SVGA over UDP

That's great to hear. Good luck :-) I hope it goes well.

Learn how to generate an SVGA display from an FPGA (it's a common project on many development boards)

Make sure your FPGA or its board has enough memory

Learn how to receive UDP packets (see

Put the two together.

Something interesting to think about is where the video frame timing and dot clock are going to come from, and how you will synchronize them. Will you buffer one or more frames or will you somehow keep the network streaming data just a bit before the electron gun needs it (or overwiting memory just after it's been read).

Also do you have the network to yourself or must you share it? How will you handle lost or corrupted packets? (simply displaying the old image segment in place of a lost/failed checksum packet might work)

800x600x4x30 = 57600000.

That's 5.76 times as many bits per second as you can fit down your

100Mbps network, ignoring protocol overhead, network congestion...

I'd have a look at VNC if I were you, to see how that 'solves' the problem (search for "VNC", there are many variants and some source code is available). Perhaps you could even use a subset of the VNC protocol as your protocol.

Tim

--
Did I really still have that sig?

VNC is a good solution, but in the case of Tim it might not be feasable. Various methods are used like compression. In some cases the interface is emulated at the end machine, so the server has to send info for window positions etc, and not a picture of the remote screen. For example the desktop background doesn't change (when we don't do any alterations), so it doesn't have to send that all the time. Only moving parts of the image could be send like the cursor or the image of a newly opened window.

Regards GM

Or look at this product:

Uses a Spartan 3 and more parts, but does video-type compression (mpeg4 like performance) for 1280x1024@30 FPS. Still a tiny board.

They promise source code for the Spartan as well.

Thomas

Care to recount the zeros?

I'll admit I mixed up bits and bytes in my first analysis (and thought

10Mbpss would do it) but your number is 57.6 million, which is only about half the capacity of a 100Mbps wire. If this is the only thing being sent over the network it should fit with careful implementation, especially of the sender (might be interesting to do the sender in an FPGA, mooching off the SVGA feature connector)

Or am I missing something?

Well, counting all the digits is more useful :)

Nope, I was off by an order of magnitude. Thanks for spotting it.

Tim

--
Did I really still have that sig?