Re: Switch and fight (leakage)?

Well, i think it is de-light-ful when someone that should know better gets startled by something known for at least 40 years.

Even those diodes (and transistors) packaged in black plastic exhibit photo effects.

TI used to have a whole series of IR diodes (for TV remote links) that had no window.

And, John, they don't "leak when light hits them"... what a dummy :-) ...Jim Thompson

I didn't say it confused *me*

I first noticed something like this when GE introduced their first plastic transistors. They had a round black plastic header (e-c-b inline) and a flatted, sort of TO-92, cylindrical body made of brownish epoxy. The top was concave, where they apparently poured the brown gunk into a mold. They would leak nA of current in healthy room light. I designed a nice 1 uV/degC opamp using them and told my boss to come over and see. Of course he leaned over and blocked the light and messed up the offset tweak.

Before long GE went all black on their packaging. I was still a teenager, so it *was* over 40 years ago.

The other plastic style back then was TO-5 looking, with a round open-top grey ceramic base and a black epoxy glob top. There was an

The latest semiconductor discovery that startled me, aside from the fA leakage of some RF transistors, is the parasitic PIN diode layer I discovered in some Optek laser diodes. It was just like having a substantial inductance in series with the part. Bad news for pulse applications. And the National substrate diode SRD thing of course.

John

I've seen some like that, a bare silicon wafer.

however photdiodes packed in black plastic are packaged in plastic that is transparent in the infrared, black in the visible wavelengths cuts down on interferance from fluorescent lights etc...

--- news://freenews.netfront.net/ - complaints: snipped-for-privacy@netfront.net ---

Let's not get carried away with generalizations. You can't whip out just any ole' 'black plastic' transistor and expect useful photo response. Those TI jobs were special 'black plastic' transparent to IR. They block the visible spectrum to reduce ambient light interference.

I tested one of those with a red superbright. Just enough attenuation that I got a useful signal with the LED positioned directly in front of the sensor (CTR on the order of 0.01%).

Tim

The old plastic transistors that were photosensitive were phenolic-encapsulated, iirc. Metal can transistors detect light that gets in through the green frit seals--it usually takes direct sunlight before that's a problem, but it can be very puzzling!

Novolac epoxy (as used in all modern plastic packages afaik) is very black at all optical wavelengths where silicon is photosensitive. I'm not sure what it's like further out in the IR.

Cheers

Phil Hobbs

Just to clarify: optical packages are _not_ made of the same stuff--ordinary plastic packages are made from Sumitomo Bakelite's Sumikon EME-6710 or its brethren. I looked for a decent datasheet but there don't appear to be any online--typical of the chemical industry.

Cheers

Phil Hobbs

Phil, Have you looked at any I/C processes, like XFAB-XH035, which has a photodiode module? They're using several "coating" layers to tailor the response curve. ...Jim Thompson

I'm currently doing an expert witness gig in a trade secret misappropriation case that centres in part on similar things, but I haven't dug into the process details.

The absorption depth of light in silicon is a strong function of wavelength, so PD response depends a lot on the epi thickness and the depth of the tub/substrate/whatever is the counterelectrode to the epi. Besides that, you can use thin metal, silicides, nitride, and other such stuff, and then there are the actual colour filters, as used in image sensors.

Cheers

Phil Hobbs

Interesting. We *need* an optocoupler with a CTR of 0.01%. The current plan is to put the LED in a cavity in an aluminum block, with the PD in the other side, and use a setscrew in the middle that can, fully scrunched down, block all the light between.

John

You got that right. Ages ago I was a test development engineer for an analog semi company. One day I get a call from product engineering saying the equipment I designed is screwed up and failing all the opamps in TO99 packages for excess Ib. I took my golden units out to production and had them tested. The data matched my old data quite nicely. After much head scratching it was determined that the package vendor had switched the glass on the frit seals and light was entering and reflecting off the lid of the can. It was puzzling but fun too as it wasn't some old problem again for the Nth time. Art

Yup. I had exactly this happening with the input Jfets of my 200pV/rtHz preamplifier built a few years ago. Measurements gave way too much of 1/f noise. A cloud passes => pb solved. Of course it was hard to sell the cloud with the preamp...

Don't you guys have a longer term cloud these days? I mean the one from way up north, not the financial one from the south :-)