What? Giving a guest lecture at the university in Durban, South Africa -- I'd love to (as long as they are paying for airface and hotel etc :-) -- I've been to India and China and Japan and ... and a bunch of other places -- I gave a lecture at the University of Oslo in Norway in Feb of this year -- but I've never been to Africa
There's a lot to think about -- one of the great things about these lectures is not that they teach you everything there is to know, but that they tell you about a lot of different things that are out there that make you think "I want to learn more about that"
Great presentations, Max. And thx again to you, Brian, and EEtimes, for making it available. I think it's time I paid more attention to eetimes and programmable planet, and read the Bebop book to find out how things are done in the years since I graduated and got sidetracked by different types of work.
There's also the question as to id a particular radiation event matters or not. Suppose you have a space system that is taking pictures, storing them in on-chip memory, performing DSP on them, and shending them to Earth. If a radiation event causes a bit to flip in the middle of a 10megabit picture ... who really cares -- that will be filtered out by the DSP anyway -- so I probably woyuldn;t worry about protecing thsi memory
By comparison, if I have another block of on-chip memory that is storing my system flight parameters -- then I will use ECC, TMR, and memory scrubbing techniques to keep tha tdata pristine...
Well, I'm being paged .. gotta run. Thanks again Max and Brian. Looking forward to the next session, and perhaps the Design West conference and thanks @jnhong ... I will. 'till next time or you can always message me ...
@Dave: I take it sram fpga's are more susceptible that non-volatile programmable devives?
All FPGAs (actually any digital chip like an ASIC/SoC) has registers and memory cells tha tcan be affected by radiation (flip bits from 0 to 1 and vice versa) -- tyh ereason SRAM-based FPGAs are especially vunerable is that their confifuration cells can also be flipped.
There are ways aroudn thsi like using triple-more redundency (TMR) and detecting when a problem has occured and then using Dynamic Partial Reconfiguration (we talked about this yesterday) to reload that part of the device
@dave: Max, are there any numbers for probability of a rad event on the typical area of an fpga on Earth surface?
I'm sure there are, but I don;t have that Data to hand. One of the bloggers on All Programmable Planet is Adam Taylor in the UK -- he's working on using FPGAs in space -- he'll be giving a talk at Design West also -- he knows all ths istuff
I have strong analog skills. What would be a great class would be a class to show example of how to take advantage of FPGA's in an all analog world. The mention of "Hampster" who created a transmitter to broadcast S-O-S is fascinnating to me. It leads directly into Direct Digital syntheses, which was starting to be a hot button at a former life.
@MaheshNarayan.Nadimpalli: (and I'm not going to typw tha tis more than once :-) Re books, there are all sorts out there -- one I woudl suggest is written by Mike Field -- it's a free intro to VHDL -- it's based aroudn the low-cost Papilio FPGA board -- the best thing is tha tMike's book is free -- mike is known as "Hamster" -- if you go to www.AllProgrammablePlanet.com and search for "Hamster" you'll see a column tha tgives details about thsi book
Thanks everyon efor the kind words. I will try to answer your questions, but some of the ones on radiation are a bit convoluted (the answers) -- what I wil ldo is write a coupel of columns on this next week or the week after and post them on www.AllProgrammablePlanet.com
It's been a great set of talks ... thanks very much! I'd like to see a "Hello World" workshop ... getting your first FPGA design working. Would have to be vendor specific but using introductory hardware is cheap.
great pictorial representations and explanations of the variations in use by many semiconductor companies, again .. . vertical integration through/across die to die ... biggest problem I could see would be cooling or heat control of typical (predictable) hot spots ...
We probably will use some programmable analog and/or all-programmible stuff at some point but in agricultural machinery we're typically late adopeters because our stuff will be expected to run 25+ years - we wait until it's well proven and available in automotive-qualified packaging.