USB VIDEO ON THE DENEB?

[platon42]

B00tDisk wrote:
USB video dongles of this kind are horribly slow on full-speed USB2 connections.

I can't imagine they'll be worth a flip on the Amiga, driver or no.  

You know how people freak out about "oooh nooo it doesn't refresh fast enough!" regarding Amiga emulators on the PC?  Yeah, this'll give them something genuine to complain about.

We're not talking about USB 1.1 full-speed. We're talking about USB 2.0 Highspeed. The Deneb is able to achieve a data rate of more than 8 MB/sec. This is surely slower by a few factors than a real Zorro III card, but still could at least be twice as fast as a Zorro II gfx board.

However, given that these adapters are normally employed on fast PCs, they usually use some kind of lossless compression on the data before sending it over the bus. If this is the case and cannot be turned off, it would really slow down the gfx update considerably.

Anyway, only a few of these adapters have been announced with Linux drivers and reverse engineering on that kind of device could be a bit arkward.

Just today, I found a better one with bith digital DVI and VGA output for 70 EUR ( http://www.pearl.de/a-PX3103-1419.shtml ), but I still consider this too much for just experimenting with it (donation anyone? )

With "current" Amiga graphic boards having problems outputting native resolutions like 1280x1024 and higher, and no-one coping with DVI or HDMI output, an USB graphics adapter might be a solution. It wouldn't be very fast, but it would provide the resolutions (up to 1680x1050 in 24 bit).

Just my two cents.

[platon42]

Jose wrote:
By the title I thought you were gonna talk about USB2.0 DVBS/T tunner cards and / or video capture cards. Having decent video on the Amiga would be nice.

There is a "new" USB Video class specification, but unfortunately, this standard is only meant for webcams. V4.1 has the class descriptors and detection for it built in, but that's currently all.

As for DVB-S/T, I'm not watching TV myself for about 10 years, so I'm not really too fond about it writing such a driver -- if anybody wants to do it, I'll happily provide Developer Specs. Also, you have to consider that a classic amiga is definately not fast enough for realtime MPEG2 stream decoding (not considering the fastest PPC amigas), so this class driver would only be reasonable for MorphOS Pegasos/MacMini machines.

[platon42]

Darrin wrote:
Thanks for that info Chris.  I figured that if Siamese could work then there was some potential here.

I'll throw 20 Euros into the pot if you want to try.  Do we have 3 more volunteers for a worthwhile project?

Speaking about potential, I think for embedded machines without onboard gfx (aka EFIKA), an adapter that uses less than 2.5W (that's the maximum allowed for USB devices) would be better suited than a Radeon whatsoever. At least for my planned Touchscreen-Car-Computer project :-)

About that "bounty" -- even if I'd get the adapter, this would not come with an obligation to complete the project. As said above, the task would be rather complex.

[platon42]

amigadave wrote:

I would be interested to read more about your Touchscreen-Car-Computer project.  I have one of my own, but it is just an Intel Mac Mini attached to a 7" Lilliput touchscreen and dual booting Vista and OSX.  It also has a GPS antennae so I can use it for voice command turn-by-turn directions.

It's still in very early state. I've got an EFIKA in a cardboard box "case" with Radeon and a PicoPSU 60W, a 8" TFT touchscreen at 640x480. It's running MorphOS 2.x and the touchscreen can be used as a mouse replacement. The touch screen and the EFIKA are powered via Y-cable from the 12V connector in the car (a nifty Toyota Aygo).

I also received my OBD2 (on board diagnosis) interface construction kit (PCB+components) yesterday that I still need to solder. It will return serial data of the ISO 9141-2 diagnosis port at the weird baudrate of 10400 baud. I still need to dig the protocol, but it will allow reading realtime information such as speed and acceleration. Unfortunately, it seems as if the EFIKA MPC5200B serial driver still needs to be written. Shouldn't be too difficult.

Ah, I need to write a nice, good looking, touch screen compatible GUI for Media Playback (only audio planned) and for diagnosis. I also have a GPS module that can be connected via USB that I might want to make use of, but open and free map material is scarce.

I need more time to finish all of that, but I hope I will some time...

[platon42]

wawrzon wrote:
im just back from an electronic shop, there are a lot pci tv_cards that are advertised as equipped with hardware mpeg2 decoder for a better performance. i didnt not really found something like that among usb dongles but i cant say i really looked close.

Sending uncompressed video across an USB link easily exhausts or exceeds the raw bandwidth of 480MBit/sec. A 50Hz true color stream would be at a sustained rate of 260MBit/sec. While this is not a problem for a PCI interface, over USB this is too much.

[platon42]

Darrin wrote:

by arnljot on 2008/9/3 5:50:16

Seems like the tally is
arnljot €20
Darrin €20
HyAmi €20
amigadave $20

So seems to me that the total is €60 + $20. Getting there Now we just need Platon42 to accept

So that's €60 + €13.86 = €73.86

I'll up my contribution to €25 and cover the remainder plus P&P (S&H).

Wow. I wouldn't have thought of people comming up with this. If somebody could collect the bucks (I have no PayPal account anymore), buy that thing and send it to me, that would be great. Thanks!

