#garfield Logs
Jan 24 2021
#garfield Calendar
12:08 AM rue_mohr: I lost everyone
12:15 AM rue_mohr: so, how can I play with 1 bit
12:15 AM rue_mohr: I suppose if I make a hardware generator, I could feed it into a software low pass and output it to a parallel DAC
12:15 AM rue_mohr: if that works, you have to be doing soemthing right
12:15 AM rue_mohr: distraction project?
12:15 AM rue_mohr: hmm
12:15 AM rue_mohr: quite probably
12:15 AM rue_mohr: its something anyhow
12:15 AM rue_mohr: ok, lets design the software fitler
12:15 AM rue_mohr: shall we use 3.5k and .0047uF again?
12:15 AM rue_mohr: how about something more reasonable, like 3k
12:15 AM rue_mohr: sure... whatever
12:16 AM rue_mohr: 0.005uF and 3k
12:16 AM rue_mohr: / SR = 500000.75800000
12:16 AM rue_mohr: / Fc = 10610.32953946
12:16 AM rue_mohr: #define k 7.9969963581
12:17 AM rue_mohr: k is 8, that makes for nice integer mathz
12:18 AM rue_mohr: ok, so set up some AVR code with a timer, causing an 500khz interrupt
12:18 AM rue_mohr: question
12:18 AM rue_mohr: yes?
12:18 AM rue_mohr: that only leaves 32 instuctions for the interrupt
12:18 AM rue_mohr: yes, it could be tight
12:19 AM rue_mohr: shall we do this on stm32?
12:19 AM rue_mohr: lets see how bad the aliasing is
12:19 AM rue_mohr: we dont need to
12:19 AM rue_mohr: ?
12:19 AM rue_mohr: we simply dont need to worry about the system thruput, its just a test
12:19 AM rue_mohr: ah if we get signal were ok
12:20 AM rue_mohr: tho we should keep tabs on the alias freq
12:20 AM rue_mohr: what is the alias freq of an oversampled 1 bit stream?
12:20 AM rue_mohr: maybe the value of the oversampling is a clue?
12:20 AM rue_mohr: huh, thats worth some thought
12:21 AM rue_mohr: anyhow!
12:21 AM rue_mohr: yes, anyhow, if its too fast we can just slow it down
12:21 AM rue_mohr: can I do the math in 32 instructions?
12:21 AM rue_mohr: maybe, there is a software divide
12:21 AM rue_mohr: we need to adjust gain too
12:23 AM rue_mohr: would it make sense to have a matching lowpass on the normal comparator input?
12:23 AM rue_mohr: hmm, I suppose
12:42 AM rue_mohr: iiinteresting
12:42 AM rue_mohr: a 1 bit running at 500khz
12:43 AM rue_mohr: breaks down at an input freq of about 4500Hz
01:03 AM rue_mohr: stm32
01:03 AM rue_mohr: m328
01:03 AM rue_mohr: stm32
01:03 AM rue_mohr: m328
01:03 AM rue_mohr: ...
01:03 AM rue_mohr: we really should use the stm32
01:03 AM rue_mohr: but I'm tired
01:03 AM rue_mohr: you know I'm right
01:03 AM rue_mohr: yes, but I'm tired
05:18 AM polprog: [02:46] rue_shop3: iirc whitequark set up a serial port on each pin, and transmitted the pin number
05:19 AM polprog: ah yes she did. that was a clever hack haha
11:41 AM rue_mohr: :)
12:06 PM rue_mohr: the other line modules for the tripple laser lens tester came in
12:07 PM rue_mohr: polprog, simple design challange...
12:08 PM rue_mohr: "small T flipflop"
12:08 PM rue_mohr: aka not a 1G SMT, but not much larger than a 8 pin dip
12:08 PM rue_mohr: https://www.electronics-tutorials.ws/wp-content/uploads/2018/05/waveforms-tim17.gif
12:09 PM rue_mohr: but maybe less parts?
