#robotics Logs
Nov 23 2017
#robotics Calendar
12:00 AM anonnumberanon: yeah that sure wont break the bank :) https://www.amazon.com/s/ref=nb_sb_noss?url=search-alias%3Daps&field-keywords=BS1362
12:08 AM anonnumberanon: man the cheap stm32f103 on ebay has like 14 16-bit timers
12:08 AM anonnumberanon: could run 20 servos at high resolution alone with nothing added..
12:43 AM anonnumberanon: obsession... :) https://www.youtube.com/watch?v=Lofbz3f_m3k
01:01 AM rue_shop3: you guys dont listen to what I say
01:01 AM rue_shop3: you can do 88 hobby servos with an stm32 before you have to start compromising resolution
01:30 AM rue_mohr: I said it so many times before I dont know why I bother stating it again, you must already know what I was on about
01:30 AM rue_mohr: you dont want to use a multichannel controll, you want to make one, I need to stay out of it
01:32 AM rue_mohr: well, didn't get he rov going, the code isn't accepting the serial stream, not sure why
01:34 AM rue_mohr: I wonder if the networked timer thing can be done by dynamically programming the processor to execute a series of instructions that create the delays be changing the path and time that instrctions at memory locations are executed
01:34 AM rue_mohr: 100G timers
01:35 AM rue_mohr: say, quad core, 3.5Ghz processor
01:35 AM rue_mohr: 7 14Ghz
01:35 AM rue_mohr: falls short
01:36 AM rue_mohr: I know, somewhere, there is somone out there
01:37 AM rue_mohr: somewhere there is someone waiting to find the someone, just like I have
01:38 AM rue_mohr: I cant be the only one
01:38 AM rue_mohr: so where the hell are they
01:44 AM anonnumberanon: since you started talking about this I've reasonably stayed quiet, but come on:
01:45 AM anonnumberanon: "I wonder if the networked timer thing can be done by dynamically programming the processor to execute a series of instructions that create the delays be changing the path and time that instrctions at memory locations are executed"
01:45 AM anonnumberanon: just re-read this please haha
01:45 AM rue_mohr: no, I'm serious
01:45 AM anonnumberanon: okay DUDE
01:45 AM anonnumberanon: it's NOT even english
01:45 AM rue_mohr: you never programmed a 6502 and had to have it modify its own instructions to do what it needed to?
01:46 AM anonnumberanon: to be lood do more timers look more like?
01:47 AM rue_mohr: about 100G timers
01:47 AM anonnumberanon: I'd like to , also wonder, like you, but the grammar of that quoted paragraph made my brain cut down on the attempt to understand
01:47 AM rue_mohr: interlinked event timers
01:47 AM rue_mohr: on a regular pc
01:47 AM Jak_o_Shadows: rue, tbh, I would favour an analytical approach
01:47 AM anonnumberanon: you need to draw this on paper and show if you want people to understand
01:48 AM rue_mohr: would require some major compression
01:48 AM anonnumberanon: or maybe I'm just that daft
01:48 AM rue_mohr: no, its a radical, insane idea
01:48 AM Jak_o_Shadows: Like, put a probability distribution on the amount of time each link takes
01:49 AM rue_mohr: and I dont even know if it could work
01:49 AM Jak_o_Shadows: I know for a fact that the telecomms people do it with probability stuffs
01:49 AM rue_mohr: it might only be solvable via an evolutionaly alg
01:50 AM rue_mohr: but how to take a standard pc, and have it simulate 100G configurable event timers, firing at a rate of up to 20hz ea
01:51 AM rue_mohr: it fits a few points
01:51 AM rue_mohr: - the code and the memory are one, being that the execution path of the code constitutes the memory
01:52 AM rue_mohr: - its compressed, and technically optimal for what its trying to do
01:52 AM rue_mohr: I need to think on this
01:53 AM Jak_o_Shadows: Going high-level, you wanted to predict delays in communication links yeah?
01:53 AM rue_bed: no
01:53 AM Jak_o_Shadows: what were you doing?
