#robotics Logs

Feb 23 2019

11:24 AM pokmo: hi
11:29 AM pokmo: if i want something faster than a raspberry pi -- probably the speed of a 5yo Intel i3 setup -- but new parts, what options do i have?
11:29 AM pokmo: i could go with a refurb PC, but i'd prefer something new so that it lasts
11:29 AM polprog: there are rugged PCs available
11:31 AM pokmo: polprog i've been looking at the NanoPC-T4, but its geekbench scores are just about 2x as good as a rpi
11:31 AM polprog: whih rpi
11:31 AM pokmo: an i5-2400 (5yrs old?) gives about 4x that
11:31 AM polprog: rpi is a computer designed to be low power
11:31 AM pokmo: polprog rpi 3
11:31 AM pokmo: https://browser.geekbench.com/v4/cpu/search?utf8=✓&q=raspberry+pi+
11:32 AM polprog: you can always take a PC motherboard with a RYZEN cpu or an i9
11:32 AM pokmo: heh i wish i could afford that
11:32 AM pokmo: i've only got a $200 budget :(
11:32 AM polprog: look around for arm sbcs
11:33 AM pokmo: NanoPC T4 https://browser.geekbench.com/v4/cpu/search?utf8=✓&q=nanopc+t4
11:33 AM pokmo: around 2900 on multicore
11:33 AM pokmo: rpi 3 around 1200
11:33 AM pokmo: i5-2400 around 9000 https://browser.geekbench.com/v4/cpu/search?utf8=✓&q=i5-2400
11:34 AM pokmo: :\
11:34 AM polprog: you cant compare an i5 to those
11:35 AM polprog: if you need power, then get an i5
11:35 AM pokmo: oh, i know, but i'm just saying that SBCs (at least the NanoPC T4) is quite far behind a 5yr-old i5
11:35 AM pokmo: and i wonder if there's anything in the sub $200 range that performs like it
11:35 AM polprog: they arent designed to be fast. they are designed to use little power
11:36 AM polprog: why do you need so much power?
11:36 AM polprog: another option is a cluster. but the efficiency of that depends on your application, as clusters need to be programmed in a different way
11:37 AM pokmo: i'm trying to build a self-driving rover that patrols around our land. the tensorflow stuff is quite power hungry
11:37 AM pokmo: a cluster of SBCs?
11:41 AM pokmo: i've thought about that, but a NanoPC-T4 costs $110 each. to match an old i5, i'd need at least 4 and heaps of extra cooling
12:36 PM rue_bed: hi
12:36 PM polprog: hey
12:36 PM polprog: im having troubles with your stm32 makefile
12:36 PM rue_bed: ok
12:36 PM rue_bed: tell
12:36 PM polprog: im missing the linker script
12:36 PM rue_bed: aha, yes
12:36 PM rue_bed: they changed that
12:36 PM polprog: i cant find it
12:36 PM rue_bed: I tried to get a new station up
12:36 PM rue_bed: yea
12:37 PM rue_bed: ok
12:37 PM rue_bed: so, it was written in _____ but is now written in python
12:37 PM polprog: mhm
12:37 PM polprog: seen that
12:37 PM polprog: you got it working?
12:37 PM rue_bed: so, to set up the new station, I carried over my _____ version
12:37 PM polprog: ah
12:37 PM rue_bed: 'didn't try the python one
12:38 PM polprog: well, ill have to turn on my workshop comp then
12:38 PM rue_bed: BUT the reason they ditched the _____ one, was because of the error in it
12:38 PM polprog: in fact, if i do that, i might just use the setup there
12:38 PM rue_bed: I wanted to go back and try using the python one
12:38 PM polprog: id have to clean *that* desk as well
12:38 PM polprog: :/
12:38 PM rue_bed: thats the story, anyhow
12:38 PM polprog: shame, i thought you got the python one working
12:39 PM rue_bed: like I say, I carried the old script up
12:39 PM rue_bed: but only because I didn't notice the python one till I was done
12:39 PM polprog: can you upload it somewhere? for backup
12:39 PM rue_bed: give the python one a try
12:46 PM polprog: hmm, i think its working
12:46 PM polprog: ie i have an elf file that has actual code inside
12:47 PM polprog: seemingly at the right address
12:47 PM polprog: let me grab the board
12:55 PM polprog: woohoo
12:55 PM polprog: rue_bed: the python version is working
12:59 PM polprog: http://polprog.net/papiery/stm32/Makefile-polprog-new
02:30 PM rue_mohr: thankyou!
