#robotics Logs

Jul 06 2017

12:10 AM rue_house: AWESOME ma had some ribbon in her art supplies that look like it will do the trick for driving the cart on the screw sorter
12:13 AM rue_house: now I just need to make up rubber pullies and... stuff
12:14 AM youngiif: ha, you have more than your share of ingenuity
12:14 AM rue_house: ? :)
12:15 AM youngiif: i'm saying you're quite clever with your materials
12:18 AM rue_house: yes yes, I'm just wondering what catches your interest
12:18 AM youngiif: im just getting into robotics so pretty much all of your stuff interests me lol
12:29 AM rue_house: :)
12:29 AM youngiif: so do you use openscad for all design work?
12:38 AM rue_house: yes, *but*
12:38 AM rue_house: I use it to extrude DXF files
12:38 AM rue_house: muuuch faster to extrude a 2d profile than to csg it
12:40 AM youngiif: yeah, i need to look more into it, i load a windows virtual machine to user inventor/autocad at the moment
12:42 AM rue_house: https://www.thingiverse.com/rue_mohr/designs
12:42 AM rue_house: check out the simpler things I'v designed
12:43 AM rue_house: I use CSG for the occasional hole, but mostly its stacked dxf ectrusions
12:47 AM youngiif: yeah I think most of the things i design right now would be simple enough I could figure it out. i did model a camera mount that snaps onto my extruder frame earlier that would probably be tough tho. one nice thing about openscad is you can make your thingiverse designs customizable https://www.thingiverse.com/app:22
12:47 AM rue_house: yup, coded model
12:48 AM youngiif: it didn't look like any of your models use that app, have you tried it before?
12:49 AM rue_house: not into it
12:49 AM rue_house: might be one day
02:05 AM Jak_o_Shadows: eyyy, nice rue
02:05 AM Jak_o_Shadows: Have you calibrated it?
02:11 AM Ubuntivity: Hello
02:12 AM Ubuntivity: I'm looking for a 2-wheeled motor module to buy online, like the front one in this picture: http://legoaces.org/MoonBots/Phase_I-Robot_Design_files/dft68c8v_75gcxq4tdw_b.jpg
02:12 AM Ubuntivity: But I do not know which term should I search for :/
02:13 AM * Ubuntivity appreciates any help
02:16 AM mrdata_: hmm
02:16 AM mrdata_: Ubuntivity, dont you like gears?
02:21 AM Ubuntivity: mrdata_: I like them as far as I can use them
02:21 AM mrdata_: so then why would a 2-wheeled motor matter?
02:22 AM Ubuntivity: mrdata_: I want to make a 8-wheeled robot, with each 2 robots 'connected' together
02:22 AM * Ubuntivity is not native English speaker
02:22 AM Ubuntivity: mrdata_: like this -> https://www.youtube.com/watch?v=YgHf9UsfCbQ
02:24 AM Ubuntivity: mrdata_: how would you describe a gear that moves 2 wheels using one motor?
02:24 AM mrdata_: think about the transmission
02:25 AM Ubuntivity: mrdata_: I'm trying to search for such a module in Aliexpress.com with no lucj
02:25 AM Ubuntivity: *luck
02:25 AM Ubuntivity: can you kindly help?
02:26 AM mrdata_: parts for transmissions are nearly always custom as far as i know
02:26 AM mrdata_: just becaus there are so many ways to do things
02:27 AM Ubuntivity: mrdata_: I used to have toys that have a generic motor-gear module that uses one motor to drive 2 wheels.
