1. Home
  2. Electronics Design
  3. Old noise generator

Old noise generator

This is an Elgenco 331A gaussian noise generator. I know very little about it, though I did see an ad in an old journal proclaiming it's good to four sigmas. It seems to be roughly "audio frequency", by which I mean, the circuits are audio quality, no particular attention paid to high or low frequency bandwidth (AC coupled, no peaking coils). I would guess 10Hz to

200kHz or so.

Front view. Some old rackmount thing:

formatting link
So I'm guessing Beloit got it from NASA surplus or something. I found this thing on the curb one day outside of a house near the college, which I'm assuming is the residence of someone interested in Beloit physics...

Other side of the faceplate:

formatting link

Behind the faceplate, the noise generator and amplifiers:

formatting link
On the far right, two 6D4 gas thyratron triodes are set in magnetic fields, generating noise per the datasheet. They look to be connected like any old gas tube, with coupling capacitors going to the rest of the circuit. I'm going to guess they use two and mix them somewhere (I haven't traced the circuit yet), to make it extra randomized or something.

Signal leaves through the coax in the top-right corner, which goes to the output stage on the left (coupled through that big fat 1.0uF). I think it's a 12AU6 cathode follower.

On the far left, two PCBs hold some germanium transistors (typical date codes are 1963), the TO-39's are 2N398As, while the TO-1's are something rather fancy (I forget the number, but they're epitaxial mesa germaniums with almost 100MHz fT). I don't really know what to make of it, but it must be in the signal path, why else use such fast transistors.

PCB closeups:

formatting link
Gotta love that clear glass body RN60B!

The power supply hangs in the back of the rackmount chassis (which must be annoying for the moment arm...)

formatting link
Two potted transformers supply everything, I think one is for heaters, the other for HV, and the thermal relay (bottom right) delays HV by a minute or whatever. Rectifiers are all silicon (I guess).

A 6BX7 and associated tubes (12AX7, and the other one may be another 12AX7) regulate high voltage. The two PCBs and heatsink appear to regulate 12.6V for heaters. The TO-3's are all 30V, 1-5A germaniums. The two pots say they're for + and - 156V, I guess one or both go to the tube regulator part. Either that or some of the TO-39's are HV germaniums, I haven't checked.

Rear view:

formatting link
Filament supply seems to be the bottom transformer, feeding a press-mounted FWB and computer-grade cap.

Tim

--
Deep Friar: a very philosophical monk.
Website: http://webpages.charter.net/dawill/tmoranwms
Reply to
Tim Williams
Loading thread data ...

Yikes, that must have cost a fortune, back when.

The gas tubes tended to make asymmetric (not prefectly Gaussian) noise, so people subtracted the output of two tubes. Nowadays, pseudo-random is easier.

Hey, we have clouds too.

ftp://66.117.156.8/c1.jpg

ftp://66.117.156.8/c5.jpg

ftp://66.117.156.8/c6.jpg

ftp://66.117.156.8/c7.jpg

ftp://66.117.156.8/Kit1.JPG

John

Reply to
John Larkin

Yup, a regular zener like a 1N758 will make pretty good gausian noise at several mA bias, ballpark 300 nV/rootHz. As the current drops, the noise starts to get asymmetric and at low currents turns into a very noisy sawtooth. Subtracting two diode signals would be good here, too, if extreme noise quality were needed.

Our house overlooks the Alemany Gap, a break in the coast range hills between the ocean and the bay. It's a major fog duct. Some days the fog settles into the gap and it looks like a river with houses on the shore. Some days we can't see the next street. We leave the heat on all year.

I spent some time in Madison, mostly summer, and it was nice, drinking beer in the student union on the lake. Lots of good coffeehouses and restaurants and such. Great Farmers' Market outside the capital building on Saturdays.

I helped this thing

formatting link

get started, a spinout of the ME department at UW.

Hmmm, it's almost time for pecan pie.

John

Reply to
John Larkin

Yep, lots of good food, good beer and most of all, great cheese. ;-)

Interesting. So, they put a big stinking pulse into it, which rips off and ionizes exactly the surface atoms, which are then detected in time and space, I suppose with microchannel plates?

Ahh, we didn't have any. But we did have pumpkin chocolate chip cheesecake, which is also quite delectable.

Tim

--
Deep Friar: a very philosophical monk.
Website: http://webpages.charter.net/dawill/tmoranwms
Reply to
Tim Williams

Yup. The idea is to rip off, statistically, maybe one ion every 5 or

10 pulses to avoid pileup confusion. The ion flies radially away from the nominally spherical tip, 100 nm diameter maybe, and it hits a microchannel plate. The field strengths are insane. The magnification is millions. The MCP makes a burst of electrons out the other side, and that hits a 2-axis delay-line detector. If we measure the four pulses that emerge from the two crossed serpentine delay lines, we can figure out the X-Y hit location and the time of flight (ie, the mass = elemental composition and isotope of that ion.) So as the tip is eroded away, we can reconstruct a 3D image of the atoms. The tip is cooled to 40K or some such to keep the atoms from wandering around. Ultra-high vacuum, too. It all gets expensive.

formatting link

For some materials, it's better to whack the sample with a femtosecond laser rather than a high-voltage pulse, to eject the ion. Did I mention expensive?

