23:04 | gwelkind | left the channel | |
23:08 | gwelkind | joined the channel | |
23:59 | gwelkind | left the channel | |
00:14 | bootstrap | joined the channel | |
00:29 | rexbron | left the channel | |
00:29 | rexbron | joined the channel | |
00:29 | rexbron | left the channel | |
00:29 | rexbron | joined the channel | |
01:02 | gwelkind | joined the channel | |
03:50 | gwelkind | left the channel | |
05:42 | Bertl | off to bed now ... have a good one everyone!
| |
06:08 | bootstrap | left the channel | |
06:52 | philippej | joined the channel | |
07:47 | jucar | joined the channel | |
08:12 | fabrizio1 | joined the channel | |
08:13 | fabrizio1 | Hi guys, just testing the client
| |
08:14 | fabrizio1 | left the channel | |
11:15 | rexbron_ | joined the channel | |
11:16 | rexbron | left the channel | |
11:38 | rexbron_ | left the channel | |
11:38 | rexbron | joined the channel | |
11:38 | rexbron | left the channel | |
11:38 | rexbron | joined the channel | |
11:43 | rexbron_ | joined the channel | |
11:45 | rexbron | left the channel | |
13:20 | philippej | left the channel | |
14:50 | philippej | joined the channel | |
15:34 | philippej | left the channel | |
15:44 | Bertl | morning everyone!
| |
16:05 | philippej | joined the channel | |
16:39 | se6astian | joined the channel | |
16:40 | se6astian | good evening
| |
16:42 | philippej | hello !
| |
16:43 | philippej | anything we need to prepare?
| |
16:43 | se6astian | I need to start my laundry but should be back just before the meeting :)
| |
16:43 | se6astian | prepare, not really I d say
| |
16:43 | se6astian | as we have no clue about FPGAs :)
| |
16:43 | philippej | :-)
| |
16:50 | fabrizio1 | joined the channel | |
16:51 | fabrizio1 | hello guys
| |
16:52 | philippej | Hello fabrizio1
| |
16:52 | se6astian | h there
| |
16:52 | se6astian | hi there
| |
16:53 | fabrizio1 | Hi Sebastian
| |
16:53 | fabrizio1 | Hi Philip
| |
16:54 | fabrizio1 | I guess we will need to wait a little for Dave
| |
16:54 | Bertl | hello fabrizio1!
| |
16:56 | fabrizio1 | hi Bert
| |
17:00 | devbisme_ | joined the channel | |
17:01 | Bertl | welcome devbisme_!
| |
17:02 | se6astian | hi!
| |
17:02 | se6astian | I think we are complete then
| |
17:03 | fabrizio1 | Hi Dave, how are you?
| |
17:03 | se6astian | is it OK for you if we continue here on the public channel or should we switch to a private one?
| |
17:03 | devbisme_ | Hi. Sorry for the delay in responding. Getting used to the web interface.
| |
17:04 | devbisme_ | Public is fine with me.
| |
17:04 | fabrizio1 | public is great
| |
17:04 | se6astian | great!
| |
17:05 | se6astian | so since we already covered the basics via email lets jump right into the core of what we wanted to talk about
| |
17:05 | devbisme_ | ok.
| |
17:05 | fabrizio1 | ok
| |
17:06 | se6astian | We want to get closer to finalizing our technical specifications for our FPGA module in the Axiom camera
| |
17:06 | se6astian | that you plan to develop something similar would be a great synergy!
| |
17:06 | se6astian | we could split development tasks and production costs
| |
17:07 | fabrizio1 | sounds god
| |
17:07 | fabrizio1 | good
| |
17:07 | se6astian | currently we have no money though, just to be clear :)
| |
17:07 | fabrizio1 | haha, no problem
| |
17:07 | devbisme_ | We want to develop a low-cost Zynq module. We thought Axiom might serve as a possible application for it.
| |
17:08 | se6astian | but we are preparing a crowd funding campaign to presell a good amount of axiom cameras soon
| |
17:08 | se6astian | and we are prepating the EU grant which you might have seen on the website
| |
17:09 | devbisme_ | What price range do you see for the Axiom?
| |
17:09 | se6astian | if that works out we will have the money to cover such developments
| |
17:09 | se6astian | our motto is "below 10.000$"
| |
17:09 | devbisme_ | So you're satisfied if it is $9.999?
| |
17:10 | se6astian | no :)
| |
17:10 | fabrizio1 | I am guessing you mean <$10k
| |
17:11 | se6astian | hopefully we will be able to reach a retail range of 6000-7000€
| |
17:11 | se6astian | but at this stage its just guessing
| |
17:11 | devbisme_ | OK, I just wanted to see where your expectations were.
| |
17:12 | fabrizio1 | for the 6k camera, you expect SD, usb3 and battery?
| |
17:13 | se6astian | battery will be third party
| |
17:13 | se6astian | which high speed interface is not yet decided
| |
17:13 | Bertl | SD as in secure digital? if so, just for the software
| |
17:13 | fabrizio1 | SD card
| |
17:13 | se6astian | micro sd card to hold the camera firmware yes
| |
17:13 | Bertl | i.e. micro SD holding the FPGA bitstream and OS
| |
17:14 | fabrizio1 | and the actual video?
