# Ideas for troubleshooting a noise generator on a PCB

I am building an electronic "dice" that uses an ATTiny84 to sample avalanche noise from a 2N2222 in reverse-bias and generate random numbers, which are then displayed on 2 7-segment displays. When I prototyped this on a breadboard, it worked just fine, so I decided to move to "production" and put it on a PCB.

Unfortunately, the noise generator on the PCB is running into some issues: It does not seem to be outputting (the "output" is the collector of the 2N3904) a noisy signal.

Now, as I am a student with limited means, I would usually go into the university lab and use one of their marvelous oscilloscopes to troubleshoot the signals themselves; however, they have locked all the labs for renovation or something and I can't use their equipment.

Therefore, I am asking for some suggestions as to how to troubleshoot with the tools I do have: soldering/desoldering equipment and a basic digital multimeter. I have already confirmed that all of my traces are intact and going where they should, and that steady voltages are what they should be. Every other component of the circuit works like a charm, except this generator. How can I glean some more information about it with what I have, and/or is there something I am overlooking that might explain the problem?

Thank you so much for your help.

• Nice question.. – Roh Aug 21 '14 at 5:22
• There's no schematic. What evidence do you have that the junction is actually avalanching? How much current? This is not a tested mode of operation, so it's certainly possible some types or samples of transistor will work better than others. – Spehro Pefhany Aug 21 '14 at 5:32
• @SpehroPefhany, yes there is a schematic in the original question. Additionally, I know that it was avalanching on the breadboard because at that point I was able to use the University 'scopes and was able to see the avalanching by myself. Don't worry, I've done my research, tried many different transistors and configurations, and this is the one that worked best :) – Asa Graf Aug 21 '14 at 16:45

Re schematic, sometimes it takes time for the images to show up I guess, and I didn't see a reference to it in the text.

Have you measured the AC voltages to see if the transistor is producing any noise? Multimeter on AC range, most are AC coupled so on a 200mV range you can see 100uV noise for 3.5 digits and 10uV for 4.5 digits basic multimeter.

You are using an ugly bias scheme for T3 (why not Q3?)- it's beta dependent.

If you can't improve it, then adjust R21 so the voltage at the collector of T3 is where it should be (looks like about +1V would put it in the middle of the available swing).

• I have measured voltages at all the points in the noise circuit: At the noise junction, I measure 8.2VDC which meshes with what the "steady" voltage should be. I can't tell if there's noise on top of that, obviously, as my multimeter isn't that precise. At the base of the 2N3904, I measure 3.5VDC and at the collector/analog input of the ATTiny I measure 2.1. These are all consistent with the "steady" voltages I want; unfortunately with my MM I am not seeing any noise. It doesn't have a 200mV AC range, I'm afraid. – Asa Graf Aug 22 '14 at 13:37
• @AsaGraf Why would you want the 2N3904 to be biased almost off? And I highly suspect you are not measuring 3.5V at the base or something is very wrong. It should be no more than 700mV DC. <rant> What #$#*&$# program is it you're using that creates schematics that look the same whether crossing lines are electrically joined or not? </rant> – Spehro Pefhany Aug 22 '14 at 13:49

OK I'm expanding a bit on Spehro's comment. I haven't looked at avalanche noise in transistors. But I have in zener diodes. (Same physics different device.) What I found was that the avalanche noise changed a lot from device to device. So one option would be to try different transistors, till you find a noisy one.
If you had some more power supply voltage, I've got a great (reproducible) noise circuit that uses a 20V zener at low current.

• You're absolutely right - in my research and tinkering I also found that to be true. When I was breadboarding this I was able to confirm its noisiness because at that point I DID have access to the University 'scopes. The components in the PCB are the exact same ones that were on the breadboard. The problem, then, in my mind, has something to do with being on the PCB itself, or the transition from breadboard to PCB. As a non-sequitur, can you show me that Zener circuit? It's too late to revise my design, but this project has really piqued my interest in intentionally noisy circuits! – Asa Graf Aug 21 '14 at 16:49
• Just to be clear, I don't only mean transistors with different part numbers. But you'll get different amounts of noise from parts with the same part number. There's not much to the zener circuit. dropbox.com/s/9hho4tfh7gqd882/DSCF0055.JPG (hmm I cut off the top of my chicken scratches...R is 1 Meg to ~200k or so, what's important is the zener current, so you'd have to adjust for different power supply voltages.) The spikey output you get is asymmetric and so you can add two opposite polarity noise sources together. The BW is not all that great, noise is flat to ~500kHz. – George Herold Aug 21 '14 at 18:10
• Thanks! Yeah, I totally got that - this was the pair of 2N2222s that performed the best out of the ones in my parts box. They all performed similarly and there was a greater difference b/w different part numbers, but yes, there was variance in the "same" component. – Asa Graf Aug 21 '14 at 20:29
• As for the Zener circuit, I may try that soon when the lab opens up again - they have the powers supplies that can give me those voltages! When I was researching for this project, I came across Zener noise but was explicitly focused on generating noise from transistors. – Asa Graf Aug 21 '14 at 20:31
• RE proto-PCB problems. The only times I've had issues with going to a pcb was in a circuit with some high impedances and a ground plane on the pcb. This caused some excess capacitance to ground. I don't see any such high impedances in your circuit. (Well the weird bias for your 2n3904, but that shouldn't matter.) Re: zener noise circuit. Two warnings. 1.) the circuit is sensitive to room light, more light less noise. 2.) I've only used Fairchild zeners. I don't know if other manufacturers would work. – George Herold Aug 22 '14 at 13:57