a thread about janitoring individual bits and using oscilloscopes and stuff

The Something Awful Forums > Discussion > Serious Hardware/Software Crap > YOSPOS > a thread about janitoring individual bits and using oscilloscopes and stuff

«‹›2 »

big shtick energy: May 27, 2004

been working with one of analog device's fattest audio DPSs (adsp-214xx) for over a year, as many complaints as I have about the libraries and support, the chip and IDE are pretty legit

at one point I rewrote a parsing routine for some encoded audio in assembly* and was proud of how many cycles it saved, until I made the same algorithmic improvements in the C code and added some compiler hints and welp the compiler beat me by a few cycles per iteration on the inner loop

*because I was a freaking nose-hair away from having enough mips for worst case

# ¿ Dec 2, 2014 01:54

Adbot: ADBOT LOVES YOU

# ¿ Apr 29, 2024 16:40

big shtick energy: May 27, 2004

Jonny 290 posted:

one of my 2015 goals is to figure out a microcontroller solution for transmitting and receiving AFSK without a dedicated modem chip. then I can strap this to a $30 chinese ham radio and have a $70 packet radio node that I can hook up to whatever and do neat things with.

there doesnt really seem to be a solution to this right now. theres an arduino modem board that'll do it but it's $$$$ and feels too Commodore-1541 to me

well some guy made an fm transmitter out of a small uc and a battery because the main pll was reconfigurable fast enough for that to be viable way to modulate the core clock with audio frequencies, so as long as the pll lock is much shorter than the symbol duration I guess you could do it. doesn't do receiving, though

# ¿ Dec 5, 2014 21:14

big shtick energy: May 27, 2004

yeah if there comes a day when using C is embedded is mostly a thing of the past I will be very sad

# ¿ Dec 5, 2014 22:56

big shtick energy: May 27, 2004

the gige switch chip I'm working with right now has a neat LED interface, to make it customizable they have all the LEDs run by this little processor that you program in assembly, it has two registers, 2 flags (carry and zero), a few dozen opcodes, and they claim the whole thing is less than 1500 gates. kinda neat.

what was not neat was not getting the documentation and having to reverse engineer those opcodes from their example code and trying to figure out what some of the custom assembly ops did. although we finally got the docs today so I guess I could have waited and saved a good few hours of work

# ¿ Jan 13, 2015 21:36

big shtick energy: May 27, 2004

eschaton posted:

now imagine using it for input devices

one of which is a mouse that's round and uses two little wheels that don't work instead of a roller ball or optical sensor

wait are people complaining about i2c? it's a pretty easy bus to use, not sure how you would get in to trouble unless you were using really wonky pull up or termination values. Although I guess sometimes it's hard to see which device is loving up the bus, which is why a little series termination on your bus can be nice because the voltage will be a little higher/lower on one side of the series terminator depending on which side is pulling down

# ¿ Apr 2, 2015 22:58

big shtick energy: May 27, 2004

JawnV6 posted:

i2c is great but i'll never understand why every device datasheet has timing diagrams explaining the protocol

like give me a register map and call it a day, there's no need to put the exact same timing diagrams on every single datasheet

some of us go *just* long enough between i2c implementations to forget. Also every so often a device fucks around with the usual setup and doesn't have sub addresses or does some other weird thing
start - address w - sub address - start - address r - sweet data - stop
managed to forget the sub address the first time so I guess I'm the guy those are for

also IMO all i2c addresses should have the msb set to protect us from the 50/50 split of people who do/don't include the r/w bit in the address when writing it in hex

# ¿ Apr 2, 2015 23:33

big shtick energy: May 27, 2004

we're doing our next product with an stm32f4, and since Kiel uvision 4 kind of sucks, it looks like we might go with gcc/eclipse/openocd. it's not a horrible idea is it?

also how does an ide in tyool 2015 only rarely have text completion actually work. and also store what windows you have open in the same file as project options you care about so you have to version control that file and then of course you have an incompatible change every time you even open a window in the bloody project. wtf keil.