01:54 AM Jak_o_Shadows: I think I might have lost it in the scrollback
01:54 AM rue_bed: its a type of neural net implementation
01:54 AM rue_bed: but I didn't say that
01:54 AM rue_bed: cause it dosn't matter
01:54 AM rue_bed: so, anyhow, super simple example
01:54 AM rue_bed: 6502 processor, 1k of ram
01:54 AM rue_bed: set up with code that just loops nops
01:55 AM rue_bed: lets say the loop takes 1 second
01:55 AM rue_bed: each address fires at 1Hz, with a sequential delay between them
01:55 AM * Jak_o_Shadows nods
01:56 AM rue_bed: if a branch instruction is placed in the middle, that checks an io port, then, concievably, the processor could increase the loop rate to 2hz if an input bit is high
01:57 AM rue_bed: sorry, pfm is part of the goal
01:57 AM rue_bed: now, hypothetically, you could have a few of those timer loops interwound and executing at the same time
01:58 AM rue_bed: this is about the limit of my mental comprehension right now, so thats where I'll stop
02:01 AM Jak_o_Shadows: ok
03:19 AM Luminax-Work is now known as LuminaxWk
03:33 PM SomeT: hi again
03:33 PM SomeT: you know on the amazon kiva robot, I can't find out or back it up with any source as to why it has so many infrared sensors on the front
03:33 PM SomeT: why does it need more than one?
03:33 PM SomeT: I thought its either a form of validation or to increase scope of the laser path / range
03:33 PM SomeT: *infra red light range not laser
03:34 PM deshipu: to know the distance in multiple directions at once?
03:34 PM SomeT: but its only on the front of the robot
03:34 PM deshipu: so that it knows at what angle it approaches the wall, and turn in the correct direction based on that
03:35 PM SomeT: ah yes
03:35 PM SomeT: of course
03:35 PM SomeT: its so obvious, lol
03:35 PM SomeT: thanks
03:35 PM SomeT: it has an encoder sensor as well
03:35 PM SomeT: and I was trying to think wtf is this sensor even doing
03:35 PM SomeT: but now that makes sense
03:35 PM deshipu: it's just a guess of course
03:36 PM deshipu: they didn't release the source code
03:36 PM SomeT: sorry rate gyro not encoder
03:36 PM SomeT: nah I am making some guesses my friend
03:36 PM SomeT: haha
03:36 PM SomeT: good guesses rather
03:36 PM SomeT: I am looking at this
03:36 PM SomeT: https://www.flickr.com/photos/txyzinfo/sets/72157666820353163/
03:36 PM SomeT: breaking down every single component
03:37 PM SomeT: then find 10 papers academically on each component
03:37 PM SomeT: and comparing them as well to academic studies of the robot
03:37 PM SomeT: I wish I was writing more words on this paper tbh, I could probably work out the source code haha
03:37 PM AchiestDragon: prob because of the sensor plain also , they look so they are a 2d plain sensor to guage hight and clearance so many because hight and clearance in a few directions
03:39 PM SomeT: I think your gonna have a whole axis of dimensions really
03:39 PM SomeT: the rate gyro should work in conjunction with them
03:39 PM SomeT: its the only thing that makes sense
03:39 PM AchiestDragon: there industrial ones for a "controled aria , so having gaps in the horisontal plain is less imprtant
03:40 PM AchiestDragon: hight of its load and clerance for it is more of an issue than things beeing in its way
03:54 PM SomeT: thanks
04:05 PM AchiestDragon: http://3dprintingforbeginners.com/how-to-make-diy-filament-for-your-3d-printer/, anyone any tips , all i can say is it seems a good idea
04:06 PM durrf: the problem with DIY filament extruders is dimensional accuracy
04:06 PM durrf: i think you want +/-0.05mm tops and its difficult to do at home
04:09 PM AchiestDragon: if i did a diy one i would prob still buy a nozle of the correct size , as left to me it would be drilled and end up with about +-25mm acuracy
04:09 PM durrf: they use the same nozzle size for 1.75 and 2.85mm filament
04:09 PM durrf: the difference is the speed at which its pulled out
04:10 PM durrf: commercial filament is run through cooling baths immidiately after extruding, and has a ton of failsafes and checks to make sure its the right dimensions
04:10 PM durrf: the filastruder project is probably significantly advanced compared to when i last looked into it
04:10 PM durrf: but its tricky
04:11 PM durrf: plus different additives and dyes change the properties of the plastic
04:11 PM AchiestDragon: k , well better to konw before investing in one , i know theres a few on the makret but no idea how good if at all
04:12 PM durrf: id read up on other peoples (recent) experiences with them, i really wanted to get into making my own filament but it didnt seem justifyable unless you can get a loooooooot of scrap or just buy fresh pellets or some shit
