#robotics Logs
Jul 14 2018
#robotics Calendar
12:13 AM Jak_o_Shadows: I suddenly realised that I only have 1 stm32 not ina project.
12:17 AM Jak_o_Shadows: so I just bought some stuff
12:17 AM rue_shop3: heh, been there
12:17 AM rue_shop3: what kinda stuff you using them in?
12:35 AM rue_shop3: ok I know what to do with the wire
12:52 AM Jak_o_Shadows: rue, I have a joystick, a clock, christmas lights controller, and a testing one. I am going to make a clock display (e.g. use a clock face & hopefuly mechanism to show a gauge).
12:55 AM mrdata: i want to generate 100 kV DC
12:55 AM mrdata: and discharge it through a perovskite
12:56 AM mrdata: piezo material
12:58 AM rue_shop3: hmm
12:58 AM rue_shop3: will it click or ring
12:58 AM rue_shop3: looking at a circuitboard I made up in 1988
12:59 AM rue_shop3: I have no idea what its for, but its some kinda 6 channel controller, I prolly still have the drawing for it
01:32 AM rue_shop3: One of the useful properties of the R C networks is that frequency is
01:32 AM rue_shop3: inversely proportional to capacitance, rather than to the square root of
01:32 AM rue_shop3: capacitance as in an LC tank circuit.
01:33 AM rue_shop3: often necessary
01:33 AM rue_shop3: to simultaneously tune two or three elements in the R C network
01:46 AM mrdata: rue claim is, it will expand and contract. but as it does this, thrust is generated
01:48 AM rue_shop3: AHA JFET PART NUMBERS!!!
01:48 AM rue_shop3: 2N5458 2n5459 2n4416 2n3819
01:49 AM mrdata: i had a 2n3819 and tried to make a Vackar oscillator
01:49 AM mrdata: but didnt get it working
01:49 AM mrdata: so i am disappointed
01:50 AM rue_shop3: I'v never had a jfet osc work
01:50 AM rue_shop3: it dosn't help when the datasheets say "general purpose amplifier"
01:53 AM mrdata: vackar oscillator was a very stable tube device in early 20th century; prior to discovery of quartz crystal properties
01:54 AM mrdata: it was like a collpits oscillator but had extra caps
01:54 AM mrdata: with particular values
01:56 AM mrdata: the idea of using 2n3819 was that the jfet was electrically very much like a tube amplifier, just working at a lower voltage
01:57 AM mrdata: so you wouldnt have to throw away old schematics
02:00 AM rue_shop3: still trying to place your knowledge
02:02 AM rue_shop3: anyhow I found what 276-1624 had in it...
02:04 AM rue_shop3: hey, wtf, the 6027 is a put not a ujt...
02:07 AM mrdata: https://physics.fullerton.edu/~heidi/JMP-MachII.pdf
02:07 AM rue_shop3: ok...
03:11 AM rue_shop3: hmm, I want to have one led drive 3 or 4 cds cells at once
03:12 AM rue_shop3: I wonder how I can do that
03:12 AM rue_shop3: there must be a good trick to make sure the distribution is even
03:14 AM rue_shop4: hey wait
03:14 AM rue_shop4: maybe the loop can be wider
03:14 AM rue_shop4: make the led-driving-op-amp get its feedback from the osc
03:14 AM rue_shop4: I would need a pretty linear freq to votlage converter
03:15 AM rue_shop4: from 1 to 10Mhz
03:15 AM rue_shop4: hmm
03:15 AM rue_shop4: huh, how does one get filter falloffs LESS than 3db/decade
03:15 AM rue_shop4: well wait
03:16 AM rue_shop4: maybe I should just work with 3db/decade
03:16 AM rue_shop4: so, filter with 1Mhz Fc, peak detector, and a mean, mean filter
03:24 AM rue_shop4: 3.3k and 100pf goes from -5db at 1Mhz to -25db at 10Mhz
03:24 AM rue_shop4: if that were a 4V p-p signal...
03:24 AM rue_shop4: :S
03:56 AM rue_shop3: from about 7Mhz, the difference in the filter output is so small it dosn't matter
04:02 AM rue_mohr: ah
04:02 AM rue_mohr: no
12:36 PM rue_mohr: so, the problem is, that if the drift in the vco your using in a controlled loop is faster than the correction time of the loop, you end up with horrid jitter
12:37 PM rue_shop3: so, what cuases the drift in the osc I have
12:37 PM rue_shop3: power line fluctuations is probably a factor
12:38 PM rue_shop3: I wonder what the power line noise is
12:38 PM rue_shop3: why did I have a dream I was out of personal cards?