02:27 AM Ubuntivity: mrdata_: I'm looking for such a ready-made module
02:27 AM Ubuntivity: mrdata_: but my language is not helping me :(
02:30 AM mrdata_: in the video you linked, he said they are using planetary gears
02:31 AM Ubuntivity: I'll search that term
02:32 AM Ubuntivity: mrdata_: it seems to describe the type of gearbox ratio
02:32 AM Ubuntivity: mrdata_: not what I need unfortunately
02:33 AM Ubuntivity: mrdata_: What I need is a 2-unilateral wheels module
02:33 AM Ubuntivity: mrdata_: that is the best description I can express
02:33 AM Ubuntivity: mrdata_: maybe it has a more technical/related name
02:34 AM mrdata_: i dont know what you found in your search, but i found this, https://en.wikipedia.org/wiki/Epicyclic_gearing
02:35 AM Ubuntivity: mrdata_: interesting
02:36 AM Ubuntivity: mrdata_: but it seems more sophisticated than what I need
02:38 AM mrdata_: there are a lot of terms to look up; https://en.wikipedia.org/wiki/Transmission_(mechanics) https://en.wikipedia.org/wiki/Differential_(mechanical_device)
02:39 AM mrdata_: http://www.tezu.ernet.in/sae/Download/transmission.pdf
02:39 AM Donitzo: Eh. Guess I'll here because it's vagely related
02:40 AM Donitzo: any suggestion for a sub 200 euro autonomous quadcopter?
02:40 AM Donitzo: I want to take some measurmenets so preferably with support for I2C
02:40 AM Donitzo: measurements*
02:41 AM Ubuntivity: thanks mrdata_. I'll dig the best term in those references...
04:14 AM Donitzo: so a cheerson cx-20 it is
04:38 AM youngiif: Donitzo: looks pretty decent for the price
04:39 AM Donitzo: as long as you don't screw up the firmware it seems to have the best features in that price class
04:39 AM Donitzo: I'd obviously love to have a crazyflie 2, but they're a bit expensive in comparison
04:41 AM Donitzo: What's most annoying is the price of positioning systems like LPS
04:42 AM Donitzo: $80 per deck, $150 per node. I assume you need at least 3 nodes for 3d positioning
04:43 AM youngiif: yeah, you got me wanting one now...
04:45 AM Donitzo: as far as I have seen there's https://www.pozyx.io/ and https://www.bitcraze.io/loco-pos-system/
04:45 AM Donitzo: both very expensive for what they are
04:47 AM youngiif: ouch, that's a bit pricey
04:48 AM Donitzo: claims you need 4 modules with pozyx for 3d localization
04:48 AM Donitzo: guess triangulation was beyond it
04:49 AM Triffid_Hunter: Donitzo: triangulation requires 4 points.. with just 3 there's an ambiguity about which side of the plane they form you're on
04:50 AM Triffid_Hunter: you can triangulate using 3 nodes in 2D, but not in 3d.. need dimensions+1 nodes
04:50 AM Donitzo: yes there is in one iteration
04:50 AM Donitzo: but I mean, you'd assume you could just check from iteration to iteration
04:50 AM Donitzo: it's not just going to teleport between 2 points
04:51 AM Triffid_Hunter: Donitzo: so what happens when you're really close to the plane?
04:51 AM Triffid_Hunter: and how do you initially establish which side you're on?
04:51 AM Triffid_Hunter: it's really not a robust solution without robust responses to those questions ;)
04:52 AM Donitzo: let me try to visualize this in my head
04:54 AM Donitzo: I'm sorry, all I know is that with triangulation you can find out that you lie on the surface of a sphere
04:54 AM Donitzo: and in my mind it seems like 3 spheres would be enough
04:55 AM Donitzo: blah, I know I'm wrong
04:58 AM Triffid_Hunter: Donitzo: put 3 beacons on a table. For every point above the table, there's also a point below the table with the exact same distances to all 3 beacons. The only way to resolve this in a mathematically robust fashion is to have a 4th beacon that is not on the table and thus can tell if you're above or below
04:59 AM Donitzo: Oh, I see
04:59 AM Donitzo: below or under the plane formed by the 3 points
04:59 AM Triffid_Hunter: yep
04:59 AM Donitzo: yeah, but that's a minor inconvience
04:59 AM Donitzo: I'm not convinced it's worth spending too much money to fix that issue alone
04:59 AM Donitzo: when you could just hard code it
05:00 AM Triffid_Hunter: well sure, that's what GPS does.. assumes you're at sea level on a simplified model of the earth's shape unless there's enough satellites to also derive altitude
05:01 AM Donitzo: you could combine the 3 points with data from for example a magnetometer
05:01 AM Donitzo: using some kind of algorithm to select the correct side
05:01 AM Triffid_Hunter: how can the magnetometer know?