| |
17:14 | Bertl | will be sent over SDI, hdmi, etc
| |
17:14 | fabrizio1 | oh, there is an HDMI out?
| |
17:14 | devbisme_ | How much buffering memory within the camera?
| |
17:15 | Bertl | we currently work with 512MB of DDR3 memory
| |
17:15 | se6astian | the storage module (SSDs) is an additional module, if we will be able to develop that in parallel to the camera depends mainly on the level of success of the crowd funding campaign
| |
17:15 | Bertl | we would prefer dedicated PL side memory with about 64M+ for buffering
| |
17:15 | fabrizio1 | what is the bit rate needed for the DDR3 memory?
| |
17:16 | se6astian | the sensor has 4096x3072 pixels and can provide up to 150 frames per second at 12 bit
| |
17:16 | se6astian | one image is 18.874,368 bytes
| |
17:16 | Bertl | we produce about 30Gbit/s at the highest framerate
| |
17:17 | devbisme_ | What does the FPGA do as the image streams by?
| |
17:18 | Bertl | whatever is necessary to get a proper image :)
| |
17:18 | Bertl | i.e. typically on a color sensor that will be:
| |
17:18 | se6astian | normal framerates like 25 FPS will result in moderate datarates of around 450 Mbytes/s
| |
17:19 | Bertl | - denoise, demosaic, color correction, optional subsampling
| |
17:19 | fabrizio1 | not sure u have this info but what is the current DDR3 memory clock?
| |
17:19 | Bertl | - image analysis (peaking, histogram)
| |
17:20 | philippej | image is then sent to sdi out to an external recorder, and at a later stage, a built in recorder will be developed
| |
17:20 | devbisme_ | Are these operations typically done within a single frame, or using image data across multiple frames?
| |
17:20 | Bertl | we currently work with the zedboard and we are almost at the bandwidth limit of the onboard DDR3
| |
17:21 | Bertl | devbisme_: probably depends on the application, but the typical use will require 2 frames max (most will work with 1 frame or less)
| |
17:21 | philippej | would it be possible to do all the processing without memory buffer at all ?
| |
17:21 | se6astian | and delay between reading image data and outputting it over an interface should be kept minimal
| |
17:22 | Bertl | philippej: yes, to some degree, but it would require synchronization of the output speed with the sensor
| |
17:22 | fabrizio1 | Dave, the current Zeb board has a 533MHz DDR3 memory
| |
17:23 | philippej | I was already wondering about that, for example a complex debayering algorithm might need info from several lines behind, but never a complete frame behind
| |
17:24 | fabrizio1 | what about a camera like the Red Epic? does it process one frame at the time?
| |
17:24 | Bertl | philippej: for example, HDMI out (think viewfinder) doesn't work with the high data rates of the sensor
| |
17:24 | se6astian | nobody knows :)
| |
17:24 | fabrizio1 | Black Magic?
| |
17:24 | se6astian | they are all closed source :)
| |
17:25 | fabrizio1 | nobody has ever looked inside to see how much memory there is?
| |
17:25 | philippej | for monitoring purpose it can be somehow dirty and won't be at faster speed than what an sdi out would provide (max 30-50 fps), we could frame skip for example
| |
17:26 | Bertl | you still need to buffer at least a frame though
| |
17:26 | philippej | for the normal case, the output would be at the same framerate as sensor imho, with the monitoring tool accepting different "standard" framerate, 24, 25, 30p
| |
17:26 | Bertl | note: the sensor sends the data at the highest speed regardless of the exposure time
| |
17:27 | se6astian | this is as much innards as anyone got to see so far I am afraid: http://cdn.cinescopophilia.com/wp-content/uploads/2012/01/RED-EPIC-X-Teardown-T7.jpg
| |
17:27 | Bertl | fabrizio1: I'm sure somebody did, but none of us
| |
17:28 | fabrizio1 | I see
| |
17:28 | Bertl | so on the image I see a number of 'memory like' chips
| |
17:28 | philippej | to say it otherwise, there are two very different use cases : sending live data to sdi, and recording
| |
17:29 | Bertl | they are not so different from the memory/buffer PoV
| |
17:29 | Bertl | anyway, not sure where we are heading ...
| |
17:30 | se6astian | yes we are drifting off topic a bit :)
| |
17:30 | fabrizio1 | we were thinking of having the SD card on the same PCB where the FPGA is. Is is bad?
| |
17:30 | Bertl | devbisme_: what kind of hardware do you have in mind?
| |
17:30 | devbisme_ | So we have an FPGA taking 30 Gbps (?) from the sensor, putting it into a frame buffer and also sending that (after some processing) out through HDMI?
| |
17:30 | se6astian | so to summarize: we need high bandwidth :)
| |
17:31 | se6astian | SD card on FPGA board is fine I would say
| |
17:31 | Bertl | fabrizio1: as I said, SD card is nice (micro sd) to hold the firmware/OS
| |
17:31 | Bertl | fabrizio1: you can't expect it to hold 4k raw data :)
| |
17:31 | se6astian | we also had in mind to put a ethernet and UART/JTAG mini usb connetor on the same board
| |
17:31 | se6astian | plus a reset button
| |
17:31 | philippej | Maybe fabrizio1 and devbisme_ you can talk a bit about the future board you have in mind ?