04:14 PM AchiestDragon: yea they sell pellets , scrap mainly dew to the fact if im making prototype parts i'd prob end up having to scrap quite a few prints ,so recycling it back to filiment use would be good
04:15 PM durrf: i just wait till MG chemicals kilograms are on sale for 14 bux at amazon
04:15 PM durrf: i have like 5 spools i havnt opened yet
04:15 PM durrf: and like 20 spools in various states of emptyness
04:26 PM AchiestDragon: if my other projects are anything to go by i usualy produce a lot of scrap , not so bad when its aliminium usualy from scrap and get about the same rate by weight ether way
04:28 PM AchiestDragon: https://drive.google.com/file/d/1rumsw_qJ8q-X8sjKIa1mRJFsiqNF2Asf/view?usp=sharing this is going to use a far bit of pla , the red lower leg parts are 300 mm long
04:29 PM SpeedEvil: A little vacuum furnace would be nice.
04:29 PM SpeedEvil: throw clean chippings and offcuts in, and get thick plate.
04:29 PM AchiestDragon: i used to weigh in the chippings also
04:31 PM AchiestDragon: if your lucky and the prices go up you sometimes go the the scrap yard see a nice block of ally , drill lots of holes and machine half of it away , and get more 6 months later when you weigh it and the bag of chippings back in when you scrap it
04:32 PM SpeedEvil: :)
04:36 PM AchiestDragon: its been a few years since i have been doing any construction but the last time i went there someone had just sraaped a whole producition line machine , tuns of usefull bits in there like industrial grade lead screws over a meter long , servos , actuators etc , if you take the time to strip the bits you get them at mixed scrap by weight from them , but depends on how well you know them
04:37 PM SpeedEvil: Or sometimes they'll dump it so you can strip it at leisure, then collect later
04:37 PM SpeedEvil: For a modest fee of course
04:39 PM AchiestDragon: well its ususaly hit and miss and they wont keep it for long , maybe 1 to 2 days till they get arround to that spot in the yard, they have to strip it down anyway to sort the scrap
05:11 PM SomeT: for my paper I have to write on sensor models and fusion techniques in relation to the amazon kiva robot, I have to write 250 words approx on this topic, am I right to presume that based on the factors of the systems the robot entails detailed here: https://pasteboard.co/GV2ExI9.png and the sensors that make them up therein, combined with the fact amazon has a computer system, the robot just
05:11 PM SomeT: uses normal computing algorithms to blend in with this system via the sensors thus sensor models are the systems in the attached image and fusion techniques are the combination of the algorithms and the systems working together?
06:35 PM SomeT: onto next part anyway now no worries
06:40 PM robotustra: AchiestDragon, cool legs
06:42 PM robotustra: AchiestDragon, next generation spider
06:43 PM SomeT: do you know what the pid controller is used in a amazon kiva robot?
06:45 PM robotustra: SomeT, do you know what is PID controller?
06:47 PM AchiestDragon: a hexapod not a spider but yea
06:49 PM robotustra: technically a spider is octapod
06:57 PM * Tom_L snickers a bit
06:58 PM orlock: i thought they were arachnids
06:58 PM Tom_L: technicalities
07:03 PM AchiestDragon: more of a dung beatle than a spider
07:27 PM AchiestDragon: re earlyer about home filiment making if you know the pull rate for the extruder , it should be posible to use a 3d printer extruder to give you a contant pull rate if the filiment is cool enough by that point , , the commertial units still seem a bit pricy
07:29 PM AchiestDragon: but something i would look at if i end up with enough scrap to make it worth doing
07:44 PM SomeT: sorry
07:44 PM SomeT: I am back
07:44 PM SomeT: I worked it out on my own lol
07:48 PM robotustra: I mean 8 legged robot can emulate 6 legged robot
07:48 PM robotustra: if pick up 2 legs :)
07:49 PM AchiestDragon: and if you turn the 2 legs it picks up though 90deg and put pincers on the end you have a crab robot :)
07:53 PM robotustra: can't argue with you
07:54 PM AchiestDragon: main issue is dynamics , and range of movment for each leg , the middle set on eather left or right and the distance it can make between steps before hitting the leg infront of or behind , with a octipod that makis it take shorter strides , although provides more traction and better load distribution
07:55 PM robotustra: it's all about the code
07:57 PM robotustra: hm, finally flatten one side of the plate
07:57 PM robotustra: 7 other sides left
07:57 PM robotustra: oh, and 16 edges
07:58 PM robotustra: it took me about 16 hours
08:33 PM Tom_L: robotustra
08:33 PM Tom_L: how old are you?
