Current Server Time: 23:15 (Central Europe)

#apertus IRC Channel Logs

2023/01/02

Timezone: UTC


01:25
anuejn
ah i see
01:26
anuejn
that again depends on the routing machxo fpgas, i presume?
03:25
Bertl
yup
07:46
illwieckz
left the channel
07:57
illwieckz
joined the channel
08:07
illwieckz
left the channel
08:19
illwieckz
joined the channel
08:39
se6astian
what exactly needs to be done on the XO2? just connecting the right in- and outputs?
12:43
Bertl
well, that would work to some degree but probably disable the sensor as the few available connections to the Zynq are already in use
12:45
Bertl
so what would be required is to handle the protocol (IIC, SPI) on the MachXO2 and somehow communicate the data to the Zynq (basically T731)
16:23
se6astian
ah, I see
17:00
se6astian
MEETING TIME, welcome to 2023, anyone around ?
17:00
Bertl
is here
17:00
se6astian
great, how are things Bertl on your end?
17:00
Bertl
pretty good so far for 2023 :)
17:01
Bertl
nothing unusual or terribly new to report yet though
17:02
se6astian
regarding the IMU CSO, it sounds like the lack of IO to ZYNQ dependency is a pretty big showstopper considering T731 has been approached a few times without real solution IIRC, is there a workaround with reduced bandwidth or some other idea how we could get the IMU data without having to jump through a lot of hoops?
17:03
Bertl
I think a quick hack solution should be doable with a fixed slot cyclic exchange
17:04
Bertl
i.e. a fixed size shared circular buffer between Zynq and MachXO
17:04
Bertl
where each bit has a well defined meaning and the buffer size is just large enough to accomodate the required functionality
17:05
se6astian
sounds interesting, would that have any drawbacks?
17:05
Bertl
probably a little trickier on the side where we only have one LVDS
17:06
Bertl
well, it would be constantly active and thus consume some power and it cannot accomodate to any changes (except for re-building the gateware for both ends)
17:07
Bertl
but I think it should work
17:07
se6astian
that sounds reasonable, will it mean a different gateware is required if no CSO is soldered on?
17:08
Bertl
probably not, but it would need to have the correct gateware on the MachXO2s
17:09
se6astian
right, that sounds like a plan
17:09
Bertl
another 'quick' alternative might be to run something like I2C (at low speeds) directly over the Zynq/MachXO2 connection
17:09
se6astian
vup/anuejn would any of this be interesting for you to work on?
17:09
Bertl
and implement the GPIO functionality as I2C slave in the MachXO2
17:10
Bertl
that would probably increase the latency somewhat but might be simpler to do
17:10
Bertl
(and with somewhat I mean significantly ;)
17:12
se6astian
latency from IMU to ZYNQ?
17:13
Bertl
latency in both directions especially for the GPIO part
17:14
se6astian
as I currently see no application where the IMU data needs to control a motor or similar and we just want to dump the motion data I guess a latency is no issue as long as we know the delay or get a timestamp so we can match footage and motion
17:14
se6astian
but maybe we also need to talk about how/where to store the motiond data?
17:15
Bertl
yes, the data could be collected on the MachXO2 side and reported back periodically
17:16
se6astian
looking at the csv format: https://docs.gyroflow.xyz/logging/gcsv/#gcsv-gyro-log-format
17:16
se6astian
the timing seems to just be listed in ms (tscale,0.001) as first parameter in the list then
17:17
se6astian
the machxo2 would get the data from the motion sensors with a fixed clk rate I assume?
17:18
se6astian
and forward that data to the Zynq also with a fixed clk rate so the zynq would need to supply the timestamp then right?
17:18
Bertl
the data rate depends on the IMU settings and how/when the data is retrieved
17:18
Bertl
timestamping depends on the transport mechanism used
17:19
Bertl
in the fixed buffer case we can assume that data arrives at precise moments based on the Zynq clock
17:19
Bertl
in the I2C case, we probably want to do the timestamping on the MachXO2 side and transfer the timestamp as well
17:19
se6astian
right
17:20
Bertl
although that would complicate things a lot compared to just a simple I2C pass through
17:20
se6astian
anyway, would be a good way forward, lets see if we can interest anyone in working on this, then we can go into more detail
17:20
Bertl
sure
17:21
Bertl
do we have anybody to do some actual testing with the IMU?
17:21
Bertl
because it might be pointless to implement communication and then not use the data for anything ;)
17:22
se6astian
well there is the https://gyroflow.xyz/ way to stabilize footage
17:23
se6astian
with the data
17:23
Bertl
yeah, but that kind-of presumes that the IMU is configured and the data retrieved is valid/good
17:24
Bertl
what I meant was, there is certainly some testing involved to make sure that the IMU is sending usable data ;)
17:24
se6astian
true, how could we verify the data at an earlier development stage?
17:24
se6astian
controlled motion with a gimbal?
17:25
Bertl
yeah, something like this is probably a good start
17:25
se6astian
I should be able to assist on that front
17:25
Bertl
have some servos/motors to move the camera in a predictable way
17:25
Bertl
then do some (data) recordings and see how that matches
17:26
Bertl
idealy some comparison device would be great, e.g. a pi-zero with an IMU or so
17:26
se6astian
good idea
17:27
Bertl
btw, that might also be a quick way to add this functionality without the IMU CSO if we only want to test whether this works or what's required setting wise
17:27
Bertl
(especially if we use a similar/identical IMU on the pi)
17:28
se6astian
I guess most use the same LSM6DS3H chip yes
17:28
se6astian
or the replacement
17:29
se6astian
https://www.sparkfun.com/products/18020
17:30
Bertl
we might even be able to simply rewire the IMU CSO to the PMOD port at the back of the camera for testing
17:32
Bertl
that is, if the main goal is to test the IMU and the usability of the collected data
17:32
Bertl
have to check schematics, but I remember that we used the I2C paddles on the camera itself, so that might work
17:34
se6astian
sounds good, lets see if this topic picks up some momentum
17:34
Bertl
yep
17:34
se6astian
anything else for today? or anyone else around?
17:35
Bertl
nothing from my side for today
17:37
se6astian
ok then! many thanks! MEETING CONCLUDED!
17:37
Bertl
thanks for moderating!
17:39
se6astian
my pleasure!
20:01
Chooks
joined the channel
20:01
Chooks
Unfortunately I do not have the knowledge to help you on the subject.
20:01
Chooks
Gyroflow has also developed an external solution for cameras without an imu (steadXP style).
20:01
Chooks
In case it doesn't work there is always this solution to integrate it on top of the camera. But let's say it's stupid if an IMU is present inside.
20:02
Chooks
https://github.com/gyroflow/flowshutter
20:02
Chooks
On our side, we will be able to start working again with you on the enclosure. We plan to pour a slab in my cellar in a few weeks to install a fablab.
20:02
Chooks
We have all the parts to install a CNC for aluminum and we will be able to help you with the protoyping of parts within 2-3 months.
20:03
Chooks
 It could be interesting to know if the enclosure can be machined on a 3-axis cnc and have your feedback from the first machining if you were present on site.
21:50
Bertl
Chooks: yay! sounds great!
22:28
se6astian
Hi guillaume, great!
22:29
se6astian
all cnc parts can be done with 3 axis, I specifically designed the parts to be "simple" and sturdy and functional
22:30
se6astian
lets have a chat about this soon as I contacted the cnc manufacturing plant for a small volume batch for early adopters just days ago
23:44
Chooks
left the channel