2 Jun
2015
2 Jun
'15
8:47 a.m.
On Jun 2, 2015, at 08:25, Alan Hightower <alan at
alanlee.org> wrote:
As far as the crazy cat lady project, as I was trying to explain
earlier.. my $.02 which maybe redundant to your own thoughts:
You did inspire my choice of project name! :)
VIDEO ADC/DECODER
You can certainly use an off-the-shelf part as there are 47 from 5
companies on Digikey alone in solder friendly packages. However if the
classic machines you are targetting are producing a clean conforming
NTSC_M/J, PAL_B/G/N, or SECAM signals, stock video decoders on most
converter boxes would work perfectly already.
Right. It's the "clean conforming" part that I have heard is the issue,
though my initial success with the first sub-$20 converter I tried makes me wonder whether
the perceived desire for Crazy Cat Lady is overstated.
I doubt you will have much
luck with the NDA process explaining a hobby cause but worth a try.
I also thought it was a long-shot, but they sent me an NDA to sign last week and now
I'm just waiting for the datasheet.
Even
if you are successful, make sure the parts you are looking at are
readily available in 10s and 20s quantity through a distributor chain at
a reasonable cost. That may be a second hurdle if it's fairly exotic.
What SiLabs part are you looking at? I can't find any recent video
decoder offerings.
The part in question is a tuner, not a decoder. It's Si2137, with 130 pieces on hand
at Mouser. But for all I know, they might have bought one tray of somebody's leftover
production parts, with no more to appear. If I use the part, I'd consider putting the
tuner on a separate board with the expectation that the Si2137 might dry up without
warning.
GENERIC ADC/DSP ANALYSIS
This is the harder path to go down. You will need a symbol rate of at
least 54 Mhz from a single or two interleaved ADCs to have a shot at at
least 50% phase accuracy (NTSC = 13.5 MHz). There are some relatively
inexpensive options, but I suggest getting a lot of input from folks and
other reference designs on the best way to build an analog front-end for
composite.
Definitely a learning curve here, and I will welcome input on the analog front end.
And only do composite/S-Video. There are already
commodity
solutions for stepping down broadcast RF frequencies to base-band.
Right. I'm looking at a $2.30 commodity part for the tuner, which is probably 1/10 the
cost I could roll it myself. But having a tuner is an important feature, as most of the
computers I'd want to target have RF/TV outputs, and I'd want to support them
without requiring modifications to pull out composite.
After
that, you will need something fairly hefty at the start to find the
characteristics of the signal and align the sampling. Then you just need
to track clock drift and adjust a VCXO.
I was wondering whether I could get away with tracking the clock drift digitally rather
than closing an analog PLL. What do you think?
Straight-forward in hardware,
but I doubt many people will have the experience to add new software
support for <insert eclectic hardware platform with slightly
non-standard composite output here> and fix bugs when they occur. So set
expectations on your time early with respect to project support.
My expectation would be that I would make the design open, but then be very surprised if
anybody rolled their own improvements rather than making sad puppy eyes and asking me to
add support for their favorite neglected computer. :)
SOC/ZYNQ/MARKET
The fact you mention ZynQ throws up a few warning flags for me.
Understood!
It's probably overkill
Certainly.
, it's expensive in low quantities through normal
distributors, and it's not very hobby friendly. $65 for a '010 alone +
DDR3 + BGA assembly would drive the resulting cost for you to make these
boards well beyond $200.
Agreed. But I do use them in my day job, so there's a strong familiarity factor. And
another thing that I find compelling is that pushing FPGA + firmware updates would consist
of instructing users to place one or more files on an SD or MicroSD card, rather than
buying a programmer cable and installing a vast FPGA development platform, without me
needing to roll my own FPGA reconfiguration solution. If I rolled my own with a cheaper
FPGA + microcontroller, it might end up cheaper in BOM cost, but maybe not by a large
margin.
And one of my challenges would be for me to try to get the BGA routed on a 4-layer board
by careful selection of unused I/O pins and the largest-pitch package available. It's
not uncommon to use 16 layers to escape those beasts with all pins used (and I've done
it myself, on more cost-insensitive designs), but that wouldn't fly for this product.
I won't know if I can do it until I try. It's promising that there are Zynq 010
dev boards in the $100 range.
--
Mark J. Blair, NF6X <nf6x at nf6x.net>
http://www.nf6x.net/