#robotics Logs
Nov 22 2015
#robotics Calendar
00:13 rue_shop3 get power supply finished
00:13 rue_shop3 take over world
00:13 rue_shop3 maybe with some distractions in the middle
03:49 rue_shop3 ouch, literraly, my inductors will not take 5A continious
03:49 rue_shop3 hot, hot hot hot
04:21 rue_shop3 2am, ok, I'll just have a nap and then keep working on it
04:21 rue_shop3 I dont ahve a simple switcher that can take 5A
04:21 rue_shop3 so I'm back to my main circuit here
04:22 anonnumberanon You are trying to make a constant current output or not?
04:41 rue_bed 250Khz 35V 5A smps for a bench supply
04:42 rue_bed I'm almost tempted to make the output isolated
04:42 rue_bed but getting the voltage and current readings back would be a pain
04:43 rue_bed hmm, unless it was done by a microcontroller on the other side and sent as serial
04:43 rue_bed no that wouldn't work, the regulator is analog
04:55 anonnumberanon I don't understand what's wrong with the other one you made. You said it wasn't constant enough?
05:03 LiohAu how can I determine which kind of timing belt should be used ? 2GT, T2.5, H ....
05:11 SpeedEvil rue_bed: you could go fully digital
05:11 SpeedEvil rue_bed: Or you could totally decouple it and cheat
05:12 SpeedEvil rue_bed: canned tiny 5V/.1A supply for the logic, high-speed isolator to directly flobble the FET
05:15 LiohAu looking at this document http://us.misumi-ec.com/pdf/fa/2010/p1095.pdf I guess that I should go on (2|3|5)GT type
05:16 LiohAu or 8YU
05:16 LiohAu now I just need to know how to select the pitch..
05:18 SpeedEvil pitch is somewhat constrained by minimum diameter
05:20 LiohAu ok, so first step is to find the motor, so I know the reduction ratio needed, and then I'll know the number of teeth of the pulleys and so the diameter of the smaller one
05:41 LiohAu SpeedEvil: since my motor will be parallel to my ball screw and driven by a timing belt, isn't there a risk that the motor shaft break ?
05:43 LiohAu I mean the belt will apply a force perpendicular to the motor shaft, so I assume that it's going to create issues..
07:07 anonnumberanon Let's do this.
07:49 SpeedEvil LiohAu: you need to add a bearing to take the load(s), yes
07:50 SpeedEvil LiohAu: If the force is high. Look at the bearing size and look it up for max side-loads.
07:56 anonnumberanon SpeedEvil, do you program/dabble in tech and stuff?
07:57 anonnumberanon i need help to see if i can use this to run my robot's walking gait: https://www.youtube.com/watch?v=pgaEE27nsQw http://www.goatstream.com/research/papers/SA2013/SA2013.pdf
07:58 anonnumberanon I need servo positions.
07:58 anonnumberanon from 0 to 180
08:28 anonnumberanon those mudules are amazing im receiving signal to my robot from the remote controller behind a wall , going for more distance
08:49 SpeedEvil anonnumberanon: I am currently unconscious and cannot respond sensibly. http://i.imgur.com/OHxaMNF.jpg Here is a picture of an amusingly named animal instead.
09:07 Snert http://openservo.com
09:44 veverak hi folks
09:44 veverak what are usually speed limits of stepper drivers/motors ?
09:45 veverak let's say steps/s it can handle
09:45 deshipu on the highway or in city?
09:45 veverak (I know it depends on stress on motor/gearbox, but I want to know how big ration in gearbox I can have for it to work correctly)
09:45 veverak NEMA17 + reprap stepstick
09:46 veverak 2,5A motor? (afaik)
09:46 deshipu I know that once the outer rim gets over the speed of sound, you may have problems...
09:47 veverak well, that makes sense..
09:47 veverak :D
09:55 SpeedEvil deshipu: easy fix though
09:55 SpeedEvil Run it in a hydrogen atmosphere.
09:56 anonnumberanon Snert, interesting that one cannot buy those yet, if ever. Everything is open though so assembly should be a quick affair.
09:57 anonnumberanon Maybe the Chinese are using this right now even to make their digital servos.
10:15 Snert yea. the buy one link takes you to a website where ya can't buy one
10:15 Snert nice.
10:32 LiohAu SpeedEvil: about the belt width
10:33 LiohAu I can't find how do you determine the width to selet
10:33 LiohAu select*
10:33 LiohAu I have the choice between 3, 6, 9
10:33 LiohAu and I don't even know what's the difference :(
10:33 SpeedEvil LiohAu: belt stiffness and load limit will scale with width
10:34 LiohAu so it depends on the torque?
10:40 SpeedEvil yes.
10:40 SpeedEvil Well - no
10:40 SpeedEvil It depends on the tension
10:40 SpeedEvil A belt of 3 width will have the same power capability as a belt of 9 width with a three times diameter pulley
10:41 LiohAu ok, but how do I determine that 3, 6 or 9 is enough?
