#robotics Logs

Mar 06 2014

20:09 rue_mohr2 but rethinking that, I think it would make the current detection and voltage monitoring problems worse than I have now
20:10 GargantuaSauce what do you mean by truly float
20:10 rue_mohr2 I'm gonna put recycling into my truck, and clear the deck for the table saw
20:10 rue_mohr2 GargantuaSauce, well, it occured to me that, because I'm using switching power regulators anyhow, I could do it thru an isolated transformer
20:11 rue_mohr2 then I could have outputs that are electrically independant
20:11 rue_mohr2 which would be nice for parallel/series combinations
20:11 GargantuaSauce so you mean multiple isolated DC-DC converters after the initial output of the big chinese power supply?
20:12 rue_mohr2 right now I'm shooting for getting atleast 180w from it
20:12 GargantuaSauce i would say use several of the latter instead
20:12 rue_mohr2 its a 360W supply, I have dialed down to 300w
20:12 rue_mohr2 the main supply has no fine grain current limiting
20:13 rue_mohr2 so I'd need two pushpull cores somewhere around 150w
20:14 rue_mohr2 which is totally within a computer power supply core
20:15 rue_mohr2 I hear things when i'm falling asleep
20:18 Triffid_Hunter computer supplies run up to a kilowatt these days
20:19 Triffid_Hunter will definitely need some capacitors between primary and secondary to keep the EMI down
20:21 rue_mohr2 Ihave 70 computer power supplies for mecha parts
20:25 rue_mohr2 http://m.eet.com/media/1156465/295533-use_a_photoelectric_fet_optocoupler_as_a_linear_voltage_controlled_potentiometer_figure_2.jpg
20:25 rue_mohr2 hmm
20:26 GargantuaSauce i didn't know there were photo-fets
20:28 rue_mohr2 there is a different schematic I'm looking for that uses a more mirrored approach
20:29 Triffid_Hunter nice, didn't know there were photofets
20:30 rue_mohr2 http://www.next.gr/uploads/135-8691.png
20:31 GargantuaSauce they are super expensive
20:31 GargantuaSauce like $2 in quantity
20:31 rue_mohr2 http://www.embedded.com/design/prototyping-and-development/4025023/Opto-electrical-isolation-of-the-I2C-Bus
20:34 rue_mohr2 if that works, I could use it for votlage and current feedback...
20:34 rue_mohr2 or could I
20:34 rue_mohr2 I'd need another supply tied to the output to power the transcievers
20:34 rue_mohr2 it would take 4 optos per channel
20:35 Triffid_Hunter rue_mohr2: if you're just transferring signal, use regular optos, make sure both sides have the same pullup arrangement
20:35 rue_mohr2 analog signals
20:35 Triffid_Hunter the fet is interesting when you can't control the pullup on the far side, eg audio volume
20:35 rue_mohr2 votlage and current feedback for the switching regulator
20:36 Triffid_Hunter yep so if you can control the downstream side, just use regular optos
20:36 rue_mohr2 if I do it completely isolated, I would have two N channel fets on the low side, driving would be easy
20:36 rue_mohr2 the trick is, its a variable supply
20:37 rue_mohr2 so analog sensors on the downstream side, and control on the primary
20:37 rue_mohr2 0-30V
20:37 rue_mohr2 I dont want the control on the primary side because I want usb control via a microcontroller off the bulk input ground
20:37 Triffid_Hunter 0 will be fun, might need an aux supply for the downstream sensing
20:38 rue_mohr2 yea, an isolated one to boot
20:38 rue_mohr2 I'v needed isolated supplies for so amny things, I'm almost thinking about making a generic baord,
20:39 rue_mohr2 have 4 or 8 isolated 5 to 9ish volt outputs few 100 ma
20:40 Triffid_Hunter heck 20mA would be enough with micropower opamp, seen those isolated fet driver things, with ultrabright LED and solar panel in a package?
20:41 rue_mohr2 I need them lots fo high side fet switching
20:41 rue_mohr2 cause P channel fets STILL suck
20:41 Triffid_Hunter they're spendy, better use a bootstrapped gate driver unless even those are unsuitable
20:41 GargantuaSauce still unhappy with the idea of bootstrap diode+cap?
20:42 rue_mohr2 if I usea pushpull int eh middle to isolate it I dont need that then do I??