| |
17:32 | devbisme_ | Our initial plans were a Zynq + LPDDR + SD card + USB 2.0 interface.
| |
17:33 | devbisme_ | Obviously, that wouldn't meet your needs.
| |
17:33 | fabrizio1 | on a system on module pcb
| |
17:33 | Bertl | hehe, right
| |
17:33 | devbisme_ | We're trying to see what it would take to do that and if it can be cost effective.
| |
17:33 | fabrizio1 | Hold on a second, I do not understand where is the final raw movie going to be stored when not leaving the camera
| |
17:34 | Bertl | in the first step, it will in 99% of all cases leave the camera via SDI
| |
17:34 | fabrizio1 | I see
| |
17:34 | Bertl | in a later step, we plan to develop a recording module
| |
17:34 | Bertl | which will store the data on SSD raid
| |
17:34 | fabrizio1 | the recoding module will have an other FPGA?
| |
17:35 | Bertl | (either SATA or custom solution)
| |
17:35 | Bertl | yes
| |
17:35 | fabrizio1 | I see
| |
17:35 | Bertl | or it might utilize the MGT ports of 7015/7030
| |
17:35 | devbisme_ | So how is the camera used if there is currently no storage unit?
| |
17:35 | Bertl | if that works bandwidth wise (we do not know yet)
| |
17:35 | Bertl | external recorder, PC card, etc
| |
17:36 | philippej | devbisme_ with something like that : http://www.convergent-design.com/Products/Odyssey7.aspx
| |
17:36 | philippej | (for cinema use at least)
| |
17:36 | devbisme_ | OK.
| |
17:36 | Bertl | one idea was to design the zynq board in such a way, that it can function as PCIe card (with a carrier)
| |
17:36 | philippej | (there are cheaper options as well)
| |
17:36 | Bertl | so basically think of the camera as the following parts:
| |
17:37 | Bertl | lens mount, sensor PCB, FPGA PCB and I/O PCB
| |
17:37 | fabrizio1 | what is the datarate of the data leaving the fpga module?
| |
17:37 | Bertl | there is also a battery part, but I leave that out for now
| |
17:37 | Bertl | the FPGA PCB receives and processes the data from the sensor PCB and streams it out via MGT ports
| |
17:38 | Bertl | the I/O PCB can be adapted to the needs, e.g. provide the drivers for SDI, have some HDMI chip, generate SATA, etc
| |
17:39 | Bertl | the FPGA board could also function as the receiver (different firmware) for a PCIe card
| |
17:39 | se6astian | I created a collaborative google doc which we can use to summarize this discussion: https://docs.google.com/document/d/1B5hAG5BJu_7NUyJa21HbjewR95UuEjKAD48wP0NoIXU/edit#heading=h.rcqm5jdjd3rv
| |
17:39 | devbisme_ | Excuse me for a sec. Someone at door.
| |
17:39 | se6astian | I added some datarates and image sensor specs already
| |
17:40 | Bertl | hmm, ends after the second line here?
| |
17:40 | se6astian | what is the datarate of the data leaving the fpga module <- SDI is limited to 3Gbit/s, SATAIII is limited to 6Gbit/s
| |
17:40 | se6astian | hmm, strange
| |
17:40 | se6astian | reload maybe?
| |
17:41 | Bertl | tried that, no change
| |
17:41 | fabrizio1 | I see
| |
17:41 | se6astian | what is the datarate of the data leaving the fpga module <- other options: SFP+ (10Gbit/s), USB3 (5Gbit/s)
| |
17:41 | fabrizio1 | for the SDI output, I am guessing we need to use some SDI chipset?
| |
17:41 | se6astian | none of that solutions can cover the full speed the sensor could provide
| |
17:41 | fabrizio1 | right
| |
17:42 | se6astian | but there might be a soltuin in the future so we do not want to limit ourselves artificially already
| |
17:42 | philippej | fabrizio1 it seems sdi can be accomplished with the high speed transceivers of the 7015 or 7030
| |
17:42 | Bertl | fabrizio1: we plan to utilize the 4 multi gigabit tranceivers present in 7015 and 7030 to transfer the output data
| |
17:43 | Bertl | so in case of SDI, those will go to 4 SDI line drivers
| |
17:43 | philippej | the IO board could provide up to 4 sdi out for instance, and the remaining lines would be used (in the future) to connect to the recording module
| |
17:43 | se6astian | another option to tackle the high speed recording is to use a much larger RAM and fill it up until full with images, then save them to recording media at a much lower speed <- this is how most(if not all) high speed cameras are doing it at the moment
| |
17:44 | fabrizio1 | I see
| |
17:44 | fabrizio1 | so the Zynq is not really doing much
| |
17:44 | se6astian | for the SDI output, I am guessing we need to use some SDI chipset? <- a cable driver or interface chip connected to one of the 4 multi gigabit tranceivers (as Bertl mentioned)
| |
17:45 | Bertl | fabrizio1: there is a lot to do actually, and I fear that the 7015 might be too small to do it, but it should work for most things
| |
17:45 | philippej | as I see it, the key is to have fast ram (if needed at all), and multigigabit transceivers (7015 or 7030)
| |
17:45 | fabrizio1 | understand
| |
17:46 | devbisme_ | What is the most demanding task for the FPGA that makes the 7015 too small?
