ATI TV Wonder USB 2.0 works, power supply issue

[jwillis84]

I also have a StarTech Empia 2860 - USB2 capture device

SVID2USB23

It works really well under Windows 10 x86/32 with no problems.

One of the things I admire about the Empia 2860 and the device drivers is that its INF file is (beautiful) and fairly easy to read. The chipset is from 2004 as well and it very widely known, the datasheet is readily available. Sometimes people find the decoder chip in cheap knock offs, or combined with dubious AC'97 or other audio encoder chips.. the knock offs aren't well known for stability or good signal handling.. but the computer side of things for windows or mac, twain or directx is very stable.

I might see if the INF can act as a template for re-writing the INF for the ATI TV Wonder USB2.0n

I don't know its an INF problem exactly, it could be a coding problem in the ATI drivers. But the Win7 Empia drivers and INF file work really well under Windows 10 without modification.

Startech generally offers some generic hard to find stuff.. but there are always datasheets.. which I like about their stuff.

I can only guess that the ATI USB drivers weak point is the INF file.. its only a guess.

The few times I've peaked at it.. it looked.. unfinished? beta.. in a state of flux? I really don't know if that's fair.

The hardware is solid, the Theatre 200 chip had a long life in many products.. but the software side of the drivers was perplexing.

[jwillis84]

gleened a little bit from the inf setup file

it says these are the drivers and what they do:

atinyuxx.sys ; ati ntsc ; TV Tuner
atinysxx.sys ; ati ntsc ; Sound Crossbar
atinyvxx.sys ; ati ntsc ; Video Crossbar
atinypxx.sys ; ati ntsc ; Parental Decoder
atinymxx.sys ; ati ntsc ; Macrovision Decoder
atinyttx.sys ; ati ntsc ; Teletext Decoder

atinymvx.ax  ; ati ntsc ; MVDetection activex filter
atinytmx.dll ; ati ntsc ; MVDetection function driver

atinyc20.cod ; ati ntsc ; DSPminicode (VideoSOAP maybe)

Only the first three are really related to AV capture, the others enable more features in the ATI MMC software.

Each has its own little stanza that builds up an interface that DirectX or another program could use.

As far as I can tell the inf file was used as a scratch pad for ideas and certain parts are commented out or look like they were "beta".

But since it was based on the use of DirectX and WDM its fairly similar to the inf for the empia 2860 capture device. I can definitely see where a little text formatting problem could be tripping it up from working well under Win7 or Win10.

This could be a dead end.. along with a modified inf file is a security .CAT (catalog) file which is suppose to be signed and act as a token to validate a driver package as valid and verified by a developer.. that's what costs a lot of money to sign and keep up to date. I am not sure how to get around that yet.

But it is interesting to read the plain text file and compare and contrast to the same plain text file from a different but similar driver.. its very educational, if academic.

p.s. One other thing

The USB device driver is in a separate driver inf and driver file completely. That driver gets loaded upon detecting a Cypress USB controller that identifies itself as an ATI TV Wonder USB2.0 with a VEN and DEV id, when that loads it enumerates the other logical devices attached to it.. which appear as the "three" hardware devices above, tuner, video xbar, sound xbar which then automatically build their interfaces and plug into the DirectX plugin gallery ready for use.

I presume the protocol then is to send USB packets of data labeled or "addressed" to one of the three devices with instructions or commands to perform. The rest of the device drivers are "virtual" devices which carry out their orders in virtual logical devices that examine the data stream coming into the operating system from the other USB devices.

amazingly

uxx is 63 kbytes
sxx is 78 kbytes
vxx is 171 kbytes

total 311 kbytes

that's the difference between a 32 bit or 64 bit driver to keep this hardware going

[lordsmurf]

Update:

Direct Win7 x86 install, no Vista start (ie upgrade Vista > 7 not done here).

Win7 x86 + AIW PCI + MMC = good!
The contrast/bright/color/tint/etc was 0 by default (0-255), so completely black. Changed to 128 in registry, all is well.

Win7 x86 + AIW PCI + VirtualDub with overlay = does not work
Only the laggy preview, which sucks, and I consider to make a card unusable, due to not being able to see output 100% motion quality.

Currently capturing VHS retail tape, with AVT-8710, at 20mbps in MMC for testing dropped frames.

This system is unstable, and refuses to shut down. Must hard power off, ignore warnings at boot that it didn't shut down correctly.

Will reformat, try again soon. Will see if direct Win7 repeatable, and hopefully without shutdown error.

[lordsmurf]

Correction: Only the video works in MMC, no audio. So not good.

This is odd: When you enable Audio Playback in VirtualDub, the overlay works, though jerky. Huh?

[jwillis84]

Its hard to comment without relating things I'm not sure about.

From rewriting the INF it looks like they were confused, or Microsoft switched the default multimedia class devices underneath in DirectX for the Audio and Video crossbar.. originally it looks like they used the same one.. later Microsoft definitely introduced a specialized one for Audio versus Video.

I am an amateur at this and DirectX technology.. so I'm very unsure of what I am looking at.

Comparing it to the empia2860 INF which does very similar things flawlessly on Vista, Win7 and Win10 it "appears" this can be sorted out.. no guarantees though.

If it works with the ATI USB 2.0n it is (possible) similar fixes might work with the PCI models as well.

I was deep diving in old MSDN archives last night and (saw) that Win10 x32 does not enforce driver signing.. so another reason to use 32 bit versions of windows.. no vendor lock-out for old hardware !

closer I get.. more demands on my personal time.. so i may be taking a break here..

i (did) finish rewriting the INF file for ATI USB2.0n but have not tested it

its not reverse engineering or code rewriting.. its just rewriting the instructions to the operating system on how to use what's there.. the drivers themselves "could be" fundamentally be broken.. but time will tell

[jwillis84]

lunch hour

.. a bit more explaining

DIrectX has well known "GUIDs" for specific interface types.

The INF file tells the Operating System this or that driver should be "hooked up" to a specific GUID.. and programs running on the inside of the Operating System "expect" specific behavior from devices hanging off those GUIDs.

When a GUID "changes" as it did in various DirectX versions, the whole behavior can be renegotiated .. its a different Interface.. and reliability goes South.

DirectX is not "supposed" to change the behavior on old Interfaces going forward.. but they stop testing them too.. so they are basically unmaintained.. also.. if the GUIDS were perfect "why change at all".. basically they were discovered not to be "perfect" and had problems.. so they made new ones to replace them.. the old ones continued with those bugs.

You can't always just "re-rig-the-plumbing" behind the scenes.. things can stop up and stop working.. but that is [kind of] what I'm trying to do..

In the ATI driver world, there are only three important drivers.. tuner, video, audio

They are hooked up by the INF file to GUIDs using the windows registry as "glue" that tell's the operating system they have a "tuner" a "video" and a "audio" device.. and that they have specific features by default.

Windows then loads the binary dll's called .sys files and starts "probing them" to make sure they are happy and sane.. that's what shows up in device manager. The result of those device "tests".

If it works then you have very common DirectX GUIDs also called Interfaces you can write programs to use.. without writing hardware specific code.

ATI drivers are 32 bit and have worked on everything from Win95 through Win7 (mostly) but the INF file for "gluing" things together was kind of not quite explicit.. it left a lot up to the imagination of the operating system to flesh out.. and sometimes the operating system guessed wrong.. and the drivers didn't quite work.