#avr Logs

Jul 22 2022

03:21 AM twnqx: today's chip quote: FT232HL, full reel, 30$/IC
03:30 AM exp: lol
03:30 AM nohit: chip of the week
03:30 AM exp: terrible chips
03:30 AM exp: terrible price
03:30 AM exp: just buy RP2040s
03:30 AM exp: you'll hate yourself less :-p
03:30 AM exp: farnell still has 120,500 at £0.77/ea
03:30 AM twnqx: eh. RP2040s don't help in that case
03:30 AM exp: why not?
04:01 AM specing_ is now known as specing
04:03 AM twnqx: incidentally, i have a few thousand spare 32MB QSPI flashs (because we couldn't get the other parts in time)
04:04 AM twnqx: too bad the rp2040 can't handle that much :(
04:04 AM twnqx: and only USB 1.1... out!
04:11 AM exp: the rp2040 can't handle the flash?
04:14 AM twnqx: nope, only up to 16MB
04:15 AM exp: that's for directly executing
04:15 AM exp: cause 3 bit PC or whatever
04:15 AM exp: 3 byte*
04:15 AM twnqx: it says "Support for up to 16MB of off-chip Flash memory via dedicated QSPI bus"
04:15 AM exp: right, there's no hard limit on flash size
04:15 AM exp: it's just where code can be executed directly
04:16 AM exp: would be very surprised to see 32mb flash modules not actually work
04:16 AM exp: but i've been surprised before :)
04:16 AM twnqx: me too
04:16 AM twnqx: e.g. the sama5d27c not being able to boot from the majority of 32MB chips
04:17 AM exp: what was the issue? timing?
04:17 AM twnqx: yey: issi, esmt ney: spansion, winbond, micron, gigadevices
04:17 AM twnqx: no idea
04:17 AM exp: that's a niche list
04:17 AM exp: something's gotta be fucked there, i wonder what though
04:18 AM twnqx: we used scopes on all lines, with a qspi protocol decoder
04:18 AM twnqx: to us it looks correct, yet the data lines float after the commands
04:19 AM twnqx: you wouldn't believe how happy we were when we found a working one
04:19 AM exp: oh i definitely would, i've been in that position
04:19 AM twnqx: especially as the issue is only with the stage 1 loader in rom
04:20 AM twnqx: when we loaded the stage 2 loader to ram and executed it, that one would happily load stage 3 from qspi
04:20 AM exp: yeah so just checking the 2040 out, it's a 2³² address space, but only 2²⁴ mapped in from the flash
04:20 AM twnqx: and u-boot (our stage 3) then the fit image, and the kernel would hyppily read/write
04:20 AM exp: hence 16 meg
04:20 AM exp: bizarre, has to be some sort of clocking issue
04:21 AM exp: would have been a fun if intensely annoying debugging session
04:21 AM twnqx: i think it's a "multitude of command variations" issue
04:22 AM exp: well forgive my ignorance, but how many different commands can a stage 1 be issueing lol
04:22 AM twnqx: 3 byte adressing? 4 byte addressing? flash descriptor tables?
04:22 AM twnqx: *a lot*
04:22 AM exp: i mean how many unique commands
04:22 AM exp: a lot of reads is still just a read :-p
04:22 AM twnqx: first, id the chip to figure out WHICH comamnd in WHICH format it uses
04:22 AM twnqx: to read.
04:23 AM twnqx: then, switch to quad mode
04:23 AM twnqx: then, start reading
04:23 AM twnqx: with the right read command.
04:23 AM exp: right, you'd expect that to be an extremely straightforward sequence to debug
04:23 AM exp: i don't think you're incompetent though so i'm sure it wasn't
04:24 AM twnqx: we stopped investigating it
04:24 AM exp: even so, i want to be clear that i'm not dismissing your efforts by saying it sounds easy
04:24 AM twnqx: i don't have the gear myself (a colleague in the use used 2 4-channel scopes to do that)
04:24 AM exp: every bit of debugging sounds easy, until it turns out it doesn't fuckin work :-p
04:24 AM twnqx: in the US*
04:25 AM exp: you should get yourself a little picoscope or something IMO, they're awesome to have just lying around
04:25 AM exp: but maybe you work in a van idk
04:25 AM twnqx: dunno if those can handle 104mhz ddr
04:26 AM twnqx: on at least 6 lines in sync
04:26 AM twnqx: or 166mhz ddr for ddr3 dram
04:26 AM exp: 104 should be reasonable
04:26 AM exp: 166 would be pushing
04:27 AM exp: you're getting into the 'maybe need an active probe' range there
04:27 AM twnqx: a little picoscope also doesn't have 6 synchronized channels, no?
