#robotics Logs

Jun 21 2018

01:43 AM rue_shop3: what I need is a current controlled oscillator
02:11 AM rue_mohr: if an avr is running at 16Mhz
02:12 AM rue_mohr: it can fill a 16 bit counter in
02:13 AM rue_mohr: 4ms
02:13 AM rue_mohr: it can fill a 10 bit counter in
02:14 AM rue_mohr: 64us
02:14 AM rue_mohr: how long does a motor take to wind up
02:16 AM rue_mohr: that'd be a task for the digital scope
02:17 AM rue_mohr: better idea, use a motor with an encoder and use the logic analizer
02:17 AM rue_mohr: guesses?
02:17 AM rue_mohr: order of magnitude?
02:17 AM rue_mohr: hmm
02:17 AM rue_mohr: 0-10000rpm in 100ms
02:19 AM rue_mohr: as an RC charge curve
02:20 AM rue_mohr: so, 16 bit, 25 samples
02:20 AM rue_mohr: 10 bit, 1560 samples
02:21 AM rue_mohr: ok range 1A
02:21 AM rue_mohr: 16 bit 15uA
02:22 AM rue_mohr: 10 bit 1ma
02:22 AM rue_mohr: question, how about 8 bit
02:22 AM rue_mohr: 4mA
02:23 AM rue_mohr: I think over a 1A range, 4mA is plenty of resolution for driving a motor
02:23 AM rue_mohr: ok
02:23 AM rue_mohr: 16us conversion time
02:24 AM rue_mohr: 62Khz
02:25 AM rue_mohr: -- analog --
02:25 AM rue_mohr: 1.1V ref
02:25 AM rue_mohr: 0.0043v/count
02:25 AM rue_mohr: no
02:26 AM rue_mohr: 10 bit adc
02:26 AM rue_mohr: ok, 0.0011v/count
02:26 AM rue_mohr: *256 = .257v range over 1A
02:27 AM rue_mohr: zippo:/tmp# ohm -v .257 -i 1
02:27 AM rue_mohr: Wattage is: 0.257000
02:27 AM rue_mohr: Current is: 1.000000
02:27 AM rue_mohr: Voltage is: 0.257000
02:27 AM rue_mohr: Resistance is : 0.257000
02:27 AM rue_mohr: so 0.27 ohms sense
02:28 AM rue_mohr: so, avr current controller then
02:28 AM rue_mohr: I suppose
02:29 AM rue_mohr: and what about hysterisis?
02:29 AM rue_mohr: I'm laughing it up now
02:29 AM rue_mohr: :/
02:29 AM rue_mohr: your two options are time and value
02:30 AM rue_mohr: ok, overview
02:32 AM rue_mohr: you take a, t13 (I dunno) and give it a serial port, it takes in two analog values, current and position, it transmitts the position, and uses the current feedback against the serial position to output a pulse for current control
02:32 AM rue_mohr: hmm
02:32 AM rue_mohr: its 2 pins in for serial, two pins for adc in, and two for pulses out = 6
02:33 AM rue_mohr: or we dont do the adc for the position
02:33 AM rue_mohr: ok, 2 pins serial, 1 pin adc, and two pins pwm
02:33 AM rue_mohr: and how do we do the bipolar current control
02:34 AM rue_mohr: ok, one avr, one flamethrower....
02:34 AM rue_mohr: so...
02:34 AM rue_mohr: yes the bipolar current feedback is a problem
02:34 AM rue_mohr: what is the sensor is absolute
02:35 AM rue_mohr: well, we are only doing charge control
02:35 AM rue_mohr: how you gonna come up with a unipolar current sense resistor, I'd like to see this
02:36 AM rue_mohr: ok, so, lets chalk that up for an advantage of analog current control
02:36 AM rue_mohr: with the disadvantage being the dac required
02:37 AM rue_mohr: can I just offset the current to 2.5V?
02:37 AM rue_mohr: how about 1.1/2 V?
02:38 AM rue_mohr: must be a more digital way of doing this
02:38 AM rue_mohr: ok, lets go there
02:39 AM rue_mohr: PFM? PWM?
02:39 AM rue_mohr: yes, well if your not playing with amplitude, thats the options
02:39 AM rue_mohr: so, we need a difference unit
02:40 AM rue_mohr: if the pulses were synchronized this might be easy
02:40 AM rue_mohr: its easy for pwm if you sync it
02:40 AM rue_mohr: so, pwm in and operate it like a hobby servo?
02:41 AM rue_mohr: half
02:42 AM rue_mohr: so, if were under current, were not going to push it down
02:42 AM rue_mohr: so the feedback pulse needs to be generated in sync with the control pulse
02:43 AM rue_mohr: they are compared with a D flipflop, and if the control is larger, the transistor is turned on
02:43 AM rue_mohr: ...aaaand dont ever stop sending pulses
02:43 AM rue_mohr: sample rate determined by pulse freq
02:44 AM rue_mohr: thats some pretty carefull pulses on the control stream
02:50 AM rue_mohr: didn't you say that digital pwm chip behaved likea monostable?
