|Oops, February is short
(28 Feb 2019 at 23:47)
|Wow, the end of February really snuck up on me! I wrote this on the morning of March 1 and backdated it. :(|
Aside from some uninteresting work travel and a bit more progress on Pac Tom, the main notable thing from February was work on my SIGBOVIK papers. It is now possible to submit, so you can too! Thank you for your suggestions for my chess paper (see previous post); it's not too late for more ideas there. I also have one non-chess paper, which turned out to be pretty fun. For that I spent a solid chunk of the weekend manually routing this bad boy:
Design rules check pass!
It may end up to be too hard to solder, but isn't it aesthetically pleasing?
|So this PCB plan, is this the one that will interface with your brain pretending it's your memory, like you suggested in the “Reverse emulating the NES” talk?
|What is that board supposed to be?|
|You have about eight hours to post the next blog entry – not that it's really needed, because we'll probably see your five SIGBOVIK articles about chess the day after.|
|One of the things that I've learned over the years is that an excellent way to get a good result at something is to "teach Tom7 the basics of it, and watch him eclipse you rapidly". In this vein, I have wondered over the years what would happen if somebody gave you programmable logic devices: surely performance of the NANDY 1000 would be greatly improved by using some CPLDs as the basic gates, and even then, you might have enough logic left over to implement the rest of the NANDY 1000! Either that, or we'll end up with a hardware description language embedded in SML, and either of those sound like good outcomes to me.
All the same, I enjoy seeing the fruits of someone's labor in a densely routed board, and am suitably impressed. I fully expect a 4-layer board  with DDR3  routed for next year's SIGBOVIK :-)
 It's more possible than it sounds! OSHPark's 4-layer process is quite good.
 It's, uh, comparably possible to how it sounds!
|Thanks! I'm using PCBWay, which also supports four+ layer, but (like basically anything that strays from the defaults) it increases the price from "suspiciously cheap" to "that's kinda what I expected." That's still somehow hard for me to justify to myself even though I'm willing to pour dozens of hours into the project, take vacation days, etc. (?). Of course the real reason to stick with two-layer is that it makes it more challenging to hand-route the board!|