I built a noise source as described by Bill Sabin, W0IYH in QST (1995)
and EMRFD. Works fine, but now it's time for calibration. Chatting with
Bill, he reinforces the notion that the best way to go is comparison
with another calibrated source, or a direct measurement with a
calibrated noise figure test set. NoiseCom used to provide this service
to hams for $50, but now they claim they have dropped that offer (it's
about $600, now). So my question is, is there anyone in the SF Bay
Area (I live in Livermore) who might be able to help? If someone owns a
source or test set, perhaps we could get together. I'm open to other
suggestions, too.

FYI, this is primarily for testing HF rcvr performance. I'm building an
R2-style DC rcvr, and it's turned into a nice research and education
project, so I go a bit overboard sometimes, but it sure is fun. I'm
also using an IQPro DDS VFO. I wrote a couple reports on its
performance that are up on the AA0ZZ website.

This is my first post to this group. There seem to be a lot of talented
folks lurking out there... Thanks in advance for any help.

73,
Gary, WB9JPS

On 29 Sep 2006 19:26:00 -0700, "Gary, WB9JPS" wrote:

FYI, this is primarily for testing HF rcvr performance.

sarcasm
I'm not sure, but at HF, can't you get a reasonable noise performance
measurement by pushing the radio off the bench and using a Radio Shack
sound level meter?
/sarcasm

Sorry.

From: Gary, WB9JPS on Fri, Sep 29 2006 7:26 pm

I built a noise source as described by Bill Sabin, W0IYH in QST (1995)
and EMRFD. Works fine, but now it's time for calibration. Chatting with
Bill, he reinforces the notion that the best way to go is comparison
with another calibrated source, or a direct measurement with a
calibrated noise figure test set. NoiseCom used to provide this service
to hams for $50, but now they claim they have dropped that offer (it's
about $600, now). So my question is, is there anyone in the SF Bay
Area (I live in Livermore) who might be able to help? If someone owns a
source or test set, perhaps we could get together. I'm open to other
suggestions, too.

FYI, this is primarily for testing HF rcvr performance. I'm building an
R2-style DC rcvr, and it's turned into a nice research and education
project, so I go a bit overboard sometimes, but it sure is fun. I'm
also using an IQPro DDS VFO. I wrote a couple reports on its
performance that are up on the AA0ZZ website.

If you can't get a calibrated comparison noise source, here's
some tips on workshop calibration techniques on random noise:

Random noise voltage or current can ONLY be calibrated by
a TRUE Root Mean Square measuring instrument. By "true" I
mean a thermionic type such as an RF power meter (thermistor,
bolometer, etc. sensor). Few voltmeters on the market have
TRUE RMS measuring capability; those that do are specified
as such and rather on the expensive side.

A very low-level random noise source can seldom be measured
directly by an available-market voltmeter or power meter.* To
calibrate it with such instruments requires amplification. A
receiver can provide that - provided - it is carefully checked
for linearity. By linearity I mean one that does not clip or
saturate the RF or IF or detector. A wideband TRUE RMS volt-
meter can measure the last IF output and do a comparison
between IF signals from a signal generator and the random
noise source into the front end. If the front end noise is
significant, the total noise input-equivalent will be the
good old square root of the sum of the squares of the noise
sources; knowing the true RMS values of both and just one
will let you find the true RMS value of input noise voltage
of the unknown. [kind of nuisance grunge work but then
most metrological work calibration is of that nature :-(]

* Jim Williams of Linear Semiconductor had a nice article in
EDN a few years back, pushing a Linear Semi true RMS IC. In
that was a list of available "RMS implied" voltmeters with
their sensor types stated. The cheapo averaging circuits in
common multimeters could be LOW on true RMS voltage readings
down to half of readings (despite their stated accuracy
specs). The Linear Semi IC apparently didn't sell well and
was taken off of active production. Perhaps only National
Semiconductor has a rough equivalent?


This is my first post to this group. There seem to be a lot of talented
folks lurking out there... Thanks in advance for any help.

Living in southern California I don't get up to mid-state
often. Difficult to drop over with a comparison source.
:-) However, a bunch of us lurking in here have been
"around the horn" a few times and we can supply a few
items of information to help. [or maybe to add to the
"random noise" level? :-) ]

wrote:
. . .
Random noise voltage or current can ONLY be calibrated by
a TRUE Root Mean Square measuring instrument. By "true" I
mean a thermionic type such as an RF power meter (thermistor,
bolometer, etc. sensor). Few voltmeters on the market have
TRUE RMS measuring capability; those that do are specified
as such and rather on the expensive side.
. . .

Not too long ago, I got an HP3400A (10 MHz bandwidth) on eBay for the
purpose of measuring noise. Don't recall what I paid, but it was very
reasonable. Since truly random noise also has low-frequency
fluctuations, an RMS meter reading wanders quite a bit while measuring
noise. I find it easier to eyeball interpolate an average reading with
the analog meter of the HP3400A than with a digital meter.