10:41 SpeedEvil find datasheets on the belts
10:41 SpeedEvil what is their load limit, and springiness
10:42 SpeedEvil Or buy some and test
10:44 LiohAu ok
10:45 LiohAu if I tell you that the ball screw will raise 10kg max, is the width choice obvious for you ?
10:48 LiohAu looks like the ball/lead screw tool here could help me : http://www.orientalmotor.com/support/motor-sizing.html
10:58 veverak don't forget reserve :)
12:12 anonnumberanon damn i think my wireless comes in only every 20ms
12:12 anonnumberanon oh stupid me. im not even sending it heh
12:13 anonnumberanon meh.. that changed nothing
12:25 rue_mohr SpeedEvil, I refuse to make a microcontroller part of the regulation loop, too much chance of failure
12:48 anonnumberanon hahAA!!
12:56 anonnumberanon man that serial comm takes a loooong time
12:58 anonnumberanon in reality the wireless commands come in every about 45 microseconds
12:58 anonnumberanon im quite blown away
12:58 rue_mohr what baud rate?
12:59 anonnumberanon no serial, but at 9600 baud and several serial statements, it would take about 17 milliseconds
12:59 anonnumberanon my next test was to higher the baud rate and see if that makes quicker serial
12:59 anonnumberanon does it?
13:00 rue_mohr character latency is 1.04ms
13:01 rue_mohr so I'm guessing your packaet is a horrid 16 characters long
13:01 anonnumberanon that sounds correct i think i do am sending about 15 chars, and it took 17ms
13:02 anonnumberanon i don't know I'm still using the arduino serial.print()
13:02 rue_mohr its position control packets?
13:02 anonnumberanon //Serial.print("X = ");
13:02 anonnumberanon // Serial.print(joystick[0]);
13:02 anonnumberanon //Serial.print(" Y = ");
13:02 anonnumberanon //Serial.println(joystick[1]);
13:02 rue_mohr your not sending position control characters...
13:02 rue_mohr OH GOD!
13:02 rue_mohr why ascii!?
13:03 rue_mohr and why spaces?
13:03 rue_mohr this is your own fault
13:03 rue_mohr what type is joystick0?
13:03 rue_mohr I suspect you should be able to do this in 5 characters/packet
13:03 anonnumberanon oh that's an int
13:03 rue_mohr yea 5
13:04 rue_mohr 10 bits of 16?
13:04 rue_mohr or?
13:04 rue_mohr 10 bits or 16 bits?
13:04 anonnumberanon will take me a bit to figure out
13:05 rue_mohr is it ADC readings?
13:05 anonnumberanon yeah
13:05 rue_mohr than its 10 bits
13:05 anonnumberanon from the remote controller
13:05 LiohAu not like if it was possible to write 10 bits anyway. I guess you'll be only able to write full bytes.
13:06 rue_mohr you can use the unused bits for sync, meaning that in 4 bytes you can send both positions
13:06 rue_mohr AAAAAABB.BBBBBBBB
13:06 rue_mohr A is the channel identifier, B is the value
13:07 rue_mohr you can use it to stream up to 64 channels
13:07 rue_mohr but I suggest sticking to 32, and make the first A always a 0
13:07 rue_mohr 0AAAAABB.BBBBBBBB
13:08 rue_mohr so that you can use it as a sync
13:08 rue_mohr or go down to 16 channels
13:08 rue_mohr 0AAAABBB.1BBBBBBB
13:08 rue_mohr posi-sync
13:08 rue_mohr from the first bit, you know if your the first or second byte, garunteed
13:09 anonnumberanon you'
13:09 rue_mohr 16 bits/channel, 1.08ms latency
13:09 anonnumberanon re saying channels of what bit size each?
13:09 rue_mohr thats 10 bit channels
13:09 rue_mohr see the B's are the channel value
13:09 rue_mohr the A are the channel number
13:14 anonnumberanon for this code I'd need to make a program for the computer end...