08:33 PM Tom_L: because at this rate....
08:46 PM robotustra: 40+
08:47 PM robotustra: I mean 8 hours on week end and 2 hours by 4 days
08:47 PM robotustra: during the week
08:48 PM z64555: 16 hours to do what?
08:48 PM Tom_L: hopefully i'll get my ballscrews and rails ordered soon
08:49 PM robotustra: to scrape the still plate flat
08:49 PM Tom_L: z64555, scrape some metal
08:49 PM robotustra: 8x8 "
08:50 PM z64555: with what, a toothpick?
08:50 PM robotustra: After scraping there are a lot of stuff to prepare, I think I'll scrape and cut metal at least 2 months
08:50 PM Tom_L: sharp piece of tool steel or such
08:50 PM Tom_L: like a file end etc
08:50 PM z64555: so uh
08:50 PM z64555: why not use an end-mill
08:50 PM robotustra: it's called scraper, with tungsten carbide
08:51 PM Tom_L: because it's not gonna be flat
08:51 PM Tom_L: and mills warp steel
08:51 PM z64555: ah
08:51 PM Tom_L: or any metal for that matter
08:51 PM robotustra: I'm within 10 microns I think
08:51 PM Tom_L: it's an exacting process
08:52 PM Tom_L: robotustra, what are you using as a test piece?
08:52 PM z64555: whats the end plate for, and whats it made out of
08:53 PM robotustra: a straight edge ruler and probes
08:53 PM robotustra: and light source
08:53 PM robotustra: the probe is 0.02 mm
08:54 PM robotustra: nut there is no visible gap between streight edge and the plae
08:54 PM robotustra: plate
08:54 PM robotustra: and also the plate hover on the stone
08:55 PM Tom_L: i think normally you start with 2 pieces to scrape
08:55 PM robotustra: https://imgur.com/JJkE4al
08:55 PM Tom_L: but i haven't actually done it
08:55 PM robotustra: you mean 3 plates technique?
08:55 PM Tom_L: yes
08:55 PM robotustra: I don't need it because I got a new flat surface
08:55 PM Tom_L: the granite is a good place to start though
08:56 PM robotustra: it's supposed to be flat within 3 microns on this size
08:56 PM robotustra: 12x18"
08:57 PM Tom_L: yeah
08:57 PM Tom_L: my bud has one ~6" thick around 4 x 6' or so
08:57 PM robotustra: for my construction it's more than enough
08:58 PM robotustra: you know, I live in appartment :)
08:58 PM orlock: robotustra: is that granite spab part of it?
08:58 PM orlock: slab
08:58 PM Tom_L: heh
08:58 PM orlock: robotustra: we have a machine here that has a granite slab about 1mx2mx60cm deep as it's "bed"
08:58 PM robotustra: it's not a granite slab, it's preceise equipment, called surface plate
08:58 PM orlock: its a gantry style "measuring machine"
08:59 PM robotustra: :)
08:59 PM orlock: in our dedicated metrology lab
08:59 PM robotustra: I use it fo flaten my metal pieces
08:59 PM Tom_L: robotustra do you have your ballscrews handy?
08:59 PM Tom_L: i wanted you to measure something for me
08:59 PM Tom_L: the driven end
08:59 PM Tom_L: how much shaft do you have for a pulley?
08:59 PM robotustra: ballscrews?
08:59 PM robotustra: ah
09:00 PM Tom_L: didn't you get ballscrews with your rails?
