8-bit Computer TV Channel Use

Mark J. Blair nf6x at nf6x.net
Sat May 23 11:57:25 CDT 2015


> On May 23, 2015, at 09:25, tony duell <ard at p850ug1.demon.co.uk> wrote:
> 
>> Handling the VHF/UHF tuner economically may be another sticky point. Maxim makes a tuner chip that's available 
>> at Digi-Key, but I refuse to design Maxim parts into anything on account of off-topic reasons. Mouser has stock of 
>> a very inexpensive ST tuner chip that looks very promising, but the full datasheet isn't openly available. I need to 
>> contact ST to see if I can talk them out of it. Their site mentions an NDA for the eval board, so it might be tough, 
>> particularly since my intention would be for my design to be open to allow off-label uses.
> 
> What I say below covers how it was generally done in the UK/Europe. No idea about elsewhere.
> 
> Building a UHF tuner (even assuming you have a working design) is non-trivial. To give you some idea, the 
> inductors are often straight metal strips, layout and length is critical. If you bend one, you throw the tuning
> way off....

Yup. I'm a radio amateur and an electrical engineer working with GPS stuff, so I understand the pain! But nowadays, a VHF/UHF tuner is a single IC, possibly surrounded by a small handful of fixed inductors. The problems are:

* Might not be able to buy it in small quantity.
* Might need to read Chinese to understand the datasheet.
* Datasheet might simply be unavailable to individuals, even if the part is available.

I haven't begun trying to crack that egg yet, but there may a successful path there. With modern parts, even a dumb digital designer like me can successfully design RF front ends operating at 1.5 GHz, and cram an entire GPS receiver onto a fingernail-sized PCB. It's black magic if somebody makes the right chip. And it's still do-able if they don't make the magic chip... it's just a lot more expensive than consumer electronics have conditioned us to expect.

I think there used to be a rule of thumb for microwave work to not try for over 10dB of gain per inch of circuit, or something like that. But now, slapping down a 20dB or greater gain LNA circuit the size of a barley corn is no big deal. Being able to make the whole circuit a lot smaller than a wavelength eliminates a lot of the pain.

As an example, Mouser carries a single-chip tuner for $2.32 at quantity 1. It's by Silicon Labs (I think I mistakenly states ST earlier). But only a short-form datasheet is available openly, and I need to contact SiLabs to see if I can get a full datasheet. I also don't know whether this analog TV tuner chip will remain in production, vs. Mouser buying some TV manufacturer's leftover parts.

> The output of the tuner fed a suitable IF filter (a SAW device was about the easiest to use) then the IF
> amplifiers (video and audio) and then detectors.

The output of a single-chip tuner might also be at IF. The Maxim part (which I will not use) outputs at 36 MHz, I think. Can't tell the output of the SiLabs part without more info. Hopefully it's either baseband or a lower IF frequency that I could sample with a cheaper ADCs for digital down-conversion. Needing to support a 36 MHz IF would probably increase ADC cost vs. using ones that just need to sample baseband or a low IF.

SAW filters are also black magic, and nowadays they are TINY!

> Oh yes, in the UK the sound carrier was 6MHz offset from the video carrier, in the rest of Europe the spacing
> was 5.5MHz. So if you want to handle sound (some computers sent their sound output over the RF output) you
> may need to cover both.

And US NTSC puts the sound carrier at 4.5 MHz, so there's another thing in favor of using SDR techniques for some portion of the demodulation if I can't find a Magic Chip that does the work more cheaply. The chroma subcarriers are also at different frequencies in the various standards.


> 
> At the output of this section you had composite video and line-level audio. What you do with those is up to
> you....

And that's where the fun begins! The plan is to infer what color the vintage computer was trying to display at any given pixel, with knowledge of the dirty tricks it used to get that color cheaply. Then cram that inferred pixel into the frame buffer, and convert the video format on the other side of the frame buffer.

-- 
Mark J. Blair, NF6X <nf6x at nf6x.net>
http://www.nf6x.net/



More information about the cctalk mailing list