12:38 PM rue_shop3: and what was Fran doing at the component store I used to go to
12:48 PM mrdata: yes, power line fluctuation will
12:49 PM mrdata: put a decoupling cap on it
01:10 PM rue_shop3: I have one, but I remembered someting
01:11 PM rue_shop3: to stabalize a pwm generator I made, I had to make the filter cap split between positive and negitive
01:11 PM rue_shop3: so, instead of 20pf to ground, 10pf to ground and 10pf to vcc
01:12 PM rue_shop3: I think it helps the circuit ride out vcc changes
01:12 PM rue_shop3: the idea of controlling the led current based on the osc freq wont work unless I can find a detector with a linear output between 1M and 10Mhz
01:30 PM rue_shop3: hmm I think ceramic caps have physical stability issues in the high freq area
01:44 PM weyland|yutani: https://electronics.stackexchange.com/questions/327975/capacitance-vs-frequency-graph-of-ceramic-capacitors
01:45 PM rue_shop3: mica caps help
01:46 PM rue_shop3: huh
01:51 PM rue_shop3: hahaha
01:51 PM rue_shop3: ok, I can see part of the problem
01:52 PM rue_shop3: if I run my finger up and down the ~3/4" ground lead on one of the 50pf caps, I can consistantly change the freq by 300Hz
01:55 PM rue_shop3: however, since I started the osc, its walked up 3Khz
01:57 PM rue_mohr: ok, I have a new direction
01:58 PM rue_mohr: if I take the 74HC14 based circuit, make it 1 stage and go for 2-20Mhz, then divide it by two, I can half a bunch of the error
01:58 PM rue_mohr: dont know if I can pull this off tho
01:58 PM rue_mohr: thats a really stupid amount of range
01:58 PM rue_mohr: do it all smt
03:08 PM rue_shop3: ok, outdoor stuff
03:28 PM Tom_L: yeah it's hot out, may as well do outside stuff
03:48 PM rue_shop4: I have my robot mowing the lawn
03:49 PM rue_shop4: so I'm working hard watching it, drinking my slushie
03:49 PM rue_shop4: need to make up the big one
03:50 PM rue_shop4: need a caster wheel sensor for it, to get vector
03:50 PM rue_shop4: maybe if I do a dual, a crude absolute encoder, with a fine incremental ganged
03:52 PM rue_shop4: OMG I'm learning!
03:52 PM rue_shop4: I have an idea
03:53 PM rue_shop4: how to make an absolute encoder thats got 3 bits, but has a resolution of more than 8
03:53 PM rue_shop4: ok get this
03:53 PM rue_shop4: you dont look at the state you wait for a transition
03:53 PM rue_shop4: ok, normal wheel would go 0,1,2,3,4,5,6,7
03:53 PM rue_shop4: there are 8 states, and 8 transitions
03:53 PM rue_shop4: but
03:54 PM rue_shop4: if you do
03:54 PM rue_shop4: 0,1,2,3,1,4,2,5
03:55 PM rue_shop4: then each state transition only has one assosciated location
03:55 PM rue_shop4: as you can see, there can be more than 8
03:55 PM rue_shop4: thinking
03:57 PM rue_shop4: 0,1 0,2 0,3 0,4 0,5 0,6 0,7 1,2 1,3 1,4 1,5 1,6 1,7 2,3 2,4 2,5 2,6 2,7 3,4 3,5 3,6 3,7 4,5 4,6 4,7 5,6 5,7 6,7
03:57 PM rue_shop4: is that all of them?
03:57 PM rue_shop4: so, 28 positions
03:58 PM rue_shop4: now, the trick is to just shuffle the deck right
04:04 PM rue_shop4: ----OMG MAKE A COMPUTER DO THIS FOR ME
04:04 PM rue_shop4: ITS LIKE THE WORST MAZE EVER
04:06 PM rue_shop4: brute force...
