Simple memory card player idea

[Karen O]

Hi,

An idea I had that would permit viewing of 405 line material is a simple box that can read uncompressed 405 line video data from a memory card and generate a video waveform for modulation.

Software would have to be written to prepare the uncompressed data for the card (which would be written straight to the physical card - i.e. no file system involved)

Memory cards are big enough now to get hours of uncompressed video onto!

[Station X]

Hello and welcome to the forums.

What sort of 405 line material did you have in mind? The problem we generally have is in converting 625 line material off air, on VCR or on DVD to the 405 lines needed for vintage TVs.

[dominicbeesley]

Hi Karen and welcome to the other side!

This might be a bit more difficult than it first appears, there are all sorts of issues reading a steady stream from an SD or USB memory card, however with a bit of buffering you should get away with it.

If you want some formatted data give me a shout I should be able to stretch the bounds of my NBTV recording software up to 405i...

Dom

[neon indicator]

It's nearly possible on a high end PIC 18F.

It can do 12MIPs.
if you use 8 bit shift register and only 2 levels (white and black) then you need 0.65 MHz data rate ( dot clock about 5.2MHz, allows external LC smoothing).

For 16 levels you need 4 bits and 2 x 4 bit DAC and 2 way switch at 5.2MHz. Data rate 2.6Mbytes /s

For 256 grey levels (quite good) you need 8 bits and data rate twice the highest frequency if you want it to look good. 5.2Mbytes per second would be OK though.

That's too ambitious for a 18F PIC. You would want an ARM or MIPS or x86 running at about 700MHz or more. Still quite cheap.

A Spartan 3E FPGA would be possibly superior and cheaper than an x86 solution. It could have RGB or Mono input at 525line or 625line and real time convert to 405. It can easily read SD cards and USB sticks and ethernet too. The 8 bit DAC for video out is cheap. The full spec ADC to read video in is more expensive item.

Note most of the cheap analogue HW on older TV tuner cards for PCs does at best 384 x 288, almost none can do 768 x 576 (ideally 625 line video needs sampled at 1440 x 576 and then downsampled to avoid aliasing).

Most Setbox and TV out graphics cards run the DAC out at 1440 pixels per line and then filter in hardware down to 720 pixels.

For simple uncompressed file the Windows BMP format works and is easy to read off a SD card. TIF, PNG and GIF need more work as they actually are somewhat compressed (without loss). JPG is lossy compression.

So for 16 grey levels from Black to White, I think it can be done with a £4 18F4550 PIC, 9 resistors for each 4 bit DAC, a two way cmos switch and a dual video opamp. Very cheap and simple. The PIC can read SD cards with socket wired direct and can connect direct to PC with USB. The software to do it with a BMP file is simple. There is even a FAT filesystem library for PIC using JAL.

Full 256 level is beyond cheap hardware. I'm not even 100% sure the PIC18F can do 16 shades.

[Karen O]

Thank you all for the keen interest and thank you for the welcome!

I worked out the number again last night:

A 16GB card could store an hour of uncompressed 405 line video. That's with 500 8 bit samples per line and 8 bit audio at 18k sample/sec (twice the line frequency).

The data would need to be streamed from the card at 40Mbit/sec. The faster ones can do this.

My plan was to use a PIC to supervise only - hardware would do the high speed shifting etc.

It looks kinda within reach but there are downsides I can see. The preparation of the data could take a prohibitively long time. Heaven knows, it took forever to move everything over to my new memory stick (admittedly, USB).

Also, these memory cards have status which has to be checked before starting transfers. I don't know if there are any reasons why the card should become uncommunicative...? If there's a micro in there it might take a nap while to does its Java garbage collection

I'll do some more research.

It's nearly possible on a high end PIC 18F.

and a PIC32 (80MHz 32bit) is only a few quid and could reconstruct an mpeg file on the fly too, do have fun.

[MikeyPP]

It's nearly possible on a high end PIC 18F.

and a PIC32 (80MHz 32bit)

http://www.linusakesson.net/scene/craft/

[neon indicator]

A PIC32 isn't a PIC in traditional 10F/12F/16F/18F sense

It's a MIPS core controller made by Microchip and "branded" PIC. I said a MIPS could do it. But I'd use a higher end ARM with Linux rather than a MIPS.

You'd be better off with a cheap Micro ITX Mobo and Linux and reprogramming the VGA timing.

[Karen O]

A little like this one:

http://www.kohsystems.dyndns.org/dok...basic_computer

(Be patient with this link - we host our own website and the ADSL link has a low upload speed).

The video generation is not the obstacle. It's getting a steady 40Mbit/sec out of the memory card that I'm not convinced of.

It's clear that this project could easily blow into something that is not simple and not cheap, in which case there is no advantage.

From what I've read the Aurora is the optimum solution and I'd prefer to buy one than spend fifty pounds on components for something that is limited to pre-recorded material.

Also there is the not insignificant matter of the modulator. That alone will involve numerous components.

I'm going off the idea
KO

Daddy, this TV takes ages to come on and show a picture. Can I use the valve telly instead?

[neon indicator]

The modulator is the EASY bit. Really.

p.s. Hosting is only £20 a year.

Your PIC box is cute.

[ppppenguin]

An Aurora really is the answer, does the whole job and it just works.

If you're interested, you might want to read the Aurora reviews and standards conversion articles on my website. There's another standards conversion article that isn't on the website yet. It's published in the latest BVWS Bulletin.