# ¿ Apr 7, 2015 18:21

big shtick energy: May 27, 2004

JawnV6 posted:

anyone have experience with RIOT, contiki, or other "iot rtos" solutions?

i need a scheduler for a wifi/tcp driver, thinking FreeRTOS will be plenty, but if one of these upstarts is really compelling I'd consider it

our next product uses freeRTOS and it seems good so far

# ¿ Aug 26, 2015 19:21

big shtick energy: May 27, 2004

Bloody posted:

if you guessed replacing the 32x8192 fifo with like 256mb of sram you guessed right!!!

how fat is your BOM that 256mbit of SRAM is no big deal?

# ¿ Dec 7, 2015 21:14

big shtick energy: May 27, 2004

Sagebrush posted:

do you have to worry about transistor switching speed? or is 200khz (i would guess, for 100kbit) just like totally bog slow as far as a transistor is concerned

i know that some types are faster than others but that's it

just do s back of the envelope calculation with the gate capacitance and gate drive current to check. rise time less than 10% of the signal period should be plenty for nice clean edges

# ¿ Dec 21, 2015 21:49

big shtick energy: May 27, 2004

just realized the webserver my former coworker wrote that assembles resources by writing them to flash before lwip sends them out is not going to work so well with the small and cheap NOR flash we specced

changing the webserver to serve from dram should be nice and fast, at least

big shtick energy fucked around with this message at 20:47 on Jan 20, 2016

# ¿ Jan 20, 2016 20:41

big shtick energy: May 27, 2004

longview posted:

too much work, just get some FRAM instead of NOR flash

haha, list price for a 1Mbyte chip is like $50

# ¿ Jan 20, 2016 22:17

big shtick energy: May 27, 2004

https://www.youtube.com/watch?v=Ac__WxROA10&t=3777s
Skip to 1h03m

Switched DC-DC power supplies completely (like, even the capacitance) on-die. Basically magic

# ¿ Feb 14, 2016 07:49

big shtick energy: May 27, 2004

debugging an occasional problem that occurs with a frequency somewhere between once per hour and one per 24 hours (depending on some conditions, maybe, I think)

madness is setting in

# ¿ Feb 15, 2016 22:42

big shtick energy: May 27, 2004

yeah we thought we had fixed it until it seemed to suddenly start happening a whole bunch to customer units in the field

# ¿ Feb 15, 2016 23:39

big shtick energy: May 27, 2004

I've now reached the stage of madness where one becomes convinced that the termination provided by attaching a scope probe is fixing everything on the bus

it almost makes sense if you forget that it's an I2C bus running at 3.3V and 10kHz and shouldn't be anywhere near that sensitive

# ¿ Feb 16, 2016 23:15

big shtick energy: May 27, 2004

yeah, a few days ago we figured out it was the capacitance

weirdly although the probes were only rated at 12pF they were more effective at preventing glitches than an 18pF cap connected closer to the problem area

# ¿ Feb 22, 2016 18:32

big shtick energy: May 27, 2004

movax posted:

add the scope probe equivalent circuit to your schematic and it should make sense

the fact you need that filter on the line is weird as hell. are your reflections / ringing that bad that devices are seeing double-clocks?

no idea, the clock/data lines always looked pretty drat clean, even with a differential probe. the lines run pretty close to a high speed lvds pair and some traces for a 12V RS232 line so there are some plausible sources of noise, but I was never able to actually see any glitches

# ¿ Feb 24, 2016 00:44

big shtick energy: May 27, 2004

looking at the captures from the diff probe, the capacitance does clean the bus up a decent amount

it's just that the dirtier signal shouldn't have been a problem unless the i2c slave devices were way shittier than advertised with much less input hysteresis than listed in the datasheet

oh well, it works now, so once the current stock is modded and the pcb design changed I'll forget about it and see if it ever blows up in my face again

