Current Server Time: 04:30 (Central Europe)

#apertus IRC Channel Logs

2019/08/01

Timezone: UTC


02:04
Fares
joined the channel
02:15
Fares
left the channel
02:15
Fares
joined the channel
02:20
Nira
changed nick to: Nira|away
02:33
Fares
left the channel
04:44
BAndiT1983|away
changed nick to: BAndiT1983
06:02
BAndiT1983
changed nick to: BAndiT1983|away
06:07
Bertl_oO
off to bed now ... have a good one everyone!
06:07
Bertl_oO
changed nick to: Bertl_zZ
06:07
apurvanandan[m]
Good night
06:41
apurvanandan[m]
Bertl, if we want to generate a divided clock like from counter which is going to drive large number of elements, so how can we improve its routing?
07:58
Y_G_
joined the channel
08:33
apurvanandan[m]
\o/
08:34
apurvanandan[m]
Bertl, I succeeded! The DDRX1 primitive worked like wonders
08:35
apurvanandan[m]
I did word alignment and 2 to 10 gearing myself using small VHDL code and now I am able to receive the counter data perfectly!
08:36
apurvanandan[m]
Here is the dump if you want to see : https://pastebin.com/g7hH33u8
08:36
apurvanandan[m]
First two columns are unnconnected, second last is the received byte from Virtex-5 and last one is generated byte in machXO2
08:37
apurvanandan[m]
And there is a constant difference of 1 between them
08:37
apurvanandan[m]
Now I will move to ber measurement :D
09:39
apurvanandan[m]
And I checked the equality of received and generated bytes on my pc using C code and for 200MBytes there wasn't any unequal value
09:40
apurvanandan[m]
Actually after that I didn't check
10:35
shebin_joseph
joined the channel
10:49
shebin_joseph
left the channel
11:33
Nira|away
changed nick to: Nira
12:09
Bertl_zZ
changed nick to: Bertl
12:09
Bertl
morning folks!
12:25
Nira
changed nick to: Nira|away
12:54
ranga_swami
joined the channel
12:55
ranga_swami
hey!! I am an open source entho!!
12:57
ranga_swami
just wanted to know, are you guys deciding to participate this year in google code in?
12:59
Bertl
hello ranga_swami!
13:00
Bertl
we haven't considered code in yet, but we are doing GSoC
13:01
ranga_swami
hey Bertl, I am in school so I can only do GCI .
13:06
Bertl
yeah, of course ...
13:17
Nira|away
changed nick to: Nira
13:33
ranga_swami
left the channel
15:19
Y_G_
left the channel
15:35
Nira
changed nick to: Nira|away
16:13
BAndiT1983|away
changed nick to: BAndiT1983
16:41
aombk2
left the channel
17:41
aombk
joined the channel
17:45
aombk
left the channel
17:53
aombk
joined the channel
18:44
apurvanandan[m]
Bertl we will also need some gearing from software side on PC for 32 to 48 gearing?
18:49
BAndiT1983
changed nick to: BAndiT1983|away
18:51
Bertl
yup
19:47
BAndiT1983|away
changed nick to: BAndiT1983
20:27
Nira|away
changed nick to: Nira
21:15
apurvanandan[m]
Bertl: I tried transfers at 350 MHz and 400MHz, at 350 MHz BER increased to 1.5% and at 400 MHz all data was corrupt
21:17
Bertl
so that won't work for reliable transfers
21:17
apurvanandan[m]
Yup
21:17
danieel
its 350 mbit? or x2?
21:17
apurvanandan[m]
What should I do with UDDCNTLN signal
21:17
apurvanandan[m]
x2 danieel
21:18
danieel
so 700 ~ 800.. the wire (connection) is properly made?
21:18
apurvanandan[m]
It is for PVT variation compensation, so I turn off PVT compensation when link is trained and I freeze the DLL
21:18
apurvanandan[m]
Is that the right way?
21:19
apurvanandan[m]
I used dupont cables danieel
21:20
apurvanandan[m]
Plus the machXO2 has a limit at 378 MHz
21:20
Bertl
so yes, that is a valid point, it might just be above the connection
21:20
apurvanandan[m]
Please plug USB module on remote beta Bertl
21:20
danieel
one thing would also tell a bit more - the BER is same for both directions, or one behaves much better?
21:21
apurvanandan[m]
I have implemented unidirectinal transfer only
21:21
Bertl
apurvanandan[m]: I'm pretty sure right now nobody is at the hub .. but if you provide me with bitfiles for the Beta, I can test it at home
21:22
apurvanandan[m]
I can't right now as code is for Virtex-5
21:22
apurvanandan[m]
It will take time
21:22
apurvanandan[m]