| |
17:46 | fabrizio1 | how many pins of the FPGA are actually needed for the SDI chip?
| |
17:47 | Bertl | depends on what chip and what solution, in the optimal case only a few, including the MGT ports
| |
17:47 | guesst | it requires only a tx pair from the MGT
| |
17:47 | Bertl | correct
| |
17:49 | fabrizio1 | A practical question, I have seen some posts about the modular design of the camera but I'd like to know the size of the PCB that needs to fit the FPGA and everything else
| |
17:49 | se6astian | currently the size of the camera enclosure is 10x11cm (looking from the front)
| |
17:50 | se6astian | so a PCB could be in the 9x10cm range
| |
17:50 | fabrizio1 | I see
| |
17:50 | se6astian | but the enclosure design is still flexible
| |
17:50 | se6astian | if we see we will not fit the parts/PCBs in that boundaries we can easily increase the size
| |
17:51 | fabrizio1 | well Dave, I dont know you but I see lots of pins and lots of very bloody fast signals
| |
17:51 | devbisme_ | I agree.
| |
17:52 | Bertl | we see the same, so we probably have a similar picture now :)
| |
17:52 | fabrizio1 | haha
| |
17:52 | rexbron_ | left the channel | |
17:52 | rexbron | joined the channel | |
17:52 | rexbron | left the channel | |
17:52 | rexbron | joined the channel | |
17:52 | fabrizio1 | question, would you guys be happy with a 2Gb RAM
| |
17:52 | fabrizio1 | ?
| |
17:52 | devbisme_ | What is the interface from the sensor? Is it 64 bits of LVDS @ 600 MHz?
| |
17:53 | se6astian | yes
| |
17:53 | Bertl | note that a preliminary check shows that the clg485 is just enough to get it done (no PL memory) and the 676pin packages of the 7030 should be sufficient for all needs
| |
17:53 | se6astian | ah sorry, 64 lines
| |
17:53 | devbisme_ | Yes, that's what I meant to say.
| |
17:53 | Bertl | yes, the full sensor connection is 64 LVDS lanes at 600MHz (300MHz ddr)
| |
17:54 | Bertl | but we can do, with 32 for a simpler/cheaper version of the FPGA board
| |
17:54 | Bertl | s/do, with/do with/
| |
17:54 | se6astian | about priorities: 30 FPS are absolutely required, anything higher we can support is nice but not demanded (in case we need to decide some compromised with the design)
| |
17:54 | se6astian | *compromises
| |
17:55 | Bertl | we can easily do that with even less LVDS, but we should be definitely fine with 32 for a 7015 board
| |
17:55 | fabrizio1 | could you send me the Zynq module you are happy with?
| |
17:55 | se6astian | http://www.xilinx.com/publications/prod_mktg/zynq7000/Zynq-7000-combined-product-table.pdf
| |
17:56 | Bertl | fabrizio1: i.e. a description for a 7015/7030 design, yes?
| |
17:56 | philippej | fwiw, a list of sdi bandwidth on page 7 : http://www.ti.com/lit/an/snla114a/snla114a.pdf
| |
17:57 | Bertl | yes, unfortunately it doesn't cover any 4k modes
| |
17:57 | fabrizio1 | yes, which one of the 4
| |
17:57 | fabrizio1 | ?
| |
17:58 | guesst | philippej: that is quite old. Today we have 6G (uhd 16:9 4:2:2 as 4 FHD quadrants)
| |
17:58 | se6astian | brb, laundry is ready :)
| |
17:58 | fabrizio1 | any I guess
| |
17:58 | guesst | and we have checked out 12G.. but that is quite short range (<25m)
| |
17:58 | philippej | guesst, if you have some docs, please add to the doc
| |
17:59 | guesst | no docs, 6G should be already at SMPTE
| |
17:59 | Bertl | note that we could potentially do 4x12G with the 7030 (with careful design), but personally I'd stop at 4x6G in this setup
| |
18:00 | Bertl | should still be more than enough to get the data out in 90% of all cases
| |
18:00 | guesst | you can get 4k over 3G, if you do not conform to those silly standards :)
| |
18:01 | philippej | the story is very different when you consider talking to to the real world (existing equipment people own trough sdi), and what we can do at a later stage with our recorder
| |
18:02 | philippej | nobody monitors on a set at higher resolution than hd 24p
| |
18:02 | fabrizio1 | this one could be ok: http://www.digikey.com/product-detail/en/XC7Z030-1FBG484I/122-1893-ND/3925766
| |
18:02 | philippej | (for example)
| |
18:03 | Bertl | this is the 484 BGA, when we pick the 7030 (or for the 7030 FPGA board) we should target the 676pin version
| |
18:03 | fabrizio1 | so the ARM will not be used?
| |
18:03 | fabrizio1 | 676 pins.... that is lots of pins
| |
18:03 | Bertl | we can always put a 7030 on the clg485 board, they are pin compatible
| |
18:04 | devbisme_ | What are all the extra pins for?
| |
18:04 | devbisme_ | I see the 64 pins for the sensor and a few MGT for the SDI output.
| |
18:04 | Bertl | mostly supply as usual, but also 50 I/O pins
| |
18:04 | fabrizio1 | (silly question) how many layers are you expecting this PCB to be?