02:50 AM rue_mohr: the pain-in-the-ass chip
02:50 AM rue_mohr: hmm
02:51 AM rue_mohr: sigma delta
02:56 AM rue_mohr: no dual slope
02:56 AM rue_mohr: ah
03:04 AM wenz: how is deep learning AI programmed?
03:05 AM rue_mohr: by using the word "NO!" liberally.
03:05 AM rue_mohr: (not played with it yet, sorry)
03:06 AM Jak_o_Shadows: Lots of data, no convergence properties, time
03:11 AM wenz: can you reference me some schoalry articles on how to do convergence properties in c++ or pseudo code?
03:12 AM Jak_o_Shadows: Do you kinda get what I mean about convergence prperties?
03:13 AM wenz: quadratic functionality?
03:14 AM Jak_o_Shadows: WEll, like, with most deep learning/machine learning, you don't have a mathematical guarantee of much
03:14 AM Jak_o_Shadows: as opposed to more traditional methods
03:16 AM wenz: so whats important
03:16 AM wenz: linear algrebra?
03:17 AM wenz: can you teach me the core principles of deeplearning
03:17 AM wenz: because i want to make a deeplearning ai
03:17 AM wenz: in c++
03:17 AM Jak_o_Shadows: yeah, i am not expert enough to teach
03:21 AM wenz: okay
03:22 AM wenz: can you show me how i can program a code that learns something
03:22 AM wenz: i want to make a robot
03:23 AM wenz: https://machinelearningmastery.com/examples-of-linear-algebra-in-machine-learning/
03:23 AM ace4016: buy a book or read up some online resources. it's a fairly extensive topic and not one for some IRC tutelage
03:23 AM ace4016: if you have something specifically you have questions about though, feel free to ask
03:24 AM Jak_o_Shadows: Honestly, I'm rather sceptical about it all, so rely on me to be unrelatingly meh
03:24 AM wenz: can you teach me some simple techniuques
03:27 AM ace4016: that website you link has said simple techniques
03:28 AM ace4016: there's also https://medium.com/machine-learning-for-humans/why-machine-learning-matters-6164faf1df12
03:28 AM ace4016: and a million other resources (it's trendy now)
03:35 AM Jak_o_Shadows: I don't hate this design: https://www.youtube.com/watch?v=zNXgT2csQ7A The video is a bit dumb though.
03:56 AM wenz: is linear algebra easy?
03:58 AM Jak_o_Shadows: I mean, the basics?
03:58 AM Jak_o_Shadows: You can go real crazy with it though
03:58 AM Jak_o_Shadows: and I don't REALLY mean "easy"
04:15 AM rue_shop4: :| wow, ceramic caps severly change value with temp
04:28 AM wenz: iif i can do 3.[(3)(2)+8(3)]-[(84+32) - (3)(3)] = -17
04:28 AM wenz: does that mean i have the potential to do linear algebra?
04:29 AM wenz: i need to hon in on my mathematical skills
04:29 AM wenz: i go expelled from school half way through year 8
04:29 AM Jak_o_Shadows: Of course you've got the potential to do linear algebra
04:29 AM wenz: thanks bro
04:29 AM Jak_o_Shadows: Linear algebra is probably a bit more complicated than that
04:30 AM wenz: ill work my way to it
04:30 AM Jak_o_Shadows: Not that i'm sure how key linear algebra is to training an off the shelf system
04:30 AM wenz: ive found a good calculator to check my awnsers
04:30 AM wenz: https://www.mathpapa.com/calc.html?q=2x+3%3D15%20@%20x%3D6
04:31 AM wenz: ive got to make my own robot this is dire
04:31 AM Jak_o_Shadows: That maths before? Probably not what i'd describe as linear algebra. However, you probably don't need to be able to do the maths just to do it
04:31 AM wenz: if i dont make my own robot i wont beable to protect my self from other robots
04:32 AM wenz: yeah
04:32 AM wenz: but its good to understand the formulars of maths to make complex mathematical expressions
04:32 AM wenz: i can make my robot
04:32 AM wenz: and i can make it upgrade
04:33 AM wenz: plug a cord into it and program it and update its system
04:33 AM wenz: wow this is really fucking cool
04:33 AM wenz: do you guys make robots?
04:34 AM wenz: i was thinking i could make my own processors and microchips
04:34 AM wenz: and put them on the robot and then program the robo
04:35 AM Jak_o_Shadows: We tend to just buy the uC's and chips
04:35 AM BitEvil is now known as SpeedEvil
04:35 AM wenz: i know about medical science so i will beable to incorporate it into a system
04:35 AM wenz: wahts the uC's?
04:36 AM Jak_o_Shadows: microcontroller
04:36 AM Jak_o_Shadows: See, cause the u is the micro symbol?
04:44 AM wenz: 6z+4x=?
04:44 AM wenz: how would you know what z is and x
04:44 AM wenz: i dont get it
04:45 AM wenz: do the numbers just magically appear?
04:45 AM wenz: 6z+4x=? 6z+4x=? 6⋅2+4⋅3=? 6⋅2+4⋅3=? 12+12=24
04:45 AM wenz: how does that work
04:45 AM Jak_o_Shadows: We're not really a general math channel
04:46 AM wenz: i know i know
04:46 AM Jak_o_Shadows: Maybe try ##math?
