Re: Archos PDA + Gumstix as core of a wearable computer

From Wear-Hard Mailing list Archive (WH) Maintained by R. Paul McCarty

Chris Saari wrote:

>
> Has anyone tried the Archos PMA430 with video input/encoding and video 
> out in a loop for EyeTap type applications? Is it fast enough to get 
> low latency video input back out to the VGA *and* do 
> archiving/encoding to disk at the same time? Is there any CPU left 
> over during this process (I know it has dedicated MPEG encoding 
> hardware DSP, but if the video caputure buffer and disk management 
> takes all the CPU due to bad software design you still can't do much 
> else for lack of cycles).

Its actually worse than that.  Archos has provided a fairly useless SDK, 
and haven't documented any of the hardware.  Most software development 
is currently restricted to porting Zaurus Qtopia apps over right now.  
We haven't even figured out what changes they made to cramfs so that we 
can burn a new system image yet, we're currently restricted to 
scribbling in the FAT32 "application data" file system.

Even assuming that you could work past the undocumented API issue (their 
apps seem to cheat and go directly to hardware), and could write a 
captured image to the framebuffer relatively snappily, it seems like 
there is some significant buffering involved internally because even 
when running their videorecorder application without recording, there is 
noticable latency between the camera and the framebuffer that would make 
an eyetap user nauseous.  I have a sneaky suspicion that the video input 
is tied directly to the MPEG hardware, and there isn't actually a way to 
intercept the raw frame data, only the encoded output stream.  Oh, and 
there isn't actually VGA out, its only analog out.  And when its 
powering the analog video output, it sucks batteries like mad.  You'd 
definitely need external power.

> The more hard data the better, like cpu load during the video in, 
> capture, out routine.

Its hard to tell, since Qtopia suspends non-foreground apps.  I'll try 
to log in via the network and check the load.

I'm convinced that the hardware internally would be powerful enough to 
act as a useful wearable, but the lack of docs and the lack of modular 
software architecture makes life difficult, so I'm currently just using 
it to record motorcycle rides.

-rg

_______________________________________________
Wear-Hard mailing list

http://www.haven.org/mailman/listinfo/wear-hard