Re: [Debconf-video] Buy some Numato Opsis boards?

On Sat, Sep 12, 2015 at 5:18 AM, Tim Ansell <mithro@mithis.com> wrote:

Hi Andy,

Carl pointed this email chain to me but it turns out I wasn't on this list previously, so sorry if this reply ends up in a weird location.

On 11/09/2015 at 5:15, Andy Simpkins wrote;

<snip>

I agree, we buy 3 of these then.
Who can authorise this spend / place order?

Thank you for your support!

I'd definitely love help making Debian the best supported HDMI2USB platform for both users and developers.

We will need to make the following hardware expansions so that we can fully replace the twinpacts (and get them supported as part of the project):
VGA input
VGA output
I guess that will be me then :-)

A good solution might be to just use cheap VGA->HDMI and HDMI->VGA adapters. On the input side they might be a bit dodgy but definitely on the output side you have a huge amount of control of the signal so even cheap adapters would work pretty reliably.

If you do decide you want to develop a VGA expansion board, my advice would be the following;

Start with the vmodvga design that Rohit did and port it to the Opsis TOFE expansion connector.

* The vmodvga works by capturing VGA passthrough in similar way to what the TwinPact does. It does have complete control over the EDID IIRC.

* The vmodvga uses an Analog AD9984 to do the VGA capture, you'll probably be tempted to use a different Analog capture chip, but I highly recommend going with something that we know already works.

* You can find the design at https://github.com/timvideos/HDMI2USB-vmodvga

* You can find more information about how to map TOFE to the Digilent VHDCI at https://docs.google.com/spreadsheets/d/15rivtOcSG4kqKf0AUK_OwBfs67XQtW8byr8kc5FUTgE/edit#gid=574892795

Help me poke Rohit about getting vmodvga support into the MiSoC firmware. https://github.com/timvideos/HDMI2USB-misoc-firmware/issues/30

We will also need the laser cut enclosure (or some other suitable box). Ideally this will also contain the PSU, and sacrificial connectors - more on this at the sprint.

See Carl's most recent email. Any Mini-ITX style case should be suitable.

And we should push for LAN functionality ASAP
- having LAN out means that we don't need the laptop that has accompanied the twinpact. (less to go wrong)

The current code for doing LAN streaming can found at https://github.com/timvideos/HDMI2USB-misoc-firmware/tree/master/hdl/streamer

We also had a student work on doing this in GSoC a couple of years ago. This was before the MiSoC firmware was developed, so it is probably not directly useful now but he did post a lot about tools he has discovered and problems he ran into. See http://hdmi2ethernet.blogspot.de/

Florent recently created this lovely system diagram (https://raw.githubusercontent.com/timvideos/HDMI2USB-misoc-firmware/master/doc/architecture.png) which shows how things fit together.

The task list in my opinion is;

* Finish and test the RTP module. I have not idea what the current state is, there is an rtp.py which has appeared since I last looked at this :). This is been tracked in https://github.com/timvideos/HDMI2USB-misoc-firmware/issues/23

* Finish and test the device operating at Gigabit speeds (most of the current stuff has been done at 100Mbit speeds but 1Gbit operation has been tested at least once). This is being tracked in https://github.com/timvideos/HDMI2USB-misoc-firmware/issues/32

* Create a module which connects the UART to a telnet / netcat adapter so you can control the device remotely. This is been tracked in https://github.com/timvideos/HDMI2USB-misoc-firmware/issues/24

Once you have that working, there are a couple of cool things you could attempt;

* See if you can fit two whole encoders + RTP streamers onto the FPGA and allow both HDMI inputs to work! (We are currently at 45% of general FPGA usage, but the problem will probably be the more specialized components and routing a 90% full FPGA).

* See if you can stream "into" the device and put the output on the screen.

Finally we don't really want to down scale slides to 480i (note: 720p is slightly lower resolution than the DV we already have from the twinpacts). Right now this just replaces DV, and gives us digital capture, but it doesn't improve on resolution.

As Kurt mentioned, DV is significantly lower resolution than 720p.

The device should also work perfectly at 1024x768 if you want to use 4:3 resolution rather than 16:9.

I would have thought that we would want to run at least 1080p as our standard capture resolution (replacing 1024x768)

If the Opsis board is successful, we will definately look at doing another board with an Artix-7 FPGA rather than the Spartan-6. That should allow us to do 1080p on all ports.

Thanks to using misoc we can reuse 95% of the code developed for the Spartan-6 on the Artix-7. I have even already have a Digilent Nexys Video which would be our "prototyping board" for any Artix-7 development we end up doing.

It still leaves the problem that you can't jam 1080p down either USB2.0 nor Gigabit Ethernet without a load of compression. We would definitely be investigating USB3.0, direct PCI-Express on such a board.

With regards to sprint / Miniconf:
It looks like the Cambridge Miniconf will go ahead. ARM will once again host, but dates have yet to be finalised.

I gave one of my Digilent Atlys boards to Stefano Rivera to bring to the Miniconf. You should be able to use it for testing.

The production Opsis boards won't be shipping out till december but I might be able to get you one of the preproduction boards sooner. Who would be the best person to "own" the board? You probably won't want to use a preproduction board in production either, so I still recommend getting 3 boards.

Hope that helps!

Tim 'mithro' Ansell