09:00 PM robotustra: for lathe
09:00 PM robotustra: yes I have them
09:00 PM Tom_L: with the bearing block in place i need to know how much shaft there is sticking out
09:01 PM robotustra: 10 mm
09:01 PM Tom_L: that's not much
09:02 PM Tom_L: the drive pulley i have is ~1" wide
09:03 PM Tom_L: not sure what to do about that yet
09:03 PM robotustra: I mean the diameter
09:03 PM robotustra: do you mean the length?
09:03 PM Tom_L: i know the diameter
09:03 PM Tom_L: yes
09:03 PM Tom_L: from the end of the bearing block to the end of the shaft
09:03 PM Tom_L: to mount a pulley
09:04 PM robotustra: 15 mm
09:05 PM Tom_L: i wonder if that's gonna be enough to hold the pulley
09:05 PM Tom_L: i wish the shaft extended to the end of the pulley but looks like it won't
09:05 PM robotustra: if it's more than a half - probably it's ok
09:06 PM Tom_L: just over half if your measurements are what mine will be
09:06 PM Tom_L: and i can't get a pulley without a hub
09:06 PM robotustra: it's a nut in here
09:06 PM Tom_L: ie the setscrew is in the pulley teeth instead of a separate hub
09:07 PM Tom_L: backlash nut?
09:07 PM Tom_L: that's what my model has
09:07 PM robotustra: no
09:07 PM robotustra: the screw itself is going throug the bearing hub and tighten with thw nut
09:08 PM Tom_L: right
09:08 PM robotustra: and the thickness of the nut is about 5 mm
09:08 PM Tom_L: double bearing block
09:08 PM robotustra: you can thread the pulley hub and use it as a nut
09:08 PM Tom_L: i'd rather not
09:09 PM robotustra: it this case yo'll have ~20mm
09:09 PM robotustra: ok
09:09 PM Tom_L: i'll give it a try and see how it works
09:09 PM robotustra: I think 15 mm is ok
09:09 PM Tom_L: haven't ordered yet but am planning the parts
09:10 PM Tom_L: that's .590", just over half the hub width at 1"
09:10 PM Tom_L: i think it should be ok
09:11 PM Tom_L: i could get a special pulley made but i don't want to go that route
09:11 PM Tom_L: like a squeeze hub that you tighten from the side instead of a grub screw
09:12 PM robotustra: hm, it mean I'll have similar problem whith it
09:15 PM Tom_L: but it only comes in nylon
09:16 PM robotustra: I'll do aluminum pulleys I think
09:16 PM Tom_itx: http://shop.sdp-si.com/catalog/product/?id=A_6T_3-H60SF3720
09:17 PM Tom_L: i'd still have to bore the center out to 10mm but i gotta do that anyway
09:17 PM Tom_L: just don't know how well the nylon pulley would hold up
09:19 PM Tom_L: the width is good on that though
09:19 PM robotustra: I have plastic pulleys on my cnc router, but they are smaller ones
09:19 PM Tom_L: probably max out around 600 rpm
09:19 PM robotustra: so the torque is not high there
09:19 PM robotustra: 600 for stepper - it's way too much
09:20 PM Tom_L: the torque would be more on the smaller pulley though right?
09:20 PM robotustra: yes
09:20 PM Tom_L: likely closer to 400ish rpm
09:20 PM Tom_L: and the small pulley would be metal
09:21 PM robotustra: oh no
09:21 PM robotustra: the torque will be smaller on smaller pulley
09:21 PM robotustra: that's why it's plastic
09:21 PM Tom_L: my small one will be metal
09:22 PM Tom_L: and i can get metal for both but not hubless like the link shows
09:22 PM robotustra: well, what force do you plannign to get?
09:22 PM robotustra: on your axis?
09:23 PM Tom_L: i haven't done any calculations on it
09:23 PM robotustra: just rough estimation
09:23 PM robotustra: 10-20 kg?
09:23 PM Tom_L: maybe
09:23 PM robotustra: it's all about max walue
09:23 PM Tom_L: right
09:23 PM robotustra: ok, you have 5 mm step?
09:24 PM Tom_L: yes
09:24 PM robotustra: the diameter is 16 mm&
09:24 PM robotustra: ?