20:42 rue_mohr2 http://www.next.gr/uploads/135-8691.png
20:42 rue_mohr2 this cricuit looks tooo simple, I'm gonna go try it
20:43 rue_mohr2 there was a cool op-amp version I cant find
20:46 rue_mohr say it again for me!
20:46 rue_shop3 http://www.next.gr/uploads/135-8691.png
20:46 GargantuaSauce it again
20:47 rue_mohr you missed for me
20:48 GargantuaSauce well i didn't say it again for anyone else
20:48 rue_mohr :P
20:54 TELunus|2 Does anyone here have tips about robots that can assemble replicas of themselvs? To my knowlage no one has really made a satisfactory one yet.
20:55 GargantuaSauce one term applied to the concept is Von Neumann Probe
20:55 GargantuaSauce and yes it is far in the future
20:56 MrCurious as soon as someone makes a replicator... someone else will hack it and makeit evil
20:57 TELunus|2 You had better hope that: A) that's not true, or B) you die before I do. Because I plan to make sure there are self replicating robots before I die.
20:57 MrCurious replicate until resources are exhausted, then leave the planet and repeat.
20:58 TELunus|2 I actually assume you leave the planet first.
20:58 GargantuaSauce pretty sure we all plan to do that
20:58 GargantuaSauce doesn't mean it's going to happen
20:58 MrCurious wait, am in in the wrong club here? is this the make the robot club?
20:59 MrCurious the real profit in robotics is in sex bots and dildonics...
20:59 TELunus|2 Yup.
20:59 MrCurious industrial robotics are already comoditized market
21:00 TELunus|2 Sure, and I want to take those industrial robots redesign them to make themselvs.
21:00 TELunus|2 Also, I don't really care much about profit, except that it could help get other people working with me.
21:02 GargantuaSauce there is nothing in the way of having a 6dof arm or whatever take a box of its parts and assembling a friend
21:04 GargantuaSauce the problem is the metal foundries, the structural part milling, motor winding, semiconductor fabrication, PCB manufacture, etc
21:04 GargantuaSauce each of those is individually automate
21:05 GargantuaSauce but making one system that combines all of them AND is composed solely of parts it itself is capable of manufacturing from raw materials
21:05 GargantuaSauce is an order of magnitude more complex of an engineering problem than anything we've approached before
21:05 TELunus|2 You say nothing in the way, except that I don't believe the gripper and parts are designed such that the robobt can actually grip each part.
21:05 TELunus|2 For example a difficult part is the assembly of the gripper it's self.
21:06 GargantuaSauce in any case you're strengthening my point
21:06 GargantuaSauce and assembly is one tiny piece of the puzzle
21:06 TELunus|2 If your point is that it's impossible, bulshit. If it's that it's difficult, I'm already well aware.
21:07 GargantuaSauce my point is it's not going to happen in our lifetimes and it'd make more sense to approach one of the subproblems so that it is more approachable in the future
21:07 rue_shop3 files/doc/pdf/62.236.156.23/pc817xx.pdf
21:09 rue_mohr TELunus|2
21:10 rue_mohr I have a project that might help
21:10 rue_mohr make a extension hand
21:10 TELunus|2 I'm young. You know what happend in my grandparents lifetime? The mechanical television was invented, and so was the roomba.
21:10 rue_mohr this is on my project list
21:10 rue_mohr basically a directly operated mechanical hand that you can use as an extention of your own hand
21:10 rue_mohr see if you can get it refined enough to build another one of them with
21:10 GargantuaSauce oh those have a cool name, what is it
21:10 GargantuaSauce waldo unit!
21:11 rue_mohr not electronic, direct mechanical , keep it simple
21:11 rue_mohr TELunus|2 then, IF SOMEONE HANDS YOU A PEICE OF METAL THEY __JUST WELDED__ you can grasp it without burning yourself
21:11 TELunus|2 We should expect there to be a similar increase, if not greater, in technology in our lifetimes.
21:11 rue_mohr AHEM
21:11 TELunus|2 Ya, ya, ya.
21:12 rue_mohr its just that you did it twice to me that bothers me
21:12 TELunus|2 No, only once.
21:12 TELunus|2 Pretty sure I would remember a second time.
21:12 rue_mohr DIP, THEN HAND OVER!
21:12 GargantuaSauce lol
21:15 rue_shop3 maybe Ian did it to me the second time then
21:15 rue_shop3 meh
21:17 rue_shop3 TELunus|2, I started a projects.hackday.com project
21:17 rue_shop3 http://projects.hackaday.com/project/296-Bench-power-supply
21:20 TELunus|2 And GargantuaSauce, thanks for the reminder. I am planning on working on a sub part of the full self replication first, it just that the sub part I'm most interested in at the moment is the assembly part.