12:09 PM aandrew: rue_mohr: sure, you can turn an FPGA into a (shitty) 100 channel ADC. I was looking at using the diffpair inputs as comparators for a similar function, but they aren't optimized for that
12:10 PM rue_mohr: what his name was pointing out that CD players are 325khz oversample with a simple filter
12:10 PM rue_mohr: oh yea, no the fpga input thing I'm gonna not go near
12:11 PM rue_mohr: performance is horrid
12:11 PM polprog: ha, if only i knew what a T flip flop is
12:12 PM polprog: shame we have exams starting in a week, i have zero time :/
12:12 PM polprog: for nothing
12:12 PM polprog: these idiots now want us to contiue doing ee labs right until the exams start
12:12 PM rue_mohr: your too busy with school to learn about the world
12:13 PM polprog: yeah
12:13 PM polprog: i mean, i still keep trying to do both
12:13 PM aandrew: rue_mohr: I was going to do something similar but with a timer instead with a bunch of analog signals feeding LVDS + inputs and the - input on an RC filter fed by PWM from another FPGA pin
12:13 PM polprog: at least the other guy is writing the lab report because he says he wants to learn
12:13 PM aandrew: but the issue is that FPGA comparators aren't designed to have precise small signal response
12:13 PM polprog: thats fine by me, last lab i was just explaining and supervising ;)
12:14 PM rue_mohr: aandrew, they do an awefull job
12:15 PM rue_mohr: but my stm32 version with the LM393 comparator didn't do awesome
12:15 PM rue_mohr: looks like I need a lower Fc on the filter
12:15 PM aandrew: possible. I haven't played around much with this idea although I'd like to
12:16 PM rue_mohr: I suppose a person needs to choose the ripple level
12:16 PM aandrew: and 16-channel comparators are expensive
12:16 PM polprog: also rue_mohr https://www.youtube.com/watch?v=ro5VE7caF8c
12:16 PM polprog: https://www.youtube.com/watch?v=7V0LCL4mL-8
12:16 PM rue_mohr: you know, 2N7000 have a really touchy Vgs
12:17 PM rue_mohr: I dont remember seeing electrolytics last time I opened it up
12:18 PM aandrew: thankfully I've more or less used discrete fets like I use discrete bipolar transistors... I slam them on and off with digital logic and almost never have to worry about the persnickettyness of them
12:18 PM aandrew: rue_mohr: very sorry to hear about your scope, hopefully that google group (or wahtever it is these days) can help out
12:18 PM rue_mohr: it works, but it cant capture
12:18 PM rue_mohr: the log says when it booted it found 0 ram
12:19 PM rue_mohr: so it put all the buffers over each other
12:19 PM rue_mohr: and when it tries to send serial it times out
12:19 PM aandrew: that's definitely an interesting failure mode
12:20 PM aandrew: it almost sounds like the capture memory is separate from the operational memory which is kind of nice, maybe there's a daughterboard you can replace outright
12:20 PM rue_mohr: I kinda wonder if I made the parallel port capture device if I could get around it
12:20 PM rue_mohr: yea, it implies there is a subprocessor having the issue
12:20 PM rue_mohr: a 68k
12:21 PM rue_mohr: oof, its got those electrolytics eh?
12:21 PM rue_mohr: ouch
12:22 PM rue_mohr: I'm worried about the cal data tho, I dont have a gbip to back it up
12:23 PM aandrew: yeah that is one thing that frustrates me about all test equipment like that -- unbackuppable (or at least not easily) cal data
12:23 PM aandrew: there are quite a few scopes that are like that
12:41 PM rue_mohr: wow, I'm having get-up motivation issues!
01:32 PM Tom_L: so no on the fpga?
01:32 PM rue_mohr: too much schmitt
01:32 PM Tom_L: did you try to optimize it?
01:32 PM rue_mohr: ?
01:33 PM rue_mohr: I was just testing the schmitt of the differential inputs
01:33 PM rue_mohr: its way too high to use for any reasonable ADC
01:34 PM rue_mohr: but my implementation with an stm32 and a LM393 didn't perform well
01:34 PM rue_mohr: I need to change the filter
01:35 PM rue_mohr: this week I'll crawl thru my scope making sure the caps aren't leaking
01:45 PM aandrew: fingers crossed
02:07 PM aandrew: I'm very curious to try out usbip; it's built into linux and might work decently well over wired connections
02:07 PM aandrew: I've tried using various types of usb-over-ip before and they were all kind of meh
02:47 PM polprog: https://twitter.com/winocm/status/1353189707669098496
03:08 PM aandrew: ... when you absolutely MUST use rambus memory?