09:24 PM Tom_L: 570 in/oz stepper at 3:1
09:24 PM Tom_L: 20:60 pulleys
09:24 PM Tom_L: teeth
09:24 PM robotustra: so you have about 3 times increase of force on your screw
09:25 PM Tom_L: yeah
09:25 PM robotustra: it means you need to have about 20/3 ~ 7 kg
09:26 PM robotustra: 7 kg * 8 mm - torque
09:26 PM robotustra: on your screw
09:26 PM Tom_L: where you gettin 8 from?
09:26 PM robotustra: what is the diameter of bigger pulley?
09:26 PM robotustra: d/2 of the screw
09:27 PM robotustra: about
09:27 PM Tom_L: 3.8" pulley diameter
09:27 PM Tom_L: pd is 3.820"
09:28 PM robotustra: ok, it's about 48 mm of radius
09:28 PM robotustra: you'll have only 1 kg of tension on your belt
09:29 PM robotustra: if you pull your timing belt with 1 kg - you'll get 20 kg force on your axis
09:29 PM robotustra: linear force
09:29 PM robotustra: let's say we have a freaction, 25%
09:30 PM robotustra: but it's way too much
09:30 PM robotustra: you'll easily get 15 kg
09:30 PM Tom_L: which has to hold the axis weight including the spindle motor etc
09:30 PM Tom_L: i put a counterweight on my sherline Z axis
09:32 PM robotustra: what is the weight of your axis and spindle?
09:32 PM Tom_L: no idea just yet
09:32 PM robotustra: approximately?
09:32 PM robotustra: are you planning to buy special spindle?
09:32 PM Tom_L: i'd guess around 5-6lbs
09:33 PM Tom_L: not for a while
09:33 PM robotustra: I think you don't need counterweight
09:33 PM Tom_L: i'll probably use the sherline spindle at first
09:33 PM Tom_L: maybe 10lbs
09:33 PM Tom_L: i'm not sure
09:33 PM robotustra: + plates + bearings
09:34 PM robotustra: let's say 20 lb ~ 10 kg
09:34 PM robotustra: so you don't need counterweight
09:34 PM Tom_L: actually, the moving weight will be less than that
09:34 PM Tom_L: http://tom-itx.no-ip.biz:81/~webpage/cnc/Mill_base_XAxis3.jpg
09:35 PM Tom_L: i was thinking the tube was part of it but it's not
09:35 PM robotustra: I got a spindle
09:35 PM robotustra: 1.5 kW
09:35 PM robotustra: if I'll not be lazy I'll try to connect it on weekend
09:36 PM Tom_L: that plate is 7 x 8 x .5" alum
09:36 PM Tom_L: so that and the spindle
09:36 PM robotustra: https://i.ebayimg.com/images/g/Uh4AAOSwePtZ5NZw/s-l1600.jpg
09:36 PM Tom_L: that the one you got?
09:37 PM Tom_L: i'd want water cooled i think
09:37 PM robotustra: yes
09:37 PM robotustra: 5.3 kg
09:37 PM Tom_L: max er11 collet size?
09:37 PM Tom_L: 8mm?
09:37 PM robotustra: I have 7 mm
09:37 PM robotustra: collet
09:38 PM robotustra: but I use 95% of the time 1/8"
09:38 PM Tom_L: i'd hope to be able to use up to around 3/8" cutters
09:39 PM robotustra: I think you can, but I prefer to cut slowly and clean
09:39 PM robotustra: small mill can do cleaner job
09:39 PM Tom_L: well things like trimming an edge square etc i have some longer 3/8" cutters
09:39 PM Tom_L: i use them on the sherline now
09:40 PM Tom_L: small cuts
09:40 PM robotustra: I use big mills only if I want to get deep
09:41 PM robotustra: I'm not sure in the squareness of my mill :) but it's another story, I tried to make it more less square with respect to Z axis
09:41 PM Tom_L: so my Z won't be as heavy as i first thought
09:41 PM Tom_L: yeah that's what i'm gonna do
09:41 PM Tom_L: just the best i can with the tools i have
09:41 PM Tom_L: this mill may build my next one :)
09:41 PM robotustra: and I didn't have tools to measure squareness
09:42 PM Tom_L: i've got a couple machinist squares
09:42 PM robotustra: now I have them
09:42 PM robotustra: and I kind of have to face the table
09:44 PM Tom_L: how much was the spindle?
