Current Server Time: 12:50 (Central Europe)

#apertus IRC Channel Logs

2019/04/28

Timezone: UTC


02:23
Bertl_oO
off to bed now ... have a good one everyone!
02:23
Bertl_oO
changed nick to: Bertl_zZ
03:08
apurvanandan[m]
aSobhy: https://github.com/Yayi98/8b10b_SERDES-Transceiver
03:09
apurvanandan[m]
Also, he has some more related repos
03:09
aSobhy
yeah thank you apurvanandan
03:11
aSobhy
:)
03:13
aSobhy
I lost the link and forgot how I reached it ^^fish memory^^
03:17
apurvanandan[m]
Np. You can fork these repos?
03:29
apurvanandan[m]
aSobhy: Can you share your proposal with me so that I can get to know your approach of the project.
04:03
aSobhy
apurvanandan[m]: sure :D here it is
04:03
aSobhy
https://docs.google.com/document/d/1gavuwFUW6bwWDHQzv-BLWdf-H1RRxQGo_mX8rc0xYB8/edit?usp=sharing
04:11
apurvanandan[m]
Thanks
04:56
aSobhy
URW
07:27
BAndiT1983|away
changed nick to: BAndiT1983
08:05
BAndiT1983
changed nick to: BAndiT1983|away
08:53
futarisIRCcloud
left the channel
09:17
Spirit532
left the channel
09:42
Spirit532
joined the channel
10:14
aman_
joined the channel
10:18
Y_G
joined the channel
10:19
aman_
left the channel
11:25
Bertl_zZ
changed nick to: Bertl
11:25
Bertl
morning folks!
11:27
apurvanandan[m]
Good morning Bertl
11:31
apurvanandan[m]
In continuation with previous talk, we will omit the metadata part of the 64-bit word coming from FIFO while sending to USB. So we can divide the 48 bits over 6 LVDS as 8 bits each.
11:32
Bertl
yes, that would be an optio
11:32
Bertl
*option
11:34
Nira|away
changed nick to: Nira
11:34
Bertl
note that for this to work, you need to regenerate the clock from the data streams
11:35
Bertl
(or find some other means to synchronize the data)
11:50
futarisIRCcloud
joined the channel
12:48
Nira
changed nick to: Nira|away
12:49
Nira|away
changed nick to: Nira
12:53
Nira
changed nick to: Nira|away
12:53
Nira|away
changed nick to: Nira
13:54
futarisIRCcloud
left the channel
14:32
Nira
changed nick to: Nira|away
14:33
Nira|away
changed nick to: Nira
14:52
Nira
changed nick to: Nira|away
14:53
Nira|away
changed nick to: Nira
15:11
lexano
left the channel
15:17
BAndiT1983|away
changed nick to: BAndiT1983
15:50
BAndiT1983
hi Y_G, please use main channel for gsoc related chats
15:51
BAndiT1983
which resources do you mean?
15:52
Y_G
Something for understanding sysfs better
15:53
Y_G
Currently I'm going through "https://embetronicx.com/tutorials/linux/device-drivers/sysfs-in-linux-kernel/" these
15:53
BAndiT1983
sysfs is the most simple type of data access in the system
15:54
Bertl
hmm, what kind of data access are we talking about?
15:54
BAndiT1983
it's like reading file data from file system
15:54
BAndiT1983
reading voltage and temperature from sysfs mappings
15:57
Bertl
sysfs entries are virtual filesystem entries ... so memory mapping might not work as expected
15:58
BAndiT1983
memory mapping?
15:59
BAndiT1983
my expectation is, that simple read() would be enough for this task
15:59
Bertl
maybe you are referring to something else with 'sysfs mappings'
16:00
Bertl
just wanted to clarify
16:00
BAndiT1983
mapping is not always related to memory mapping, but to simple placement of data, as the system grabs the data somewhere and places it in sysfs
16:01
Bertl
let's avoid confusing terminology then
16:02
BAndiT1983
not confusing to me at all
16:02
danieel
there is no data in sysfs, it is created on the fly
16:03
BAndiT1983
oh well, nitpicking continues
16:10
BAndiT1983
changed nick to: BAndiT1983|away
16:27
lexano
joined the channel
16:58
BAndiT1983|away
changed nick to: BAndiT1983
17:04
Nira
changed nick to: Nira|away
17:04
Nira|away
changed nick to: Nira
17:10
se6astian|away
changed nick to: se6astian
17:19
Nira
changed nick to: Nira|away
17:20
Nira|away
changed nick to: Nira
17:23
Nira
changed nick to: Nira|away
17:24
Nira|away
changed nick to: Nira
18:10