03:35 PM polprog: dont ask me
07:32 PM rue_mohr: omg, one day after the next
07:33 PM rue_mohr: should I got the hospital tonight and pick up all the old printers and montiors?
07:33 PM Tom_L: no
07:33 PM Tom_L: you got enough to keep you busy for a lifetime
07:38 PM rue_mohr: some of those wide screen monitors work...
07:48 PM aandrew: man that would be *awesome* to have a steady supply of dead printers, especially if they're the old dot matrix ones
07:49 PM aandrew: when my 8yo and I tear apart a modern printer we're lucky to find two good motors in it. I mean it's *amazing* how much motion they can squeeze out of a single motor and a bunch of clutches/gearing, but shitty because the motors are weaksauce and few.
07:49 PM Tom_L: "Being a little older I am very fortunate to have someone call and check on me every day. He's from India and is very concerned about my car warranty."
07:49 PM aandrew: about 10y ago I was on a binge to get lots of LCDs from old/dead laptops. Do you know what I've done with them? abso-fucking-lutely nothing. nada.
07:49 PM aandrew: Tom_L: lol
07:50 PM rue_mohr: ha
07:51 PM rue_mohr: TIME!
07:51 PM rue_mohr: at 5:39 I finished lunch
08:02 PM rue_mohr: is it bad to put a heartbeat monitor on a computer to reset it when it crashes?
08:09 PM aandrew: if you must keep that specific computer running, no, but my first instinct would be to replace it
08:09 PM rue_mohr: ok I guess I better go get them
08:09 PM rue_mohr: well its got bad caps AGAIN, I'm sure
08:09 PM rue_mohr: its the shop fileserver
08:09 PM aandrew: ... how does it have bad caps *again* ?
08:09 PM aandrew: are you replacing the caps with more bad caps?
08:10 PM aandrew: I'm also willing to bet that the fileserver has NO serious CPU need so why not replace it with a $20 orangepi lite2 or something similar?
08:10 PM aandrew: (can't remember if that one has SATA)
08:10 PM rue_mohr: well it was WAY better with the pentium 150
08:11 PM rue_mohr: I need the parallel and serial ports
08:11 PM rue_mohr: BUT
08:11 PM rue_mohr: you might be on to something
08:11 PM rue_mohr: bbl, I need to go get more junk
08:11 PM rue_mohr: need?
08:11 PM rue_mohr: hmm
08:12 PM aandrew: why do you need the parallel and serial ports on a file server? maybe break that apart into a file server that is tiny and sips power and sits in a corner away from abuse, and then something replaceable with real lpt and usb serila
08:14 PM aandrew: not what I was thinking but not bad: https://wiki.radxa.com/Dual_Quad_SATA_HAT
08:19 PM aandrew: https://www.cnx-software.com/2017/01/19/orange-pi-zero-nas-expansion-board-with-sata-usb-and-av-port-sells-for-10-shipped/ is what I was thinking but that's only one sata port, you'd need a port multiplier which sucks
08:20 PM aandrew: https://shop.allnetchina.cn/collections/sata-hat is where you find the first one but sold out
08:20 PM aandrew: btw rockpi is awesome. I got the pro one as it's got a PCIe slot on it
08:22 PM aandrew: aha even better: $25 ali: https://www.aliexpress.com/item/33033052388.html
08:27 PM aandrew: that one uses a nanopi m4 which is discontinued (m4v2 is there, but $80)
08:28 PM aandrew: but for $105 you can get them both from friendlyarm
08:28 PM aandrew: https://www.friendlyarm.com/index.php?route=product/product&product_id=268
08:32 PM aandrew: that'd take care of your fileserver. it looks like the rock64 has the same pinout which should work, but you'd have to check for sure.
