[sdiy] Uniformly distributed noise generator?

cheater00 . cheater00 at gmail.com
Mon Jun 10 20:36:22 CEST 2013

Hi Scott,
I'll first give you a more meta-topical (but more important) answer
and then a few examples.

In physics, experiments are reflections of conceptual models. So you
have an idea and you test it. Or you make a test and then try
different ideas.

When doing this, you want your conceptual framework to be as simple as possible.

The conceptual frameworks in statistics always assume that your
sources of randomness are true, not pseudo-random.

Above I have given you an example of where this true randomness
doesn't work (with the two pseudo-random units that are in close

In the end, I can't follow through with my assumption that the source
is truly random, if it's only pseudo-random. This means I cannot use
this conceptual model, and I have to get a much more complicated one.
This means I can also not use mathematical frameworks. I have to use
much more complicated maths that considered correlation between
randomness sources; if at all they exist, and a lot of problems have
likely not been considered in such a way. This means that I have to
throw away centuries of advancements by the brightest minds the world
has ever seen. This is a huge disadvantage.

Whether or not it actually matters for a specific measurement is a
different thing; it's that I can't undeniably prove that my experiment
reflects the conceptual framework I'm using on paper.

Below is an email which I've sent to a member of this list off-list.
It describes a few simple things I'd like to try. Certainly there are
much more complicated things which rely on randomness much more, but
there's no time to come up with examples.



the project is actually to come up with a measurement methodology. In
physics, Monte Carlo and generally statistical methods are very
popular and they give great results in measurement and simulation.

In normal measurements, you're trying to measure some quantity that
changes as a variable sweeps. So for example, you change the voltage
across the transistor junction from 0 to 1V, and then you look at how
the current across that junction changed. You get a curve that
characterizes the transistor.

When applying the Monte Carlo method, you do "random sampling". This
means that you set the variable you control to a random quantity, and
read the output off. Then you set your variable to another random
quantity, and read off again. And so on until you've done enough.

So for example, you'd roll the dices and randomly select a voltage
across the transistor junction between 0 and 1V. Then you'd check the
current. Then you'd select another voltage randomly, etc. Each of
those measurements is denoted as a single point on some current vs
voltage graph. As you go, you will have made many measurements that
are so close together that they form a curve.

I would like to experiment with this measurement methodology because I
haven't seen anyone use it in real-world measurements and it seems
like a shame. I would like to see what possibilities there are.

One example of where random measurements could be better follows: they
assume that the measured device has no state. This means, for example,
no hysteresis. Hysteresis is not visible in a normal sweep like I
described in the second paragraph. However it would be immediately
visible with random measurements - you wouldn't get a fine curve;
instead you'd get a much thicker line (if it's a line at all).

Another example experiment I'd like to perform is: if you sample a
periodic waveform at uniformly random intervals, how difficult will it
be to reconstruct the original waveform? This can be attempted on an
oscilloscope in the following way:

1. Feed the waveform into the oscilloscope as you normally would. Get
a stable waveform display, a good "lock" (in measurement nomenclature,
you say your time base is triggering correctly).

What happens is that the time base, which is a saw-up oscillator,
sweeps the electron beam across the phosphor from left to right (as
the saw rises). As it moves the electron beam horizontally, the
amplifier ("vertical amplifier") moves it up and down, so if it's a
sinewave then it's going to move the beam up and down and together
with the horizontal movement the electron beam will actually draw a

So now do the following: turn off the electron beam, and only turn it
on while the time base voltage (saw wave) is exactly equal to your
source of randomness. You do this via so called "Z modulation", which
is just turning the brightness of the electron beam down or up.

Will you see the same waveform as if you weren't doing the comparison?
This can be easily checked in software on a computer, but I'd like to
build up a methodology where I could integrate this with other
measurements one can do with an oscilloscope (and there's a whole
world of things you could come up with!)

On Mon, Jun 10, 2013 at 8:07 PM, Scott Gravenhorst <music.maker at gte.net> wrote:
> "cheater00 ." <cheater00 at gmail.com> wrote:
> Damian,
>>That's one reason why for me pseudorandom is not necessarily random enough.
> And when I say "EXACTLY", I mean, fully describe the test, it's conditions and
> everything else about it, otherwise all we are doing is playing mental mumbly-peg.
> -- ScottG
> ________________________________________________________________________
> -- Scott Gravenhorst
> -- FPGA MIDI Synth Info: jovianpyx.dyndns.org:8080/public/FPGA_synth/
> -- FatMan Mods Etc.: jovianpyx.dyndns.org:8080/public/fatman/
> -- Some Random Electronics Bits: jovianpyx.dyndns.org:8080/public/electronics/
> -- When the going gets tough, the tough use the command line.
> _______________________________________________
> Synth-diy mailing list
> Synth-diy at dropmix.xs4all.nl
> http://dropmix.xs4all.nl/mailman/listinfo/synth-diy

More information about the Synth-diy mailing list