Nira
changed nick to: Nira|away
18:10
Nira|away
changed nick to: Nira
18:14
Nira
changed nick to: Nira|away
18:15
Nira|away
changed nick to: Nira
18:21
apurvanandan[m]
Bertl, we have data_clk at 160MHz, so can we give 5 times the clock ie 640MHz from the mmcme to the OSERDES for the serial side of output?
18:23
Bertl
we can do a lot with the MMCM and OSERDES, given that they are in the right place :)
18:24
apurvanandan[m]
This means the 48 bit words will be transfered in one 160Mhz cycle.
18:25
apurvanandan[m]
And you told earlier today that there are other options also.
18:25
Bertl
best make a simple diagram with bandwidths and clock rates
18:26
apurvanandan[m]
Ok , I will make it and share it here.
18:27
apurvanandan[m]
Bertl, do you want to tell some other ways of distributing the data over LVDS?
18:28
Bertl
well, here some options as I see it:
18:29
Bertl
- have a dedicated clock line to synchronize with (leaves 5 LVDS channels for data)
18:30
Bertl
- use a periodic coding to allow for clock recovery and dynamic synchronization
18:30
Bertl
- use an external clock from one of the routing fabrics to synchronize to
18:31
Bertl
and there are several options on the coding side, e.g. various 8/10 codings and the use of CRC/ECC bits
18:36
apurvanandan[m]
OK, I will research about these (clock recovery) and revert back to you to decide which method will I use.
18:36
Bertl
it's probably best to plan for testing them on the hardware
18:37
Bertl
there might be advantages or disadvantages we do not see atm
18:37
Bertl
note that this is also valid for aSobhy and the bidirectional packet protocol
18:39
apurvanandan[m]
You mean to implement all these methods and then choose the best one?
18:39
Bertl
at least to implement a test setup which can evaluate them
18:39
felix_
Bertl: i wrote two emails about the idea from november in vienna maybe 15 minutes ago; might be useful to forward that to apurvanandan[m] and aSobhy
18:40
Bertl
please do so
18:40
felix_
i don't have their mail addresses, so it would probably be easier if you forwarded those emails
18:41
Bertl
okay
18:41
apurvanandan[m]
felix_: I already PM my email to you.
18:42
felix_
would be also ok for me to paste that somewhere and link it here. i mean there's no secret stuff in there
18:42
felix_
yeah, i think that's better
18:43
felix_
https://paste.fedoraproject.org/paste/accJwB7GYeTJiN5J27yuzw
18:45
felix_
might also be a good idea to see if modifying the 64b/66b code to a lenght that fits better for the 10 bit color formats would keep the nice properties of the code
18:45
Bertl
forwarded as well, you are in CC
18:45
felix_
thx
18:59
apurvanandan[m]
felix_: Just to veriify, you plan to use 4 LVDS pairs for data, 1 pair for clock, and 1 pair for back communication?
19:00
Bertl
the SDI plugin is a dual plugin module, so it has 2x6 LVDS channels
19:01
apurvanandan[m]
ah, Okay
19:03
felix_
10 lvds pairs for data
19:03
felix_
4 pairs should still be sufficient for 1080p60
19:04
Bertl
also note that the MachXO2 is not capabale of the datarates we expect from the Artix 7
19:05
Bertl
(just as clarification for the USB 3.x plugin and the RF protocol)
19:09
apurvanandan[m]
Ok, will keep this in mind. Going to sleep, Good night.
19:09
Bertl
nn
19:15
BAndiT1983
changed nick to: BAndiT1983|away
19:32
felix_
yeah, my email was focused on the sdi module with an artix 7
19:36
BAndiT1983|away
changed nick to: BAndiT1983
19:40
Nira
changed nick to: Nira|away
19:41
Nira|away
changed nick to: Nira
19:51
Bertl
off for now ... bbl
19:51
Bertl
changed nick to: Bertl_oO
20:33
Nira
changed nick to: Nira|away
20:33
Nira|away
changed nick to: Nira
20:59
BAndiT1983
changed nick to: BAndiT1983|away
21:38
Y_G
left the channel
21:45
lexano
left the channel
21:48
lexano
joined the channel
22:15
se6astian
changed nick to: se6astian|away
23:31
futarisIRCcloud
joined the channel