03:30:48
##forth
<KipIngram>
Yes, exactly.
03:31:11
##forth
<KipIngram>
And the bit error rate goes up because it takes that much less cell voltage drift to cause a data read error.
03:31:27
##forth
<KipIngram>
The math behind these error detection/correction codes is just fascinating.
03:32:24
##forth
<KipIngram>
And all that is smack in the data path, so it has to be fast. That's why they have this guy putting it all in hardware - it would be so much easier to code in software.
03:33:21
##forth
<X-Scale>
Coding Theory is an amazing subject.
03:33:59
##forth
<X-Scale>
Hardware is software that can't change.
03:59:00
##forth
<tpnix>
KipIngram, incredible tech, I could happily live another lifetime just to see what unfolds!
04:00:07
##forth
<tpnix>
X-Scale, hardware at the technology we have now is capable of change, ie determine bad cells and switch them with good ones ?
13:44:37
##forth
<KipIngram>
Well, these days hardware is more or less more software, at least on the digital front. It wasn't always that way - it wasn't when I first started out. Hardware was done with schematics, not with code. I hated to see it change - I think something was lost along the way. The "shape" of the schematic conveyed information and was a communication channel in your brain all on its own.
13:44:48
##forth
<KipIngram>
I was very good at it too - it was the thing I did best.
13:44:53
##forth
<KipIngram>
Then the world moved on.
13:47:02
##forth
<KipIngram>
tpnix: That's another thing those background scanners do. NAND flash pages do eventually become so error prone that they're deemed unusable, and they get marked bad and their data error corrected and moved to other pages.
13:49:33
##forth
<KipIngram>
Also the programming of those cells is really an analog thing - you're trying to put the cell voltage right in the middle of its target range. The flash changes gradually over time and the parameters for that get gradually evolved.
13:59:32
##forth
<KipIngram>
Regarding hardware that changes, I've always loved that idea. I feel like the FPGA vendors have somewhat held us back on that front by refusing to release documentation on their configuration bitstreams. It's hard to morph your hardware when you don't have a clear picture of what the bits in the stream actually do.
14:00:39
##forth
<KipIngram>
They tell themselves that's some incredible "secret sauce," and maybe in rare instances it is, but my bet is that 90+ percent of it is fairly mundane.