13:14 anonnumberanon roll me up a bit of getchar()
13:15 rue_mohr void recieveByte(unsigned char b) {
13:15 rue_mohr static char b0;
13:15 rue_mohr unsigned int value;
13:15 rue_mohr unsigned char channel;
13:15 rue_mohr
13:15 rue_mohr if(0) {
13:15 rue_mohr } else if ((b & b1000000) == 0) { b0 = b;
13:15 rue_mohr } else {
13:15 rue_mohr channel = (b0 >> 3);
13:15 rue_mohr value = b & b01111111;
13:15 rue_mohr value |= (( (unsigned int)b0 & b00000111 ) << 8);
13:15 rue_mohr dothing(channel, value);
13:15 rue_mohr }
13:15 rue_mohr
13:15 rue_mohr }
13:16 anonnumberanon when i looked at this stuff i found: http://www.cyberciti.biz/hardware/5-linux-unix-commands-for-connecting-to-the-serial-console/
13:16 anonnumberanon but that's programs, not my own program
13:16 anonnumberanon gotta think about this more
13:17 anonnumberanon i could write a terminal program but i don't know how you access the serial port from a C program
13:17 rue_mohr use ssls2
13:18 rue_mohr http://ruemohr.org/~ircjunk/programming/c/ssls2.c
13:18 rue_mohr http://ruemohr.org/~ircjunk/programming/c/ssls2.h
13:21 rue_mohr sendPos(unsigned char channel, unsigned int value) {
13:21 rue_mohr unsigned char b1, b2;
13:21 rue_mohr value &= 0x3F;
13:21 rue_mohr channel &= 0x0F;
13:21 rue_mohr b2 = (value & b01111111) | b10000000 ;
13:21 rue_mohr b1 = (value >> 8) & (channel << 3);
13:21 rue_mohr
13:21 rue_mohr send(b1);
13:21 rue_mohr send(b2);
13:21 rue_mohr }
13:21 rue_mohr there, I did all the work for ya :)
13:21 anonnumberanon you take all the fun out of it!
13:22 anonnumberanon just kidding
13:22 rue_mohr I have examples of using ssls2 if you need
13:22 rue_mohr there are enough syntax corrections in there to keep you busy for a half hour or so
13:23 e_mohr does a lot of bit-di
13:23 anonnumberanon man i actually lost my train of thought with all this
13:23 rue_mohr you were complaining that all the useless packing in your data stream causes too much latency
13:24 anonnumberanon i was thinking negatively and thought wireless transmission would take too long and that i was gonna have to rewrite the wireless library but actually it's blazing fast
13:24 rue_mohr you managed to turn a 2ms packet into a 17ms one
13:24 anonnumberanon and now I'm wayyy ahead of schedule
13:25 onnumberanon looks at his list of things to do for the
13:25 rue_mohr - write linux Super Simple Linux Serial library
13:25 rue_mohr - write binary packet encoder
13:26 rue_mohr - write binary packet decoder
13:26 rue_mohr - done
13:26 rue_mohr :)
13:26 anonnumberanon heh that sure is on the list but
13:26 maiden i hope the serial transmission/reception is thread based or interrupt handled ;)
13:27 rue_mohr usually interrupt
13:27 rue_mohr I didn't spool the output, but thats usually not an issue
13:27 rue_mohr and it can be done easy enough
13:27 maiden got a spi implementation in one of our products with "while(!gotResponse);"
13:27 rue_mohr most of the time I'd return the packet bytes, but its good to save ram on avrs
13:28 rue_mohr eeeek
13:28 maiden i am going to hunt down the author and.....
13:28 rue_mohr that could take an undefined amount of time
13:28 maiden found a "sizeof(void)"
13:28 maiden in a if statement!!!
13:28 maiden grr..
13:29 maiden if ( a > b || sizeof(void))
13:29 rue_mohr for(timeout = 255; !gotresponse & timeout != 0; timeout--, sleep());
13:29 maiden some people is better then others....
13:29 rue_mohr what the hell, sizeof(void)!?
13:29 maiden indeed
13:29 maiden and -Wall did not pick it up.
13:29 rue_mohr what the hell was the intent!?
13:29 anonnumberanon ah yeah, i think next step was comparing between using 10 digital ports (for servo positions) and using the decade counter instead
13:29 maiden found it when i enabled --pedantic
13:30 rue_mohr anonnumberanon, io count
13:30 anonnumberanon like 13 or something
13:30 maiden rue_mohr, probably lazyness, went from sizeof(x) -> sizeof(void) instead of removing the code...
13:30 rue_mohr a > b || 2
13:31 rue_mohr I'm not sure how the order of ops happens on that
13:31 maiden neither am i. but i guess it returns zero?
13:32 maiden the funny part is that it was in production code.. the product is in the hands of about 7000 units.
13:32 maiden omg, that sentence..
13:33 maiden s/units/people/g
13:35 maiden i cant decide if i am going to spend some time on learning KiCAD or creating a biped model for simulation with ROS/Gazebo
13:35 rue_mohr http://paste.debian.net/335006/
13:35 rue_mohr yup, always 1
13:36 rue_mohr I hope its not the safety stop.
13:36 anonnumberanon maiden, I'll give you some useful specs for that model...
13:38 maiden i am currently playing around with building a replica of this: http://www.jsk.t.u-tokyo.ac.jp/~report/proceeding02/icra02/nishiwaki_forcesensor.pdf
13:38 maiden "A Six-axis Force Sensor with Parallel Support Mechanism to Measure the Ground Reaction Force of Humanoid Robot"
13:39 rue_mohr http://paste.debian.net/335007/ <-- this is MY favorite abuse of code :)
13:39 anonnumberanon ah, so a human-sized robot..
