#robotics Logs

Apr 15 2014

00:00 GargantuaSauce now i don't bother because i can cram a megabit down the wire and it works fine
00:00 rue_more well, your crazy.
00:00 GargantuaSauce anyway, write() blocks
00:01 rue_more I have ti nonblocking
00:01 GargantuaSauce well do it my way if you want blocking
00:01 rue_more and linux will, just cache up what dosn't fit
00:01 GargantuaSauce yeah
00:01 rue_more in 2 seconds you can buffer up a half hour of serial data to a port
00:02 rue_more IN THE TIME IT TAKES TO THINK "OH GOD!" AND PRESS CTRL-C
00:02 GargantuaSauce yeah.
00:02 GargantuaSauce so don't write that much that fast
00:02 GargantuaSauce oooor use blocking calls.
00:02 GargantuaSauce oooooooooooor flush however you flush doing it your way
00:03 GargantuaSauce easiest thing would definitely be to just update at a fixed rate well below the max bitrate
00:03 rue_more hahah the code was backwards
00:03 GargantuaSauce like 50hz * 8 commands
00:03 rue_more http://osdir.com/ml/linux.serial/2005-05/msg00015.html
00:03 GargantuaSauce heh
00:04 rue_more sweeet
00:04 e_more watches a servo wash a sine wave back and f
00:04 GargantuaSauce now do 32!
00:05 GargantuaSauce https://www.youtube.com/watch?v=1HPVbuaONO4
00:05 rgantuaSauce phantom j
00:05 rue_more cant, without things stepping on the timing interrupts and possibly delaying them, I can only do 8
00:06 rue_more but I can put it on something smaller than a mega32
00:06 rue_more these are 8 channels with 24000 counts of resolution from .5 to 2ms
00:06 rue_more I just cant bare 8 bit for robot positioning
00:07 GargantuaSauce the servos' adc is probably only 10
00:07 GargantuaSauce MAYBE 12
00:07 GargantuaSauce but probably 10.
00:07 GargantuaSauce but yeah 8 sucks
00:08 rue_more its waaaay more
00:08 rue_more caue there is no adc
00:08 rue_more its analog
00:08 GargantuaSauce oh hurr
00:09 rue_more what you CAN argue is that 1/256th of error may not generate enough drive to cause the motor to actaully turn :)\
00:09 GargantuaSauce yeah that's where I is nice
00:10 rue_more but it does notice the ns of time difference
00:10 rue_more even if thats drowned out in noice
00:10 rue_more Vc s^
00:11 GargantuaSauce noice brah
00:11 rue_more I should get another servo and I can watch my cosine channel
00:12 rue_more ok, the 8 servo controller is working, the servo command spooler is working
00:12 rue_more next layer is the linear interpolator
00:13 GargantuaSauce lerp it good
00:13 rue_more it should just drop in easy
00:13 rue_more its super-generic
00:13 rue_more I wish I had a spline interpolator :(
00:14 rue_more hey, you might like this trick
00:15 rue_more will work even for your digital servos
00:15 rue_more run the servos power, from positive, thru the EB junction of a PNP transistor
00:16 rue_more shunt the EB with a resistor such that the idle power of the servo does not trigger the transistor
00:16 GargantuaSauce what why
00:16 GargantuaSauce oh for current limiting?
00:16 rue_more then put a votlage divider to ground such that the resulting pulses, caused by motor activation generate 0-5V pulses
00:17 rue_more now, measure the length of the current pulses to the mtoor, (which is what you get) and by the length of them, you know the stress on the servo
00:17 rue_more no current measurement required
00:17 GargantuaSauce ah ok
00:17 rue_more :) I cant wait to try it
00:17 rue_more I want to use that to do force control on the gripper
00:17 GargantuaSauce not super useful wihtout also knowing position
00:18 rue_more its just force feedback
00:18 GargantuaSauce yeah and you can't distinguish between external resistance and inertia
00:18 rue_more if you sweep the positions on the servo faster than it can move, but fast enough to catch short current pulses, you can work out where the servo acutally is
00:19 GargantuaSauce which involves changing the position to measure it!