06:36 PM robotustra: evening
06:44 PM theBear: hmm... i usually pronounce it more like 'morning!', but nevertheless <ACK><OBJ data='hurro fellow' alt='one cannot simply walk into the new millenia and the pointy-bracketed-syntax styles that come with it!'>
07:00 PM rue_shop4: the bender for the bot is rolling the pipe, cant see whats causing it
07:01 PM rue_shop4: one of the rollers must be tilted, measured them every way I can think of, its not showing
07:21 PM rue_shop4: 8-| oh, I could make systems that required two transitions too,
07:21 PM rue_shop4: but, eckhg,
07:34 PM rue_shop4: arg, recursive code.... .MY BRAIN...
07:43 PM Tom_shop: videos to come after bit
08:14 PM rue_shop4: hmm, it dosn't work with 2^2 options
08:16 PM robotustra: need to buy round file tomorrow
08:16 PM robotustra: mine is dull
08:22 PM rue_shop4: this can be done with more bits tho
08:22 PM rue_shop4: with 3 bits, I can do the following
08:23 PM rue_shop4: 0,1,2,0,3,2,4,0,5,1,4,5,2,6,5,3,1,7,4,6,7,(0)
08:23 PM rue_shop4: thats 21 positions from 3 bits
08:24 PM rue_shop4: and I'm just huffing it, if I work out this software, there are gonna be longer seqences
08:24 PM rue_shop4: then you just have a state machine work it out
08:42 PM robotustra: what are you doing with state machines?
08:47 PM robotustra: some progress today https://imgur.com/dLDacrP
08:51 PM rue_shop4: I'm making a 3 bit encoder
08:52 PM rue_shop4: but I realized that by using the state changes and not the states themselvs as the positions, I can get a LOT more than 8 positions from the encoder
08:52 PM rue_shop4: the seqence above demonstrates
08:52 PM rue_shop4: thats 21 positions from a 3 bit encoder,
08:52 PM rue_shop4: as soon as the state changes, you know exactly where you are
08:53 PM rue_shop4: a state machine can be used to decode it, that bit is easy
08:53 PM rue_shop4: the question is, given 8 states, what is the longest bidirectional sequence that can be made
08:54 PM rue_shop4: er, unidirectional?
08:55 PM rue_shop4: writing a program to brute force it
08:55 PM rue_shop4: cool? or over your head?
09:03 PM robotustra: what is bidirectional sequence?
09:06 PM rue_shop3: 0,1,2,0,3,2,4,0,5,1,4,5,2,6,5,3,1,7,4,6,7,(0)
09:06 PM rue_shop3: see how nomatter which direction you go, any two numbers only come up once?
09:06 PM rue_shop3: so, if I see 5 and then 1, I know I'm in the same position as if I saw 1 and then 5
09:12 PM rue_shop4: eeek, how do I get my result out of this recursive code...
09:14 PM robotustra: it's called grey code
09:14 PM rue_shop4: nope, each number is grey is unique
09:15 PM rue_shop4: I'm talking about an encoder where the transitions between the states are whats unique, not the states themself
09:15 PM robotustra: ok, you are using algorythm which is used in archivers
09:15 PM rue_shop4: as that sequence demonstrates, you can get 21 positions out of a 3 bit encoder wheel
09:16 PM rue_shop4: and probably more
09:16 PM robotustra: it's because you detect the position from 2 neighbour states, not one
09:16 PM rue_shop4: yes
09:17 PM robotustra: that's why you can encode 8*2*2 states
09:17 PM rue_shop4: the robot I want to use it on will be all over the place, so needing a change wont matter
09:17 PM robotustra: 32 states
09:17 PM rue_shop4: dunno, it has to go back to the same state
09:17 PM rue_shop4: 32 transitions
09:17 PM robotustra: the save as 5 bits encoder
09:17 PM robotustra: same
09:23 PM robotustra: it's interesting which type of encoder they use in here http://www.centroidcnc.com/centroid_diy/acorn_cnc_controller.html
09:27 PM rue_shop4: thats just a standard one
09:27 PM Tom_L: ok, video uploading
09:29 PM Tom_L: had to edit a couple spots in the gcode
09:33 PM rue_shop4: ok, the recursive code works, I just have to get the results out of it somehow
09:35 PM rue_shop4: oh god, there are like, billions of sequences for 8 symbols
09:39 PM rue_shop4: I suppose this is a tree with a fanout of ~7
09:39 PM rue_shop4: depth of over 21
09:39 PM rue_shop4: :D
09:39 PM rue_shop4: computers gonna hate me
10:13 PM rue_shop4: mrdata, so, all this stuff with the oscillators has me wondering just how much the timing in a computer drifts
10:14 PM robotustra: I need an encoder with 2.5" hole in it
10:14 PM rue_shop4: let me guess, it has to be a 25mm OD encoder
10:14 PM rue_shop4: ;)
10:16 PM robotustra: no
10:16 PM robotustra: it has to be 64.5 mm ID encoder
10:16 PM rue_shop4: but what about compact construction!
