13
\$\begingroup\$

I discovered today that a glass packaged axial leaded 5V Zener diode will become a source of about 0.450 Volts when the glass package is held in the beam of a low-power purple (405nm) laser pointer.

The test setup: Scope probe (with ground clip) attached across the zener. With laser turned off, scope reads zero volts as expected. Turning laser on and aiming it at the glass package of the diode, the scope reads a fairly stable 450mv (noisy though: 30mv p-p ~100kHz). (edit: this noise could be a product of the laser-driver step-up circuit)

The laser is a cheap one and purports to be 1mW rated.

Interrupting the beam with opaque materials instantly stops the voltage reading from the diode. Modulating the laser with a 5kHz square wave causes the diode to exhibit a 5kHz response (in phase with the laser's modulation as far as my scope can tell).

I realise that this is rather unscientific but my question is this:

Is this typical of glass zeners and if so, should a designer avoid using glass zeners in sensitive analog circuits. Or is this too specific to be a real-world problem?

\$\endgroup\$
  • \$\begingroup\$ Related: Quartz windowed UV eraseable EPROMS used in office environments lost their program in time due to light from std fluorescent tubes. Long before detectable gross program failure program operation could be disturbed by changing the light entering the EPROM by eg just waving a hand over it in a lit office. Ask me how I know :-). FIX: Apply light proof stickers as one was meant to do. \$\endgroup\$ – Russell McMahon Nov 17 '16 at 8:22
  • \$\begingroup\$ People reported attempting ceramic bodied 1-time EPROM erasure using XRays but I never heard of this being successful. It was however found that this was one way of implementing a replacement for the very hard to source Signeics 25120 WOM \$\endgroup\$ – Russell McMahon Nov 17 '16 at 8:23
  • \$\begingroup\$ @RussellMcMahon, a friend told me the reason why he had never had a single UV-EPROM failure in the field - because he put two layers of thick self-adhesive copper tape over the quartz windows and wrote "DO NOT REMOVE TAPE" on top in permanent marker. :) \$\endgroup\$ – Wossname Nov 17 '16 at 8:50
  • \$\begingroup\$ Lonnnnnng ago we used to erase them in the sun. Quite effective - NZ sun is some of the mist US rich on earth (really) but any old sun will do if you apply enough of it. AFAIR 20-30 minutes was the minimum but we used ~= half a sunny day through an office window. The building had glazed brick window sills inside and we'd place the EPROMS on that to erase. A substantial %age (10%?) turned in Write-only-memories in the process. After due thought we started putting them in conductive foam while erasing and had no more losses. ESD? Oh yes. DO look at the WOM datasheet above, and then see .... \$\endgroup\$ – Russell McMahon Nov 17 '16 at 10:44
  • 1
    \$\begingroup\$ .... The real story of the Signetics 25120 Write Only Memory \$\endgroup\$ – Russell McMahon Nov 17 '16 at 10:45
16
\$\begingroup\$

Diodes of all sorts, including the ubiquitous 1N4148, packaged in transparent packages tend to have some sensitivity to light (both photoconductive and photovoltaic as you have observed). The 1N4148 can apparently produce 10nA in direct sunlight.

I rather suspect your zener diode when used normally with several mA flowing would have negligible response to normal room light. Zeners are not terribly precise devices in the first place. However, say you are using it as a noise source, say for audio or cryptography, you might want to keep it dark or use a plastic packaged device.

It's worth considering such effects if you have a very sensitive circuit and it is exposed to light, either from openings in the enclosure or because some designer has peppered the PCB with highly luminous LEDs that are modulated or blink.

That includes glass MELF packages as well as axial-lead packages (photo from Digikey).

enter image description here

\$\endgroup\$
  • 1
    \$\begingroup\$ "tend to have some sensitivity to light" All PN diodes are potentially photodiodes. In some sensitive circuits this can show up during test as the circuit board is illuminated by room lighting. Fluorescent is worse because of its 60 Hz component. \$\endgroup\$ – WhatRoughBeast Nov 9 '16 at 20:44
  • 1
    \$\begingroup\$ @WhatRoughBeast ITYM 120Hz (or 100Hz), assuming ye olde magnetic ballasts. \$\endgroup\$ – Spehro Pefhany Nov 9 '16 at 20:46
  • 2
    \$\begingroup\$ @SpehroPefhany, I wasn't deliberately trying to use it as a noise source, but rather just experimenting purely out of curiosity. I'm glad I did. :) \$\endgroup\$ – Wossname Nov 9 '16 at 20:49
12
\$\begingroup\$

"Or is this too specific to be a real-world problem?" Not at all. It's a problem for me as I use them for cryptographic random number generation. I've recently been using BZX85C24 Zener diodes. Running it at 30uA can create a noise level of 1V peak to peak (if you measure it enough times). But that's in total darkness. Get some sunlight on it and the noise drops dramatically to a quarter or less. Even worse is getting mains powered lighting on it like incandescents. You just pick up a mass of mains hum all over the signal that's totally trashes the entropy output.