00:19 rue_more but I dont know if that would time correctly, might not work
00:19 GargantuaSauce not a solution!
00:19 rue_more not if you pulse it fast enough
00:19 rue_more the new pulse on an analog servo will reset the drive cycle
00:19 rue_more an analog servos pot runs off a constant 2.5V reference
00:20 GargantuaSauce well i dunno about these. i am hesitant to try weird stuff with them because sending a super short pulse already made one fry itself
00:20 rue_more I'm gonna bet the digital one uses similar
00:20 GargantuaSauce also these servos definitely pulse at several KHz
00:20 GargantuaSauce and accept control pulses at hundreds of Hz
00:20 GargantuaSauce soooo i dunno what to conclude from that
00:20 rue_more yea, they might not like that
00:20 GargantuaSauce i am going to make real servo electronics instead of trying to squeeze gold out of these turds
00:21 rue_more the trick would be to have a pulse long enough to know if there is more than a minimal amount of current but short enough to not really move the servo
00:21 GargantuaSauce then i can use the current hardware with the new electronics, and eventually replace them with better gearmotors and encoders
00:21 rue_more you, of course dont get polatiry on that current sense trick
02:51 tex83 okay I have about $75 to spend on a bunch of students getting a bunch of really interesting sensors to use as robotics control devices. I'd like to have them arrive in Texas inside of two weeks. Robotics people, tell me which things to get (remembering that I need 6 of each), and where to get them.
02:52 tex83 I can probably raise that budget if something really awesome is happening, but I wanted to save that for strange chassis designs and stuff, exotic things to play with at the end.
03:25 joga $75 is honestly not much if you need a bunch and not a single sensor...
03:26 joga and may take more than two weeks to ship things from china
03:26 joga (at least to where I live, which is in europe but anyway)
03:27 joga but I guess at least basic ultrasonic sensors you can get quite a few for that money
03:30 joga well, here's 5 at for relatively cheap at least and from US http://www.ebay.com/itm/5-Pcs-HC-SR04-Ultrasonic-Module-Distance-Measuring-Transducer-Sensor-for-Arduino-/141253820401?pt=LH_DefaultDomain_0&hash=item20e3623ff1 ...
03:36 joga seems there are several similar batches though, I'd just search ebay for stuff located in the US, but I'm not sure how you define "really interesting" :)
03:38 joga these are perhaps more interesting but cost about six bucks a piece http://www.ebay.com/itm/MPU-6050-3-Axis-Accelerometer-Sensor-Gyroscope-6DOF-Module-3-3V-5V-For-Arduino-/161108492494?pt=LH_DefaultDomain_0&hash=item2582d080ce
03:40 joga basic temperature and light sensors you can get for almost no cost at least
03:45 joga depending on other parts you have, some cheap webcams may be nice, but that's assuming these robots have a computer to do CV on, specialized vision sensors are usually quite expensive
03:47 joga (students always want to make their robot 'see' anyway) :)
03:49 joga also some simple bumper switches may be nice to have
10:35 MrCurious one for GargantuaSauce http://hackaday.com/2014/04/15/automated-bed-leveling-for-3d-printers-is-now-solved/
13:27 GargantuaSauce MrCurious: pretty dumb solution compared to https://www.youtube.com/watch?v=HcnhCNh2Ln4
14:56 GuShH_ rue_more: got bit? http://gushh.net/tmp/dscn5299f.jpg heh scored a lot for relatively cheap
14:57 GuShH_ the guy had bigger ones, mostly blank... but I have no use for anything bigger, although it was a good price... if you want to buy them "brand new" they are expensive in big sizes
14:58 GuShH_ 3/4" and 1" dimension of about 6 inches long, not cheap.
14:58 GuShH_ then again, you need a massive lathe to use those
15:24 password2 mmmm
15:25 password2 is that tool die for lathes?
