On 12/10/2012 03:46 AM, Dieter BSD wrote:
Jack writes:
We might find that the cost of DACs and cable drivers explodes
above 2560x1600, if available at all.
That's why I asked "Is 4096*2160 unreasonable?". From reading
press releases, it sounds like they drive 4K displays with either
2 HDMI or 1 Displayport. So I predict a new revision of HDMI
to drive 4K with a single cable.
We only need the DAC for analog (VGA) right? Are there going to be
analog 4K displays? I'm guessing no. So we only need the cable
drivers to be fast.
Please do not go into VGA. I know it works the best for some, but it's
fate is same as parallel PCI's - it is way outdated.
You can do 4K/25p in HDMI 1.4 (with 225 MHz transmitters). Currently
there is no 297MHz transmitter, just receiver. If you try to do it in a
"software way" in FPGA, you will need HDMI license ($10K annually!).
4K can be driven by one HDMI, but there is no such source. Current
models present multiple HD streams on one display, that is the reason
for multiple HDMI inputs.
Displayport 1.0/1.1 (the basic 2.7Gb variant) can handle up to 2560x1600
@ 60p or 4K up to about 40fps (or lower fps with higher bit depth)
Displayport 1.2 (5.4Gbps per lane) can handle twice of that, making
4K/60p possible.
The 3D television fad is dying. It looks like there is going to be
a push towards 4K displays as the next reason to get everyone to
buy new displays again. There is also talk about 8K displays,
but I suspect (1) that they will push 4K first, then push 8K later.
That way they get 2 sales instead of just 1. (2) attempting to
support 8K might be too difficult/expensive. Although it is just
a matter of multiple cables rather than obscenely fast signals
(bleeding edge) it might be doable?
Yes, it is and it seems it alwyas will be a choice to do it with
multiple links. The current 4K display modules are anyway interfaced in
tiles, so even if it is possible to use one cable on input, it will be
broke up into several links going to the display.
The today's usual configuration is four dual-lvds links - 4x (1
clock+2x4 pairs data) or two embedded displayport links.
Things it would be useful to know:
- Did I get anything wrong in the above?
- Is there a new revision of HDMI coming out? When can we get specs
so that we can support it? It wouldn't hurt to check on DVI and
Displayport as well.
Just check the pixel rate. Old hdmi/dvi interface which use 165MHz
hardware is insufficiet, a newer hdmi (225/297M) or a duallink DVI might
be a choice.
HDMI 1.3 increased the bandwidth
HDMI 1.4 is the proper standard for 4K
- cost of DAC to support at least 2560x1600.
Are there such displays? Connect them rather digitally.
- cost of cable drivers to support at least 2560x1600, per head.
A) Displayport implemented using GTP in Spartan6.. just the price of
the plug. The rest is IP you have to code.
B) I've got some sample chips (LVDS to DP converter), about $5 - $10. It
is fed by 10 LVDS pairs where you serialize the pixels. For low volume
and fast delivery that might be an option.
- cost of cable drivers to support at least 4096x2160 or 4096x2560, per head.
At low framerate, both of the above A/B. Or there is an hdmi transmitter
which could do it on hardware side, but it is not certified to do so.
Would need a try.
At decent framerate you have to go into high-end FPGA ($500-$1000) to be
able to use 5.4G signalling defined in DP1.2.
- Can we use the same cable drivers for DVI, HDMI, and Displayport,
putting a LFH connector or 2 on the bracket and then adding a
LFH-to-whatever cable to get DVI, HDMI, or Displayport?
In custom silicon Yes.
In discrete devices No.
You can do DVI/HDMI as they are both the same on electrical side.
Or you can do DP - cheaper to develop, not constrained by television
resolution and color standards.
Maybe I'm asking the wrong questions. Maybe I should be asking
what are the latest, greatest revisions of DVI, HDMI, and Displayport,
what resolution, refresh rate, and color depth do these support, and
what does it cost, per head, to support them?
Go into implementing DP 1.1. You can have it for <$10 as a chip, or if
you code the IP, you can "save" that on every board. The savings are
only when you have spare GTP oututs on the FPGA. Otherwise the
transceiver equipped FPGA's have higher price than these external chip
would cost.
Daniel
_______________________________________________
Open-graphics mailing list
[email protected]
http://lists.duskglow.com/mailman/listinfo/open-graphics
List service provided by Duskglow Consulting, LLC (www.duskglow.com)