I expect that not many people use analogue noise sources for testing, as digitally generated sources are available. But for cryptography, you absolutely need the analogue variety. You can use light tight enclosures, but I prefer to use heat shrink tubing on the diodes themselves. If you don't take precautions against the photo electric effect in these applications, the whole device can fail to provide secure random numbers.

\$\endgroup\$
  • \$\begingroup\$ Very interesting. I have long been interested in cryptography and also in the related field of random noise synthesis. In my day-job I have seen diode-noise sources used in EMC laboratories as wide band stimulus in immunity testing. Randomness is one of those ethereal goals (like stable frequency generation) that evades us at every turn. Cursed physics! \$\endgroup\$ – Wossname Nov 9 '16 at 22:11
  • \$\begingroup\$ @Wossname: well, randomness as such is easy to obtain. Randomness with a well-controlled distribution, that's more difficult. Fortunately you can “repair” uneven distributions quite well by following a physical entropy source with some digital pseudorandom step, but I suppose this may be not stand up to some standards of cryptography. \$\endgroup\$ – leftaroundabout Nov 9 '16 at 23:10
  • \$\begingroup\$ What I wonder is: if diode shot noise has such problems, then why isn't Johnston noise used instead? Resistors surely aren't photosensitive in this way, and temperature variations shouldn't be fast enough to cause much trouble. \$\endgroup\$ – leftaroundabout Nov 9 '16 at 23:14
  • \$\begingroup\$ @leftaroundabout -- it tends to be a problem of "not enough of it" -- it's a pain in the arse getting enough resistor noise to not get swamped by other noise sources... \$\endgroup\$ – ThreePhaseEel Nov 9 '16 at 23:29
  • \$\begingroup\$ @leftaroundabout It's actually fairly easy to get full uniformly distributed randomness from any entropy source. The trick is to compress the entropy beyond a measured threshold so that your entropy in is >2x the entropy out. It's like whisky distillation. You start off with 2 bits /byte source entropy, compress (refine) 5 times and you'll get 8 bits /byte entropy. That's a pure uniform random distribution of cryptographic quality. For compression I use a Pearson hash, but the traditional SHA functions will do too. \$\endgroup\$ – Paul Uszak Nov 9 '16 at 23:58
5
\$\begingroup\$

All semiconductors

... have a photo-electric effect including LED's which can be used as ambient light detectors.

So if you are operating in high ambient light and low current affects your operation, simply block the light.

Laser induced arcs are possible in small air gaps which also have negative resistance like a semiconductor during ionization.

\$\endgroup\$
  • 3
    \$\begingroup\$ We once demonstrated how a UV-EPROM could be used to capture a bar code in close range with good lighting from scanning the memory for charge voltage. Then soon later digital cameras chips came out \$\endgroup\$ – Sunnyskyguy EE75 Nov 9 '16 at 21:33
  • 1
    \$\begingroup\$ that sounds fascinating, do you have any links relating to that? By a weird coincidence I've got an ST brand "M27C2001-12F1" (1994 vintage) on my desk as an ornament as it is so beautiful. \$\endgroup\$ – Wossname Nov 9 '16 at 21:40
  • 1
    \$\begingroup\$ no, we did this before the internet was born \$\endgroup\$ – Sunnyskyguy EE75 Nov 9 '16 at 21:51
  • 4
    \$\begingroup\$ Yes, especially DRAMs in ceramic packages were useful for those purposes: they had small metal covers that could be easily removed exposing the die to light. The regularly arranged memory cells could be used as image sensor (long before CCDs were available or cheap). \$\endgroup\$ – Curd Nov 9 '16 at 22:00
  • 2
    \$\begingroup\$ I remember a project in one of the hobby electronics magazines in the late 80's or so that used a dram as a camera and timed the decay of the cells following a refresh to produce output. And yes, cryptographic quality noise is hard. \$\endgroup\$ – Dan Mills Nov 10 '16 at 10:59

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.