| |
18:05 | Bertl | at least 6 I guess, more like 8 or 10 probably
| |
18:05 | se6astian | back
| |
18:05 | fabrizio1 | hum.. that is lots of layers
| |
18:05 | devbisme_ | What size is the BGA on the Zed? That uses 8 layers.
| |
18:05 | Bertl | well, if you can do it realiably with 4, I'm fine with that :)
| |
18:06 | devbisme_ | No way we'll route those internal pins with four layers.
| |
18:06 | fabrizio1 | not even with 6
| |
18:06 | fabrizio1 | especially if we talk about 700 pin
| |
18:07 | Bertl | the zedboard is a clg484
| |
18:07 | devbisme_ | What are the 50 I/O pins for? Are they single-ended or differential (thus, 100 pins).
| |
18:08 | Bertl | in our case, they would either be for additional sensor LVDS or for PL memory, depending on the 7015 PCB
| |
18:09 | fabrizio1 | the 700 pin Zynq is $300
| |
18:09 | Bertl | between 300 and 400 USD in the single piece price range
| |
18:10 | Bertl | they get significantly cheaper after 500 (we were told)
| |
18:10 | fabrizio1 | guys the problem is that I and Dave would be interested in making and selling this fpga board to other people too. But with a $340 FPGA, I am not sure how many people would be interested. Dave what do you think?
| |
18:11 | Bertl | what about the 7015 board then?
| |
18:11 | devbisme_ | There is a market at the high end, but I'm not sure we can access it.
| |
18:12 | Bertl | that would be clg485, already provide the MGT ports, and work as low end board for our purpose
| |
18:12 | fabrizio1 | $150
| |
18:12 | philippej | sorry if it's a silly question, is it possible to find two 7xxx, one cheap and one expensive with the same pin count/layout that could be put on the same board ?
| |
18:13 | fabrizio1 | not silly at all
| |
18:13 | guesst | i do not think so
| |
18:13 | Bertl | yes, the 7030 in clg484 works on the clg485
| |
18:13 | se6astian | Zedboard is 400$ retail price btw and it sells pretty well
| |
18:13 | fabrizio1 | but if you guys are talkin about 700 pins...
| |
18:14 | guesst | Zedboard is underpriced as all devkits (usually the cost of fpga on them is about the price of the whole board)
| |
18:14 | Bertl | from xilinx DS190: The Z-7015 device in the CLG485 package and the Z-7030 device in the SBG485 package are pin-to-pin compatible
| |
18:14 | fabrizio1 | XC7Z020-1CLG484C Z
| |
18:15 | devbisme_ | No MGT on that one.
| |
18:15 | Bertl | the 7020 is the zed chip
| |
18:16 | fabrizio1 | yeah at 485 pins
| |
18:16 | Bertl | IMHO there is no point in creating a new zedboard
| |
18:16 | Bertl | there are more than enough 7010/7020 boards out there
| |
18:16 | fabrizio1 | we are making a system on module board not a protoboard
| |
18:16 | Bertl | the MGTs are new in the 7015, so that would be an interesting system
| |
18:16 | fabrizio1 | good point
| |
18:17 | Bertl | fabrizio1: there are many 7010/7020 boards out there as well
| |
18:17 | Bertl | (just not open hardware)
| |
18:17 | fabrizio1 | good point Bertl
| |
18:17 | fabrizio1 | guys the ping count scares me not only because of the pop price for the Zynq but also for the gazillion layers of this PCB
| |
18:18 | devbisme_ | It would be challenging.
| |
18:19 | Bertl | the 676 pin version I presume? because the 484 shouldn't be much different from the 400 (routing wise) and I doubt you want to go for anything less
| |
18:20 | Bertl | the 225 package has only 54 PL I/Os, so not much one can do with that
| |
18:20 | se6astian | agreed just 192 pins more in same grid
| |
18:20 | fabrizio1 | ahaha well yes routing 700 pins is something I would never do
| |
18:20 | guesst | depends on where to route it :)
| |
18:21 | fabrizio1 | :)
| |
18:21 | fabrizio1 | XC7Z030-1FBG484C would not fit the bill?
| |
18:21 | Bertl | would be an unfortunate choice
| |
18:21 | guesst | which one can do 8 MGT ?
| |
18:21 | fabrizio1 | I see
| |
18:21 | Bertl | nothing available under the WebPack license
| |
18:21 | devbisme_ | 7045
| |
18:22 | Bertl | fabrizio1: mostly because the 485 package wouldn't be different, but allow for the 7015 as well
| |
18:22 | Bertl | fabrizio1: but it wouldn't give the benefit of having more sensor LVDS available
| |
18:23 | Bertl | (or PL memory, or whatever you want to do with those I/Os)
| |
18:23 | fabrizio1 | I am afraid it wont be easy to pull those pin out of the board
| |
18:24 | fabrizio1 | 7045 is >$1k
| |
18:26 | guesst | so thats out even for me :)
| |
18:26 | fabrizio1 | so if I understand well there is not Zynq with less than 680 pins that would make you guys happy?
| |
18:26 | devbisme_ | So is it the Axiom position that 676 pins is a necessity at this point?