13:40 maiden making my life hard ;)
13:40 anonnumberanon I was thinking a small one, 16".
13:41 maiden rue_mohr: this is mine: http://www.ioccc.org/2000/anderson.c
13:41 rue_mohr hahaha
13:42 maiden http://www.ioccc.org/2000/anderson.hint
13:42 rue_mohr did you write that?
13:42 maiden hell no, i am not that insane...
13:43 maiden "The program is obfuscated without the use of the preprocessor or any
13:43 maiden special compilation parameters. In fact, the program doesn't even
13:43 maiden bother using any keywords, save the single 'char' at the beginning.
13:43 maiden This program was crafted using only operator precedence rules,
13:43 maiden short-circuit evaluation and recursion."
13:43 rue_mohr it uses putchar
13:44 rue_mohr ok I'm gonna wake up
13:45 maiden talking about insane code, i have a friend who wrote a RPG in Befunge, where the sourcecode was the map, and by playing the game you changed the code :p https://en.wikipedia.org/wiki/Befunge
13:46 LiohAu maiden: isn't that if() generated code?
13:46 maiden "The language was originally created by Chris Pressey[3] in 1993 as an attempt to devise a language which is as hard to compile as possible — note that the p command allows for self-modifying code."
13:47 LiohAu when you generate conditions inside a if adding an "|| false" you don't have to check if you have conditions to put in your if..
13:48 maiden LiohAu, sadly not, its just plain bad code.
13:48 LiohAu ok :d
13:50 maiden makes me think about when i was looking on writing a linux driver for a graphics accelerator i built, found this: https://github.com/torvalds/linux/blob/v4.1/include/linux/fb.h#L534
13:51 maiden i know that feeling that generates those fantastic defines ;)
13:53 anonnumberanon "Again, if it's OK with K&R, it's OK with me."
13:53 anonnumberanon Heh, hopefully it's not UB. UB is not okay in K&R.
13:53 maiden aaw, you hate syntax then :p
13:53 maiden not syntax, i was thinking of code formatting ;)
13:54 anonnumberanon Well uh, yeah as a matter of fact i do hate syntax but i don't know why you'd think I do.
13:54 anonnumberanon I don't know too much about code formatting except I like indentation with 1 tab
13:55 maiden anonnumberanon, i mean, i think this is OK with K&R: <anonnumberanon>
13:55 rue_mohr never use tab characters
13:55 maiden http://www.ioccc.org/2000/anderson.c
13:55 maiden doh..
13:55 rue_mohr they are different on every platform and it never works out
13:56 maiden rue_mohr, python with \r and \n mixed and with mixed tabs and spaces edited with notepad ;)
13:57 rue_mohr tabs can either represent a number of spaces OR advancement to the next tab-stop, which is why the idea dosn't work
13:57 rue_mohr everyone implements tab differently
13:57 rue_mohr a space is always implemented in the same way
13:58 anonnumberanon wait why shouldn't you use tabs?
13:58 anonnumberanon i only care about C
13:59 rue_mohr becasue every editor implements them differently
14:00 rue_mohr unless you wrote your own editor, you better start reading its source when it goes obsolete
14:00 rue_shop3 Im watching you
14:00 rue_mohr and you find... hey dude!
14:05 maiden anonnumberanon, what humanoid specs are you referring to?
14:05 anonnumberanon weight of servos, sizes and weights of frame parts
14:06 maiden ok, i am building a ultra light robot about 180cm high ^_^
14:07 deshipu how heavy is "ultra-light"?
14:07 maiden lighter then this: https://www.unirel.vt.edu/CHARLI.html
14:07 deshipu and that's how heavy?
14:07 ace4016 ~6' tall robot that's considered an ultralight...
14:07 ace4016 must be made mostly of air :P
14:08 maiden that one is 4.7 feet high and weighs 27 pounds
14:08 maiden but that is measured in bodyparts and not in a international standard, so i do not understand it :p
14:09 deshipu ace4016: like Baymax? ;)
14:10 maiden 12.3kg and 1.43m high.
14:10 ace4016 deshipu, yea :P
14:10 deshipu what's with people making huge robots? :/
14:11 ace4016 presence
14:11 maiden deshipu, we are masochists ;)
14:11 deshipu https://hackaday.io/page/1252-small-robots
14:11 ace4016 also, small robots mainly seem to be toys; or so people seem to think?
14:11 deshipu ace4016: fuck people
14:12 deshipu small robots will be the next buzzwords, you will see
14:12 deshipu instead of talking about "the cloud
14:12 anonnumberanon I am building a Robo-One fighter.
14:12 deshipu " everyone will talk about "the swarm"
14:12 deshipu anonnumberanon: awesome, pics?