05:02 AM Triffid_Hunter: all it does is tell you direction (relative to the sensor's orientation) and strength of local fields
05:02 AM Donitzo: for example, you place the anchors so that the clockwise facing normal goes east
05:02 AM Donitzo: well okay, a gyroscope then
05:03 AM Triffid_Hunter: Donitzo: so how does a mag or gyro tell you which side of a vertical plane you're on?
05:03 AM Triffid_Hunter: they sense orientation, not position or distance
05:03 AM Donitzo: direction traveled
05:03 AM Donitzo: combined with direction
05:04 AM Donitzo: okay fine, I should just have said accelerometer
05:04 AM Triffid_Hunter: accelerometers can't tell velocity worth a damn, let alone position
05:04 AM Donitzo: you don't need velocity, just acceleration
05:05 AM Donitzo: since both can be detected by the positioning system
05:05 AM Triffid_Hunter: ok so accel is zero, you're measuring approach then growing distance, which side of the plane are you on?
05:05 AM Donitzo: the system would have to store the last known acceleration and position
05:05 AM Donitzo: I mean, it's not perfect
05:06 AM Triffid_Hunter: but it doesn't know position until you pass through the plane and it still doesn't know which direction the velocity points in
05:06 AM Donitzo: magnetometer + accelerometer gives you acceleration in heading
05:07 AM Donitzo: you'd hard code the direction your anchor plane is facing
05:07 AM bordeaux_facile is now known as toinetoine
05:07 AM Triffid_Hunter: see I've done this math with quadcopters which can accelerate in any direction without changing their heading
05:08 AM Donitzo: if increasing distance in same heading as anchor plane, it's one one side, otherwise the other
05:08 AM Donitzo: well, something along those lines
05:08 AM Donitzo: I mean, it can intepret the side based on if it's getting closer or further away from the plane + the heading information
05:09 AM toinetoine is now known as bordeaux_facile
05:10 AM Donitzo: I have not done anything like this but I don't see why it wouldn't work
05:11 AM Triffid_Hunter: Donitzo: only if you can safely assume that your thing can only move forwards along its heading vector
05:12 AM Donitzo: the anchors points would have two guesses where the bot can be right?
05:12 AM Donitzo: One would be in agreeance with the heading, and the other mirrored
05:13 AM Triffid_Hunter: Donitzo: so your thing can't move in reverse? or sideways?
05:13 AM Donitzo: even if the heading data is not perfect, there is a large margin for error
05:13 AM Donitzo: it should be able to move in any direction
05:13 AM Donitzo: magnetometer + xyz acceleration still translates to another 3d vector
05:14 AM Donitzo: of the heading you're accelerating in, right?
05:15 AM Donitzo: rotate the acceleration around the z axis with the direction of the magnetic field
05:15 AM Triffid_Hunter: Donitzo: yeah but you're assuming that your velocity vector is parallel to heading vector which is a very tenuous assumption
05:16 AM Triffid_Hunter: unless your robot has wheels and can't reverse or slide
05:17 AM Donitzo: Does it matter what direction the robot is facing? The point of using the magnetic field is to get more or less absolute acceleration
05:17 AM Donitzo: So it wouldn't mattered if you picked up and carried the bot from side to side
05:18 AM Donitzo: Well, I'm a game developer, not someone with knowledge in robotics so I'm probably missing something
05:20 AM Triffid_Hunter: Donitzo: unless the robot is physically constrained in some way the heading vector and velocity vector are not related
05:20 AM Triffid_Hunter: Donitzo: accelerometers have noise and drift, and thus typically are almost totally useless for integrating to find velocity
05:21 AM Donitzo: Yeah but the velocity isn't really important. It's just a side effect of acceleration
05:21 AM Triffid_Hunter: Donitzo: so you have no way to detect current velocity, yet your 3-beacon triangulation requires a reliable knowledge of velocity
05:22 AM Donitzo: But do you need velocity at all? The positioning system can predict the acceleration of the bot on both sides of the plane
05:22 AM Donitzo: so you're just comparing one acceleration against another
05:23 AM Donitzo: maybe even just comparing the angle of the acceleration
05:23 AM Donitzo: the angle between the accelerations*
05:24 AM Donitzo: the positioning system says, at this side the bot_acc = (10, 0, 0). on the other bot_acc = (-10, 0, 0)
05:24 AM Donitzo: the bot says, my bot_acc = (8.67456, 0.34554, 0.6564)
05:24 AM Donitzo: in the coordinate system aligned with the magnetic field
05:24 AM Triffid_Hunter: well you can try it if you like.. throw together a simulation that moves your robot along some path.. extract whatever a local accelerometer and mag would read, take distances to each beacon, plug them into your algorithm then draw the predicted position. then try a hundred random paths and see if it stays good :P
05:25 AM Donitzo: when I get the chance to buy one of those systems
05:26 AM Triffid_Hunter: buy? can do this purely in software in your favourite language
05:26 AM Triffid_Hunter: when you add a ton of sensor noise and systemic errors from a real system, it only gets exponentially harder. best not to try to debug the math at the same time as sensor issues
05:27 AM Donitzo: true
05:27 AM Donitzo: but in a perfect game world, it should also not fail I believe. Let me draw a quick image
05:29 AM Donitzo: after coffee
08:14 AM Sol is now known as Guest25075
09:37 AM z64555: velocity is important?? what?