15:26 password2 damn , now i remember how badly i want a lathe of my own
15:37 GuShH_ password2: hss bits and some carbide
15:37 password2 *jealousy intensifies*
15:38 GuShH_ why, they are "used" don't buy that stuff brand new, costs too much
15:38 password2 i can't even afford lathe
15:38 password2 not even a mini one
15:38 GuShH_ look for used
15:38 password2 10" is all i want
15:39 GuShH_ that's what she said
15:39 password2 i tried
15:39 password2 hah
15:39 password2 here where i live is only big one
15:39 GuShH_ where is here
15:40 password2 i gound one second hand one listed im my entire country
15:40 password2 RSA
15:40 password2 *found
15:40 GuShH_ Hrmm
15:40 password2 and the shipping would have been twice that of a new one
15:40 GuShH_ :(
15:40 GuShH_ I've been looking for a milling machine for a loooooong while now
15:40 GuShH_ so it's not just you :p
15:41 password2 i believe here are a few
15:41 password2 but the boiler makers who use them have not discovered the internet
15:41 GuShH_ heh
15:41 password2 so its mostly word of mouth
15:41 GuShH_ see if any shops around know of anyone selling something
15:41 password2 you should see the add they post
15:42 password2 i tried
15:42 GuShH_ I need to find flea markets...
15:42 GuShH_ (where they sell used tools)
15:42 password2 the one oadd said "lathe for sal , neg"
15:42 password2 thats it
15:42 GuShH_ lol
15:42 password2 no number , no email
15:42 password2 no specs
15:42 password2 no size
15:42 password2 nad
15:42 password2 a
15:42 password2 and its common
15:43 GuShH_ "lathe for sale, if you find it it's yours!"
15:43 password2 or they just sa $40000
15:43 password2 lol
15:43 password2 also , wanted to ask
15:43 password2 is it possible to do wood on a metal lathe?
15:43 GuShH_ you can
15:44 password2 good
15:44 GuShH_ but you'll need a toolrest
15:44 password2 ok
15:44 password2 just wondered
15:45 GuShH_ there's also the fact that wood contains highly abrassive particles within, so you would have to cover the lathe bed and clean it up
15:45 password2 as some said it would screw the machine up , and i was sceptical
15:45 password2 ah
15:45 GuShH_ it's similar to surface grinding on a lathe
15:45 GuShH_ it produces a lot of abrassive dust
15:45 GuShH_ if you don't clean it up, it will eventually cause wear and damage.
15:45 password2 i geuss i will mostly be making shafts
15:46 password2 and winding coils
15:46 GuShH_ wood lathes are much cheaper
15:46 password2 yes , but much less precise
15:46 GuShH_ not precise at all :p
15:46 password2 doing wood would be a bonus
15:47 GuShH_ I have a rule not to ever put wood on mine, likewise not to grind on it
15:47 password2 somehow i have no one wound chain mail coild on a lathe
15:47 GuShH_ and not to let anyone use it.
15:47 password2 i would also want to add a strobe light
15:47 GuShH_ heh
15:47 password2 but i realize in dangerouse that is
15:48 GuShH_ for a wood lathe it's handy
15:48 GuShH_ no point for a metal lathe, I can't see any benefit
15:48 password2 true
15:49 password2 also if i do it i would use a switching color like red/green , so that i see a red stationary image , but the green image would remind me its spinning
15:52 password2 this is the y one i want http://www.adendorff.co.za/ProductDetails.aspx?id=466
16:57 GuShH_ heh actually that lathe is cheaper in south africa than it is here (567 USD in SA vs over 1k here)
16:57 GuShH_ but everything costs crazy amounts here... stupid country
17:00 GuShH_ password2: that lathe costs twice as much here :p
17:01 password2 really?
17:01 password2 wow
17:01 GuShH_ yeah
17:01 password2 ididn't realize it would be more expensive than here
17:02 GuShH_ everything is :(
17:02 GuShH_ but machining stuff is always expensive
17:02 password2 where are you then?
17:02 GuShH_ Argentina
17:02 password2 ah
17:03 ssword2 opens a
19:08 rue_more wow
19:08 rue_more 9.5m grade difference in my lot
19:20 Tom_itx good drainage
19:22 rue_more yup, drains into the house great
19:23 rue_more says from the fireshed the line is 2.5m over....
