Current Server Time: 23:10 (Central Europe)

#apertus IRC Channel Logs

2020/07/29

Timezone: UTC


01:58
karl[m]
left the channel
01:58
holger[m]
left the channel
01:58
apurvanandan[m]
left the channel
01:58
bluez_[m]
left the channel
02:01
holger[m]
joined the channel
02:01
karl[m]
joined the channel
02:01
apurvanandan[m]
joined the channel
02:01
bluez_[m]
joined the channel
02:03
hans[m]1
left the channel
02:03
fabian[m]1
left the channel
02:03
elkos
left the channel
02:03
pedro[m]
left the channel
02:03
francois[m]
left the channel
02:03
gretax[m]
left the channel
02:03
robert[m]1
left the channel
02:03
oscar[m]
left the channel
02:03
apurvanandan[m]
left the channel
02:03
holger[m]
left the channel
02:03
karl[m]
left the channel
02:03
bluez_[m]
left the channel
02:03
Guest93381
left the channel
02:03
promach3
left the channel
02:03
joe[m]1
left the channel
02:03
panintended
left the channel
02:03
pierre[m]5
left the channel
02:03
davidak[m]
left the channel
02:03
preetimenghwani[
left the channel
02:03
alexandros[m]
left the channel
02:03
metal_dent[m]
left the channel
02:03
MarkVandenBorre[
left the channel
02:03
john_the_savage[
left the channel
02:06
WalterZimmermann
joined the channel
02:10
apurvanandan[m]
joined the channel
02:11
joe[m]
joined the channel
02:11
davidak[m]
joined the channel
02:11
karl[m]
joined the channel
02:11
preetimenghwani[
joined the channel
02:11
aleb
joined the channel
02:11
oscar[m]
joined the channel
02:11
elkos
joined the channel
02:11
alexandros[m]
joined the channel
02:11
john_the_savage[
joined the channel
02:11
robert[m]
joined the channel
02:11
bluez_[m]
joined the channel
02:11
holger[m]
joined the channel
02:11
francois[m]
joined the channel
02:11
pedro[m]2
joined the channel
02:11
metal_dent[m]
joined the channel
02:11
pierre[m]
joined the channel
02:11
fabian[m]1
joined the channel
02:11
hans[m]2
joined the channel
02:11
panintended
joined the channel
02:11
gretax[m]
joined the channel
02:42
promach3
joined the channel
02:42
MarkVandenBorre[
joined the channel
05:06
intracube
left the channel
05:07
intracube
joined the channel
06:08
BAndiT1983|away
changed nick to: BAndiT1983
06:52
flyinprogrammer
joined the channel
08:13
flyinprogrammer
left the channel
08:13
flyinprogrammer
joined the channel
08:46
kbeckmann
joined the channel
09:31
kbeckmann
se6ast1an: hi! about the ft60x and throughput.. the stuff that i've done successfully with real-time data and ft600 has been working great up to around ~70 MB/s. Above that, I really need a larger FIFO.
09:35
flyinprogrammer
left the channel
09:50
kbeckmann
if i generate data from the fpga, i can stream at 194MB/s, however there will be a lot of random, but short, delays which is why a larger FIFO is needed when you want to stream real-time data.
09:54
mumptai
joined the channel
10:00
WalterZimmermann
left the channel
10:06
se6ast1an
hi kbeckmann!
10:06
se6ast1an
you are very quick :)
10:06
se6ast1an
too quick for me
10:06
se6ast1an
but now I am back
10:06
kbeckmann
haha i just had some spare time now.
10:08
se6ast1an
what transfer mode are you using with USB
10:08
kbeckmann
245
10:09
se6ast1an
we want to get to the USB3 bandwidth limit so 70MB/s or even 194MB/s is still not really what we aim for
10:09
kbeckmann
i have not tried to use the fancier mode, it seems a bit complicated. but makes sense if you are pumping data in both directions simultaneously.
10:09
kbeckmann
well the ft600 has a theoretical max of 200MB/s
10:10
kbeckmann
it has only 16 bits in contrast to tf601/tf602 that has 32 :)
10:10
kbeckmann
ft*
10:10
kbeckmann
so 194 / 200 is probably as good as it gets.
10:10
se6ast1an
ah, we use the FT601
10:11
se6ast1an
but Isochronous transfer is not supported from what we gathered so far IIRC
10:12
kbeckmann
right. and bulk will always have some random delays/jittar to it.
10:13
kbeckmann
i don't know how the ft602 works, but my guess is that it's just ft601 with some extra descriptor glue and some more configuration functionality.
10:13
kbeckmann
i'm not even sure if you can have isochronous mode at these speeds.
10:17
kbeckmann
honestly I'm not sure what is causing the random jitter. it could be the driver. i should try again with this driver that is a kernel module, it might work better actually. https://github.com/lambdaconcept/ft60x_driver
10:20
kbeckmann
se6ast1an: what data rates were you able to get with your ft601 adapter? were you limited by your fifo filling up and having to drop input data, or were you not able to reach a high data rate at all when sending generated data from the fpga?
10:22
se6ast1an
we need to wait for Bertl_zZ, he will be able to share details
10:22
kbeckmann
if you were having issues with your gateware, you could look at https://github.com/lambdaconcept/usbsniffer/blob/master/gateware/ft601.py for a good implementation.
10:22
kbeckmann
cool
10:23
se6ast1an
the large fifo should not be an issue, the fpga side will very likely also not be the bottleneck
10:23
kbeckmann
which fpga are you using again?
10:24
se6ast1an
from what I remember Bertl_zZ said the FT601 had eratic behaviour and higher datarates worked only unreliably...
10:25
se6ast1an
Lattice MachXO
10:25
se6ast1an
on the usb3 module
10:25
se6ast1an
zynq 7020 on the camera core side
10:26