# ¿ Feb 24, 2016 20:31

big shtick energy: May 27, 2004

JawnV6 posted:

a datasheet lied

well yeah but you would think "can handle like <100mV of noise on a 3.3V bus" would be something you could trust

# ¿ Feb 24, 2016 21:52

big shtick energy: May 27, 2004

Barnyard Protein posted:

it looks like a peanut-butter and germanium sandwich on rye.

intel getting into some exotic materials to keep moore's law going

there's been a lot of articles on the death of moore's law recently, but it's sounding more plausible that they're right. this time it's not so much based on the physics but on the economic argument that no one can afford the fabs anymore

then again there was all the angst about 28nm being the cheapest per transistor and things going up from there and yet now there is tons of poo poo using 14/16 finfet

supposedly the reason people want iot to be real is that without an absolutely ginormous market for chips (like an order of magnitude bigger than smartphones) there wouldn't be enough customers to fill a fab a few nodes from now

# ¿ Mar 4, 2016 18:51

big shtick energy: May 27, 2004

Star War Sex Parrot posted:

this thing is my enemy

tools, documentation and the hardware are pretty good with our analog devices dsp, unfortunately some product-ruining problems in certain software libraries has kind of led me to hate them

# ¿ Mar 5, 2016 06:32

big shtick energy: May 27, 2004

is the place you're getting the certificate a for-profit institution? are you using loans to pay for it?

# ¿ Apr 8, 2016 16:16

big shtick energy: May 27, 2004

Mr Dog posted:

ya risc-v is extremely ftw but i asked rjmcall about it in a couple of other threads and he said it's an academic wanking exercise because function call prolog/epilog is comically bloated or something like that

also sorry about your job

compared to what? aren't you saving a bunch of registers in most architectures (except for SPARC as described in the paper jawn linked)

# ¿ May 3, 2016 20:17

big shtick energy: May 27, 2004

Storysmith posted:

a joke yes but in order to get modelsim to work with cvs for a group project in school i actually had to add a bunch of stuff to the ignore file and write tcl that just blew away the existing project and readded everything because otherwise it would just merge conflict on everything when you tried to commit

ise was not appreciably better at this to the point that I stole and tweaked some open source dude's makefile and just did everything outside of the gui

i don't know what Actual Companies developing Actual Designs in teams of more than 4 people would do other than drink heavily and hate everyone they work with

pretty sure it's massive stacks of arcane Tcl scripts (and drinking and hate of course)

# ¿ May 8, 2016 22:57

big shtick energy: May 27, 2004

some people using some really scrub tier micros when Stm32F4 discovery boards (with built in debugger) are only $20 and have everything you need, including the ability to develop with eclipse and gcc

maybe I'm biased because it's what we use at work and is by far the least awful tool chain and kit I've used for embedded work

# ¿ May 9, 2016 21:07

big shtick energy: May 27, 2004

I'm getting the sense that the RISC-V guys will need some help from an adult if they want their poo poo to be used in the real world.

# ¿ May 9, 2016 22:44

big shtick energy: May 27, 2004

gently caress consumer grade flash memory

myself and others, through a series of seemingly reasonable decisions, made our product completely dependent on consumer grade flash memory working properly

on the plus side we can afford to get industrial grade microSD cards for the product at least (at like $50 a card)

# ¿ May 26, 2016 22:56

big shtick energy: May 27, 2004

BobHoward posted:

probably "does it poo poo itself into oblivion soon as you accumulate a few write passes"

consumer grade tlc nand flash is poo poo tier stuff and lol if u think the ftl (translated: error correction, wear leveling, etc) implemented in a microsd card is good enough to compensate under any serious write load, especially random write. sd cards are for temporarily storing ur digicam photos until you copy them to a real storage device, no fucks are given about making it good enough to reliably boot a linux from or w/e. it has to be cheap, it has to be super low power, all else is a distant second

yeah basically

write load is not that high but it's basically 1 cluster every 10 seconds which is enough I guess. also the occasional random power loss which isn't really their fault since consumer cards aren't specced for that

crazy part is that a bunch of different systems have started failing within a week of each other despite being manufactured and put into service on wildly different dates (same batch of sd cards tho)