But tommorow will be fine
21:23
danieel
and the ddrx1 is used in this?
21:23
apurvanandan[m]
Yup
21:24
danieel
that is up to 300 or 250mhz only (-6 and -5) xo2 speedgrades
21:24
danieel
mbps!!
21:24
apurvanandan[m]
But later I combine words to 10 bits encoded words
21:24
danieel
*xo3
21:25
Bertl
yes, it might be interesting to see if DDR x2 can get higher rates or bails out at the same point with similar error rates
21:26
apurvanandan[m]
Ok I will try with that also
21:27
danieel
what speedgrade is the xo2?
21:30
RexOrMatrix[m]
Note: I'm not sure there's a USB module at the Hub anyways.
21:30
danieel
the XO2 supported speeds are in DS1035 page 3-19 ... and for 700mbit+ you need DDR4 and -6 part
21:31
danieel
i wonder how your design could pass timing checks :)
21:33
Bertl
RexOrMatrix[m]: there should be two, one with FT601 and one with FT602
21:33
RexOrMatrix[m]
Noted.
21:47
illwieckz
joined the channel
21:55
apurvanandan[m]
I used DDR x4 earlier, but there were lot in error in the received and it was sort of blackbox, there was no way to debug word alignment by the primitive
21:55
apurvanandan[m]
@ danieel
21:56
Bertl
but you might want to check your constraints in case you didn't get any timing errors (at least for 400MHz)
21:56
Bertl
because as danieel mentioned, you should get a lot of warnings/error there
22:02
BAndiT1983
changed nick to: BAndiT1983|away
22:05
danieel
apurvanandan[m]: these "phy" serdeses do not deal much with word alignment
22:06
danieel
its just the 4 consecutive bits, and then another 4... if you have some framing, you have to do it in logic
22:06
danieel
8b10b is only hardwired in multi-gigabit transcievers
22:07
apurvanandan[m]
I am not using transceivers
22:08
illwieckz_
joined the channel
22:11
apurvanandan[m]
I am using the primitve IDDRX4B which does the word alignment using a bit slip input
22:12
illwieckz
left the channel
22:13
apurvanandan[m]
danieel: The speedgrade is -6
22:19
apurvanandan[m]
Alignment signal is ALIGNWD (mentioned in DS1035)
22:31
uberardy_
joined the channel
22:36
apurvanandan[m]
Thanks danieel for telling about DS1035, I didn't knew that so much description was there about DDR primitives
22:37
uberardy
left the channel
22:37
apurvanandan[m]
aSobhy, you should have a look over DS1035
22:41
apurvanandan[m]
Bertl, what do we do when we want to generate exact 378 MHz from 100 MHz clock by PLL?
22:42
Bertl
find a common clock rate and adjust the PLL, but I doubt that the MachXO2 PLL can do that
22:42
apurvanandan[m]
No I meant for Virtex-5
22:43
Bertl
ah, well, have to check the MMCM/PLL for the Virtex-5
22:47
apurvanandan[m]
Actually VCO freq has a limit between 400 MHz to 1GHz, and we need to multiply by 15 and divide by 4 the 100MHz clock
22:47
apurvanandan[m]
For generating 375 MHz
22:48
apurvanandan[m]
So VCO freq can't be 1.5GHHz
22:49
Bertl
got a link to the virtex 5 PLL/DCM documentation?
22:50
apurvanandan[m]
https://github.com/apurvanandan1997/Systems_Reading_Material/blob/master/XUPV5%20Virtex-5%20Documentation/ug190.pdf
22:50
apurvanandan[m]
This is the user guide
22:54
Bertl
and where does it explain the DCM?
22:55
apurvanandan[m]
From page 89
22:55
apurvanandan[m]
Chpater PLL
22:55
Bertl
ah, under PLL, tx
22:59
Bertl
so you probably need to chain two PLLs to create something in the correct range
22:59
apurvanandan[m]
:(
22:59
apurvanandan[m]
ok
23:05
apurvanandan[m]
So first multiply by 3 in first PLL, then multiply by 5 and divide by 4 in next PLL
23:05
Bertl
or 3/2 and 5/2
23:06
apurvanandan[m]
great
23:09
apurvanandan[m]
'not of locked' signal of first PLL goes to second as Reset?
23:15
Bertl
why not
23:15
apurvanandan[m]
Just comfirming
23:29
apurvanandan[m]
Should I stop the delay updation based on PVT variation when link is trained?
23:29
apurvanandan[m]
Or should I keep updating?
23:38
Bertl
PVT can change over time, so ideally you want to update continuously, not sure how well that works on the MachXO2 though
23:39
apurvanandan[m]
Ok great
23:45
apurvanandan[m]
And what about freezing DLL after link trained?