Confirmed it's a silicon bug.
Notices by Andrew Zonenberg (azonenberg@ioc.exchange), page 14
-
Embed this notice
Andrew Zonenberg (azonenberg@ioc.exchange)'s status on Wednesday, 27-Mar-2024 03:35:07 JST Andrew Zonenberg -
Embed this notice
Andrew Zonenberg (azonenberg@ioc.exchange)'s status on Wednesday, 27-Mar-2024 03:35:07 JST Andrew Zonenberg Sooo I guess new line in my pcb checklist: stm32 jtrst pin should be dedicated if you're using jtag mode. Don't use it for anything else
-
Embed this notice
Andrew Zonenberg (azonenberg@ioc.exchange)'s status on Wednesday, 27-Mar-2024 03:35:06 JST Andrew Zonenberg Progress! The spaghetti bowl is even larger.
All of this cabling is in use for this test setup. Currently troubleshooting the SPI interface between the supervisor and main MCU.
Once that's done I'll have full connectivity between all of the various subsystems so I can control and monitor health from one spot.
-
Embed this notice
Andrew Zonenberg (azonenberg@ioc.exchange)'s status on Wednesday, 27-Mar-2024 03:35:06 JST Andrew Zonenberg And a second new line in the PCB checklist: "read errata of every chip in the design"
-
Embed this notice
Andrew Zonenberg (azonenberg@ioc.exchange)'s status on Friday, 08-Mar-2024 10:58:44 JST Andrew Zonenberg @dalias Yep I saw some of those recently too. It's something buttcoin related right?
-
Embed this notice
Andrew Zonenberg (azonenberg@ioc.exchange)'s status on Thursday, 07-Mar-2024 03:50:38 JST Andrew Zonenberg @acsawdey @MLE_online @madrush Except then you get a deathtrap of a body not designed to modern safety standards.
Additional (probably impossible) bonus: Something as safe as a modern car in a high speed crash, but also designed with repairability in mind (e.g. avoid use of disposable composite panels on fenders and other areas likely to suffer minor damage, prefer sheet metal that can be hammered back into shape and returned to service).
-
Embed this notice
Andrew Zonenberg (azonenberg@ioc.exchange)'s status on Thursday, 07-Mar-2024 03:50:38 JST Andrew Zonenberg @acsawdey @MLE_online @madrush What I want is an EV that's built like an 1980s American / recent-ish European pickup (big enough bed to carry a sheet of plywood or drywall, but not 3m high and weighing as much as a herd of elephants).
Something able to drive on dirt logging / forest service roads without getting stuck or scraping the underside, but low enough that I won't run over a small car without seeing it.
Ideally with a front bench seat so my wife and kid (when she's old enough to be out of a child seat at least) can all fit without needing an extended cab.
And as dumb and mechanical as possible (mechanical steering, no / easily removable telematics, ideally no touchscreen, as a minimum no touchscreen interaction required for normal in-motion vehicle controls like climate control).
Too bad it doesn't exist and probably never will.
-
Embed this notice
Andrew Zonenberg (azonenberg@ioc.exchange)'s status on Thursday, 07-Mar-2024 03:50:37 JST Andrew Zonenberg @acsawdey @MLE_online @madrush Exactly.
We need vehicles that won't be totaled or generate hundreds of pounds of plastic garbage when somebody bumps a shopping cart into you at the grocery store.
With a bit more time and initial investment you could probably make vehicles consistently last - and be maintainable for - 20+ years. Just look at the US post office and their "long life vehicle" delivery trucks. Last units produced in 1994, still the mainstay of the fleet today although a replacement is under development.
But there's no money in making a family car that lasts 30 years.
-
Embed this notice
Andrew Zonenberg (azonenberg@ioc.exchange)'s status on Wednesday, 28-Feb-2024 20:41:45 JST Andrew Zonenberg @aeva @mjg59 "feed the program to itself"
That's a new take on the classical halting problem paradox.
-
Embed this notice
Andrew Zonenberg (azonenberg@ioc.exchange)'s status on Tuesday, 06-Feb-2024 07:30:32 JST Andrew Zonenberg @jmorris @oshpark @rgilton Yep.
I mean, this is the FPGA I'm putting on the board. $249 @ qty 1. https://www.digikey.com/en/products/detail/amd/XC7K70T-2FBG484C/3641737
It has 31 of these connectors https://www.digikey.com/en/products/detail/amphenol-rf/925-197J-51PT/10712133 on it. At the qty 25 price break that's another $175.
The power supply is five of these https://www.digikey.com/en/products/detail/murata-power-solutions-inc/MYMGK00504ERSR/9748947 ($6.42 each) plus one beefier one for a higher current rail.
I could go on and on, but it adds up.
