Dylan Patel: Hi. Thanks for taking my question. I wanted to ask about how I think through the number of wafers per tester, right? So there’s all the various different configurations that you’ve sort of options you’ve had for various different customers. Some people want to do the bipolar voltage channel modules and some people want to do high voltage, some people want to go negative. I’m curious, can you sort of outlay how to think through what is the test time for these various systems, for various options? And then sort of how many can be tested for FOX XP, right, because some of these are 9 per system versus 18. I would love to hear that sort of rationale.
Gayn Erickson: All right. Well, as I described this, I simply want you to listen in terms of just the simple math because mostly what I’ll describe as a simple math. But obviously, the piece is, well, what is customers A, B and C’s test time. And I want to give a little bit of color on that when I’m done. From a simple math perspective, the NP systems have two wafers. They’re usually used for engineering. And if you want to do small lot production, you can test two wafers at a time. So if your test time was 24 hours, you would get two wafers per day off of that. If your test time or, call it, cycle time is 12 hours, you could get four wafers a day off of them. The XP, which is fully compatible blades, which are effectively that each tester uniquely can test a wafer has either 9 or 18, following the same math, you would either get 9 or 18 wafers a day at 24 hours or 2x that at 12 hours or 4x that at six hours.
So I’m not trying to be coy, but that’s the way of thinking about the capacity off of it. Now the real debate is and the discussion is, well, what’s the test time? Well, if you get into the what’s called the backup curve of reliability, and if you go type in bathtub curve and reliability, you can find all kinds of articles out there to talk about it. Basically, when a device is first manufactured, all semiconductor devices, as soon as they are functionally good, they have a likelihood of failure at that point in time. And as time goes on, the likelihood that they fail actually decreases. This was observed a long time ago and is very consistent across all semiconductor processes. So what that looks like is the likelihood of failure drops as time goes, and then at some point, it stops dropping and it’s the bottom of the bathtub curve.
20 years from now, they start failing again, and that’s the other end of the bathtub curve. What’s important is depending on the expectation of the customer or the application is whether or not the failure rate upon shipment is good enough. And if it is too high, you do things like stress and burn-in, which is what we supply to decrease and move it down the curve, the higher the quality, the more screening or burn-in you need to do to move it down the curve. And that is generally energy and time. So an example we’ve tried to use in certain applications like the inverter of the automobile, probably the highest reliability requirement out there because on an inverter, it might have 12 or 24 chips in a single module or 48 in a single inverter on the Tesla, for example, and Tesla Plaid has three of those.