14:13 maiden I will not use any servos ;) will use stepper motors and or brushless motors with hall effect sensors and angle measurements ;)
14:13 ace4016 so...a servo
14:14 maiden but cheaper and with more custom options ;)
14:14 ace4016 just not an R/C servo
14:14 maiden or those crazy expensive robotic servos
14:15 ace4016 i think people forget servos are more general than the canned options :P
14:15 deshipu stepper + gearbox + driver + encoders > servo
14:15 maiden have you seen this: http://letsmakerobots.com/robot/project/dc-motor-biped-12-dof-legz-3d-printed
14:15 deshipu maiden: are you sure it will be cheaper?
14:16 SpeedEvil rue_bed: I diddn't say that.
14:17 SpeedEvil rue_bed: Digital control simply means to turn the FET on and off.
14:17 SpeedEvil Instead of analog feedback signals for voltage and current across the barrier, you instead do simply FET on/off
14:18 maiden deshipu: you know any servos that are in the 2.5KW range? like this: http://www.hobbyking.com/hobbyking/store/__18180__Turnigy_Aerodrive_SK3_6364_213kv_Brushless_Outrunner_Motor.html
14:19 deshipu good luck with a gearbox for that
14:19 maiden deshipu: we used one of those in this: http://www.idt.mdh.se/rc/page_id1/dasher
14:19 SpeedEvil Do note that you will need massive amounts of forced air to keep that cool at speed
14:19 deshipu also, that's not a stepper, that's a brushless motor
14:19 SpeedEvil At high torque, ratehr
14:20 maiden see the back of it in mid waist in the picture
14:20 deshipu maiden: the website says it's hydraulic
14:20 anonnumberanon deshipu, yeah here, this is when my desk was still pretty clean, how I almost cannot move my mouse anymore from parts and wires http://img.ctrlv.in/img/15/10/20/562601d43035f.png
14:21 anonnumberanon *now
14:21 maiden it is :D and guess what drives the hydraulics
14:21 deshipu anonnumberanon: double servos for the knees?
14:21 deshipu anonnumberanon: I always wondered why people do that -- is that for greater speed?
14:22 deshipu maiden: well, that's not like a servo
14:22 maiden the motor is for the arms
14:22 SpeedEvil anonnumberanon: there is an easy solution for that. http://www.ebay.co.uk/itm/USB-Mouse-3D-mouse-Controller-CAD-Mouse-3Dconnexion-Space-Navigator-used-TOP-/251984278068?hash=item3aab6ef234:g:e7wAAOSwqv9V6EzF
14:22 maiden *there are motors for the arms as well :p
14:23 deshipu maiden: so what are the gearboxes made of?
14:23 deshipu maiden: to handle 2.5KW
14:23 anonnumberanon deshipu, i don't think it even has knees, it has weird legs
14:23 maiden that one was acctually 2 KW
14:24 maiden as for the gearbox, i can ask if i get hold of the one who was responsible for the mechanics
14:25 maiden my contribution to this project was only some minor FPGA coding for the sterio vision..
14:26 maiden the mechanics was acctualy built the year before, they first thought they could get it to run 100m in a straight line without any vision system :p
14:27 deshipu with a compass?
14:27 anonnumberanon deshipu, i do know the answer for the double knee question though, you see them bouncing up and down to rotate , the bouncing is done with the knees
14:27 maiden gyro, magnetometer and accelerometer
14:28 maiden but the thing about running is that you spend some time in the air and reality is reality :p
14:28 deshipu anonnumberanon: how are you going to control all those servos?
14:28 deshipu I always hated reality
14:28 deshipu but it's the only place where you can get a decent steak
14:29 deshipu anonnumberanon: also, my little eye spies a cool keyboard in there...
14:29 deshipu anonnumberanon: that's a 60%?
14:29 anonnumberanon deshipu, ;)
14:32 anonnumberanon deshipu, to control the servos I'm planning on recording a walking gait by hand for the beginning
14:32 deshipu anonnumberanon: but what electronics?
14:32 anonnumberanon deshipu, atmega328p
14:32 deshipu a classic
14:32 deshipu but can it handle that many
14:33 anonnumberanon yeah
14:33 deshipu the default arduino servo library only can handle 12
14:33 anonnumberanon the question is rather, the memory size
14:33 anonnumberanon im not using servo lib
14:34 deshipu well, for hardwired stuff you can always put things into progmem
14:34 anonnumberanon im toggling pins on and off instead, much better
14:35 maiden anonnumberanon: regarding walking, take a look at this: http://www.mate.tue.nl/mate/pdfs/10796.pdf
14:35 deshipu anonnumberanon: https://github.com/thebiguno/stubby/blob/master/avr/lib/pwm/pwm.h
14:36 anonnumberanon atmega328p has limited pins so gonna use decade counters and take up only like 3 pins
14:37 anonnumberanon 2 decade counters, if i had hands and other things, three decade counters and 4 pins used (1 for clock, the rest for each decade counter)
14:37 anonnumberanon and thanks for that lib but ill write my own
14:37 deshipu anonnumberanon: I ended up using one atmega as the servo controller, and the second one to just send servo positions to it over i2c
14:38 anonnumberanon for your 4 legged?