04:27 AM exp: they're usually MSOs
04:27 AM exp: so if you don't need eye diagrams, you can use digital inputs
04:27 AM twnqx: 104 ddr equals 208mhz data rate, so.. 500 mhz at least?
04:27 AM exp: yeah these things go up to about 200mhz bandwidth w/1gsps or so
04:27 AM exp: not too bad
04:28 AM exp: certainly not for the price and openness
04:29 AM twnqx: i have the smallest picoscope (2 channels, 10mhz) for some simple measurements, e.g. supply noise
04:29 AM twnqx: never managed to get it running on linux
04:29 AM twnqx: also they dropped winblows support for anything lower than malware 10
04:30 AM exp: well that's not surprising, w7 is eol
04:30 AM exp: i'm trying to think if i've run this scope on linux, it's plugged into a windows laptop at the moment
04:30 AM twnqx: yeah sure, but 10 is not usable in a professional environment IMO
04:30 AM exp: but i swear i have
04:30 AM twnqx: too much uncontrolled sending data to MS
04:31 AM exp: anyhow it sounds like you already know baout picoscopes so i'll just shut up :)
04:31 AM twnqx: i know them, and i have a love/hate relationship with them
04:31 AM twnqx: also they don't work greatly with openscopehal, which would be my preferred system thanks to the decoders
04:32 AM twnqx: or what was it called...
04:32 AM exp: azonenberg is working directly with picoscope on this and future product devs i think
04:32 AM exp: so any issues like that i'm sure they'd both love to hear
04:32 AM twnqx: andrew is also our designer/anylyst in the US :P
04:33 AM exp: aah fair enough, he has a lot of gear and a lot of ideas
04:33 AM exp: i'd like to get to his stage in a few years time
04:33 AM twnqx: and that's not even his day job...
04:33 AM exp: well it is isn't it
04:33 AM twnqx: nope
04:33 AM exp: it's just he has his own side hustle
04:33 AM twnqx: that IS is side hustle
04:33 AM twnqx: a bit of chip design, pcb design, software
04:34 AM exp: what's his day job then?
04:34 AM twnqx: Associate Principal Security Consultant at IOActive, Inc
04:34 AM exp: meh same core job really
04:37 AM twnqx: meh i wish i could stab the authors or rfc8152
04:37 AM twnqx: but when i tried to reach out to one of them, his brother answered to say that he'd died in 2020 :(
04:37 AM exp: yes it's horrible when that happens
04:38 AM exp: haven't come across CBOR before
04:38 AM twnqx: it's neat for some things
04:38 AM twnqx: best summarized as "binary json"
04:38 AM exp: i hope not because that makes it the unholy arse of satan
04:38 AM exp: (i dislike json lots)
04:39 AM twnqx: well, it's better specified
04:39 AM exp: also any rfc that splits its EBNF or whatever up into sections i can't automatically parse out can eat a dick
04:39 AM twnqx: 8152 has them in a aplit-out git repo
04:40 AM exp: that's nice at least
04:40 AM twnqx: it just doesn't help as they miss some key tings
04:40 AM exp: programmatically parsing RFCs into grammar is a thing on my todo list
04:40 AM exp: brb 5 mins
04:40 AM twnqx: e.g. if you encrypt something with AES-GCM you get a "tag" for verification purposes in the end
04:41 AM twnqx: the rfc does not really specify what to with that, but at least the reference code just appends it to the ciphertext, and splits it out from decrypting before
04:41 AM twnqx: would have been nice to have that written out
11:50 AM SamantazFox_ is now known as SamantazFox
04:02 PM specing_ is now known as specing