Noisy NPN SOT-23 reverse breakdown BE junction (Audio)

What's the quality like? Zeners tend to get asymmetric and pulsey at low currents. What RMS amplitude are you seeing?

Here's a BFT25, which is an RF part with really tiny chip:

A real zener diode can make pretty good noise, about 340 nV/rtHz

Most are about 1mV RMS over perhaps 1MHz BW. The best one was about 3x higher.

I have not done much evaluation of the quality yet, I was going to run it into an NI 24-bit data acquisition box and do a detailed evaluation when there's enough signal to make it worthwhile. I can hook it up to my DSA815 spectrum analyzer but that's 50 ohm input which would require a driver.

This is really just a baseline generator that I'm making to compare some better ways of generating noise.

Thanks, John. Maybe I'll order some BZX84 which will fit the footprint well enough.

Some random sources suggest that noise emplitude increases with lower currents (like 10uA) but I suspect the quality as well as the bandwidth will suffer at very low (

Am 09.02.21 um 22:06 schrieb Spehro Pefhany:

Here are some low-voltage BZX84 . Sorry I was looking for low noise, not maximum noise. On the pics left/right, there are some LEDs and regulators.

0dB is 1 nV/rtHz. Look how the noise grows towards larger breakdown voltages. Bias was 14V from NiCd and 1 or 2K wire resistor. The bias contributes nothing.

cheers, Gerhard

That last one is 3 uV per root Hz. Huge. 3000x a 50 ohm resistor.

Something like a 2SD2704K, maybe. It has a very high BE breakdown voltage, so once it gets going it might be quite vigorous.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC / Hobbs ElectroOptics 
Optics, Electro-optics, Photonics, Analog Electronics 
Briarcliff Manor NY 10510 

http://electrooptical.net 
http://hobbs-eo.com

Thanks very much for the ZIP. I can unzip it, stick it in XnView, and flip back and forth to study the results. Very good!

What happens above 3 mA where a zener normally operates? There is such a paucity of information that your measurements would be invaluable.

--
The best designs are no accident - sw

Thanks.

I only have 30V to play with, so it probably won't even break down. (25V minimum). Very high hFE which is interesting- I guess it's not symmetrical.

It doesn't matter in this case, but this looks like just the kind of part that would have niche applications and then be discontinued.

--
Best regards,  
Spehro Pefhany

It's used for audio muting. I've used them for a decade or more. The

2SD2114K is a lower-voltage, higher-current version. Both are great for cap multipliers.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC / Hobbs ElectroOptics 
Optics, Electro-optics, Photonics, Analog Electronics 
Briarcliff Manor NY 10510 

http://electrooptical.net 
http://hobbs-eo.com

Would you get more signal if you used emitter and collector instead? the noise could be amplified by the reverse beta

--
  Jasen.

30V. perfect for 20V zener and 100k ohm resistor... try it, (if a glass encapsulated zener shield it from any light.)

GH

What you are looking at is avalanche multiplication in a tiny gap.

Multiplication is a statistical process, so if the current is low enough th at you've mostly only got one charge carrier in the gap, you have ocasions when the charge carriers gets through the gap without generating a new char ge-carrier pair, and the avalanche stops for a bit until thermal noose (or a cosmic ray) generates a new one.

We had a long thread about this some twenty years ago, and all theory got d redged up and thrashed out - a lot of it by Winfield Hill - with Tony Willi ams posting observations.

It's a perfectly fine random noise source, even if it is asymmetrical. Dira c spikes have a flat frequency spectrum.

--
Bill Sloman, Sydney

I'd guess that noise amplitude drops with current, and bandwidth and statistical quality improve. But that's a guess.

For better symmetry, use two and subtract.

Noisecom sells specific noise diodes and probably has data sheets and appnotes.

A small FPGA or uP can make an essentially perfect pseudorandom bit stream that will filter nicely to gaussian noise. AoE talks about that.

People used to use gas tubes and thyratrons for noise too. Magnets helped somehow. WWII radar jammers used photomultiplier tubes (in the dark) to make noise pulses. I used to buy surplus chassis, with a 931A PMT, for a couple of dollars.

--

John Larkin      Highland Technology, Inc 

The best designs are necessarily accidental.

Am 10.02.21 um 04:00 schrieb Bill Sloman:

Yeah. Visual symmetry across the samples is about as un-random as it gets. And in the long term, there cannot be continuous asymmetry because we would see outrageous voltage peaks then.

Do you see any peaks in my Zener spectra above?

Gerhard

Yeah. Unless you need noise from a real physical process for some reason there are so many more repeatable ways of making audio-frequency white noise in 2021 that aren't much more expensive than the circuitry you'd use to do it the old fashioned way

A waste of time. What you get may be more visually appealing, but won't be any more random.

notes.

am that will filter nicely to gaussian noise. AoE talks about

Except that it isn't random. This rarely matters, but there are situations where it can.

somehow. WWII radar jammers used photomultiplier tubes (in the dark) to make noise pulses. I used to buy surplus chassis, with a 931A PMT, for a couple of dollars.

It's all thermionic noise pushing the occasional electron over some potenti al barrier or other. Better photomultipliers than the RCA 931A can react to passing cosmic rays as well, but that's typically around one per second, s o it isn't often enough to be useful.

--
Bill Sloman, Sydney

Not fine for non-linear devices. For example I have used Skyworks SMP330-085LF limiter diodes. The data sheets don't give away all the secrets of how they operate, but it looks like forward conduction generates floating charge carriers that support reverse conduction. So an assymetric noise source could cause them to conduct more if inverted than non-inverted, for example.

CH

Sorry that should have been SMP1330-085LF

CH