14:38 deshipu yeah, actually, now I'm replacing the second one with rpi
14:38 deshipu just because it's so popular
14:38 anonnumberanon so far i have one atmega for the wireless controller, one for the servo controller
14:39 deshipu pro minis are so cheap, you could even have one per servo ;)
14:39 anonnumberanon funny you mention that i just bought 5
14:39 anonnumberanon was $17 in USA
14:40 deshipu $1.7 in China
14:40 deshipu with free shipping
14:40 deshipu worldwide
14:40 anonnumberanon yeah ordering from China takes too long
14:40 anonnumberanon i waited forever for the servos
14:40 deshipu nah, you just have to plan ahead and keep a stock
14:41 maiden pro minis are cheap, but maple mini is not that much more and then you get a full 32bit cortex m3 :D http://www.ebay.com/itm/1pc-NEW-STM32-ARM-Cortex-M3-Leaflabs-Leaf-Maple-Mini-Module-for-Arduino-/141404280807?hash=item20ec5a17e7:g:4PEAAOSwxH1UErbR
14:41 SpeedEvil deshipu: I ordered 20 of everything on aliexpress.
14:41 SpeedEvil deshipu: works well.
14:42 deshipu maiden: I actually prefer the other stm32 board
14:42 maiden what board? discovery?
14:42 anonnumberanon i got 20 servos so I can break 3 at least, not counting 2 super strong servos that cost a lot
14:42 SpeedEvil http://www.fleasystems.com/forums/showthread.php?tid=60 - or you can go FPGAduino
14:42 deshipu maiden: http://www.ebay.com/itm/STM32F103C8T6-ARM-STM32-Minimum-System-Development-Board-Module-For-Arduino-DHUS-/321569700934
14:42 deshipu maiden: but since I'm lame, I program it with stm32duino
14:42 maiden oh, sweeet!
14:43 anonnumberanon those very strong servos will never break, have 30 kilogram.centimeters of torque
14:43 deshipu maiden: it doesn't have the maple's bootloader, but you can program it over serial
14:43 deshipu anonnumberanon: everything breaks, don't use glue :P
14:44 maiden deshipu: but now days its http://www.ebay.com/itm/1PCS-ESP8266-ESP-03-Serial-WIFI-Module-Wireless-Transceiver-Send-Receive/171556622269?_trksid=p2047675.c100005.m1851&_trkparms=aid%3D222007%26algo%3DSIC.MBE%26ao%3D1%26asc%3D20131003132420%26meid%3Daa5327cb01724c189b264810ff8a742a%26pid%3D100005%26rk%3D1%26rkt%3D6%26sd%3D181647832581 that the boss :D
14:44 anonnumberanon deshipu, i may or may not try and create metal cases for the servos i already have, that's what breaks on those servos
14:45 deshipu maiden: yeah, I'm just fighting with one right now
14:45 akem deshipu you use arduino sdk? if yes C only or you also got python through it one way or another?
14:45 deshipu akem: wait, can you rephrase?
14:45 maiden deshipu: https://github.com/espressif/ESP8266_RTOS_SDK ?
14:46 deshipu maiden: nah, arduino prostethics :P
14:46 deshipu akem: I use arduino, nodemcu and micropython on the esp8266
14:46 deshipu akem: in different projects
14:46 maiden i see, sweet :D
14:47 deshipu right now I'm working on this: https://hackaday.io/project/8521-katka
14:47 maiden prototyping goes so much faster with scripting then using c :p
14:47 deshipu but for some reason the Arduino's Servo library for ESP8266 generates wrong signal for my servos
14:47 deshipu and I don't have a logic analyzer to see what is wrong
14:47 anonnumberanon Because I'll be fighting this robot, i expect a lot of mechanical failure all the time, and I'm kinda looking forward to it, just for the sake of reengineering them, end goal being having a method to have cheap strong servos so that a lot more people can build these fighters.
14:47 deshipu maiden: it does, especially when you have an interactive console to try things
14:48 maiden deshipu: aha going the swarm way? http://www.swarmrobot.org/
14:48 maiden deshipu: thats just awesome!
14:48 deshipu well, if you can call a single robot "swarm"...
14:49 deshipu I want a small robot, because they tend to be harder to break
14:49 deshipu and you can take them with you everywhere
14:49 maiden it have to start somewhere ;)
14:49 SpeedEvil deshipu: square cube is magic
14:49 akem deshipu ok nice, stripped down version of python i guess.