10:17 PM robotustra: I actually can put it iside my spindle
10:17 PM robotustra: small one
10:17 PM robotustra: but I would like to have it outside
10:18 PM robotustra: and I don't want to run extra belt for encoder
10:18 PM rue_shop4: wtf, my code is on acid
10:20 PM rue_shop4: [0,1]->[1,2]->[0,2]->[0,3]->[1,3]->[1,4]->[0,4]->[0,5]->[1,5]->[1,6]->[0,6]->[0,7]->[2,7]->[2,3]->[3,4]->[4,5]->[5,7]->[6,7]->[3,6]->[3,7]->[4,7]->[4,6]->[2,6]->[2,5]->[3,5]->End of list.
10:20 PM rue_shop4: I have millions of transistion strings like that
10:21 PM rue_shop4: need to have the code filter it now
10:21 PM robotustra: it looks like I have to do plan B
10:22 PM robotustra: install encoder on the end of the spindle
10:22 PM robotustra: coaxial
10:23 PM rue_shop4: most solutions are 25 transitions
10:26 PM robotustra: what is the max PRM encoders have
10:26 PM robotustra: can handle
10:26 PM rue_shop4: depends on resolutions
10:27 PM rue_shop4: a 2 pulse encoder can usually take on about 500k
10:27 PM rue_shop4: Hz
10:27 PM rue_shop4: well, just draw a line at 500Khz output signal
10:28 PM robotustra: 2000 pr
10:28 PM rue_shop4: a good one should do 500Khz, in a differential signal
10:28 PM robotustra: https://www.ebay.com/itm/NEW-Incremental-Rotay-Encoder-2000P-R-AB-phase-encoder-6mm-Shaft-5-24V/232364729921
10:29 PM robotustra: it means it's ok up to 250 rpm?
10:29 PM robotustra: per second
10:29 PM robotustra: rps
10:30 PM robotustra: 1500 rpm
10:30 PM rue_shop4: response frequency: 0-50KHz
10:31 PM rue_shop4: 2000 ppr
10:31 PM robotustra: I mean if it will be 3000 rpm it's 50 turns per second
10:31 PM rue_shop4: root@freebee6:/etc# calc 50000/2000
10:31 PM rue_shop4: 50000/2000 -->> 25
10:31 PM rue_shop4: so 25rps
10:31 PM rue_shop4: root@freebee6:/etc# units 25rps rpm
10:31 PM rue_shop4: * 1500
10:32 PM Tom_L: ok, having trouble with the first video but here's the 2nd one
10:32 PM Tom_itx: https://www.youtube.com/watch?v=KIWKo-1HP3c
10:32 PM robotustra: it means I can't put 2000 ps encoder
10:32 PM rue_shop4: robotustra, "resolver"
10:32 PM robotustra: and I shouldnt
10:33 PM rue_shop4: Tom_itx, you must have one hell of a compressor...
10:33 PM Tom_L: i do
10:33 PM Tom_L: 5hp 80gal
10:33 PM Tom_L: 2 stage
10:34 PM Tom_L: didn't run that often really
10:34 PM Tom_L: compared to when i was using it for finishing
10:36 PM Tom_L: i got a little too agressive on a couple of the cutters in a couple spots, i need to edit the program in those areas
10:36 PM Tom_L: but this is a first test run anyway
10:37 PM Tom_L: having trouble rotating the 1st video
10:38 PM rue_shop4: is the motor the limit? or were you pushing the bit
10:38 PM Tom_L: motor is the limit
10:38 PM Tom_L: it was getting pretty warm even with the water cooler
10:38 PM Tom_L: it helped a whole lot though
10:38 PM rue_shop4: resin :)
10:39 PM Tom_L: it did the job like it is but i was pushing the motor to it's physical limits
10:40 PM Tom_L: i almost stalled it a few times but the PID kicked in and it recovered
10:40 PM Tom_L: slowed the feed down a bit in those places
10:41 PM Tom_L: that's partially where i need to edit the program
10:41 PM rue_shop4: ... this program isn't finding a solution...