-
Embed this notice
Andrew Zonenberg (azonenberg@ioc.exchange)'s status on Tuesday, 06-Feb-2024 07:30:31 JST Andrew Zonenberg @jmorris @oshpark @rgilton So we're at the point that I don't even blink at the cost of throwing a STM32L031 on there just to manage sequencing of the dozen-odd power rails and resets for everything.
And the STM32H735 running the management interface is barely visible in the overall BOM cost either.
-
Embed this notice
Andrew Zonenberg (azonenberg@ioc.exchange)'s status on Tuesday, 06-Feb-2024 06:52:57 JST Andrew Zonenberg @jmorris @oshpark @rgilton This is actually "cheap". I was quoted something like $6.9K for an American fab to make a very similar specced board (but I think that one was only 8 layers not 10).
But when you add in the additional $1K-ish of components I'm putting on the board, the custom rackmount chassis, another $1K or so of custom semi rigid cable from PCB to front panel, etc. this is going to end up being one of my most expensive projects to date.
-
Embed this notice
Andrew Zonenberg (azonenberg@ioc.exchange)'s status on Tuesday, 06-Feb-2024 06:48:16 JST Andrew Zonenberg Got the QA report. I love this shop, they do this as standard not an extra-cost option.
-
Embed this notice
Andrew Zonenberg (azonenberg@ioc.exchange)'s status on Tuesday, 06-Feb-2024 06:48:16 JST Andrew Zonenberg Trigger crossbar board shipped before CNY fab shutdown! Hoping it gets out of China before all of the FedEx staff leave for vacation.
Stencil is already here, parts should be coming in the next day or two and should arrive before the PCB.
This will be a fun design to bring up.
-
Embed this notice
Andrew Zonenberg (azonenberg@ioc.exchange)'s status on Tuesday, 06-Feb-2024 06:48:11 JST Andrew Zonenberg @rgilton This was for five boards 165x121 mm, 10 layers ENIG with impedance control on 4 layers and filled ViP (full depth vias only, no microvias or blind/buried).
Mixed Taiwan Union TU872SLK for signal layers and Shengyi S1000-2M between power planes and their respective grounds.
-
Embed this notice
Andrew Zonenberg (azonenberg@ioc.exchange)'s status on Tuesday, 06-Feb-2024 06:20:15 JST Andrew Zonenberg The RX pairs are easy to route on the top layer. I wasn't expecting a challenge there.
This is using 200 μm traces with 150 μm gaps: pretty tight coupling but that's necessary to avoid excessively large traces on OSHPark's 200 μm thick FR408HR prepregs. A loosely coupled pair with 50Ω single ended impedance for the legs would be 380 μm trace and space probably of 500μm or larger!
I could plausibly pull the gap all the way down to the fab limit of 125 μm but that felt like it might be pushing things just a bit too far in terms of repeatability. This is already a non-controlled-impedance process where I'm just hoping I get close enough that the SERDES work decently.
-
Embed this notice
Andrew Zonenberg (azonenberg@ioc.exchange)'s status on Tuesday, 06-Feb-2024 06:20:15 JST Andrew Zonenberg Starting initial testing to see how in the world I might try to fan out the GTYs of an UltraScale+ B676 package on OSHPark 4 layer.
I don't have the full schematic done, just doing initial layout feasibility checks.
The goal is to hook up "as much as reasonably possible". First non-negotiable requirement is that it has to fit on an OSHPark 4L board and it has to provide minimum power and JTAG connections so I can verify the reballed chip is alive.
After that, anything is a bonus.
Here's an initial quadrant dogbone fanout on 0.5mm legs for all of the power/ground balls in one corner of the GTY area. This will likely not be the final via location.
-
Embed this notice
Andrew Zonenberg (azonenberg@ioc.exchange)'s status on Tuesday, 06-Feb-2024 06:20:14 JST Andrew Zonenberg Gonna need to throw every trick in the book at this layout.
This book, that is. Luckily the little one is asleep so she's not trying to read it :p
-
Embed this notice
Andrew Zonenberg (azonenberg@ioc.exchange)'s status on Tuesday, 06-Feb-2024 06:20:14 JST Andrew Zonenberg The TX pins deeper into the array aren't so lucky. This is KiCAD's initial routing.
Let's just say I don't need to throw this in Sonnet to know that 25G NRZ isn't going to be happy on THIS. I could go fully single ended and route separate tracks for each leg but I still couldn't make them 380 μm for 50Ω single ended impedance, and the transition back to closely coupled would also be a pain.
-
Embed this notice
Andrew Zonenberg (azonenberg@ioc.exchange)'s status on Tuesday, 06-Feb-2024 06:19:03 JST Andrew Zonenberg Have other stuff to do today so might not get back to this until tomorrow.
But good progress! A bit more schematic stuff to hook up and then final design review (although I've checked most of it pretty thoroughly as part of GROVER, the schematic is largely copy pasted except for the top level).