Roy Lewallen, W7EL

Gary, WB9JPS wrote:
I built a noise source as described by Bill Sabin, W0IYH in QST (1995)
and EMRFD. Works fine, but now it's time for calibration. Chatting with
Bill, he reinforces the notion that the best way to go is comparison
with another calibrated source, or a direct measurement with a
calibrated noise figure test set. NoiseCom used to provide this service
to hams for $50, but now they claim they have dropped that offer (it's
about $600, now). So my question is, is there anyone in the SF Bay
Area (I live in Livermore) who might be able to help? If someone owns a
source or test set, perhaps we could get together. I'm open to other
suggestions, too.

FYI, this is primarily for testing HF rcvr performance. I'm building an
R2-style DC rcvr, and it's turned into a nice research and education
project, so I go a bit overboard sometimes, but it sure is fun. I'm
also using an IQPro DDS VFO. I wrote a couple reports on its
performance that are up on the AA0ZZ website.


Sorry I can't help with a calibrated noise source from this side of the
Atlantic. The amateur microwave community hold regular meetings that
always include a measurement session using calibrated professional test
equipment. Given a calibrated professional noise source and a NF test
set, it is very easy to transfer the ENR calibration to your HB source
(especially if the room temperature happens to be 290K :-)

However, you don't really need a professional test set. You can make
very accurate measurements of relative RMS noise power levels using a PC
sound card, and this opens the way to NF measurements. The only hardware
accessories needed are a simple linear downconverter of the type
described in EMRFD, and a noise source that can be gated by an output
from the PC. For details and software, see:


The only other issue is that almost all professional noise sources are
designed and calibrated mainly for UHF/microwave use. For your HF
application, you'll typically be limited to just one transfer
calibration point at 10MHz. You can rely on a professional source to
have no nasty resonances between calibration points, but the ENR will
probably roll off rapidly below 10MHz owing to the relatively low value
of the DC blocking capacitor.



--
73 from Ian GM3SEK 'In Practice' columnist for RadCom (RSGB)
http://www.ifwtech.co.uk/g3sek

Roy Lewallen wrote:
wrote:
. . .
Random noise voltage or current can ONLY be calibrated by
a TRUE Root Mean Square measuring instrument. By "true" I
mean a thermionic type such as an RF power meter (thermistor,
bolometer, etc. sensor). Few voltmeters on the market have
TRUE RMS measuring capability; those that do are specified
as such and rather on the expensive side.
. . .

Not too long ago, I got an HP3400A (10 MHz bandwidth) on eBay for the
purpose of measuring noise. Don't recall what I paid, but it was very
reasonable. Since truly random noise also has low-frequency
fluctuations, an RMS meter reading wanders quite a bit while measuring
noise. I find it easier to eyeball interpolate an average reading with
the analog meter of the HP3400A than with a digital meter.

I agree there on analog display versus digital.

Dug out my CD of old "saves" and found the Jim Williams article
in the 11 May 2000 edition of EDN (www.ednmag.com). In that
he lists the following as having zero error when compared to using
the Linear true-RMS IC: HP 3400, 3403C, 3478; Fluke 8920A.

By the bye, the 1967-era HP pseudorandom noise generator has
the following "calibration" procedure to check RMS output voltage:
Connect it to an HP digital voltmeter-recorder, set it to slow
speed and record 1023 voltage readings (corresponding to the
slow-speed 'flat' pseudorandom waveform parts), then take the
sum of the squares of all those readings and find the square root
of it. Heh heh heh...I can just see a calibration technician doing
all that work with a mechanical calculator in the 60s...with the QC
chief storming in after an hour or so asking "You STILL working
on that?!?" :-)

Gary, WB9JPS wrote:

I built a noise source as described by Bill Sabin, W0IYH in QST (1995)
and EMRFD. Works fine, but now it's time for calibration. Chatting with
Bill, he reinforces the notion that the best way to go is comparison
with another calibrated source, or a direct measurement with a
calibrated noise figure test set. NoiseCom used to provide this service
to hams for $50, but now they claim they have dropped that offer (it's
about $600, now). So my question is, is there anyone in the SF Bay
Area (I live in Livermore) who might be able to help? If someone owns a
source or test set, perhaps we could get together. I'm open to other
suggestions, too.

FYI, this is primarily for testing HF rcvr performance. I'm building an
R2-style DC rcvr, and it's turned into a nice research and education
project, so I go a bit overboard sometimes, but it sure is fun. I'm
also using an IQPro DDS VFO. I wrote a couple reports on its
performance that are up on the AA0ZZ website.

This is my first post to this group. There seem to be a lot of talented
folks lurking out there... Thanks in advance for any help.

73,
Gary, WB9JPS

For very low noise frequencies, perhaps you could use a pseudo-random
sequence generator made from fast logic clocked at say a hundred MHz, and
calculate what the "noise" power spectral density in the output signal is.
My only other suggestion is if you can make a 50 Ohm resistive terminator
on the end a teflon cable that will go into an oven, then you can just
measure the noise temperature directly... as long as you don't want an ENR
over 3dB.

Chris