09:31 PM rue_mohr: Life would be complicated if I cant run fullblown linux on it
09:31 PM rue_mohr: BUT I have some micro machines that can do that
09:36 PM rue_mohr: 7:30pm, the truck is full of printers and monitors
09:36 PM rue_mohr: and 2 laptops
09:38 PM aandrew: rue_mohr: ... that *is* fullblown linux
09:38 PM aandrew: or do you mean x86
09:39 PM rue_mohr: nyam...
09:39 PM aandrew: hm?
09:50 PM rue_mohr: there are a few reasons its easier for me to deal with it as a normal atx in a case
09:50 PM rue_mohr: I do have a bunch of motherbaords I could swap in
10:26 PM rue_shop1: 3k 0.047uF
10:42 PM rue_mohr: / SR = 500000.75800000
10:42 PM rue_mohr: / Fc = 1128.75846164
10:42 PM rue_mohr: #define k 70.9829647473
10:43 PM rue_mohr: what school lacks is lots of practical application
10:43 PM rue_mohr: a course should start off with a project goal that encompases everything that the course covers
10:43 PM rue_mohr: and the course be an overexplained group build
10:44 PM rue_mohr: or multiple projects
10:44 PM aandrew: rue_mohr: that's pretty much the only way I can learn
10:44 PM aandrew: I am a terrible academic
10:45 PM rue_mohr: piles and piles of theory are just about pointless
10:46 PM rue_mohr: this filter stuff is crazy
10:46 PM rue_mohr: there are so many things to use it on
10:46 PM rue_mohr: I told myself filters were super important a long time ago
10:47 PM rue_mohr: things like -6db/decade messed me up for who-knows how long
10:48 PM rue_mohr: hmm do I change the fileserver motherboard or move the drives to another machine
10:48 PM rue_mohr: 4 drives
10:48 PM rue_mohr: 2 raid 1 for the os, 2 raid 1 for the data
10:50 PM rue_mohr: I should take and do a torture test on a board
10:51 PM rue_mohr: tho both the last boards have had lockups on lots of drive activity
10:51 PM rue_mohr: too many threads
10:51 PM rue_mohr: I wonder how I can focus me
11:48 PM aandrew: rue_mohr: I got a 16-bay 3.5" SATA thing with dual power supplies for like $50 last year. took forever to find a SAS controller I was happy with (HP P210 eventually) but it's so nice now... all the drives are in their own thing, it's got redundant power and even SAS if I wanted it
11:48 PM aandrew: no more trying to find a case that can store drives and then have to worry about powering things down to swap a drive
11:49 PM aandrew: I've only got 6 drives in it right now but it's futureproof
11:49 PM aandrew: at least until SATA drives go extinct
11:49 PM rue_mohr: heh
11:49 PM rue_mohr: I do not disagree
11:49 PM aandrew: I always buy the cheapest spinning rust I can find, I don't need speed, just redundancy and space
11:49 PM rue_mohr: but I have a hardware availability thing
11:50 PM aandrew: software raid, LVM and XFS. I know everyone loves zfs but it's just too weird for me
11:50 PM aandrew: oh this was just a lucky find
11:50 PM aandrew: I wasn't actually looking for it, the local computer recycler had it and didn't know what to do with it
11:50 PM rue_mohr: I can also stick a 1G card in a PC and get up to the 200-300Mbps it can do
11:51 PM aandrew: yep that's what I did - the blade server is stupid cheap old tech but waaaaaay more powerful than I need
11:51 PM rue_mohr: I'm trying to expand the path between the internet and the file servers to 1G
11:51 PM aandrew: but having "industrial grade" hardware is nice if you can avoid the price tag which I was able to do
11:51 PM aandrew: anyway - gotta go to bed
11:51 PM aandrew: it's late here
11:52 PM aandrew: not like you westerners and your -3h timezones :-)
11:52 PM rue_mohr: I have that poweredge T610 going, 2 procs (16 v cores) 24G ram
11:52 PM rue_mohr: nothing I have can use that much threading
11:53 PM rue_mohr: well, not counting firefox
11:54 PM rue_mohr: this machine is using 3.9 of its 7.8G of memory, with 209M in its 20G swap space
11:54 PM rue_mohr: huh 110 days uptime
11:55 PM rue_mohr: I need to rework all my tech
11:55 PM rue_mohr: :(