Ohm metal film resistor discussed above. Use the gain
and frequency settings typical of data collection for the nucleus of
interest. Take one scan of data with the resistor at room temperature
and save the file with a file name like "HOT.NMR". Cool the resistor in
liquid nitrogen and take another scan and save this file with a file
name like "COLD.NMR". Enter the NF command and follow the directions.
NUTS will load the two data files and calculate the system noise figure.
In the non-2-lettered command mode the user can also enter the name of
the hot file as argument 2 and the name of the cold file as argument 3.
woody_at_acornnmr.com
-----Original Message-----
> From: rajan [mailto:rajan_at_cabm.rutgers.edu]
> Sent: Thursday, September 19, 2002 6:42 PM
> To: ammrl_at_chemnmr.colorado.edu
> Subject: requesting info. about a noise source...
Hello all
I have been looking for some time to get a good white noise source
(typically 10 MHz -1 GHz) to test the noise figure of
my Varian INOVA spectrometers. I have contacted couple of vendors, but
1.) they have a very large band width source than the one I really need
and 2.) they are quite expensive. Has any of you rigged a reasonable
one in the lab yourself ? If not, any good place to go to?
Thanks a lot
rajan paranji
=====================================
RAJAN K. PARANJI, Ph.D.
NMR FACILITY MANAGER
Center for Advanced Biotechnology and Medicine
Rutgers University
679, Hoes Lane
Piscataway, NJ 08854
ph: 732 235 5327
=====================================
Received on Mon Sep 23 2002 - 10:37:57 MST