# ¿ May 27, 2016 18:27

big shtick energy: May 27, 2004

movax posted:

jesus christ please god no

seriously, why would they? do they want xilinx's serdes's that badly to build some kind of giant backbone/virtex ultrascale like part that xilinx already basically produces just for cisco/juniper/etc?

why should anything make sense, just buy buy buy

intel has still yet to release any altera chips made by intel fabs

# ¿ Jun 1, 2016 01:44

big shtick energy: May 27, 2004

Bloody posted:

that's probably a multi year transition modern processes aren't that simple

yeah I guess it's only been 1 year, thought it was longer for some reason

# ¿ Jun 1, 2016 02:30

big shtick energy: May 27, 2004

randomly came across this pro click: http://newsletter.sigmicro.org/sigmicro-oral-history-transcripts/Bob-Colwell-Transcript.pdf

sorry about the formatting I'm on mobile

quote:

Anyway, for some reason, there was an organizational meaning at which Albert Yu could not

appear. He designated Fred Pollack, but Fred could not appear, so Fred designated me, and I

showed up. So first of all I am two organizational levels down from who is supposed to be

sitting there and I ended up sitting next to Gordon Moore. This was probably about 1994 or

so. The presenter happened to be the same guy who was in the front of the car from when I

interviewed with the Santa Clara design team; same guy. He's presenting and he's predicting

some performance numbers that looked astronomically too high to me. I did not know

anything about how they expected to get there, I just knew what I thought was reasonable,

what would be an aggressive boost forward and what would be just wishful thinking. The

predictions being shown were in the ludicrous camp as far as I could tell. So I'm sitting and

staring at this presentation, wondering what are they doing, how is it humanly possible to get

what he's promising. And if it is, is it possible for this particular design team to do it. I was

intensely thinking about what's happening here. Finally I just couldn't stand it anymore and I

put my hand up. There was some discussion, but you have to realize none of these people

were really chip designers or computer architects, with the exception of Gelsinger and Dadi

Perlmutter.

...

Yeah, his real name is David, he�s an Israeli. Everybody calls him Dadi. And then

Pat Gelsinger who was the chip architect, designer in 386 and 486. But most of those guys at

this presentation haven't designed anything themselves, they know how to manage

complicated large expensive efforts, which is a different animal. Anyway this chip architect

guy is standing up in front of this group promising the moon and stars. And I finally put my

hand up and said I just could not see how you're proposing to get to those kind of

performance levels. And he said well we've got a simulation, and I thought Ah, ok. That shut

me up for a little bit, but then something occurred to me and I interrupted him again. I said,

wait I am sorry to derail this meeting. But how would you use a simulator if you don't have a

compiler? He said, well that's true we don't have a compiler yet, so I hand assembled my

simulations. I asked "How did you do thousands of line of code that way?" He said �No, I did

30 lines of code�. Flabbergasted, I said, "You're predicting the entire future of this

architecture on 30 lines of hand generated code?" [chuckle], I said it just like that, I did not

mean to be insulting but I was just thunderstruck. Andy Grove piped up and said "we are not

here right now to reconsider the future of this effort, so let�s move on". I said "Okay, it's your

money, if that's what you want."

Suddenly this came up again later in another guise but again Andy shut me off, he said

"we're not here to discuss it". Gordon Moore is sitting next to me and hasn�t said a word, he

looks to all intents and purposes like he's asleep. He's got his eyes closed most of the time,

you think okay, the guy's tired, he's old. But no, 20 minutes into this, he suddenly opens his

eyes and he points to me and he asks, "did you get ever get an answer to your question?" and

I said, "actually no, none that I can understand". Gordon looked around and says, "how are

we planning to move ahead with this, if the answers don't make sense?" and this time Andy

Grove said to him "We�re not here to discuss that, Gordon". [laughter]. Gordon really

impressed me with his ability to cut right to the heart of the matter. I saw him do that more

than once, you think he's not paying attention, but when he pipes up, what he's about to say

is dead on; he's not only been following the discussion but he sees where the real issue is, he

points at it. That guy was impressive.

