jbeale
Well-known member
ah yes, mysterious hardware problems...
ah yes, mysterious hardware problems...
Since you mention that- I have no idea what's inside the RED box so I really shouldn't be writing this.
But I will say that about 6 years ago, I was involved with a OEM product that used some large FPGAs (programmable logic- sort of a flexible, dedicated high-speed CPU). At that time, and I believe still, there were two main vendors of these chips: Xilinx and Altera. They both had competitive parts, so at the outset your choice might be based on design tools or how nice the company reps are, but later once your design is locked and the boards are fabbed it's not easy to change vendors.
Anyway, after some mysterious field failures we determined that the FPGA we used had some subtle flaws. I think it was the timing of some gates was not quite what the model said it was. Everything worked fine in our initial testing, but some of our firmware builds (after shipping) used different connections inside the chip and exposed the problem. We found that some chips worked and some didn't. When we complained to the company (I forgot which one it was... let's call it "Company X"
at first they ignored us (some speculated because we were a small customer.) The problem was really killing us and we kept at them, and eventually they admitted that there was a problem in manufacturing which might possibly have caused the problem but the newer date codes were good because they silently fixed it at some point (without mentioning the issue to anyone, or at least to us.) Gotta love those customer relations. 
inch:
Problems related to marginal timing (eg. internal "logic race") have interesting consequences. They can be temperature-dependent (timing generally changes with temperature). If you are right on the edge, you can get big increases in current consumption and heat generation of the component, when there is a transient driver conflict (two logic outputs fighting each other on the same bus line). If I had two supposedly identical FPGA-based systems, and one runs hotter than the other in the same conditions, I would start to suspect something like this (of course there are many other causes too).
ah yes, mysterious hardware problems...
Since you mention that- I have no idea what's inside the RED box so I really shouldn't be writing this.
But I will say that about 6 years ago, I was involved with a OEM product that used some large FPGAs (programmable logic- sort of a flexible, dedicated high-speed CPU). At that time, and I believe still, there were two main vendors of these chips: Xilinx and Altera. They both had competitive parts, so at the outset your choice might be based on design tools or how nice the company reps are, but later once your design is locked and the boards are fabbed it's not easy to change vendors.
Anyway, after some mysterious field failures we determined that the FPGA we used had some subtle flaws. I think it was the timing of some gates was not quite what the model said it was. Everything worked fine in our initial testing, but some of our firmware builds (after shipping) used different connections inside the chip and exposed the problem. We found that some chips worked and some didn't. When we complained to the company (I forgot which one it was... let's call it "Company X"
at first they ignored us (some speculated because we were a small customer.) The problem was really killing us and we kept at them, and eventually they admitted that there was a problem in manufacturing which might possibly have caused the problem but the newer date codes were good because they silently fixed it at some point (without mentioning the issue to anyone, or at least to us.) Gotta love those customer relations. inch:Problems related to marginal timing (eg. internal "logic race") have interesting consequences. They can be temperature-dependent (timing generally changes with temperature). If you are right on the edge, you can get big increases in current consumption and heat generation of the component, when there is a transient driver conflict (two logic outputs fighting each other on the same bus line). If I had two supposedly identical FPGA-based systems, and one runs hotter than the other in the same conditions, I would start to suspect something like this (of course there are many other causes too).
