00:38 | intracube | left the channel | |
02:11 | dmj_nova | left the channel | |
06:58 | ApertusWeb0 | joined the channel | |
06:58 | ApertusWeb0 | left the channel | |
07:57 | dmj_nova | joined the channel | |
08:12 | dmj_nova | Bertl: hmm...if you were interfacing with a sensor, what would you want that interface to be?
| |
08:53 | FergusL | dmj_nova: if I can allow myself to answer, the direct interface to the sensor varies
| |
08:53 | FergusL | it might be analog values to read directly or sometimes I2C
| |
08:55 | dmj_nova | FergusL: yeah, first time designing an interface for a sensor if you can't tell
| |
08:56 | FergusL | hehe
| |
08:57 | FergusL | I've been using Arduino for about a year, it was a great first step
| |
08:57 | FergusL | (though I already had learned quite a lot about electronics at school)
| |
08:58 | dmj_nova | I've had a fair bit of experience with other sorts of electronic stuff
| |
08:58 | dmj_nova | but it can be interesting when one realizes a gap in experience
| |
09:10 | dmj_nova | FergusL: probably a serial interface would be less of a pain to work with if you have many sensors
| |
09:11 | dmj_nova | and each sensor communicated a 10-12 bit code
| |
09:12 | FergusL | yes, definitely, but sensors don't generally have directly serial outputs, most of the time they're connected to a microcontroller which in turn process the signal and sends serial messages
| |
09:15 | dmj_nova | I suppose it's the same here
| |
09:15 | dmj_nova | 10 "digital" lines per sensor
| |
09:16 | dmj_nova | (okay the number is rather arbitrary and based on how long and how precise the encoder is)
| |
09:16 | FergusL | what is the sensor you're looking it ?
| |
09:16 | dmj_nova | Basically trying to figure out how best to make that useful
| |
09:17 | dmj_nova | FergusL: I'm building a new kind of linear encoder with absolute positioning via graycode
| |
09:18 | FergusL | okay, I see
| |
09:23 | FergusL | is it for something apertus-related or ?
| |
09:23 | dmj_nova | motion control camera rig related :)
| |
09:25 | FergusL | oh nice !
| |
09:26 | dmj_nova | I found that linear encoders are stupid expensive
| |
09:26 | dmj_nova | and most of them can't do absolute positioning so you've got to recalibrate every time you start it up
| |
09:29 | FergusL | yes, but that's what gray code fixes iirc
| |
09:29 | dmj_nova | yep
| |
09:29 | dmj_nova | which is why the graycode in my design is built into the physical nature of the sensor
| |
09:30 | FergusL | hm, some encoders do that
| |
09:30 | dmj_nova | yep, but totally not cheap
| |
09:31 | dmj_nova | (well, cheap by industrial robot standards, yes)
| |
09:31 | dmj_nova | but still like ZedBoard expensive
| |
09:35 | FergusL | ha yes...
| |
09:35 | FergusL | though the Zedboard isn't that expensive for what it is
| |
09:35 | FergusL | FPGA programming is *always* expensive anyway
| |
09:35 | dmj_nova | quite true
| |
09:35 | dmj_nova | but it's still a big chunk of change for an individual
| |
09:36 | dmj_nova | and when your design needs at least 3 of them, it adds up
| |
09:36 | dmj_nova | BTW, I'm trying to design a moco rig for indie film
| |
09:39 | dmj_nova | so cost is a big deal when you have lots of parts added together, and it would be rather silly for the encoders to double your rig cost
| |
09:40 | dmj_nova | heck, my stepper controllers are $35 bucks
| |
09:41 | FergusL | is it a 3 axis controlled head ?
| |
10:09 | dmj_nova | Well, at the moment the only thing that's been built is an arm
| |
10:10 | dmj_nova | but the intent is for something like a 6-axis rig
| |
10:10 | dmj_nova | so rotation and translation
| |
10:28 | FergusL | I see
| |
10:28 | FergusL | what range to you plan to achieve ?