Kids these days!

John

Reply to
John Larkin

Sorry. I thought this was a thread about our resident right wing nuts.

Never mind.

-- Paul Hovnanian mailto: snipped-for-privacy@Hovnanian.com

------------------------------------------------------------------ Nondeterminism means never having to say you are wrong.

Reply to
Paul Hovnanian P.E.

Ah, good point. They're a little slow for this application, but you might tempt some cryptologists into downloading their entropy. Run the archives through a hash function and you've got a lovely source of random numbers.

Tim

--
Deep Friar: a very philosophical monk.
Website: http://webpages.charter.net/dawill/tmoranwms
Reply to
Tim Williams

On Thu, 26 Nov 2009 19:02:08 -0800, John Larkin

Would adding do something similar? One example I found on the net suggest 24V zeners with a current source of about 20mA. However I have plenty of 9.1V/500mW smd ones in the cupboard that I could use in series.

Joop

Reply to
Joop

They're fine at a few mA. 9.1V might be a little low; I forget how the noise of zener effect compares to avalanche. Generally, the zener effect ends around 6V (the crossover point having an interestingly low tempco right around 6.2-6.8V), so >6V zeners are actually avalanche diodes.

Tim

--
Deep Friar: a very philosophical monk.
Website: http://webpages.charter.net/dawill/tmoranwms
Reply to
Tim Williams

The Central Limit Theorem says that summing a number of imperfect noise sources trends towards Gaussian. I don't know if putting zeners in series is equivalent to summing noise sources, or if three 9 volt zeners in series behave like a single 27. I doubt it.

Just because something is on the net doesn't make it sensible. 9.1 volt zeners, biased to maybe 5 mA, make nice noise. Sum a few or better yet sum and difference an even number of them... that's probably a faster approach to good gaussian noise than straight summing.

Summing does push the probability distribution towards gaussian, but it doesn't much improve spectral flatness.

John

Reply to
John Larkin

I think it was only mentioned that the 24V zener produced more noise power than a 9V one. On flatness the different voltages were not compared. The current had some influence on flatness though. I guess I will just experiment and see what it brings.

Thanks for responding.

Joop

Reply to
Joop

Gain is cheap. The noise from an opamp or a mmic will tend to be of lower quality than the noise from a zener, so all you need is enough zener noise to swamp the noise of whatever amplifier you'll use after the zener. A 9.1 volt zener will make several hundred nv/rthz, plenty to blast past the noise of a good opamp or mmic, which would be a few nv/rthz.

If you do get any data on zener noise for various diodes at different currents, please post it.

John

Reply to
John Larkin
.

How high in frequency does the noise extend with reasonable flatness, John?

Jim

Reply to
RST Engineering

Offhand, I once hooked up a 1N4742 at around 5mA and 'counted' the frequency. It was somewhere in the 160-240kHz range, so I guess bandwidth won't be far past there. Zeners have pretty fat junctions, so you can't expect much. Low wattage zeners might have some RF performance.

Tim

--
Deep Friar: a very philosophical monk.
Website: http://webpages.charter.net/dawill/tmoranwms
Reply to
Tim Williams

Don't know. It's clearly limited by the corner formed by the junction capacitance and the zener dynamic impedance.

Here's some good stuff:

formatting link

They have diodes rated from 100 KHz to 100 GHz.

John

Reply to
John Larkin

At one time they would sell off-spec diodes to the experimenter for dirt cheap. Last time I bought a NoiseCom diode was back when I was spending Other Peoples' Money and didn't care the expense.

What I'm trying to make is a noise source fairly flat from roughly 1M to 1G to use as a poor man's sweep generator to use with a spectrum analyzer to examine various filter topologies.

My first thought was to use a zener or perhaps a hot carrier in breakdown and amplify the noise using a cascade of two or three of the little Mini-Circuits Labs mmics in series to get me up to some decent power out -- say 0 dBm or so.

Got any other thoughts on how to generate flat noise to a Gig on the cheap?

Jim

Reply to
RST Engineering

Photocurrent shot noise. Get a 2.5 Gb/s PD off eBay and shoot a milliwatt or two of light into it. The photocurrent noise is given by the shot noise formula, so you know it a priori.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal
ElectroOptical Innovations
55 Orchard Rd
Briarcliff Manor NY 10510
845-480-2058
hobbs at electrooptical dot net
http://electrooptical.net
Reply to
Phil Hobbs

A small zener driving a MMIC chain should work. Can your spectrum analyzer normalize?

But noise is noisy! How about a nice stable picket-fence spectrum, a line every 50 MHz maybe, from an SRD or a fast logic gate?

An EclipsPlus gate, driven by a square wave, should make fairly flat odd harmonics up to a GHz or so. Adjust the input frequency to cover various ranges of interest.

John

Reply to
John Larkin

Would you elaborate just the littlest bit?

Jim

Reply to
RST Engineering

Normalize what?

That's not going to work so awfully well for a 21.4 MHz. IF filter that is only 50 kHz wide to start with. I need instantaneous bandwidth of a Gig or so. Now a line every 50 mHz. should work well {;-)

Never heard of EcipsPlus. Is it ECL?

Jim

Reply to
RST Engineering

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.