| |
18:26 | Bertl | what I'm trying to say is, if you don't want to go for the 676 pin version, go for the 485 pin 7015 and design for the 485 pin 7030
| |
18:27 | Bertl | in this case it can be used with the 7015 _and_ 7030, allowing for low price/high performance versions
| |
18:27 | philippej | another point is that it would be wasted efforts to not work on this together imho :-)
| |
18:27 | Bertl | and if designed properly, we can also use it as low pin count version for the axiom
| |
18:27 | fabrizio1 | absolutely
| |
18:28 | fabrizio1 | so what's the answer to Dave's question?
| |
18:28 | devbisme_ | It looks like a 485-pin version would serve?
| |
18:28 | Bertl | the 676 pin question? No at this point it isn't a necessity
| |
18:29 | Bertl | doesn't mean that we won't plan for a 676pin version as well
| |
18:29 | fabrizio1 | what advantage would u get from the 676?
| |
18:29 | Bertl | i.e. interfaces to sensor board should reflect that in some way
| |
18:29 | philippej | and let's not forget that we can at an higher level (carrier board / connectors) allow for further upgrades with higher pincounts
| |
18:29 | Bertl | as I said, more LVDS connections to the sensor board
| |
18:30 | Bertl | we probably can
| |
18:30 | fabrizio1 | hum.. the sensor board is?
| |
18:30 | Bertl | 't do 64LVDs on the 485 package
| |
18:30 | fabrizio1 | sorry Bertl but I am a little lost
| |
18:30 | Bertl | the board connecting to the FPGA board, and carrying the sensor chip :)
| |
18:31 | Bertl | for example, the CMV12000
| |
18:31 | devbisme_ | Aren't you using 64 lanes with the Zed board?
| |
18:31 | fabrizio1 | oh, how many lines has this chip? I thought it had only 64 lines
| |
18:31 | Bertl | devbisme_: no, the FMC doesn't allow for that many
| |
18:32 | guesst | you havent got the HPC variant :)
| |
18:32 | se6astian | LVDS are pairs around 144 lines to the image sensor in total
| |
18:32 | Bertl | we are using only 32LVDs for data, plus 3 more for clock/control
| |
18:32 | Bertl | guesst: correct, the zedboard doesn't feature a HPC variant
| |
18:33 | devbisme_ | So we would be looking at a higher-density connector (more expensive) if we went to the 676-pin package?
| |
18:33 | guesst | i can see here an option how can be everybody happy to the max given the constraints involved
| |
18:34 | se6astian | We do not plan to use FMC connectors for the final axiom camera
| |
18:34 | se6astian | rather we are thinking about 2x PCIe connectors
| |
18:34 | se6astian | 164 pins each IIRC
| |
18:35 | se6astian | and a dual backplane to connect the boards with the image sensor and the FPGA together in a stack
| |
18:35 | guesst | inspired at freescales elevator ? :)
| |
18:36 | Bertl | or just simple FCI connectors with 150/200 pins each
| |
18:37 | fabrizio1 | does the CCD board have more than 150 pins?
| |
18:37 | se6astian | Added an old concept image to google doc: https://docs.google.com/document/d/1B5hAG5BJu_7NUyJa21HbjewR95UuEjKAD48wP0NoIXU/edit#heading=h.9ovdvybs2w0j
| |
18:37 | se6astian | it shows the dual backplane concept with SODIMM connectors instead of PCIe
| |
18:37 | fabrizio1 | I see
| |
18:38 | Bertl | fabrizio1: as we already have 140 pins for LVDS + Control, yes, we will need a lot more to transfer power or a separate connector
| |
18:38 | Bertl | and even with power on a separate connector, you will require shielding, so 200pin is probably the minimum
| |
18:39 | se6astian | another concept image uses two FCI connectors as Bertl mentioned: https://www.apertus.org/sites/default/files/open-modules-concept-front-V1.jpg
| |
18:39 | Bertl | and it is a CMOS sensor in our case :)
| |
18:39 | fabrizio1 | very cool
| |
18:39 | fabrizio1 | Red will definitely try to buy us !
| |
18:40 | fabrizio1 | or maybe Arris
| |
18:40 | philippej | here with lateral connectors only for sensor front-end <-> fpga connection : http://public.philippejadin.be/axiom.png
| |
18:41 | se6astian | at least they should be scared a bit ;)
| |
18:41 | fabrizio1 | is it impossible to think of doing the image compression on the same FPGA?
| |
18:41 | philippej | at least there are options for routing all those pins
| |
18:42 | se6astian | image compression would be one of the things on the list of things that are "nice to have" in the future but not essential in the first stage
| |
18:42 | philippej | the recording format would cinema dng : http://en.wikipedia.org/wiki/CinemaDNG
| |
18:43 | philippej | (be)
| |
18:43 | Bertl | fabrizio1: note that you usually do not want lossy compression
| |
18:43 | philippej | which is basically raw sensor data in a nicer wrapper
| |
18:44 | fabrizio1 | OK
| |
18:44 | fabrizio1 | so same datarate? 30Gbps?