09:38 AM z64555: no
09:38 AM z64555: nononononono
09:38 AM z64555: *not really important. sorry
09:38 AM z64555: that's what I was objecting to
09:39 AM z64555: Donitzo: You can derive position, velocity, from just knowing acceleration. BUT. you'll run into a lot of edge cases where that becomes a huge problem
09:39 AM z64555: First of which is sensor error. You get a white (or pink) noise spectrum on _every_ sensor, regardless of how top-grade it is
09:41 AM z64555: If your dead reckoning is based on just the acceleration alone, this amount of error can seriously throw off your estimates.
09:42 AM z64555: Next is quantization error. Your floating point arithmetic is not 100% correct, even you use double-precision
09:43 AM z64555: This is not a major problem for short simulations, but the longer the simulation runs, the higher the probablility of your derived values drifting away from their actual.
09:44 AM z64555: After that, is processor performance, which may or may not affect the accuracy of your simulation depending on what you are using for your time reference
09:47 AM z64555: Performance tends to vary based on your sensor fusion algorithm, ranging from the simple and fast derivation of accelerations to the more complex Kalman filters
09:47 AM rue_house: https://www.youtube.com/watch?v=AeCLbhPHltw
09:54 AM veverak: rue_house: that looks cool
04:51 PM anniepoo: making up leads for capacitive touch system
04:53 PM Guest25075 is now known as Solgriffin
05:58 PM Sol is now known as Guest30584
05:59 PM Guest30584 is now known as SolG
05:59 PM SolG is now known as SolGr
06:17 PM SolGr is now known as Solgriffin
07:27 PM rue_shop3: ok, hwere is my todo list
07:27 PM rue_shop3: quick before I fall asleep
07:28 PM ferdna: what would you name Q on a pcb? (A1241-YK803)
07:29 PM ferdna: looks like a transistor
07:29 PM rue_shop3: its a transistor
07:29 PM rue_shop3: K is relay
07:29 PM rue_shop3: prolly 2SA1241
07:29 PM ferdna: really? i can't find datasheet on it
07:30 PM ferdna: coool
07:31 PM rue_shop3: flat, with a metal tab
07:31 PM rue_shop3: bipolar
07:31 PM rue_shop3: 50V 2A
07:31 PM rue_shop3: 100Mhz bandwidth
07:32 PM rue_shop3: gain between 70 and 240
07:32 PM rue_shop3: the datasheet is everywhere
07:32 PM ferdna: cool thanks
07:32 PM rue_shop3: 2SA1241, how much can I sell it to you for?
07:36 PM rue_shop3: A, B, C, D are prefixed by 2S
07:36 PM rue_shop3: N is prefixed by a 2
07:36 PM rue_shop3: OKI parts use a MSM prefix
07:37 PM rue_shop3: NEC prefixes are UPD, uPc
07:37 PM rue_shop3: JRC is NJM
07:38 PM rue_shop3: there is something else I cant remmeber
07:38 PM rue_shop3: I just know it when I look at the parts
08:06 PM rue_house: I got the thing with all the tubes out of the medical machien
08:06 PM rue_house: the hoses were all disconnectable, so thats all safe now
08:49 PM rue_house: http://ruemohr.org/~ircjunk/tutorials/elex/micropwm/p1080966.jpg
08:54 PM Administrator is now known as Guest70931
08:54 PM Guest70931 is now known as Corey
08:55 PM Corey is now known as The_Jester
08:55 PM rue_house: http://ruemohr.org/~ircjunk/tutorials/elex/micropwm/p1080966.jpg
08:56 PM Tom_itx: rue
08:56 PM Tom_itx: you don't have to install that saleae as root do you?
