Does this quote from the documentation mean that I will cause problems if I reverse biase the base-emittter juntion of an NPN transistor in Sky130?

The docs say "The SKY130 process offers "free" NPN devices. The NPN uses the deep n-well as the collector. The device is not optimized, and must be used in the forward-active mode." and "Reverse-active mode operation of the BJT's are neither modeled nor permitted."

https://skywater-pdk.readthedocs.io/en/main/rules/device-details.html

How are the diodes used? Is it to set/limit a voltage or short out current? Have you looked in the standard I/O library to see what types of diodes are used there? Some of the diodes may operate in reverse breakdown.

As far as I know both the diodes and BJTs aren't modelled for reverse operation. I tried modeling the reverse junction operation, to gauge their voltage threshold to design esd protection, and the results I got were pretty much nonsensical. No reverse current, no reverse breakdown, no leakage.

I want to use the diodes with a high reverse bias to generate voltage noise. I had a quick look at the documentation: https://skywater-pdk.readthedocs.io/en/main/contents/libraries/sky130_fd_io/docs/user_guide.html but I must admit I find it difficult to search the documentation for what I am looking for. My current conclusion is that I should get hold of a test chip and measure some devices using a probe station to create my own model for the IV characteristics of the sky130_fd_pr diode_pw2nd_05v5 diode at breakdown

Do you care about the type of noise you’re generating? Thermal, 1/f, popcorn? I don’t know enough about the type of noise generated during reverse breakdown but if you don’t care about the noise type, simply amplifying the noise of an amplifier by grounding the inputs, and then cascading a bunch of x10 stages in series works well. I’ve breadboarded this circuit before and was able to match output noise to hand calculations.

I don't particularly care what type of noise I get as long as my output signal is dominated by the noise and can't easily be influenced by other signals external to my circuit. What part number devices could I buy to try out the x10 stages in series idea that you mentioned on a breadboard?

Do you care if the amount of noise level is repeatable from part to part? If not, any off-the-shelf opamp would work, like 741. It’s not part specific.

Not really, I will try it with the 741

@Leonardo Gomes I mentioned this in the #sky130 channel but it does look like there is some data about reverse operation in there--it's either hidden by or disabled with a typo but you can 'break' at least some diodes--they have breakdown current specified and... some data... about breakdown voltage. See: https://open-source-silicon.slack.com/archives/C016UL7AQ73/p1714242971403829

Thanks for pointing this out, I should have looked inside the Spice model!

I replied to Pat's Issue on the sky130_fd_pr repository issue tracker. The upshot, though, is that the official definition of the level 3 diode model is supposed to take

vb

or

bv

(or

vrb

or

var

) as the name of the reverse breakdown voltage. The ngspice level 3 diode model, which was added only recently, apparently allows only

bv

(at least it doesn't allow

vb

). So while somebody ought to create a bug report for ngspice, I think the quick solution is for me to change the breakdown voltage parameter name for all the diode models in sky130, as that should make them compatible with ngspice without breaking compability with other simulators (except I'm not sure about Xyce, which didn't have any support for level 3 diodes last time I checked, but that was a while back).

Saw you fixed it Tim. Thank you.

Would the guard around the wire be there to stop latch up when connecting 15 V to your chip? I wanted to do something similar on a Tiny Tapeout analog tile but I couldn't because I would have caused latch up!

@Luke Harold Miles: Yes, although you don't need a "big wire" as presumably you are not drawing much current through it. You would need to use the Caravan version of the chip and connect to a pad that does not have any ESD protection. But it is doable. I have seen 50V brought on-chip that way (but only to a floating gate, where there is no conductive path anywhere). Protecting against 15V on chip will be a challenge, but there are 16V and 20V transistors in the PDK.