On Aug 22, 2007, at 6:26 PM, Dieter wrote:
Yes, it does. Of course, FireWire is more reliable and works better.
Does Firewire retransmit if data gets corrupted on the wire?
Hmm, I assume it is up to the applications involved -- it is a memory
mapped interface. At the very least it can run IP over FW and not
worry about it.
I think that we should not worry too much about FireWire just yet.
Yes, our big worry is decoding the video.
1394c supports autonegotiation of Ethernet or FireWire and runs on CAT5.
so we are probably going to pay around $50- to get the
processing from an FPGA to implement 720p without a DSP.
I was under the impression that FPGAs were hundreds if not thousands
of dollars? And yet $50 worth of FPGA can decode 720p video? Is this
H.264? I presume we also need to decode 1080i which would bump the
requirements up a bit.
Yes _good_ FPGAs are thousands of dollars ;-)
There is a large (mobile phone/PDA/etc) market for lower end FPGAs that
provide a processor bus for ARM and StrongARM CPUs, so they have gone
really far down in price.
Yes, decoding 1080i would require more processing power. The real
difference
is that 1080p has 129600 4x4 Macroblocks/sec, where as 720p only has
57600.
I am still trying to figure out if we can half that number for 1080i,
which means that
it should easily be doable, or if MB bandwidth is the same for 1080i
and 1080p (if
it was, there would be less 1080i and more 1080p support, so I have a
feeling it is
not, but I am not sure).
Could the OGD board be used to test the decoder?
Yes. It will have an S4000 on it, much more than we need. We could
either try to
make the design entirely in the FPGA (better idea) or use the IDC
expansion bus
for the DSP (we should probably only do this once we are sure a design
will not
fit in the 1000, 1500, 1200E, or 1600E, since it would really reduce
our BOM cost
to only have the FPGA).
eek! turns out that is quiescent current, so the actual (using as much
information as I could from that Ocean Logic article) estimate is more
like 600 mA, not including I/O! (700 mW)
IIRC some of the dedicated decoder chips are in the 0.75-2 Watt range.
I was under the impression that FPGAs were power hogs?
Yes, they are also decoding 1080p, over twice the bandwidth. I based my
power on the Spartan power calculator, which is not too accurate, and
the
110Mhz clock speed noted in the Ocean Logic article. The DSP draws
around
800 mA for DSP and ARM at 1.2V and 100 mA at 1.8V for IO.
This doesn't include any sort of memory, which adds up pretty quickly.
Assuming we go with 256MB of DDR2 SDRAM, there are 512Mb (32MB x 16)
667 Mhz DDR2 chips that sell for about $7-8 Qty 2 on Digikey which use,
when
all banks are open and a read or write is going on, about 150 mA each.
We need
four of those, so it would be about 600 mA (RAM) * 1.8V + 700 mA (FPGA)
* 1.2V +
100 mA (DSP IO) * 1.8V + 800 mA (DSP core) * 1.2V is about 3 Watts, so
the whole
system would probably run in at about 5 Watts - not bad.
The DSP also has some DACs and video encoders, so we don't have to
worry about
that.
What sort of outputs should we have?
composite/component/HDMI/DVI/S-PDIF/?
If the total power consumption is less than 10 Watts or so I'll be
happy. I want to avoid the 100 Watt sort of power consumption that
a CPU or GPU solution burns, along with the noisy unreliable fan.
Agreed.
} The DM6443 looks like a better fit than the DM6446.
} The 6443 doesn't have the camera controller, but is (I think)
} otherwise the same as the 6446.
The VICP is also not there, so I don't have to worry about what it does
;-)
The 6446 is designed for transcoding, so it has a resizer, etc, but I
think
you are right -- the 6443 should be fine.
} The DM6443 is less expensive than the DM6446, and there is a
significant
} price drop going from 1000 to 10,000. The DM6443 is $29.95 in
quantity
} 10,000. I wonder what the price is for 100,000?
}
} TI has a development system for $2000.
}
} http://linuxdevices.com/news/NS9968931411.html
IIRC there was a mention of a newer model a bit faster than the
DM6443/DM6446?
(720 MHz vs 600?)
I am not sure about this.
Do we have any idea how many of these we could sell?
The Turbo.264, a Mac H.264 encode hardware accelerator the size of a
USB Flash drive
(with USB), supports up to 800x600 resolution out and costs $99. If
ours costs in the same
range (assuming we don't need the DSP, we might be able to charge such
a low amount),
supports encode/decode of 1080p, and can run Linux/custom FPGA and DSP
apps, I think
that we can get a lot of sales. Add in the built in TV out and
networking (along with the USB
we get for free with the 644*), and we should be able to sell at least
as many as that Turbo.
Remember, we should be able to support Linux, Windows, and Mac OS X,
which should
increase our sales.
I don't know how many Turbo.264's have been sold.
Since the DSP has an ARM on it, we could actually run X11 on the
device, thereby completing
our original goal...
_______________________________________________
Open-hardware-ethervideo mailing list
[email protected]
http://lists.duskglow.com/mailman/listinfo/open-hardware-ethervideo
