This Is Why We Do Not Have Vga On The Pandora!

[darkblu]

Umm, no. USB can transfer at 480Mb/sec, which at best gets you 48MB/sec, but it is a half duplex shared serial bus. This means that between each block of data, the bus must change direction so the device can send an acknowledge at least back to the master. Also there is other stuff going on on the bus, even if only one device is connected. There is no way you are going to be sending uncompressed VGA data at 60 hz refresh across a USB bus.

my knowledge of the matter is cursory, at best, but everything i've ever read on the subject indicates that there's no ack phase in isochronous usb. neither is the integrity of the data guaranteed, for that matter.

http://techon.nikkeibp.co.jp/article/HONSHI/20080529/152601/?P=3

anyway, i was just speculating on the possibility. not referring to existing products - usually those're aiming at higher-than-VGA resolutions, ergo sending raw framebuffer streams would not be an option for them.

 

[Loonie]

Of course it will, through S-Video...
My TV is a CRT TV (I know, I know, time to replace it), and it has an S-Video-in...

Was my sarcasm not sarcasm-y enough? Damn, I need to work on that.

 

[Enverex]

Umm, no. USB can transfer at 480Mb/sec, which at best gets you 48MB/sec, but it is a half duplex shared serial bus. This means that between each block of data, the bus must change direction so the device can send an acknowledge at least back to the master. Also there is other stuff going on on the bus, even if only one device is connected. There is no way you are going to be sending uncompressed VGA data at 60 hz refresh across a USB bus.

my knowledge of the matter is cursory, at best, but everything i've ever read on the subject indicates that there's no ack phase in isochronous usb. neither is the integrity of the data guaranteed, for that matter.

http://techon.nikkeibp.co.jp/article/HONSHI/20080529/152601/?P=3

anyway, i was just speculating on the possibility. not referring to existing products - usually those're aiming at higher-than-VGA resolutions, ergo sending raw framebuffer streams would not be an option for them.

True, but assuming 100% throughput (which is unrealistic even without ACK packets) is only 60MB/s on USB2. 90% of that (your estimate) is 54MB/s. But that's all assuming it's sending the full VGA payload over the USB and not doing something with it in the process. Guess someone would need to test one to find out how (and how well) it actually works rather than just speculating raw statistics.

 

[darkblu]

True, but assuming 100% throughput (which is unrealistic even without ACK packets) is only 60MB/s on USB2. 90% of that (your estimate) is 54MB/s. But that's all assuming it's sending the full VGA payload over the USB and not doing something with it in the process. Guess someone would need to test one to find out how (and how well) it actually works rather than just speculating raw statistics.

of course, the proof would be in the pudding. but as you note, we've been discussing the heaviest-payload scenario so far. drop those 24bits/pixel to 18bits/pixel (which is as much as many an LCD can actually show) and all of a sudden we don't need to speculate of the possibility already.

 

[fischju2000]

In any case the Pandora just does not have the horsepower to render games at higher resolution, or upscale standard definition games. A VGA output on the Pandora, no mater how it is implemented would be good for static displays and not much else. It would also increase the cost of the system, use more power, and only a few people would benefit from it.

That probably could have been worded better, unless you really mean to suggest that the Pandora can't do anything at high resolutions. The board itself could drive a full 1920x1200 display (or two) from the LCD output that the screen is attached to. The beagleboard doesn't have any problems with this high definition on DVI or HDMI.

 

[rv6502]

In any case the Pandora just does not have the horsepower to render games at higher resolution, or upscale standard definition games. A VGA output on the Pandora, no mater how it is implemented would be good for static displays and not much else. It would also increase the cost of the system, use more power, and only a few people would benefit from it.

That probably could have been worded better, unless you really mean to suggest that the Pandora can't do anything at high resolutions. The board itself could drive a full 1920x1200 display (or two) from the LCD output that the screen is attached to. The beagleboard doesn't have any problems with this high definition on DVI or HDMI.

unless the frame buffer is contained in a separate memory connected directly to the video circuit it wont leave much main memory bandwidth for anything.
1920x1200 x 60Hz = 138.24 mega pixels/sec, at 16bpp that's 276.48MB/sec (263.67MiB/sec)

and that's just for DISPLAYING the frame, not even rendering it.

I don't know why someone would want to plug his Pandora to a monitor let alone an ANALOG VGA one (and where would the plug be? ) but its still a fun topic.

ok, let's go crazy: how about using a resistor ladder on the 2nd LCD digital data lines ?
that would make a really horrible non-linear analog output but it would work

 

[WizardStan]

The <edit>second</edit> digital lines also serve as GPIO, and I'd be very surprised if MWeston didn't find a use for them.

 

[darkblu]

ok, let's go crazy: how about using a resistor ladder on the 2nd LCD digital data lines ?
that would make a really horrible non-linear analog output but it would work

of course it would.

ps: did you just call that home-made LPT covox from my childhood 'really horrible' and 'non-linear'?

 

[Exophase]

I personally tend to think there are satisfying compromises between 60Hz and "static presentations".. I think that 30Hz is somewhat sufficient for anything, even if it's a significant quality compromise for fast moving games.

At least it's not like the OpenMoko Neo1973 Freerunner phone that has so little bandwidth to the display controller that you can't even update the whole framebuffer at anything faster than 11Hz.

 

[rv6502]

ok, let's go crazy: how about using a resistor ladder on the 2nd LCD digital data lines ?
that would make a really horrible non-linear analog output but it would work

of course it would.

ps: did you just call that home-made LPT covox from my childhood 'really horrible' and 'non-linear'?

well, mine was.. but back then I made it with old +/-15% resistors recuperated from some junk and huge globs of solder.
but it worked

 

[rv6502]

I personally tend to think there are satisfying compromises between 60Hz and "static presentations".. I think that 30Hz is somewhat sufficient for anything, even if it's a significant quality compromise for fast moving games.

At least it's not like the OpenMoko Neo1973 Freerunner phone that has so little bandwidth to the display controller that you can't even update the whole framebuffer at anything faster than 11Hz.

you can't feed a VGA monitor at 30Hz directly, its 50-70Hz depending on the resolution.
outside of that the protection / power saving will kick in on just about any CRTs less than ~15 years old and any LCDs, besides, you'd get huge flickers if it worked.

if you mean by using a USB VGA adapter and updating the USB device at 30Hz and less, yeah, that would work, but that's not built-in and you can do it already anyway.

for games, the PowerVR drivers usually works in deferred rendering with 1 or 2 frames of delay before being able to grab the frame, diff it, compress it, and upload it to the USB VGA device.
I would be surprised if you can get a playable experience with a USB VGA adapter, but its something I'd try just for fun.