21:21 GargantuaSauce as for that i think the best approach is having a variety of end effectors
21:21 GargantuaSauce don't expect one gripper to be able to work with all the parts
21:23 rue_shop3 maybe Ian did it to me the second time then
21:23 rue_shop3 meh
21:23 rue_shop3 TELunus|2, I started a projects.hackday.com project
21:23 rue_shop3 http://projects.hackaday.com/project/296-Bench-power-supply
21:23 rue_shop3 had ate my main text entry tho
21:23 GargantuaSauce freenode is shitting the bed tonight
21:25 GargantuaSauce TELunus|2: have you looked at big expensive multiaxis CNC machines much?
21:26 TELunus|2 Afraid not.
21:26 TELunus|2 What are you thinking?
21:26 GargantuaSauce well they have 2-3 effectors and each has a variety of tooling which can be automatically swapped out
21:26 GargantuaSauce i'd say there's inspiration to be had there
21:27 GargantuaSauce have a 6-axis arm or whatever with a standard socket for different manipulators to plug into precisely
21:28 TELunus|2 Oh, right. It was only a few axis, but I used to program a cnc punch press ocaisionally at one of my previous jobs.
21:28 TELunus|2 It did automatically switch effectors.
21:29 GargantuaSauce in general i'd say modularity is key. the whole system cannot be expected to meet all of the requirements
21:29 TELunus|2 Of course it's effectors were baisically just punch and die, but still.
21:29 GargantuaSauce but different bits and pieces could collectively do so
21:29 TELunus|2 Ok, thanks.
21:30 GargantuaSauce and i guess the socket is equivalent to abstraction and i may as well be talking about software design
21:30 TELunus|2 Hardware design could probably actually learn a thing or two from software design.
21:31 GargantuaSauce yeah for sure
21:31 GargantuaSauce i think a modern operating system is the most complex single construct ever produced by our species
21:32 GargantuaSauce so obviously we've come up with some tools for dealing with that complexity which could definitely be applied in other domains where we want to move forward
21:32 GargantuaSauce in any case if you ever find yourself approaching this as a funded or commercial affair, i want in :V
21:35 TELunus|2 I'm thinking of trying it as a non
21:36 TELunus|2 -profit.
21:36 TELunus|2 Which doesn't mean I can't pay employees if I get money.
21:36 TELunus|2 Which is still a big if.
21:37 GargantuaSauce yeah
21:37 TELunus|2 It's been pointed out that I baisically need visible progress before anyone will be interested in either joining my cause or funding me.
21:37 GargantuaSauce i am doing a software startup now and even with the approach of having a *product* the notion of getting funding is super imposing
21:38 TELunus|2 Ya.
21:39 TELunus|2 I say "me" and "my", but I actually have another friend who was pretty much on board the moment I mentioned the idea. Although the original idea was not this in particular, and not fully even relevant to robotics.
21:40 orlock software design could learn from hardware design
21:40 GargantuaSauce this is also true
21:40 orlock who would you trust to build your bridges, skyscrapers and planes?
21:40 orlock real engineers, or software guys?
21:41 GargantuaSauce well the dynamics of the things we engineer are a bit more conducive to rigorous testing
21:41 GargantuaSauce there are more states in a kilobyte of memory than there are elementary particles in the universe
21:43 GargantuaSauce and church and godel and turing showed us like 80 years ago that it's fundamentally impossible to prove software to be perfect
21:43 TELunus|2 So you baisically said that the number of combinations of elements in a kilobyte is more than the number of things to combine in the universe.
21:43 TELunus|2 But the number elements in a kilobyte is much much less.
21:43 TELunus|2 Which is why we can impement them at all.
21:57 Triffid_Hunter GargantuaSauce: klee can go a long way
21:57 Triffid_Hunter and similar tools
21:57 GargantuaSauce oh yes for sure
21:58 GargantuaSauce static and dynamic analysis, test-oriented development, so on and so forth
22:02 TELunus|2 What's klee?
22:04 GargantuaSauce it's an llvm target that automates testing using symbolic magic
22:05 GargantuaSauce http://www.doc.ic.ac.uk/~cristic/papers/klee-osdi-08.pdf
22:15 GargantuaSauce http://www.youtube.com/watch?v=imVNg9j7rvU