| |
18:45 | se6astian | yes
| |
18:46 | guesst | when your hardware offers at most 4x 6G links, you might have a little dificulty storing 30G
| |
18:47 | Bertl | we'll use the half bit technology :)
| |
18:47 | fabrizio1 | I see
| |
18:47 | fabrizio1 | well I hope one day Tarantino will appreciate all this effort
| |
18:47 | philippej | the practical limit for now would be two fast ssd's together. Beside that, it won't be that useful imho
| |
18:47 | guesst | Bertl: like the one in MLC chips right ? :)
| |
18:48 | Bertl | yes
| |
18:48 | philippej | (or short burst stored in ram, and slowly copied to disk)
| |
18:48 | philippej | (which is an even more specific use case)
| |
18:49 | fabrizio1 | Dave are you there? or you passed out thinking about all these pins to route?
| |
18:49 | devbisme_ | Just listening.
| |
18:49 | guesst | philippej: 30Gb/s at 2Gbit memory is.. a 66 msec burstie :)
| |
18:50 | guesst | can anybody of you be a little bit more real
| |
18:50 | Bertl | we do not plan to use less than 512MB :)
| |
18:50 | se6astian | yes, if we decide to go the burst route we would need to significantly increase the amount of RAM
| |
18:50 | se6astian | to at least 2049 MB :)
| |
18:51 | fabrizio1 | that is lots of RAM
| |
18:51 | guesst | that is not significant. you have to do double/quarduple of what competition got to make it equal when you release
| |
18:51 | Bertl | fabrizio1: don't worry, 512MB seems to be working fine, 1GB would be nice to have
| |
18:51 | philippe_ | joined the channel | |
18:51 | philippe_ | left the channel | |
18:52 | fabrizio1 | 1GB is 10 Gb ... that is lots of RAM too
| |
18:52 | Bertl | guesst: we'll quadruple everything every two years
| |
18:52 | philippe_ | joined the channel | |
18:52 | fabrizio1 | I see that there is a 900 pin FPGA
| |
18:52 | fabrizio1 | just kidding
| |
18:52 | se6astian | :)
| |
18:52 | Bertl | fabrizio1: the microzed features 1GB already
| |
18:53 | Bertl | that's two tiny DDR chips
| |
18:53 | guesst | i think the discussion about ram size is insignificant. you can be fine with 288mbit of memory given that it got the bandwidth
| |
18:54 | fabrizio1 | well the 600 MHz RAM worries me as well
| |
18:54 | guesst | next step is about 32+ GB to make it worth, in between - no gain
| |
18:54 | Bertl | lets say 432mbit, but yes, I agree, for the PL side, this is probably the essential miminum
| |
18:55 | guesst | Bertl: my I present a workable solution for all?
| |
18:55 | devbisme_ | Go ahead!
| |
18:55 | Bertl | go ahead, we are always open to suggestions
| |
18:55 | guesst | so listen
| |
18:56 | guesst | plan mechaniclally for a big fpga, and 4 channel memory, but actually do the first design with the smaller one
| |
18:56 | guesst | also a mechanical reserved space should be present at sides of this fpga/ram core
| |
18:56 | guesst | that is for adjusting the module to various form factors and customizing i/o interfacing
| |
18:56 | guesst | as for axiom, choose the max zynq you can code in webpack
| |
18:56 | Bertl | so far sounds like we are planning to do
| |
18:57 | guesst | and plan the bandwidth to match the equality of sensor interface vs. ram (vs. gtp out)
| |
18:57 | guesst | that will give you the max framerate you can do.
| |
18:57 | guesst | the guys from hardware will have a core which can be reused for other projects - when they customize the pcb
| |
18:58 | Bertl | nothing new yet, let's hear your solution
| |
18:58 | guesst | e.g. add whatever plugs or interfaces it needs
| |
18:59 | Bertl | I presume you are referring to the 'core' as routed layout, yes?
| |
18:59 | guesst | yes.
| |
18:59 | guesst | so they can form a dimm card from it, or a pcie card - on single board
| |
19:00 | guesst | or just your apertus interconnect system
| |
19:00 | Bertl | well, this is how you can work with open designs
| |
19:00 | Bertl | no need to build everything from scratch there
| |
19:00 | guesst | also the core could be replaced with the 7045 once software is freed or somebody got a license
| |
19:01 | guesst | there is currently no high performance core (like a 4x 16bit ddr3)
| |
19:01 | Bertl | only in the smaller footprint if available
| |
19:02 | Bertl | but yes, there is no reason not to utilize a 7045 once it is covered by the WebPack
| |
19:02 | Bertl | but you cannot design a layout for a 676 pin package and use a 485 pin chip instead :)
| |
19:03 | guesst | you can reserve space for it
| |
19:03 | guesst | and if you have the outer design
| |
19:03 | guesst | you just replace the center of the board.. without doing a complete modification
| |
19:04 | Bertl | okay, that might be feasable, if you consider the additional I/O lines properly
| |
19:04 | guesst | also doing a fixed core (group of chips) would mean, that when one uses that, it could be populated at those designers... with the rest done by hand on each custom layout
| |
19:04 | se6astian | so in summary I think we discussed the essential questions and defined requirements to a certain extent now
| |
19:04 | se6astian | are there any further questions?
| |
19:05 | devbisme_ | This discussion is in a Google doc somewhere?