19:23 rue_more I'm sure its not that far
19:23 Tom_itx so umm maybe the house should be at the top of the grade?
19:24 Tom_itx or under it
19:27 rue_more yup, 2.5m
19:30 rue_more looking at the lot drawing for the shop, I didn't check to see if the house wasn't straight on the lot
19:30 rue_more I have them both drawn in straight
19:30 rue_more ok, interpolator
19:30 rue_more tea
19:30 rue_more then interpolator
19:33 rue_more :) #less interlin.c &
19:34 rue_more axisAdd(axies_t * this, int start, int end, void (*StepFor)(void), void (*StepBak)(void))
19:34 rue_more hmmmm
20:03 rue_more hmm I dont know that I want minimum step based interpolation for this
21:25 rue_more funny
21:26 GargantuaSauce https://www.youtube.com/watch?v=yn3FWb-vQQ4
21:29 rue_more the knock detection code I was playing with used an arduino function called map()
21:29 rue_more this lead me to work on making one of my own
21:30 rue_more in the process I realized that my interpolator used the same process to do the calculations
21:30 rue_more so I wrote it from that
21:30 rue_more now I'm using it to make a new interpolator
21:31 rue_more so it seems
21:56 rue_more ok, complete rewrite of interlin
22:15 rue_more hmmm, I wonder if I can pull off passing a paramiter to a callback function
22:16 rue_more is 256 axies on one interpolator enough?
22:16 rue_more what if I have a robot with 6000 axies?
22:17 rue_more oooh nevermind, I put a 65535 axis limit on it
22:17 rue_more the char is used for hte axis NAME...
22:17 rue_more 65536, 0 is a valid axis
22:18 rue_more ..but SHOULD it be?
22:18 rue_more and what if I have 65537 axies on one robot?
22:18 rue_more ya know, my 12' mecha only has 46 axies, maybe I wont worry about this today
22:19 GargantuaSauce 32 bits or bust
22:19 GargantuaSauce you never know when you'll want to deal with 4 billion degrees of freedom
22:20 rue_more damnit, you DO havea point
22:20 rue_more I dont need tea, I have wind!, not THAT kinda wind, I mean like a second wind, or a third. how do you know what wind your on?
22:22 GargantuaSauce i tally up the day's caffeine intake
22:23 rue_more I'm at 700, but there are no units!
22:24 rue_more I dont know if it was the tea of the sci-fi trailers & shorts
22:24 rue_more I can think! I have an hour and 45 mins till I terminate
22:25 rue_more I need a file format that stores a) multiaxis points, b) a sequence between the points
22:25 rue_more points can be reused
22:25 rue_more is this relivent enought o your stuff to put thought into?
22:26 GargantuaSauce uhh
22:26 GargantuaSauce gcode?
22:27 rue_more good idea, but no
22:27 rue_more too much implied path control
22:30 rue_more where did my function declerations go/
22:30 rue_more ?
22:30 rue_more I need a program that parses a C file and spits out the declerations forme
22:30 rue_more sounds like a job for perl
23:35 rue_more hmm
23:35 rue_more if I have a set distance to travel, and I have an increment step size
23:35 rue_more that dosn't divide evenly
23:42 armyofevilrobots @rue_more: Bresenham
23:42 armyofevilrobots bonus points, very simple math that can be driven during interrupts.
23:42 rue_more I suppose its not really a forward idea to use one stepsize and apply it to all axies
23:42 rue_more I have floating point at my control
23:43 rue_more #define RangeRemap(v,Il,Ih,Ol,Oh) (((((v)-(Il))*((Oh)-(Ol)))/((Ih)-(Il)))+(Ol))
23:44 armyofevilrobots New addition to 3d printer: video time lapse: https://www.youtube.com/watch?v=jazPNQH29j4
23:45 armyofevilrobots ALso, printing a crapton of little dudes for the coastmakers night.
23:47 rue_more hmm