looking for article in EDN: Test Whether a Noise Source is Gaussian

[[email protected]]

I'm looking for a copy of an article which appeared a long time ago in
EDN:
"Test Whether a Noise Source is Gaussian", EDN, 11 Dec. 1, pp.
272-274, S.T. Michaels,

Is this article in someone's collection? I would really appreciate if
I could get a copy of it

regards,
Hugo Coolens

 

[NotShown]

The correct reference seems to be "Test Whether a Noise Source is
Gaussian", EDN, 11 Dec. 1986, pp. 272-274, S.T. Michaels.

I would be interested in a copy of the article as well.

I have posted this correction 3 days ago, but it did not show up
on my server or the WEB. Sorry for any ghostly duplicates that
might show up.

Gerhard van den Berg
CSIR

Hi Gerhard,

Most people who had copies probably got rid of them when the
internet arrived. You might try a local library. According to the
Invention Registration below, the author is S.R. Michaels, not S.T.
Michaels. That would make a difference in the library search.

There is a pretty good description in the US Statuatory Invention
Registration H1458. You can download the pdf with schematic and
complete description. Go to

http://www.google.com/patents

and enter "H1458".

The first item in the result is "SIGNAL AMPLITUDE DISTRIBUTION
ANALYZER", by Robert A. Slack. This is the one you want.

 

[NotShown]

[...]

Thanks for the references. We have a reasonable science library that
might have
the EDN article or access to it.
I have already persued patent H1458 and the excellent and interesting
paper by Phil Hobbs http://electrooptical.net/www/isicl/isiclAO.pdf.

My challenge is to measure the degree of Gaussianicity (sounds like
insanity -
I prefer Gaussian or non-Gaussian) or not, of various noise
sources/designs.

Many thanks for the feedback.

Gerhard van den Berg

You might be interested in work done by Terry Ritter on semiconductor
noise generators.

Here's the start page:

http://www.ciphersbyritter.com/RADELECT/MEASNOIS/MEASNOIS.HTM

Next page is noise measurements of zener, bipolar, led, diodes, ic's:

http://www.ciphersbyritter.com/RADELECT/MEASNOIS/NOISMEA1.HTM

Next is "Analyzing Noise Generators". His comments are very true:

"Any electronics experimenter can build a noise source.
Unfortunately, building a good noise source is harder than it looks.
Correctly using such a source is harder still."

http://www.ciphersbyritter.com/NOISE/NOISRC.HTM

Check the waveform of zener noise near the bottom of the page. The
plot is annotated "PCM - F:\Noise\NoiseWav\zcc1.wav". It is hard to see.

He remarks "See how the waveform looks more "fluffy" on the bottom?
It is possible that we are somehow clipping some of the signal."

Actually, this is the normal output of a zener noise generator. The
noise is caused by micro-avalanche discharges, which reduce the
voltage across the diode. This was discussed by Win and others in
great detail some years ago. The solution is to combine the signals
from two zeners and invert one in a differential op amp.

Next is "Statistical and Graphic Views of Noise". This is the most
interesting part. He shows graphs of Amplitude vs Frequency,
Autocorrelation, and Amplitude Distribution of 25 different methods:

http://www.ciphersbyritter.com/NOISE/NOISCHAR.HTM

This shows how difficult it is to make a good random noise generator.

 

[NotShown]

Thanks for all the feedback.
I am well aware of the excellent contribution by Terry Ritter on
various issues with regard to 'generating' and 'measuring'
true randomness.

My problem is that I need to measure up to 1 MHz and
possibly up to 10MHz. That is why I looked into possibly of
using the comparators technique. I hope using 2 to 5 ns comparators,
to do the distribution sampling, might give me reasonable
answers without blowing the budget.
The other issue is digitising such wideband analogue noise.
The experiments with the comparators might also
provide some insight into employing them for that purpose.

I might still look into the SpectraPlus software and compare the
results of the comparator technique with the results of the
sampling and signal analysis techniques.
It would also be a nice analysis tool to have for other experiments.

Gerhard van den Berg
CSIR

You need a really fast comparator to handle the peaks in 10MHz random
noise. I've had very good results with the Motorola MC100EP16 line
receiver. The device is temperature compensated, so you won't see much
drift on a long test run. Here's some more info:

220 ps Typical Propagation Delay
Maximum Frequency > 4 GHz Typical
200 fs RMS Jitter
4V p-p max input at 5.5V VEE

A small amount of hysteresis will kill any oscillation. You need a good
ground plane, very good bypassing, and very short leads. The output must be
50 ohm transmission line or coax. Use both outputs differentially to
minimize ground bounce when switching.

You need two separate comparators for positive and negative peaks, so you
can arrange the hysteresis so it doesn't affect the desired voltage level.

I've tried many other comparators, but none have come close to this one.