| |
19:05 | se6astian | or should we conclude the continue via email/google doc after this first successful get-together
| |
19:05 | se6astian | her: https://docs.google.com/document/d/1B5hAG5BJu_7NUyJa21HbjewR95UuEjKAD48wP0NoIXU/edit#
| |
19:06 | fabrizio1 | please lets continue in the next days writing down all the components we like
| |
19:06 | se6astian | yes that would be great
| |
19:06 | guesst | se6astian: what is the point of always calculating with full aperture ? anamorphic is expensive and not that used these days
| |
19:06 | fabrizio1 | lets go for the 700 pin for now? Dave?
| |
19:06 | se6astian | btw do you have any experience/preference with the interfaces: USB3, SATA, SDI or SFP+ ?
| |
19:07 | devbisme_ | 700 pin means a lot more layers.
| |
19:07 | philippej | left the channel | |
19:07 | guesst | we have working SDI :) as you may know
| |
19:07 | se6astian | guesst, what do you mean with "full aperture"?
| |
19:07 | guesst | 4:3
| |
19:07 | se6astian | aperture is a lens element
| |
19:07 | Bertl | guesst: we haven't seen any open hard or software regarding SDI
| |
19:07 | fabrizio1 | I know....
| |
19:08 | guesst | Bertl: xilinx core is bound to a license?
| |
19:08 | fabrizio1 | SATA would be good i think
| |
19:08 | fabrizio1 | Let's list a 450 pin option too
| |
19:09 | fabrizio1 | guys it was great to talk to you
| |
19:09 | Bertl | fabrizio1: why not plan for a 676pin version space wise (as guesst suggested) and go for the 485pin for now?
| |
19:09 | guesst | Bertl: that is what i suggested
| |
19:09 | guesst | "but actually do the first design with the smaller one"
| |
19:09 | devbisme_ | Fabrizio, we should go through the discussion, pull out the agreed-upon specs and then run two versions using 485 and 676 and see what the cost implications are.
| |
19:09 | guesst | you can plan for 676+4 ram chips
| |
19:10 | guesst | and then see what you can do with the smaller fpga
| |
19:10 | fabrizio1 | I did not expect to talk about such a large ping count
| |
19:10 | guesst | i am worried about ram channel count when you use a massive amount of pins for sensor (unnecesarily)
| |
19:10 | Bertl | devbisme_: also any high speed buffer memory on the PL side might be a big win
| |
19:11 | fabrizio1 | lets all try to come up with actually components
| |
19:11 | Bertl | as guesst stated, about 64MB would suffice
| |
19:11 | fabrizio1 | on a shared document
| |
19:11 | Bertl | sounds good
| |
19:11 | se6astian | great
| |
19:11 | fabrizio1 | 64 MB?
| |
19:11 | se6astian | please bookmark the google doc: https://docs.google.com/document/d/1B5hAG5BJu_7NUyJa21HbjewR95UuEjKAD48wP0NoIXU/edit#
| |
19:11 | Bertl | PL side, i.e. for the fabric
| |
19:11 | se6astian | it can freely be access and edited by anyone who has the link
| |
19:12 | guesst | i thought of PL mem only :)
| |
19:12 | se6astian | *accessed
| |
19:12 | Bertl | the cores need memory as well, not using the PS memory controller would be a waste
| |
19:12 | fabrizio1 | ok, give us sometime to do some reading on what it takes to route 600 pins
| |
19:13 | Bertl | but it could be reduced to low bandwidth low pincount memory
| |
19:13 | se6astian | sure, many thanks for your time and being here
| |
19:13 | devbisme_ | Bye, guys. Thanks!
| |
19:13 | fabrizio1 | buy guys
| |
19:13 | Bertl | feel free to ask specific questions regarding axiom anytime here on the channel or via email
| |
19:13 | fabrizio1 | ok i sill
| |
19:14 | fabrizio1 | i will
| |
19:15 | fabrizio1 | buy guys
| |
19:15 | fabrizio1 | left the channel | |
19:16 | philippe_ | bye!
| |
19:16 | se6astian | bye!
| |
19:16 | devbisme_ | left the channel | |
19:18 | guesst | what memory speed can be achieved on zynq?
| |
19:28 | philippe_ | left the channel | |
19:34 | Bertl | Speed
| |
19:34 | Bertl | of up to 1333 Mb/s for DDR3 is supported.
| |
19:35 | Bertl | (on the PS controller)
| |
19:36 | Bertl | the PL could go up to 1866Mb/s DDR3 on higher grade zynq
| |
19:36 | Bertl | the PS has 32 or 16 bit width
| |
19:44 | guesst | i hate these ps/pl :)
| |
19:44 | guesst | arm/fpga should be easier
| |
19:47 | philippej | joined the channel | |
19:47 | guesst | probably then a 4 chip design, 1+3 should do the full bandwidth when not used for same speed output, just caching, right?
| |
19:48 | guesst | by chip i mean 16bit, never saw a wider one
| |
19:49 | philippej | left the channel | |
19:54 | philippej | joined the channel | |
19:55 | FergusL | left the channel | |
20:31 | philippej | left the channel | |
20:37 | FergusL | joined the channel | |
21:47 | se6astian | tme f bd
| |
21:47 | se6astian | gd nght
| |
21:47 | se6astian | damn keyboard
| |
21:48 | se6astian | left the channel |