[platon42]

amigadave wrote:

As one of the at least two contributors from the land of NTSC to help with the purchase of the USB video adapter, I must ask a stupid question.  Will your work to write a driver include both PAL and NTSC devices.

Will the driver work with just one specific adapter?

First, we're talking here about a driver for USB graphics adapters. This thread is not about writing a driver for a USB *digitizer* dongle that would digitize and stream an external video source to the host computer. In a previous post I made clear that this would not make sense for 68k Amigas, and probably also not really for PPC Amigas. The MPEG1/2 or MPEG4 stream send by these adapters is far to CPU intense to decode to the graphics card in realtime (e.g. to watch TV).

The output format is probably freely configurable within the limits of the DAC. I don't know if PAL or NTSC can be driven by the chipset. The main use is for high-resolution TFTs though.

And regarding your question with the specificy: Yes, the driver would only work with one adapter, with the same kind of chipset.

[platon42]

wawrzon wrote:
i can try to get hold of hardware in question and send it to chris. btw i do not need no refunds in this case. but what i wonder is what kind of gfx card are we getting here, does it have some 3d capabilities such as could be used by w3d or so?

No 3D support planned.

[platon42]

bloodline wrote:
Does chris have any experience writing P96 drivers?

No, I don't, and I wouldn't write a P96 driver rather than a cybergraphix driver because I have access to the development material (and sources) there and the developer of CGX is part of the MorphOS team that I can frequently ask, if I'm stuck somewhere.

[platon42]

wawrzon wrote:
ah, and how about this:
http://www.diwa-shop.de/idealo.jsp?artId=0-61-DA-70820
would it be too much for a classic (say even with a ppc)?

Probably.. If I'll ever write the UVC webcam class driver and you find a grabber that also uses the USB Video Class standard, we will find it out.

[platon42]

arnljot wrote:
@bloodline
I guess Platon42 needs a P96 driver if he wants to test this, and has thought of this when he mentioned the product which he'd looked at. Although, I think he can write such a driver, Chris himself is the one most qualified to answer that question. So you should direct your question to him, and not the thread

@Platon42
I guess/hope too that the Z3 DMA speed should be sufficient, as the driver would use the draw functions in the USB card and only copy rasters as needed from the system memory. But alexh's friends should confirm how these cards work. The full screen raster of what's projected to the user is composed and built by a graphics card, and hopefully this is how those USB thingy-madingies work... :-P

I indeed think that I would be able to write such a driver. But I'd make it CGX, not P96.

While I am sure the device has its own frame buffer, I'm not sure about the blitter capabilities of the device. In the worst case, all changes to the framebuffer have to be made locally and then transferred back to the USB adapter. But I really guess that there is at least some support there.

I'm not sure if this the worst case, but I expect the adapter to "simply" expose the chipsets register set and some kind of memory bus over USB control and bulk transfers. Maybe even ISO transfers, but that's no problem for the Deneb at least.

[platon42]

wawrzon wrote:
ok, the piece has been ordered. i hope it arrives soon. i'll keep u posted.
@platon42: is this ur current adress i in my deneb manual? otherwise pm me where to send it, please.

Which piece exactly did you order? Hopefully not the video grabber... :-(

My current address is not in the manual, please PM me, in case it is needed.

[platon42]

wawrzon wrote:
the video adapter has just arrived. im going to forward it to chris as soon as its possible.

The eagle has landed. wawa personally came along at my office and handed the box over. But it is uses same DisplayLink technology (quick test on my office's laptop) that alexh told we won't get our holds on. I will try a short analyser run on it anyway.

But about false advertising: The adapter sold at Pearl only goes up to 1280x1024 and not to 1600x1200 as they say on the website. wawa should report that.

It uses the LibUSB, which is LGPL & GPL, so at least here we could demand the source code for. I doubt, it will be any use though.

[platon42]

Well... what can I say, probably *all* USB Display Adapters use the same kind of chipset: The DisplayLink DL-120 or DL-160 (latter has the higher resolutions). I opened the box and the thing consists of only three chips. The very fat ASIC BGA DL-120, a DVI encoder chip and 16MB of DDR memory. 16 MB, seems a lot for an adapter that can only display 1280x1024 in 24 bit. But then again, it surely has at least two frame buffers (double buffering) to avoid tearing effects and probably a rather large decompression buffer for the compressed data. The device only has one bulk output with 512 bytes per packet and on interrupt in endpoint. The rest is done via control transfers.

This means, there is no way the to read back pixels from the framebuffer on the adapter, only a "write to memory" feature. This means the Amiga has to have a copy of the current framebuffer in memory. One could save the memory for the delta buffer, if the MMU would be used to calculate accesses (like the Savage Shapeshifter driver).

Except for the decompression, I expect the device to be rather simple hardwarewise -- maybe the registers of the DVI encoder chip are directly accessible over control transfers. If that's the case, luckily the docs of the chip are available on the net.

I'll keep you informed how it turns out. But I really need some more sleep...

[platon42]

Zac67 wrote:
Just got a DeLock 61644 USB-DVI adapter for evaluation with our notebooks (all are docking and TFT equipped, but DVI-capable machines are out of the budget).

The DeLock looks *exactly* like the one wawa provided. If can determine the DL-chipset by opening it. I would like to know if it is the DL-160, because according to the specs, it should support 1600x1280, which mine doesn't :-(