09:46 PM robotustra: 295 usd
09:46 PM Tom_L: not so bad
09:46 PM robotustra: it's spindle + inverter
09:46 PM robotustra: https://www.ebay.com/itm/1500W-1-5KW-Air-Cooled-CNC-Spindle-Motor-Inverter-Converter-13pcs-ER11-Collet/302490507750
09:46 PM Tom_L: right
09:47 PM Tom_L: 1.5kw is about max for 110v
09:47 PM robotustra: it's usually needed about 400-600W for the lathe
09:48 PM robotustra: of 100 mm chuck
09:48 PM Tom_L: have you powered it yet?
09:53 PM robotustra: no, may be weekend
09:53 PM robotustra: I have similar one
09:53 PM robotustra: on the router
09:53 PM robotustra: but 220 v
09:54 PM Tom_L: what do you make with the router?
09:55 PM robotustra: cutting plastic, aluminum
09:55 PM robotustra: brass
09:56 PM robotustra: tried to cut some steel, but it requires a cooling
09:56 PM robotustra: actually I made it to make robots :)
09:57 PM robotustra: "general purpose cnc mill"
09:57 PM robotustra: ah, I make pcb's on it
09:58 PM Tom_L: same here except the pcb's
09:59 PM Tom_L: headin to sleep here..
10:03 PM robotustra: good night
11:05 PM anonnumberanon: Iv been thinking of servo deadbands.
11:05 PM z64555: they've been thinking about you, too
11:06 PM anonnumberanon: If for instance a servo has a deadband of 5 microseconds, that means you dont need a 1 microsecond resolution to drive them, but scaling to 5 would be good.
11:07 PM anonnumberanon: However it dawned to me one thing. There should be a way to synchronize the pulses to place them exactly in the center of each deadband, and assuming that the deadband is constant...
11:08 PM anonnumberanon: Send 100 to 104 microseconds will keep the servo at the same position
11:08 PM anonnumberanon: 105 to 109 would move it to the next position
11:09 PM z64555: whats this a measure of? the PWM on time?
11:09 PM anonnumberanon: Both orders should thus be 102 and 111 respectively
11:09 PM anonnumberanon: Yeah the pulsewidth of the pwm sent to the servo
11:11 PM z64555: Nah, don't do that. You'll probably end up with the servo twitching
11:11 PM z64555: I mean, there's nothing stopping your from trying it
11:11 PM anonnumberanon: I suspect though that every single would need to be deadband-synchronized, potentially requiring circuitry for each one of them, for instence a current sensor that detects when the servo actually moves!
11:11 PM z64555: but placing the waveform in the middle of the deadband will have the same affect as placing it at the begging or end
11:12 PM z64555: since the deadband is a delta
11:12 PM * z64555 may or may not be awake
11:13 PM anonnumberanon: Oh yeah ur right. But maybe the middle exactly, of the deadband would have twitc-reducing properties!
11:14 PM z64555: apply scientific method :P
11:14 PM anonnumberanon: I will try and find info on servo deadbands, i suspect also it may come from the gears themselves
11:14 PM z64555: Thats gear backlash, which is a factor, definitely
11:14 PM anonnumberanon: Yeah i want to understand the deadband's origins...
11:15 PM z64555: but there's also limitations of the servo control circuit
11:15 PM anonnumberanon: The potentiometer maybe?
11:15 PM z64555: like if its digital, it'll have its own resolution
11:16 PM z64555: the pot doesn't have any major deadband effects, provided that it hasn't "rusted"
11:17 PM anonnumberanon: In anycase the byproduct of this hypothesis veing true would remove the problem of timing overhead when you send 2 orders being spaced out by only a few microseconds
11:18 PM z64555: erm
11:18 PM z64555: they still need a minimum amount of PWM off time to register the command
11:18 PM anonnumberanon: In the new system, they would now either be sent at the same time, or sent with 5 microseconds from each other! Allowing us to have more time to deal with such challenge of making interrupts send orders so close to each other!