| |
10:30 | dmj_nova | about a 6 foot cubed area?
| |
10:30 | dmj_nova | ish
| |
10:30 | dmj_nova | meaning about 6ft on each dimension
| |
10:30 | dmj_nova | perhaps more on one translation axis
| |
10:31 | dmj_nova | I'm hoping to be able to get repeatable "handheld" or "steadicam" type shots
| |
10:32 | dmj_nova | which should be good for having actual ground truth 3D coordinates for each frame (rather than having to spend hours and hours matchmoving in blender)
| |
10:34 | FergusL | that and more importantly the ability to shoot multiple passes
| |
10:38 | dmj_nova | FergusL: and potentially be able to film the same scene simultaneously in multiple locations
| |
10:39 | dmj_nova | (assuming one had the software to do real-time sync...which is really trivial)
| |
10:40 | FergusL | the usual technique is to shoot one or two passes at a location
| |
10:40 | FergusL | and then doing others on green screen for example
| |
10:41 | dmj_nova | yep, though for example you could film a location and two actors in separate locations and have the rig respond to both
| |
10:42 | dmj_nova | so you can play the actor's timing off each other
| |
10:42 | dmj_nova | but the bigger deal is enabling multiple passes and allowing freer use of vfx in smaller productions
| |
10:43 | dmj_nova | also getting background plates for compositing with
| |
10:44 | dmj_nova | often you can get away without much masking by diffing against the background plate, but that only works for tripod shots
| |
10:47 | FergusL | ToS has proven 3D tracking are okay, also it might have been on tripod and/or travelling rails only
| |
10:48 | dmj_nova | FergusL: ToS enabled us to do good tracking at all
| |
10:49 | dmj_nova | But, keep in mind that tracking still takes significant human effort and requires that the shot have good references to track in the first place
| |
10:49 | dmj_nova | I could give you plenty of shots that are untrackable
| |
10:51 | dmj_nova | and the tracking has to be done separately for every shot
| |
10:51 | dmj_nova | one tracking shot is fine, and entire scene...takes a lot of work
| |
10:52 | dmj_nova | whereas setting up a moco rig and tracking that way gets more efficient the more shots you need
| |
10:54 | FergusL | absolutely yes, that's what is used on bigger setups
| |
10:55 | dmj_nova | moco rigs are really expensive though
| |
10:55 | dmj_nova | not something an indie production could use for half their film
| |
10:55 | FergusL | yes
| |
10:55 | FergusL | and I think you can't avoid expensive sensors which are a crucial point
| |
10:56 | FergusL | or rather shouldn't !
| |
10:57 | dmj_nova | well, as long as you stay within the stepping capacity of the motors you don't need hardly any sensors for moving the rig accurately
| |
10:57 | dmj_nova | (ie, as long as you aren't skipping steps)
| |
10:57 | dmj_nova | the sensors are just an input mechanism
| |
10:58 | dmj_nova | so your operator can follow the actors rather than planning the shot out ahead of time in the computer
| |
10:58 | Bertl | morning everyone!
| |
10:59 | dmj_nova | morning Bertl
| |
10:59 | Bertl | dmj_nova: well, the sensor interface largely depends on what kind of data (amount, interval) you need to transport over what distance and how reliably
| |
11:00 | Bertl | there are quite a number of standartized bus systems for transferring data digitally (which usually is preferable over analog)
| |
11:01 | dmj_nova | yeah, and the sensor is pretty much inherently digital
| |
11:01 | Bertl | simple/cheap systems include IIC, one wire and SPI
| |
11:02 | dmj_nova | IIC == I2C?
| |
11:02 | Bertl | more expensive but also more reliable ones include rs-232, rs-485 and CAN
| |
11:02 | Bertl | yes I²C = I2C = IIC
| |
11:03 | dmj_nova | hmm...I wonder if mocobus would be a good idea
| |
11:03 | dmj_nova | seeing as how the sensor will necessarily be used with mocobus steppers
| |
11:05 | Bertl | it seems to be a modified rs-485+ttl system (never heard of mocobus before_
| |
11:05 | Bertl | s/_/)
| |
11:05 | dmj_nova | mocobus is what's used by the nanomoco stepper controllers made by dynamic perception
| |
11:06 | dmj_nova | The Axiom is supposed to eventually have a mocobus interface.