14:49 deshipu akem: micropython
14:50 deshipu akem: only stripped of its standard library
14:50 deshipu akem: but the esp8266 version is not ready for serious use
14:50 deshipu missing to much
14:50 deshipu and nobody is really working on it
14:52 akem deshipu so you upload the interpreter when you got to update or rebuild it and the scripts to the chip something like that?
14:52 maiden well i got plenty of boards that are ~5x3cm connected to 600mAh batteries with a 50 Mhz cortex M4, GPS and GSM with prepaid data ^_^
14:52 maiden so i have plenty of "dev boards" to play with
14:52 anonnumberanon isn't there a computer program simulation to do what that paper says deshipu ?
14:52 maiden oh, they got a magnetometer, accelerometer and gyro as well :D
14:53 deshipu anonnumberanon: which paper?
14:53 anonnumberanon you put your shapes, weights and DOFs and can start playing with walking gaits, that PDF you linked me for walking
14:53 deshipu maiden: sounds sweet
14:53 deshipu anonnumberanon: that was maiden, I think
14:53 anonnumberanon oh
14:53 maiden anonnumberanon, there is some i think, i have just not found them yet :p
14:54 maiden have not found any in ROS, but there ought to be one or two..
14:54 deshipu zmp is very popular
14:54 maiden look at open robot projects, Atlas? Pepper?
14:54 maiden indeed, so there ought to be some good "generic" implementation
14:54 deshipu but doesn't that require pressure sensors in the feet?
14:55 maiden nope
14:55 maiden not really
14:55 maiden you could work around it with gyro/magacc data
14:55 deshipu wouldn't that be too slow?
14:55 deshipu I mean, accelerometers are noisy
14:55 deshipu by the time you have enough data to filter it, it's too late?
14:56 anonnumberanon nah they aren't very noisy
14:56 anonnumberanon you use a complimentary filter
14:56 anonnumberanon getting data and computing the PID corrections takes about 1.5 milliseconds
14:57 anonnumberanon (on the 328)
14:59 maiden anonnumberanon: see: https://www.youtube.com/watch?v=nsul_eZwTm0
15:00 anonnumberanon the challenge for a robot (as I see it, now before I even have any kind of walking implemented) is to chose when to make corrections, when to incorporate it into the walking/running gait
15:00 deshipu I have an accelerometer on one of my quadruped robots
15:00 deshipu but never managed to use it for anything
15:01 maiden deshipu: a random number generator comes to mind ;)
15:02 deshipu so, I tried to make the robot stop and curl up its legs when you pick it up
15:02 deshipu and to stretch it legs when you are putting it down
15:02 deshipu and then resume walking
15:03 anonnumberanon maiden, that's not very good
15:03 maiden and if you get bored: https://www.youtube.com/watch?v=4ETpaqN5pqE
15:03 anonnumberanon (minute 1:10) shows utter failure of that guy's implementation
15:04 anonnumberanon and Dr Guero 's solution is state of the art
15:04 maiden do he release any papers or code?
15:04 deshipu don't get me started on Dr Guero
15:05 deshipu but that's neural networks, right?
15:05 anonnumberanon who knows, the way it responds it feels the same as a basic quadcopter's implementation
15:05 maiden or simple fuzzy logic or q-learning :p
15:05 anonnumberanon good luck asking him in Japanese
15:05 maiden google translate :D
15:06 maiden i can haz your helps please
15:06 deshipu anonnumberanon: well, there are some videos from tunning it
15:06 anonnumberanon he has probably been contacted by Kondo already
15:06 anonnumberanon deshipu, tuning? not on the same channel?
15:07 deshipu https://www.youtube.com/watch?v=xUDIYxYhd2o
15:07 deshipu https://www.youtube.com/watch?v=OzsnUPLY-nI
15:07 deshipu I assume that's related
15:07 deshipu of course that's a large assumption
15:07 anonnumberanon i forgot about that one
15:08 anonnumberanon and: Oh god, what a freaking genius.
15:08 deshipu I think he just has too much time on his hands ;)
15:09 anonnumberanon "It is an old book, but [PARALLEL DISTRIBUTED PROSESSING David E.Rnmelhart] is the book you want to read."
15:10 anonnumberanon he said that in the comments
15:10 maiden that reminds me of this: https://www.youtube.com/watch?v=u9UpCgkxh_o
15:11 maiden if you really want to play around with neural networks and how the brain works, i strongly recommend: http://nengo.ca/
15:13 anonnumberanon what's neural networks about?
15:13 anonnumberanon some field of AI?