10:41 PM Tom_L: still figuring out what this thing can do
10:42 PM Tom_L: i have yet to stall a stepper
10:43 PM Tom_L: yesterday the X axis bumped against the table vise and i thought i was a gonner but it just started pushing the mill across the table instead of stalling
10:43 PM rue_shop4: heh
10:43 PM rue_shop4: "will make its own space"
10:44 PM rue_shop4: only enough room for the table and the mill moves itself back and forth
10:44 PM Tom_L: i thought i had it positioned where it would clear everything but i added a "table" to the vise to hold the keyboard and mouse and that's what it hit
10:44 PM Tom_L: clamped a board in the vise as a table
10:44 PM rue_shop4: so, done a pcb with it yet?
10:44 PM rue_shop4: I have that on my list of things to do
10:44 PM Tom_L: i won't
10:45 PM rue_shop4: heh, I have to get my etching setup back going
10:45 PM Tom_L: how do i rotate a yt video?
10:45 PM rue_shop4: oh dont look at me
10:46 PM Tom_L: i found it
10:46 PM Tom_L: it's gonna take a while
10:54 PM rue_shop4: ok I have to pass a BEST param thru this recursive thing... arg
10:54 PM rue_shop4: why do recursive fn's always need 100000 params?
11:01 PM rue_shop4: I dont know if I should leave this other copy trying to find an ideal solution...
11:15 PM Tom_itx: https://www.youtube.com/watch?v=VUDHCXu4Pwo
11:17 PM rue_shop4: [0,1]->[1,2]->[0,2]->[0,3]->[1,3]->[1,4]->[0,4]->[0,5]->[1,5]->[1,6]->[2,6]->[2,3]->[3,4]->[4,5]->[2,5]->[2,7]->[3,7]->[3,6]->[4,6]->[4,7]->[5,7]->[5,6]->[6,7]->[0,7]->[0,6]->End of list, 24 usefull transitions
11:18 PM rue_shop4: +- a bit, it cant tell if its comming or going
11:19 PM rue_shop4: even 23 positions is pretty good for a 3 bit encoder
11:19 PM rue_shop4: seeing as its absolute
11:21 PM Tom_L: i'd never have tried that on the sherline
11:28 PM rue_shop4: I think thats the limit
11:28 PM rue_shop4: I'v seen a few 1000 results go by with 24 usefull transitions
11:29 PM rue_shop4: which means I did pretty damn good just fudging it there
11:31 PM Tom_L: what's the 3 bit encoder for?
11:42 PM rue_shop4: large mowing robot
11:43 PM rue_shop4: caster wheel
11:43 PM rue_shop4: and probabyl everything else
11:43 PM rue_shop4: geez man, its a 3 bit encoder with a resolution of like 23 lines
11:43 PM rue_shop4: ok, I watched myfordboy do a 3d print lost investment casting
11:44 PM rue_shop4: once again he has done something with no effort that I could possibly never pull off trying as hard as I can
11:44 PM Tom_L: who?
11:44 PM Tom_L: oh
11:44 PM rue_shop4: myfordboy
11:44 PM Tom_L: nm
11:44 PM Tom_L: i may have seen that
11:45 PM rue_shop4: he make it look so effortless, you should see the results _I_ get
11:45 PM Tom_L: he's probably done it more too
11:46 PM rue_shop4: yes
11:46 PM rue_shop4: https://imgur.com/gallery/mFm9Xyu
11:46 PM rue_shop4: I am not level 100
11:59 PM rue_shop4: [0,1]->[1,2]->[0,2]->[0,3]->[1,3]->[1,4]->[0,4]->[0,5]->[1,5]->[1,6]->[2,6]->[2,3]->[3,4]->[4,5]->[2,5]->[2,7]->[3,7]->[3,6]->[4,6]->[4,7]->[5,7]->[5,6]->[6,7]->[0,7]
11:59 PM rue_shop4: (some may be represented backwards, right now that requires human repair
11:59 PM rue_shop4: )