Back
[03:29:09] <rue_shop3> so I want to make a duty, as opposed to frequency, pll
[03:29:49] <rue_shop3> if I take two square waves
[03:29:54] <rue_shop3> and delay one
[03:30:03] <rue_shop3> I can make pwm out of the difference
[03:32:43] <Jak_o_Shadows> So I just realised that nobody has found the forward kinematics of a rotary stewart platform
[03:32:59] <rue_shop3> haha
[03:33:02] <rue_shop3> its all simple
[03:33:19] <rue_shop3> I have my brain into a problem of mine tho
[03:33:38] <Jak_o_Shadows> The rotary part makes it harder
[03:33:47] <rue_shop3> only if your an ammature
[03:34:32] <rue_shop3> why do people have problems with kinematics?
[03:35:33] <Jak_o_Shadows> Go:
http://jak-o-shadows.github.io/picturedump/3dp/sg/7.jpg
[03:35:58] <Jak_o_Shadows> Only sensors are the pots, the top joints are 3DOF, the lower joints are also 3DOF, then obviously the pot rotates
[03:36:56] <rue_shop3> you break the elbows down into linear distances and just solve it
[03:37:09] <rue_shop3> use the library I wrote
[03:38:37] <Jak_o_Shadows> Yup, I can find the location of the lower joint. That is easy. But that joint isn't rigid
[03:38:49] <Jak_o_Shadows> Knowing that end point doesn't give you a closed form for the top joint
[03:39:27] <rue_shop3> yea, but the points on the table at the top are fixed distance from each other so your golden
[03:40:45] <rue_shop3> it prolly turns into an ugly 6 way simotanious equation if you do it right
[03:41:05] <Jak_o_Shadows> Yeah, they are.
[03:42:11] <rue_shop3> I assume you want the result to be the position and vector of the top centre of the table?
[03:42:24] <rue_shop3> 2 3d vectors
[03:42:42] <Jak_o_Shadows> yeah.
[03:42:49] <rue_shop3> I hate 3d rotation vectors
[03:42:57] <Jak_o_Shadows> haha. Yeap.
[03:43:01] <rue_shop3> definition of the rotations gets evil
[03:43:24] <rue_shop3> have you solved the position yet?
[03:43:41] <rue_shop3> yea, 3 of those are redundant
[03:44:18] <Jak_o_Shadows> Nah, haven't solved for position.
[03:44:47] <rue_shop3> what are you using for the centre of the joint?
[03:44:54] <rue_shop3> looks like metal
[03:45:20] <rue_shop3> ... is it just a filed nut!?
[03:45:22] <Jak_o_Shadows> What I have done is guess a platfomr pos/rot. Then I can work backwards and find the supposed length of the popsticks. Then I was going to (newton-iteration) on that to.
[03:45:26] <Jak_o_Shadows> Nah, it's a craft bead
[03:45:45] <rue_shop3> heh
[03:45:49] <rue_shop3> have a pic of one out?
[03:45:51] <Jak_o_Shadows> 3D printer isn't nice enough to print it
[03:45:53] <Jak_o_Shadows> yep
[03:46:25] <Jak_o_Shadows> Or, wel, a blurry close-up:
http://jak-o-shadows.github.io/picturedump/3dp/sg/4.jpg
[03:46:43] <Jak_o_Shadows> Wait, here it is:
http://jak-o-shadows.github.io/picturedump/3dp/sg/5.jpg
[03:46:47] <rue_shop3> I mean just the bead
[03:47:07] <rue_shop3> heh
[03:47:10] <rue_shop3> er, huh
[03:47:22] <rue_shop3> not quite spherical, but whatever
[03:47:36] <rue_shop3> you snapped them in?
[03:49:16] <rue_shop3> are the pots 10k?
[03:51:43] <Jak_o_Shadows> Yeah, push fit
[03:51:44] <Jak_o_Shadows> 100K
[03:52:14] <rue_shop3> good, if you were also buying that many 10k pots from china, I'd be worried about the supply
[03:53:02] <Jak_o_Shadows> Huh?
[03:53:19] <rue_shop3> I buy loads of them for my robots
[03:54:25] <Jak_o_Shadows> ah
[03:54:33] <Jak_o_Shadows> I'm decently happy with those joints, I should put it online
[03:54:58] <rue_shop3> its made me realize that a rounded nut might work out ok too
[03:55:21] <rue_shop3> tho I could make a form cutter for the lathe
[03:55:58] <Jak_o_Shadows> I just didn't want to spend so much money on ebay ones
[03:56:25] <rue_shop3> yea, really overpriced
[04:00:30] <Jak_o_Shadows> I don't see how you can find position separetly than rotation
[04:01:18] <rue_shop3> solve the radii for the top plate with offsets that put them all in context of the centre of it
[04:02:01] <Jak_o_Shadows> That I do understand
[04:02:35] <rue_shop3> for any 3 of them, there should only be 1 solution
[04:02:57] <Jak_o_Shadows> Yep yep
[04:03:22] <rue_shop3> maybe I'm missing something, but I think the other 3 are redundant for position calculation, you would need more than 3, (maybe just 4) for rotation
[04:04:01] <rue_shop3> but once you know the position, the rotations would solve quickly
[04:04:07] <Jak_o_Shadows> Yeah, for radius only, you only need 3
[04:04:54] <rue_shop3> the problem is that I'm inputting votlage and not freq
[04:05:26] <rue_shop3> so its almost more a matter of converting a variable freq to a pwm, which.... isn't hard...
[04:06:08] <rue_shop3> oh, but I want the freq stable...
[04:06:52] <rue_shop3> so I can go from FM to PFM
[17:20:34] <veverak> fuuu
[17:20:45] * veverak can't emulate rpi in virtual nah
[17:20:53] <veverak> propably bad qemu setup?
[17:58:36] <RiotingCherry> Hello
[18:00:24] <veverak> deshipu: btw: any progress with controlling servos from rpi?
[18:00:29] * veverak figured out he need 2 more servos
[23:46:58] <RyanS> is it problematic to mount encoders on a gearmoter when you have a 3:1 reduction (its an RC 4wd car with bevel gear diffs), backlash, efficiency loss etc?