15:15 maiden i wrote a paper that introduces neural networks :p
15:16 maiden here you go: http://www.idt.mdh.se/kurser/ct3340/ht10/FinalPapers/4-Lenander_Fosselius.pdf
15:16 maiden read from introduction ;)
15:17 anonnumberanon nice, thanks
15:17 maiden oh wait, i think this is better: https://www.youtube.com/watch?v=bxe2T-V8XRs
15:17 maiden depends on who you are ;)
15:18 anonnumberanon yeah im that guy
15:18 maiden me2, but never leave a chance for some bragging ;)
15:19 anonnumberanon my big struggle nowadays is: "what do I do myself, what do I rip"
15:20 maiden thats easy, "FUN, BORING"
15:20 anonnumberanon as a compromise I think do everything yourself once for the learning and then the rest is up to how fast you want to finish a goal
15:21 anonnumberanon bbl
15:21 maiden i rip it off, then learn it in order to get the shit to work, then when it works i realise the implementation sucks, then i rewrite the entire thing...
18:06 Anniepoo 8cD noodling about with an arduino and servos just for kicks
18:06 niepoo noodle noodle no
18:37 rue_mohr what you got happening so far?
18:39 maiden oho, found a "better" way to simulate :D
18:39 maiden panda3d with bullet physics, its quite straight forward :D
18:41 maiden are building a simple biped now, then i can set all the lenghts, widths etc. etc as parameters ^_^ then hook it up to an api for controlling all the joints and we have a simple generic simulator for bipeds :D
19:13 robotustra night
21:12 rue_mohr so, 2 surprising uses of 555's that I'v found are: overload controller, and level shifting fet driver
21:13 orlock I've never used one
21:13 rue_mohr the fet driver was funny
21:13 drebo what makes the 555 so versatile?
21:13 orlock but i just got one last week!
21:13 rue_mohr I needed to take the ttl signals, boost them to 12V and drive a mosfet
21:13 orlock drebo: becauses pulses are handy
21:14 rue_mohr if you tie 2 and 6 togethor, make it your input, and put a 1k pulldown on pin 5 (voltage adjust), then it'll take in the ttl signals, and the driver is quite happy driving a mosfet base
21:14 orlock rue_mohr: i ordered a box of random odds and sodds from the local geep-bits-vendor
21:15 rue_mohr I tried a totem pole driver after the 555, but it didn't help much
21:15 orlock rue_mohr: ended up wiht an 8x8 LED array, a maple mini ARM board, 100 assorted LED's, 1280 arduino mega, 4x arduino pro mini's
21:15 rue_mohr not bad
21:15 orlock rue_mohr: 128x160 SPI lcd displau with sdcard slot
21:16 mrdata 555 is verstile because it's easy to build and offers .. what like 200mA current or something on its output
21:16 orlock easydriver stepper motor controller, and a 555 timer board - it has a pair of pot's and some other components on it
21:16 rue_mohr and drive 1A mosfet gates just fine
21:16 mrdata and takes 3 to 15 V power iirc
21:16 mrdata i got one running at 2.5V last year\
21:16 orlock And one of these: http://www.freetronics.com.au/products/170mm-electronics-ruler
21:16 rue_mohr I hope you didn't design an electric fence with it
21:17 mrdata lol no
21:17 mrdata it was a servo controller
21:17 mrdata for a hobby servo
21:17 rue_mohr ah, hard to get the timing right
21:17 mrdata nah
21:18 mrdata works great
21:18 mrdata i did find, tho in the process that a bunch of capacitors i have are all 20% low
21:19 mrdata so ended up with three in parallel to get things all right
21:20 mrdata (decreasing size)
21:21 orlock started rewiring my nightlight
21:21 robotustra night
21:21 mrdata for more night or more light?
21:21 orlock less wires bulk
21:22 robotustra evening
21:22 orlock 19 LED's, i had them wired up to a PCB via .1" plugs
21:22 orlock so 5 each 0f 8 pin plugs on a PCB - ended up quite bulky
21:23 orlock so i trimmed the wires and i've spliced them together into groups of three, so 6 groups of 3 for +ve and 6 groups of 3 for -ve
21:23 orlock i'll splive them together again so theres no PCB in it
21:23 orlock mrdata: http://en.crypt.net.au/TStar/ https://www.youtube.com/watch?v=Slq_ag38eKc
21:26 orlock ahh my thumbnails are all broken
21:29 mrdata yeah, but the large images are too large
21:29 data reads them sl
21:30 mrdata printed with PLA?
21:32 orlock yeah
21:32 robotustra any new robots been built?
21:32 orlock not by me, waiting till this is done
21:33 robotustra this means what?
21:33 robotustra star?
21:33 orlock yeah
21:34 orlock just needs PCB soldered up, wiring redone, and a stand cad'd up and printed
21:34 robotustra what is the size of it?
21:34 orlock about 24cm tall across the largest points
21:34 robotustra good, what printer?
21:35 orlock prusa i3
21:35 orlock tool up whole build plate
21:35 orlock took up
21:35 robotustra nice
21:36 robotustra preparing to xmass?
22:13 orlock robotustra: nah, daughter is a MLP fan
22:43 orlock http://hackaday.com/2015/11/21/fleafpga-arduino-uno-fleafpgauno/
22:43 orlock Good to see my old DSO getting some use!