08:58 PM rue_house: yup
09:12 PM Tom_itx: says it cannot execute binary file
09:16 PM zhanx: chmod a+z filename
09:16 PM zhanx: err
09:16 PM zhanx: chmod a+x
09:25 PM Tom_itx: i get nothing
09:26 PM Tom_itx: i've installed these before and never had this problem
09:31 PM zhanx: hmm chmod 777 filename (it might be a bad download)
09:32 PM Tom_L: http://support.saleae.com/hc/en-us/articles/210245603-Older-Saleae-Logic-Beta-Downloads
09:32 PM Tom_itx: just found that
09:33 PM Tom_L: Linux Compatibility release, for users with older libraries:
09:34 PM Tom_itx: there's a bunch of em there if you need one
09:34 PM zhanx: k
09:35 PM Tom_itx: the one you had was the last 32 bit
09:38 PM zhanx: thought it was
09:38 PM Tom_itx: grabbed it again and it still won't execute
09:39 PM zhanx: let me get it
09:39 PM Tom_itx: how can i tell if this is 32 or 64bit linux?
09:39 PM Tom_itx: wheezy
09:40 PM zhanx: lm /proc/cpuinfo
09:41 PM zhanx: weird logic works fine for me
09:41 PM Tom_itx: command not found
09:42 PM zhanx: i downloaded it extracted it and ran it fine
09:42 PM zhanx: granted i am on mint 17 64 running a 32 bit programm
09:44 PM zhanx: tom you are 32 bit then
09:44 PM zhanx: that command only works on 64
09:45 PM Tom_itx: ok
09:46 PM zhanx: see i solve it the easy way, run a 64 command only and if it fails its 32 bit
09:52 PM Tom_itx: odd
09:53 PM Tom_itx: i downloaded the compatibility one and it loaded
09:58 PM Administrator is now known as Guest3733
10:49 PM rue_shop3: almost polished off the first box
10:49 PM rue_shop3: just 11 more boards after this one
10:50 PM rue_shop3: it turns out a bunch of these supplies use up to 3 TL431
10:50 PM rue_shop3: !!
10:56 PM youngiif: what boards you working on rue? i was disconnected for a bit so missed out
11:01 PM Administrator is now known as Guest79911
11:01 PM rue_shop3: 3 boxes of computer power supply boards
11:01 PM rue_shop3: it took 22 boards to get to the top of the first box
11:02 PM rue_shop3: which is odd, cause they aren't large boxes
11:03 PM youngiif: ah
11:22 PM rue_shop3: 1 more board and I'll call it a night on that
11:22 PM rue_shop3: prettymuch bedtime anyhow
11:22 PM rue_shop3: almost 3kg of parts
11:22 PM youngiif: what time zone are you in?
11:23 PM youngiif: did you get some free parts somewhere?
11:23 PM Snert: rue_bed2: what are you doing? Hoarding used psu parts?
11:39 PM rue_shop3: I'm stripping down a surplus of baords to make some space
11:39 PM rue_shop3: I think there were about 100 computer powers upply boards
11:39 PM rue_shop3: the rubbermaid component tub is about 1/3 full
11:40 PM rue_shop3: it'd be 2/3 if I'd not seperated out the transformers and heatsinks
11:40 PM rue_shop3: ...no, It'd be full
11:45 PM rue_shop3: and I sorted out some silicon... about a kg or so...
11:45 PM rue_shop3: SO MANY COOL PARTS
11:46 PM rue_shop3: I think I can turn lots of the smps transformers into awesome soldering guns
11:58 PM rue_house: so about 1 or 2 guage
11:58 PM rue_house: how many 22 guage do I have to gang?
11:59 PM rue_house: how does that work now
11:59 PM rue_house: I think you go up 3 guages by doubling the wire size?