04:46 AM wenz: if you were to help me what would you think about that just for the time being
04:47 AM wenz: u stands for micro and C stands for controller
04:47 AM wenz: is that correct?
04:48 AM wenz: its one of them special u that have a longer flick tail
04:48 AM Jak_o_Shadows: µ is the micro symbol - but that's harder to type
04:48 AM Jak_o_Shadows: yeah
04:48 AM wenz: like the unit of mesuerment
04:48 AM wenz: howd did u do that
04:49 AM Jak_o_Shadows: Copy and paste from wikipedia
04:50 AM wenz: how would you make a robot that could connect to wifi that would have a set of instructions to login to this channel and send a test signal?
04:50 AM wenz: would that be possible?
04:50 AM Jak_o_Shadows: That's more of an IRC Bot - and we're not huge fans of spammy IRC bots.
04:50 AM wenz: we can have robot communications
04:51 AM wenz: no no
04:51 AM wenz: im not talking about virtual AI
04:51 AM wenz: im talking about a movable robot that could connect into irc
04:51 AM wenz: not a bot
04:51 AM Jak_o_Shadows: The IRC side of things wouldn't be too different?
04:52 AM wenz: i was thinking something
04:52 AM wenz: maybe the robots could collect data and we cold relay are robots info in exchange for stuff
04:52 AM wenz: like a trade or something
04:52 AM wenz: i mean the robots could only get smarter
04:53 AM wenz: like u could talk to my robot
04:53 AM wenz: and i could talk to your robot
04:53 AM wenz: and we could inspect the robots brain
04:54 AM wenz: like the robots could share infomation with eachouter and could embed commands to allow the information to update to robots firmware
06:38 AM wenz: 6<8→true is this true or false?
09:07 AM z64555: right pointing arrow usually means "Results in" or "evaluates to"
09:09 AM justan0theruser is now known as justanotheruser
09:42 AM veverak: ok
09:42 AM veverak: my physics knowledge just got fucked
09:42 AM veverak: I have two water storage boxes
09:42 AM veverak: each 2l
09:43 AM veverak: they are connected with 4mm ID hoses (3x)
09:43 AM veverak: at the level of the "bottom" side of the boxes, so they maintain shared water level
09:43 AM veverak: on one box
09:43 AM veverak: there is a 4mm hole for hose -> pump umps water here from big box that is 0.5m below the two boxes
09:43 AM veverak: (that works)
09:44 AM veverak: there is also a hole for 4mm hose at around 5cm "water height level" in one of the boxes
09:44 AM veverak: that hose is for the big box under the two small boxes
09:44 AM veverak: and
09:44 AM veverak: when the level of water is above those 5cm... (2cm above it!)
09:44 AM veverak: the water os not flowing to the bix box
09:44 AM veverak: I can't comprehen
09:44 AM veverak: d
09:45 AM veverak: NOW IT STARTED
09:45 AM veverak: after the two boxes are almost overflowing
09:45 AM veverak: like wtf
09:46 AM Phreaked: airlock somewhere?
09:52 AM veverak: nope
09:52 AM veverak: I believe it's caused by water properties
09:59 AM veverak: yeah
09:59 AM veverak: after calming down, it actually somehot makes sense
10:00 AM veverak: the system has minimal pressure and the hole is actually 3mm wide at certain moment
10:00 AM veverak: I believe that's the phase where water starts to do it's dark magic properties
10:00 AM z64555: "dark magic properties"
10:01 AM z64555: transfer rate of the water is governed by both the height of the water columns and the properties of the plumbing between the two boxes
10:02 AM z64555: the plumbing has a sort of resistance factor that is both a factor of its material/smoothness as well as the diameter of the pipe
10:03 AM z64555: remember, fluids still have both a dynamic and static friction to them
10:04 AM veverak: bigger hose
10:04 AM z64555: are tbe box tops uncovered?
10:05 AM z64555: if so, you could have interference from air currents pushing down on one of the boxes
10:05 AM veverak: they are
10:05 AM veverak: https://imgur.com/a/MQ28C53
10:05 AM veverak: both boxes were overflowing
10:05 AM veverak: output hose is -2cm from the 'top of the box'
10:07 AM z64555: blah, I'll have to look at this later. Data is slow until the 24th and I've got places to be
10:07 AM z64555: sorry m8
10:07 AM veverak: np
10:07 AM z64555: I should have time again in. hm.
10:07 AM * z64555 looks at watch
10:07 AM z64555: 8 hours
05:28 PM z64555: alright, I'm back
05:28 PM z64555: so its a little hydroponics thing
05:31 PM z64555: veverak: would you please draw a diagram of your set up? The image you shared doesn't show how the hoses are connected in relation to the basins / water boxes
05:35 PM veverak: will do tomorrow I hope
05:35 PM veverak: z64555: late night here and I have exam tomorrow
05:35 PM veverak: so sleepy as hell and tired :)
05:36 PM z64555: ah, maybe we can pick this up tomorrow, then
05:37 PM z64555: good luck on your exam!