11:18 PM z64555: which is a different specification
11:19 PM z64555: stop it, you're sounding more like rue :P
11:19 PM anonnumberanon: Lol
11:19 PM anonnumberanon: I just needed to let it all out
11:22 PM anonnumberanon: Lets continue. We had a pulsewidth range of 500 to 2500 us. Divide this by the deadband which is 5, giving 200 possible orders, or a resolution of a little less than 8 bits!
11:23 PM anonnumberanon: So why are we trying to be able to handle 2000 orders than?
11:23 PM anonnumberanon: Wait sorry not 200, 400, with deadband synchronizing.
11:24 PM anonnumberanon: Which is little less than 9-bit.
11:25 PM anonnumberanon: So help me God.
11:27 PM anonnumberanon: Feel free to complain more if you start detcting babbling.
11:34 PM anonnumberanon: "The deadband was introduced as a way to keep a servo from dancing itself to death. "
11:34 PM anonnumberanon: WOW
11:34 PM anonnumberanon: It may be by design
11:34 PM anonnumberanon: How how do we determine its center point...
11:34 PM orlock: yeah thats what i always thought
11:35 PM rue_mohr: hey hey hey
11:35 PM orlock: anonnumberanon: I used a protractor and sat there fiddling with pulse timing values
11:35 PM anonnumberanon: Haha
11:35 PM rue_mohr: using 16 bits lets you use those bits to nudge the servo horn on the edges of the deadband
11:36 PM anonnumberanon: The trick would be to use circuitry, maybe a current sensor. Because that way you plug in a servo and no matter what it gets deadband-synchronized
11:36 PM rue_mohr: ANd I told ya that the servo dosn't run the motor till AFTER the control pulse
11:36 PM rue_mohr: it multiplies the difference of the control pulse and the internal ref pulse to get the time to drive the motor
11:36 PM rue_mohr: is anyone listening this time?
11:37 PM rue_mohr: http://ruemohr.org/~ircjunk/tutorials/elex/hobbyservo2/p1080887.jpg
11:37 PM rue_mohr: bottom trace is the control pulse
11:37 PM rue_mohr: top trace is the current draw from the motor
11:37 PM orlock: awesome CRO
11:37 PM orlock: i think i saw it in an eposide of the simpsons
11:38 PM rue_mohr: rue_mohr, its hopeless, shut up and put the keystrokes into a new C library
11:38 PM rue_mohr: but, I want to help them understand
11:38 PM rue_mohr: rue, no, its not gonna happen
11:38 PM rue_mohr: nothing is impossable!
11:38 PM rue_mohr: it is, your patience if finite
11:38 PM rue_mohr: no....
11:39 PM rue_mohr: yea, sorry, see this? its the last of what you have left
11:39 PM rue_mohr: why is it in a ketchup jar?
11:39 PM rue_mohr: has to do with the liberal way you have used it
11:39 PM rue_mohr: oh
11:39 PM rue_mohr: but see almost nothing left
11:40 PM rue_mohr: I been pouring patience all over everything?
11:40 PM rue_mohr: yea, and its gonna lead to your death
11:40 PM rue_mohr: but I dont wanna die
11:40 PM z64555: careful with that stuff
11:40 PM rue_mohr: you doddled too long, its becoming your only option
11:40 PM anonnumberanon: So lets see
11:40 PM anonnumberanon: Order received
11:41 PM rue_mohr: http://ruemohr.org/~ircjunk/tutorials/elex/hobbyservo/servo101.html
11:41 PM anonnumberanon: Servo responds immediately
11:41 PM anonnumberanon: What about the deadband
11:42 PM rue_mohr: I wonder where they do that
11:43 PM rue_mohr: must be the RC between 9 and 11
11:43 PM orlock: https://meblogwritegood.files.wordpress.com/2011/08/screen-shot-2011-08-30-at-12-09-46-pm.png
11:43 PM orlock: No, not quite the same
11:43 PM rue_mohr: its better thana digital scope
11:44 PM rue_mohr: no aliasing and your always getting a min and max
11:44 PM rue_mohr: I dont think any digital scope gives you a min-max trace for each sample period
11:48 PM rue_mohr: google all you want
11:49 PM rue_mohr: I can make 0-100% duty pwm in 5 transistors, you need 5000