#robotics Logs
Apr 24 2017
#robotics Calendar
12:02 AM z64555: I really need to dig out some components and build these circuits to practice on. Theory don't hold too well in memory as well as hand-on experiance
12:53 AM Guest58934 is now known as CrazEd
12:53 AM CrazEd is now known as Guest68493
01:54 AM Guest68493 is now known as CrazEd
01:54 AM CrazEd is now known as Guest25339
02:11 AM pokmo: hi
02:12 AM pokmo: i'mr trying to mount a micro servo perpendicular to the housing. does anyone know if there are L brackets or something for mounting at right angle?
02:19 AM gottaname|wurk: pokmo, I just print the bracket
02:19 AM gottaname|wurk: 3d*
02:19 AM gottaname|wurk: 3d printers are useful things
02:20 AM pokmo: gottaname|wurk, yeah, but i don't have one :(
02:20 AM gottaname|wurk: pokmo, then you have to improvise then
02:21 AM gottaname|wurk: brass sheet, bend it then drill holes
02:21 AM gottaname|wurk: go search your nearest large DIY hardware store
02:25 AM pokmo: hmm
02:55 AM Guest25339 is now known as CrazEd
02:55 AM CrazEd is now known as Guest41738
03:07 AM mrdata_: https://www.youtube.com/watch?v=5Mf0JpTI_gg
03:56 AM Guest41738 is now known as CrazEd
03:56 AM CrazEd is now known as Guest49225
04:39 AM rah_: I want a hackable home vacuum robot
04:39 AM rah_: I'm looking at the Neato XV series as it runs Linux and has had some work done with it
04:40 AM rah_: however, they seem hard to come by except the newer XV Signature but I'm not sure whether this runs (the GPL-covered and hackable) Linux like the XV-11/XV-25/etc. or if it runs RTOS like their newer Botvac series
04:41 AM rah_: can anyone shed any light on this?
04:41 AM rah_ is now known as rah
05:32 AM rah: s/RTOS/QNX/
07:48 AM veverak: deshipu: found cheap brushless motors
07:48 AM veverak: 10$ for set of motor/eletronics to have brushless motor with encoder and driver
07:49 AM SpeedEvil: ?
07:49 AM * veverak really thiks about building one leg for robot from these
07:49 AM veverak: cheap leg with full position feedback? I kind of like that
07:49 AM veverak: :)
07:49 AM robotustra: veverak: now extract rare earth magnets from it and sell it for $20 and buy 2 motors
07:50 AM SpeedEvil: where from?
07:50 AM veverak: P.S: somehow sounds more interesting than adding feedback to sg90
07:50 AM SpeedEvil: Or are there only a few
07:50 AM veverak: SpeedEvil: aliexpress 5PCS mitsumi three-phase outer rotor micro
07:50 AM veverak: (aliexpress has horrible urls, so not pasting it)
07:50 AM veverak: http://www.dhgate.com/product/brand-new-28mm-mitsumi-dc-3-phase-outside/388356100.html here is charasteristics of the motor
07:50 AM SpeedEvil: just cut after the first html
07:51 AM SpeedEvil: then it's short
07:51 AM veverak: https://www.aliexpress.com/item/5PCS-Japan-MITSUMI-With-Nine-pole-coil-Three-phase-Outer-Rotor-Micro-Brushless-Dc-Motor-For/32600538825.html
07:52 AM SpeedEvil: That doesn't have a driver.
07:53 AM veverak: that can be bought
07:53 AM SpeedEvil: Also, 8mNm
07:53 AM veverak: SpeedEvil: they make driver for these as single circuit
07:53 AM SpeedEvil: yes.
07:54 AM veverak: and you can buy gears to make worm wheel drive
07:54 AM veverak: (which is quite nice mechanism for robotic arm)
07:54 AM SpeedEvil: That is 8 grams at 10cm torque
07:54 AM veverak: where wrom wheel gives you 1:30 ratio (or, the one I found and like)
07:54 AM veverak: *worm
07:54 AM SpeedEvil: And yes, but once you start buying gears, drivers, ... it starts to greatly increase the overall price
07:55 AM robotustra: this motor is not very good for static applications like legs
07:55 AM robotustra: too low torque at 0Hz
07:55 AM veverak: SpeedEvil: you can buy set of gears for 5 sets for like 5$ ?
07:55 AM * veverak will again do the calculations
07:56 AM veverak: but yeah, 3d printing required
07:56 AM SpeedEvil: And the shaft, and the gearbox, and the assembly, and the grease, and ...
07:57 AM SpeedEvil: If I'm doing the maths right, at 6000RPM, you're getting about 2.4W mechanically out of that
07:57 AM SpeedEvil: And it has quite a lot of inertia
07:59 AM veverak: SpeedEvil: compare the motor with https://tkkrlab.nl/wiki/Arduino_towerpro_sg90_9g_servo with 1:30 gearbox
08:00 AM SpeedEvil: That is the wrong thing to be comparing it to.