# ¿ Jun 3, 2016 20:04

big shtick energy: May 27, 2004

is this a group project? how long do you have

edit: does the business case for it matter at all or are you just marked on it as a technical exercise?

# ¿ Jul 14, 2016 19:19

big shtick energy: May 27, 2004

longview posted:

it's really not that difficult with modern circuitry and well designed standards

alternatively you can always decide to save like $2 on a $5k BOM by making the daughter board for your >100MHz SERDES chip 2-layer (neither of them having much ground plane on them) instead of 4-layer. there was no amount of eldritch patterns made with copper tape that could make that poo poo pass FCC and we had to just get a 4-layer board

also for the guy who wants to parallel DC supplies: if you want to cowboy it, you can always just use regular supplies and put like a 0.1-ish-ohm, many-loving-watt resistor between each supply and the voltage rail. i think the sense node should still be on the power supply side of the resistor but if it smokes then probably it should have been the other way

# ¿ Jul 20, 2016 01:25

big shtick energy: May 27, 2004

Zopotantor posted:

One day, when the stars are right and your signal traces are routed just so, the many-angled ones will rise and rule again. I�, cross-talk Cthulhu noise floor wgah-nagl clock-domain crossing fhtagn!

when standing inside a giant EMI test chamber in the middle of the night that's roughly what goes through your head

# ¿ Jul 20, 2016 21:40

big shtick energy: May 27, 2004

movax posted:

http://www.bloomberg.com/news/articles/2016-07-26/analog-devices-said-in-advanced-talks-to-buy-linear-technology

fuuuuuuuuuuuuck

i love ltc parts, i'm ehhh on ADI's stuff (outside of SDR)

many, many parts are gonna get EOL'd like a motherfucker

Linear's datasheets/whitepapers are amazing, especially the ones by that Jim something guy (someone must remember who I mean, he had a datasheet where he put "they said I could put anything I wanted to on this page so here are a bunch of doodles")

but we gotta have more semiconductor mergers for the semiconductor merger throne

# ¿ Jul 27, 2016 01:24

big shtick energy: May 27, 2004

https://en.wikipedia.org/wiki/Jim_Williams_(analog_designer) was the guy

anyway, not looking forward to 2025 when the whole semi industry becomes one big company

# ¿ Jul 27, 2016 01:29

big shtick energy: May 27, 2004

Mr Dog posted:

don't most large asics have power supply sequencing requirements that you have to follow in order to avoid latch-up? i know at least back in the day the TI OMAP had to have a second companion chip just to switch poo poo on in the right sequence with the right delays

yeah it's pretty common but it should be explicitly stated in the datasheet (where it should tell you the correct sequence and time delays)

# ¿ Sep 9, 2016 19:46

big shtick energy: May 27, 2004

is jffs2 a horrible decision that will haunt me forever? it seems alright but I haven't actually looked through the source yet

# ¿ Sep 27, 2016 22:47

Adbot: ADBOT LOVES YOU

# ¿ Apr 29, 2024 16:40

big shtick energy: May 27, 2004

on second thought the available jffs2 code is all for Linux and I don't know if wrapping/rewriting hundreds of system calls and dynamic memory allocation is a path to a happy life

otoh YAFFS looks pretty good but a commercial license is $10k which is probably 3x the price at which we say "hmm let's think about this"

there's an active github project that looks like exactly what I want with a permissive license but who knows what rare bugs could be lurking within that

# ¿ Sep 29, 2016 20:39

The Something Awful Forums > Discussion > Serious Hardware/Software Crap > YOSPOS > a thread about janitoring individual bits and using oscilloscopes and stuff

«‹›2 »