| |
11:07 | Bertl | yeah, well, I do not see a problem with that interface per se, no idea why I would want a mixed bus though
| |
11:08 | dmj_nova | you mean mixed with both sensors and motors?
| |
11:09 | Bertl | no, I mean rs-485 and ttl (for whatever purpose)
| |
11:09 | dmj_nova | ah
| |
11:09 | Bertl | because that will severely limit the usability IMHO
| |
11:10 | dmj_nova | how so?
| |
11:11 | Bertl | rs-485 is a standard defining the electrical characteristics of a serial connection
| |
11:12 | Bertl | termination, voltages, etc, but not the actual protocol
| |
11:12 | Bertl | (so you can do whatever you like with it)
| |
11:12 | Bertl | and rs-485 can be used over quite a distance
| |
11:13 | Bertl | 35MBit/s over 10m, 100kBit/s over 1km
| |
11:13 | Bertl | now TTL levels on the other hand will not be reliable over more than a few meters at most
| |
11:14 | Bertl | (their level will fall below anything which can be recognized on the other end)
| |
11:14 | Bertl | they also suffer from EMI and offset voltage issues
| |
11:14 | Bertl | (they are not differential signals)
| |
11:16 | Bertl | so, at first glance that looks to me like you are building a high security transport vehicle but you decide to send half of the data via pigeons
| |
11:17 | Bertl | but maybe I'm just misinterpreting the design, haven't found a conclusive documentation
| |
11:18 | dmj_nova | could be, I haven't tested mocobus with anything longer than a couple meters
| |
11:20 | dmj_nova | seems to used twisted cat5 cable for all the connections
| |
11:21 | dmj_nova | I could ask Chris about the TTL thing though
| |
11:26 | dmj_nova1 | joined the channel | |
11:26 | dmj_nova1 | hooray, I just had a power failure
| |
11:26 | dmj_nova | left the channel | |
11:26 | dmj_nova1 | changed nick to: dmj_nova
| |
11:27 | FergusL | hu
| |
11:48 | dmj_nova | I suppose I'm rather limited to the constraints of mocobus anyway :/
| |
11:49 | dmj_nova | well, unless I also build my own motor controllers
| |
11:49 | dmj_nova | which would be a huge waste of effort and a giant fuck you to dynamic perception
| |
11:53 | dmj_nova | might be able to design it so it can be used as regular RS-485 except when interfacing multiple devices on a single bus
| |
11:57 | christiankrupa | left the channel | |
11:59 | christiankrupa | joined the channel | |
12:35 | Bertl | rs-485 is not limited to 2 endpoints
| |
14:09 | jucar2 | left the channel | |
14:23 | jucar1 | joined the channel | |
15:36 | intracube | joined the channel | |
16:08 | jucar1 | left the channel | |
16:21 | jucar1 | joined the channel | |
16:28 | jucar1 | left the channel | |
16:45 | jucar1 | joined the channel | |
18:22 | dmj_nova | Bertl: there must've been some good reason for the TTL logic
| |
18:23 | Bertl | maybe, if you feel like, you might ask the developers :)
| |
18:24 | dmj_nova | I might do that when he gets back to me with my last set of questions
| |
18:24 | Bertl | no doubt there has been a reason for adding it, not sure it was a good one though :)
| |
18:25 | dmj_nova | so far their engineering has seemed sound to me (though everyone makes mistakes)
| |
18:26 | dmj_nova | the stepper controller is quite good, and the software for it is also well-done (though it didn't work on Linux until I got my hands on it)
| |
20:09 | jucar1 | left the channel | |
20:22 | jucar1 | joined the channel |