08:00 AM SpeedEvil: You need to at the least estimate the whole cost of the whole system, and compare it to something of similar cost
08:01 AM SpeedEvil: Note, for example, that the above motor alone weighs 31g
08:01 AM veverak: in progress
08:01 AM veverak: ...
08:01 AM veverak: like I said, I have to do the math
08:01 AM veverak: :)
08:02 AM SpeedEvil: Meaning, you also need to account for the extra weight - if you bring the servo mass in at 50g each, then that's if you have ten of them an extra 400g of payload your robot can have if you just buy the servos (if tehy are equivalent)
08:03 AM veverak: true
08:03 AM veverak: and all parameters
08:03 AM veverak: but comparing that takes time
08:04 AM veverak: and prototypes
08:04 AM veverak: it's utterly pointless to do it without prototype of each version
08:04 AM veverak: I just believe that it's in category "it's worth it propably"
08:05 AM SpeedEvil: Also, this is misleading
08:05 AM robotustra: veverak: do you have your design already?
08:05 AM SpeedEvil: these are second-hand motors, and pulls.
08:05 AM veverak: robotustra: had ideas, stopped developing it because best motor I had before was 10$/piece
08:06 AM veverak: will try again
08:06 AM SpeedEvil: Unless you're going to do a lifetime buy, these may end up horribly expensive as you need to entirely redesign
08:06 AM robotustra: I would not base my design on cheap motors
08:07 AM SpeedEvil: https://www.banggood.com/Racerstar-BR2212-2450KV-2-3S-Brushless-Motor-For-RC-Airplane-p-1085432.html?rmmds=search
08:07 AM SpeedEvil: (probably not particularly suitable in this instance)
08:08 AM veverak: that's overkill
08:08 AM veverak: :)
08:08 AM veverak: hmm
08:08 AM robotustra: it's light weight
08:08 AM robotustra: but you need controller
08:08 AM veverak: still weighs more than motors I found
08:08 AM veverak: robotustra: you can buy IC for controlling these cheaply
08:08 AM robotustra: and make your own position feedback
08:08 AM veverak: but tehy are up to 2A tops
08:09 AM robotustra: solder it yourself?
08:09 AM veverak: I know, guys in hackerspace are working on software for that
08:09 AM veverak: but their target are bigger motors
08:09 AM veverak: but ... one processor can handle multiple motors
08:09 AM robotustra: I don't know what is you idea, but if you want be cheap - it's better to use brushed motors
08:09 AM veverak: and you get position/load feedback, which sg90 does not have by default
08:10 AM SpeedEvil: you do not simply get position/load feedback
08:10 AM SpeedEvil: You need external sensors
08:10 AM veverak: ...
08:10 AM veverak: I know
08:10 AM SpeedEvil: Or a more complex driver
08:10 AM veverak: :)
08:10 AM SpeedEvil: I mean for the brushless
08:11 AM veverak: again
08:11 AM veverak: I KNOW
08:11 AM robotustra: actually I got 3phase motor controller but it's not cheap
08:11 AM SpeedEvil: you can also get feedback from the servo nearly free with just a resistor
08:11 AM veverak: not fun in that
08:12 AM veverak: and that's what I said some lines above
08:13 AM robotustra: http://www.ebay.com/itm/DC-12V-36V-15A-500W-Brushless-Motor-Controller-Hall-BLDC-Driver-Board-LJ-/112320972101?hash=item1a26da0545:g:XEAAAOSwhlZYt8vb
08:13 AM robotustra: they have hall sensors connections
08:13 AM veverak: yep, it's commonly used solution
08:13 AM veverak: (hall sensors)
08:13 AM robotustra: didn't connect it yet
08:15 AM veverak: 6.4$ per "joint" in case I buy 5pcs
08:15 AM veverak: motor + driver + gears + hall sensor
08:15 AM veverak: without: magnest, 3d printed plastics, 2mm shaft
08:16 AM veverak: and CPU controlling that
08:35 AM veverak: CPU added 4$ to cost of 5 units, but I doubt it is able to drive up to 5 of them
08:35 AM veverak: hmm
08:35 AM veverak: magnets/shaft/3d printed plastic should be somewhere in hackerspace in a high amount
08:37 AM veverak: so, 35$ for batch for 5 units, where I can have ARM Cortex controlling multiple joints with feedback (I think that IC driver I found has load feedback pin )
08:37 AM veverak: +- my human error and unexpected things not taken into account
08:37 AM veverak: oh, yeah, IE pcb that I can create in hackerspace
08:39 AM * z64555 wonders if he should pick up a few hall-effect sensors
08:42 AM veverak: z64555: they sounds interesting
08:42 AM veverak: some people use them as replacement for joytisck for flying simulators
08:42 AM veverak: I mean, as replacement for interla sensing of those devices
08:42 AM z64555: you mean position or displacement sensing
08:42 AM veverak: yep
08:43 AM z64555: hall-effects are very much nonlinear, though. which can cause issues if you have it set up too far
08:43 AM z64555: *too far from the magnet
08:43 AM veverak: yeah, that can be "tricky" part