02:30 PM rue_mohr: like i say, it wasn't till it was too late I realized they had switched to python
02:31 PM rue_mohr: oh polprog I think I have a mod
02:31 PM rue_mohr: do you have an issue where that script sometimes wont flash, you have to try it twice?
02:31 PM polprog: no, although i havent used that makefile in a long time before today
02:32 PM rue_mohr: the make install *might* snag up every second run, I modified the openocd command
02:32 PM rue_mohr: to fix
02:36 PM polprog: whats the fix
02:37 PM Bloop is now known as Guest93791
04:10 PM Tom_itx is now known as Tom_L
04:19 PM Moistmelon is now known as Wetmelon
05:56 PM rue_mohr: I altered the openscad command I'll find it
09:06 PM rue_mohr: polprog,
09:06 PM zhanx: rue_mohr, last 4 parts are on the printer to flip the motors from top to bottom
09:06 PM rue_mohr: openocd -f $(OPENOCD_SCRIPTS_DIR)/interface/stlink-v2.cfg -f $(OPENOCD_SCRIPTS_DIR)/target/stm32f1x.cfg -c init -c "reset halt" -c "flash write_image erase ./$(BINARY).elf" -c "verify_image ./$(BINARY).elf" -c "reset run" -c shutdown
09:26 PM zhanx: wow rue_mohr must be busy to completely ignore that
10:48 PM rue_mohr: putting the z motors on the bottom is worse
10:49 PM rue_mohr: the shafts on the motors are sprung against the front of the motor, so pulling on them, they wont move, but if you push on them, the bearings slip and they push in about 1/16 of an inch
10:50 PM zhanx: pillow blocks to prevent the push in
10:50 PM rue_mohr: I dont know how they get away with it on the i3
10:51 PM rue_mohr: ok, I'm going to replace them with one motor on the top running a gang shaft that operates the Z via belt, cause threaded rod was a stupid choice
10:51 PM zhanx: k
10:52 PM zhanx: well i can do this you do that, see which one breaks first and fix from there
10:52 PM rue_mohr: its been on my list for about a year
10:52 PM zhanx: right now i am printing a tool at .1mm res
10:52 PM rue_mohr: I can recover 4 motors when I do it
10:54 PM Tom_L: rue_mohr, been doing any more with the sprayers?
10:54 PM rue_mohr: slowly, I think there were 6 orders so far this year, the other guy is doing it all up
10:55 PM rue_mohr: I been keeping the machiens running and modifying the design to suit production issues
10:55 PM rue_mohr: I cant shake this exhaustion
10:56 PM zhanx: rue_mohr, i noticed while it takes 10x longer to print in smaller layers they don't peel up
10:56 PM Tom_L: how many printers do you have working now?
10:57 PM zhanx: knowing rue 15
10:57 PM Tom_L: zhanx makes sense
10:57 PM Tom_L: thinner layers cool quicker and probably have less warpage stress
10:58 PM zhanx: Tom_L, you have to remember i am 2 years behind rue on the learning curve of it
11:02 PM zhanx: rue_mohr, other than academic purpose what do you think of me trying to print at 0.001/0.002 mm layer height
11:08 PM rue_mohr: the goal is 4 normal and one special
11:08 PM rue_mohr: no special one yet
11:08 PM rue_mohr: 4 printers is enough that by the time I design the 5th part, the first printer is finished
11:09 PM zhanx: rue do they all have separate pcs?
11:13 PM rue_mohr: ?
11:13 PM rue_mohr: whe I'm designing something complex with lots of peies
11:13 PM rue_mohr: 4 printers can keep up
11:14 PM rue_mohr: with me, thats why 4
11:14 PM rue_mohr: the 5th will be for finite special materials
11:15 PM rue_mohr: silicone, wax, chocolate, sugar, clay
11:15 PM rue_mohr: wax, silicone and sugar form a neat triangle
11:16 PM rue_mohr: when you look at the matrix of water solubility and heat behaviour