kbeckmann
alright
10:26
kbeckmann
looks like lcmxo2-2000hc-6tg100c from the photo
10:30
kbeckmann
looks like it doesn't have a huge amount of embedded ram. are you able to buffer the frame on the camera module and consume the framebuffer asynchronously? in my case i am receiving a HDMI stream that doesn't wait for anybody, so I have to buffer as much as possible (because the ft60x will randomly stall because of how USB works).
10:34
kbeckmann
sorry for all the questions.. i don't know a lot about your architecture, should read up on that :)
10:36
se6ast1an
no worries
10:36
se6ast1an
but I am afraid we have to wait for Bertl_zZ to wake up :)
10:36
se6ast1an
for these technical details
10:56
Bertl_zZ
changed nick to: Bertl
10:56
Bertl
morning folks!
10:59
Bertl
kbeckmann: we basically retrieve the data from DDR memory on the camera side and send it directly to the MachXO2
10:59
Bertl
so buffering/feedback is possible, but we need a 'guaranteed' throughput as we are transporting live data as well
11:01
Bertl
kbeckmann: why should 'USB' randomly stall on a dedicated USB connection of good quality?
11:27
Bertl
off for now ... bbl
11:27
Bertl
changed nick to: Bertl_oO
11:28
kbeckmann
I see, that's great. This means that even if the transfers stall for short periods of time, this should be fine. As far as I understand it, bulk transfers don't have any guarantee of when they will occur, so your OS might handle some interrupt and not do another USB transfer for a while.
11:50
kbeckmann
Bertl_oO: something that i also noticed, was that it's really important to transfer large chunks of data to the FT60x. The datasheet mentions this as well. If you transfer less than 1024 bytes, this will lead to a small USB bulk transfer which will lead to increased overhead which in turn will lower your throughput. Maybe this is what happened to you? I worked around this by using ping-pong
11:50
kbeckmann
buffers and only initiate a transfer when i had a full 4096 bytes to transfer.
11:58
Nick123
joined the channel
12:25
Nick123
left the channel
12:34
kbeckmann
It seems that the USB spec allows for isochronous transfers utilizing up to 80% of the available bandwidth, which for SuperSpeed means arond 500 MB/s. However your device still needs to support and configure this, which the FT60x doesn't seem to do, not even the ft602 that uses the UVC (USB Video Class). So you'd have to go for the Cypress FX3 or similar if you want this.
12:51
Bertl_oO
well, the OS (on the USB controller side) is Linux, so I think it will not just ignore USB transfers because some interrupt is handled ... at least we get around 420MB/s transfer via USB storage which also uses block transfers
12:53
Bertl_oO
the FT60x is limited to 32bit so the maximum data rate we can produce there is 3.2 Gigabit/s but I do not see why that should not be doable with a 5Gbit/s connection
12:54
Bertl_oO
and yes, we figured out (too late) that the FT60x (not even the FT602) officially support isochronous transfers ... which seems especially strange as the FT602 is intended for UVC video
12:56
Bertl_oO
the main problem so far seems to be to figure out the perfect package size as the payload on the bus is limited and to find a driver which is not based on the FTDI libraries which seem to do an awful job at receiving the data reliably :)
12:57
Bertl_oO
so we are very much interested in alternative/known good solutions which can unlock the full potential of the FT60x chips
13:24
se6ast1an
and yes we are also having our eyes on the FX3 now
13:57
kbeckmann
Bertl_oO: regarding drivers, it might be worth checking out https://github.com/lambdaconcept/ft60x_driver
14:22
RexOrMatrix[m]
joined the channel
14:22
RexOrMatrix[m]
Is the imec Temp Cam on schedule?
14:39
oscar[m]
left the channel
14:40
se6ast1an
RexOrMatrix[m]: wrong channel probably but yes
14:42
BAndiT1983
changed nick to: BAndiT1983|away
14:50
flyinprogrammer
joined the channel
15:20
MichaelSchwarzma
joined the channel
15:22
BAndiT1983|away
changed nick to: BAndiT1983
15:33
BAndiT1983
changed nick to: BAndiT1983|away
15:33
BAndiT1983|away
changed nick to: BAndiT1983
17:41
flyinprogrammer
left the channel
18:58
MichaelSchwarzma
left the channel
19:29
BAndiT1983
changed nick to: BAndiT1983|away
20:06
BAndiT1983|away
changed nick to: BAndiT1983
21:39
Bertl_oO
off to bed now ... have a good one everyone!
21:40
Bertl_oO
changed nick to: Bertl_zZ
21:49
panintended
Hi all
21:49
panintended
Here's a PR for T1203 (the observer pattern for subscribing to attribute changes)
21:50
panintended
https://github.com/apertus-open-source-cinema/AXIOM-Remote/pull/20
21:50
panintended
Sorry it took a while, but today I had finally some breathing room with my day job :)
21:51
panintended
Still no tests to go with it, but se6ast1an said it might still be worth merging it to the main repo
21:53
se6ast1an
thanks!
21:53
BAndiT1983
hi panintended, just merged, thanks for your support!
21:53
BAndiT1983
will go through it soon, maybe tomorrow
21:56
se6ast1an
great
22:05
panintended
cool, any questions please let me know and I'll get back to you the soonest possible!
22:05
panintended
have a good night guys
22:05
panintended
talk to you soon
22:28
BAndiT1983
changed nick to: BAndiT1983|away
00:20
mumptai
left the channel