08:44 AM veverak: anyway, I need 120 degree real usage movement for joint
08:44 AM veverak: 180degree hall sensor should handle that
08:44 AM veverak: (in theory if course, needs prototype)
08:45 AM z64555: isn't that the aperature of the sensor, meaning it can sense 180 degrees around its forward vector
08:45 AM z64555: or whatever its called. Just started on my first cup of coffee
08:45 AM z64555: Hdd's spinning up
08:45 AM veverak: yep
08:45 AM veverak: or, I think so :
08:45 AM veverak: :)
08:47 AM veverak: A1302KUA-T is the one that seems interesting
08:54 AM * z64555 looks at a "Hotend"
08:54 AM veverak: hmm, have to go to hackerspace
08:54 AM veverak: discuss if I don't miss some component and if code will be usable on this thing
08:55 AM z64555: it's a block of metal, tapped for a brass nozzle and hole drilled for a heater and a thermocouple
08:55 AM z64555: opposite of the nozzle is a tube with a heat sink
08:55 AM veverak: yep
08:56 AM z64555: the heat sink is a radial type, could probabaly make one on a lathe
08:57 AM veverak: yep but it can test your engineering skills
08:57 AM veverak: *lathe
08:58 AM z64555: something about this design seems inefficient in terms of heat transfer
09:01 AM z64555: the feeding mechanism is also off-site of the hotend
09:01 AM veverak: z64555: depends
09:02 AM veverak: z64555: you want short area where temperature goes from low to high temp
09:02 AM veverak: the shorter the better
09:03 AM z64555: that's what I mean, you shouldn't need that heat sink, or at least for it to be very large
09:03 AM veverak: you need it to make the area of heat change shorter
09:03 AM veverak: :)
09:04 AM z64555: https://images-na.ssl-images-amazon.com/images/I/61Ct%2B6AWX0L._SL1200_.jpg
09:04 AM veverak: you heat the "neck" tube from one side and cool it from other side
09:05 AM veverak: so the "heat change lengh" stabilize at quite short space
09:05 AM veverak: if you remove the radiator and fan
09:05 AM veverak: that lengh increases a lot
09:05 AM veverak: which is really bad for usage
09:05 AM z64555: that's my point. The heat shouldn't be traveling up the nexk
09:05 AM z64555: *neck
09:06 AM veverak: z64555: ABS/PLA has one bad property, there is temperature range where it is still stiff, but soft enough to adhere to the walls of the tube where it is
09:06 AM veverak: that is problem, because it increases resistance for pushing it quite a bit
09:06 AM veverak: you need to decrease the space where that happens as much as possible
09:06 AM veverak: z64555: why not?
09:06 AM veverak: heat travels where it can
09:08 AM veverak: z64555: https://umgallery.ultimaker.com/uploads/gallery/album_33/sml_gallery_3723_33_14479.jpg
09:08 AM z64555: Yes, I see what you're trying to tell me, I already know that
09:09 AM z64555: I'm stating that there should be a combination of materials for the feeding tube/neck that would prevent heat trasfer
09:09 AM veverak: some people used that approach
09:09 AM z64555: such as, glass, or a ceramic
09:10 AM veverak: but these materials are fragile
09:10 AM veverak: the nozzle can hit the 3d printed things sometimes
09:10 AM veverak: z64555: you are right about that, it's just that there are other problems arising with changing the materials
09:10 AM z64555: so make a replacable fail point
09:10 AM veverak: like several times during print
09:11 AM veverak: and "I hit something with the nozzle, I have to replace it" would be solution that is too unpractical
09:11 AM veverak: trust me
09:11 AM veverak: :)
09:11 AM veverak: z64555: I don't say that there is not better solution, it's just that nobody came with that yet
09:11 AM z64555: or better yet, not hit the thing in the first place :P
09:12 AM veverak: easier said than done
09:12 AM veverak: :)
09:21 AM z64555: they do a little bit of material science to reduce the transfer, it looks like. the neck tube on the hotend I showed you looked like it was made out of mild steel
09:23 AM veverak: z64555: btw: https://wiki.e3d-online.com/wiki/E3D-v6_Documentation
09:25 AM SpeedEvil: stainless is even better
09:25 AM SpeedEvil: There are very few metals with lower heat conduction than stainless
09:25 AM veverak: I think it's used for the neck
09:26 AM SpeedEvil: SS about a third of steel
09:27 AM z64555: gtg, hope to catch up with ya'll later :)
09:37 AM rue_house: someone who does not have/have_used a printer should not speculate on the design problems, when you get into it (for even less than $100) you realize that its a totally different game than it should be
01:42 PM nbro: since there are other robotics channels on freenode, what’s this channel exactly about (even though this question may seem so trival)?
02:06 PM SpeedEvil: Everything from little things with servos made from kits to much larger robots
02:06 PM SpeedEvil: basically construction or design of any size or sort of robot
02:08 PM nbro: ok
02:17 PM deshipu: nbro: there are?
07:30 PM z64555: deshipu: yeah, like quadrotor for the quadrotors. Although last I checked they were mostly competition flyers with potty mouths
07:31 PM deshipu: not sure if I would categorize a quadcopter as a robot
07:32 PM SpeedEvil: In what way is it not a robot.
07:32 PM SpeedEvil: IMU, controlled, heavy amount of onboard processing ,...
07:32 PM deshipu: in the sense that it doesn't mimic any animal or human in any way
07:33 PM SpeedEvil: hummingbird
07:33 PM deshipu: well, my laptop also does heavy onboard processing, but it's not a robot
07:35 PM jandor: you could add come actuators
07:35 PM jandor: and it will be
07:35 PM deshipu: does a cd tray count?
07:35 PM jandor: besides of cdrom lid opener, ofcourse
07:35 PM jandor: lol
07:35 PM jandor: it might
07:36 PM jandor: if it can do something with it
07:36 PM deshipu: if you make it do something that a human or animal would do
07:36 PM deshipu: hence my personal definition of a robot :)
07:36 PM jandor: so you define robot as thing that does human work?
07:36 PM jandor: we type simultaneously
07:37 PM jandor: lol
07:37 PM deshipu: not necessarily work
07:37 PM jandor: well sometimes I sit and do nothing
07:37 PM deshipu: a device that is designed to mimic some aspect of humans or animals
07:37 PM jandor: can a stone mimic some of my aspects?
07:38 PM deshipu: yes
07:38 PM jandor: can we count is a robot?
07:38 PM deshipu: and those plastic figurines of robots from japanese movies can too
07:38 PM deshipu: and you would still call them "robot" in everyday language
07:38 PM jandor: hm
07:39 PM deshipu: of course, an engineer could choose to use a definition that more closely defines the set of skills necessary to build it
07:39 PM deshipu: so, anything that has sensors, actuators and a feedback loop, for example
07:40 PM deshipu: but that doesn't agree with common language
07:40 PM deshipu: you don't call your toilet flusher a robot
07:40 PM jandor: what are you doing in robotics?
07:41 PM jandor: is it a hobby or a work?
07:41 PM jandor: or something else?
07:41 PM deshipu: just simple hobby toys
07:41 PM jandor: okay
07:41 PM deshipu: walking quadrupeds, mostly
07:41 PM jandor: with some math?
07:41 PM deshipu: yeah, some basic geaometry and linear algebra
07:41 PM deshipu: nothing too fancy
07:42 PM jandor: I'm trying to understand the math behind bipedal walking
07:42 PM jandor: no success yet
07:42 PM z64555: inverse kinematics
07:42 PM deshipu: jandor: I can recommend the "underactuated robotics" course for that
07:43 PM jandor: where?
07:43 PM deshipu: https://www.edx.org/course/underactuated-robotics-mitx-6-832x-0
07:43 PM deshipu: here
07:43 PM deshipu: it's free
07:43 PM jandor: ah
07:43 PM jandor: cool
07:43 PM jandor: thanks
07:44 PM deshipu: the math is a bit daunting, but even if you skip most of it, the course still gives you a great introduction into the nature of the problem
07:51 PM deshipu: jandor: here's one of the lectures: https://www.youtube.com/watch?v=ZetCOQS4b9c
07:51 PM jandor: oh god
07:51 PM jandor: 10 weeks course
07:53 PM deshipu: yeah, it's a bit too condensed and too fast
07:53 PM deshipu: but you can do it in your own pace if you don't care about the grade
07:58 PM jandor: a quick recap
07:59 PM jandor: it is underactuated because we can't control contact points with ground
07:59 PM jandor: amirite?
07:59 PM deshipu: kinda
08:00 PM deshipu: it has more degrees of freedom than you have actuators
08:00 PM jandor: yep
08:00 PM deshipu: in practice, any robot that is not bolted down to the floor is underactuated
08:00 PM jandor: ok seems like I can get through it
08:01 PM jandor: thx
08:15 PM jadew` is now known as jadew