[evingar]

In some respects, it depends on whether the project is a means to an end or whether an engineering challenge is sought.

If the former, I would agree with Jeffrey. The Aurora is a very cost effective way to get first rate results for providing signals for 405 tellys.

[murphyv310]

Personally I can not seeing it working. Virtually no Video card will work on 405 and you wont be able to encode either. Windows is a no no and even Linux wont work with certain cards.
As Jeffrey says dig deep and buy an aurora, there is "no contest"!

[Kat Manton]

Originally Posted by murphyv310

Virtually no Video card will work on 405

Well, hardly any except for several fairly common nVidia cards which can be had s/h for about a tenner...

... which I've persuaded to produce 240-line, 405-line, 819-line, CBS field sequential... pretty much anything, as long as video timings are specified explicitly and are based on a pixel clock frequency which the card genuinely can generate.

You're still left with having to build a dedicated PC with all the right bits in it, set it up as a MythTV system (rather than as a general-purpose desktop PC), build a circuit to generate the appropriate composite video signal from separate RGB video and syncs, build a modulator...

Many hours of fun if you regard playing around with computers, operating systems and software as entertainment; a frustrating exercise if you just want to watch telly.

I agree; an Aurora "really is the answer, does the whole job and it just works." And keeps working...

(I still fancy designing and building a pure hardware standards converter, just for the hell of doing it, though.)

Back on topic... this idea sounds conceptually similar to the R T Russell Test Card Generator but with considerably more storage.

Producing the data to write to the card would be fairly straightforward. FFmpeg can decode video files, scale to whatever resolution is required (I'm using 504x376 as 'digital 405'); convert to greyscale and write individual uncompressed frames (one frame per file, sequentially numbered) in .bmp or whatever. It'd then be fairly trivial to read the files, strip out the data, re-arrange it as required then write it directly to the card. Piping the output of FFmpeg into the card-writing software would be better (it wouldn't require disk space for intermediate storage.)

So IMO getting the data onto the card is the easy bit.

It's interesting to think about how to do this regardless of whether or not you choose to build it

Kat

[julie_m]

I agree with Kat.

It's ultimately a matter of whether you like the experimental, "saw a bit off, weld a bit on" side of things for its own sake, or whether you just want stuff to work.

[neon indicator]

I like doing stuff partly because it's there. But in this case I'd only build it for 2 level B&W, no greys, and just caption overlay (i.e. as addition to Aurora) with PS/2 keyboard socket (only needs 2 pins on PIC, simple SW) and 8 bit parallel to Serial shift register to AND or NOR into the video stream. I'd stick a resizeable and positionable Analogue clock in too as I have done the 16F/18F code and it uses a look up table for sin/cosine so is very fast.

Cheap PIC18F4550 on veroboard with USB for PC, video pass through on RCA jacks and PS/2 socket for captions/titles. simple 4 x 4 matrix keypad built in with letter input via numpad.

As soon as you add greyscales the complexity and CPU power need rises so rapid you end up using a high end MIPS/PIC32, ARM, x86 or cheap FPGA and a competitor to Aurora. Also to convert live video you then double the complexity.

A modulator is simple. Video & sound, on veroboard is two transistors. I've built up to 3GHz on veroboard (glue copper foil on blank side and use SM parts on track side). VHF is pretty easy on Veroboard. I've made a UHF 625 modulator, 405 is lower sound oscillator AM instead of FM (no varicap needed) and the video modulator is just opposite sense. You modulate sound (3.5MHz?) and add that to video, then modulate both on mixer/osc at 45MHz or whatever. For 625 you use a 6Mhz IF filter as oscillator pulled by varicap. but 405 you can likely just use a free running osc and "tune it" with a frequency counter or with the target TV set.

[Karen O]

Oh, I'm definitely doing all this for the challenge. But I think too it's a kind of resistance fighting against the tyranny of commercial electronics which seems determined to lock us hobbyists out! I will find a way to infiltrate their impenetrable protocols (cue evil cackling).

It peeves me that there's so much power in modern consumer electronics but without documentation we can make little headway in re-applying it. I would not be surprised if your average set top box were capable of generating old TV standards with just a firmware change.

Anyway, the only way I can see of making my idea work cheaply is to desolder the flash memory chip from a USB flash drive and put it in a custom plug-in module - based say, on a twenty pin header socket (there are only actually twenty used pins on a flash memory). A mating header would be fitted to the flash drive to form a programming unit for my custom memory card

[ppppenguin]

I'd be tempted to start with a flash card, rather than a USB stick, because the interface is simpler. Then use an FPGA to do all the logic. You can get various modules and evaluation boards, some quite inexpensive, that make it easy to use the ball grid and other SM FPGA parts.

As for reprogramming a set top box to do 405, I think not. The NTSC/PAL outputs are usually generated by a dedicated chip such as the ADV7311. Actually one engineer did manage to make the earlier ADV7171 coder chip do 405 NTSC. I once tried to do it myself and failed so all credit to him. I doubt if the later and more complex coder chips will be so amenable.

Another friend with little design experience successfully built a test card generator using a CPLD and some EPROMs. He was thinking about modifying it to use flash memory. I persuaded him to use a Xilinx Coolrunner CPLD rather than TTL/CMOS logic. He was very dubious at first but I helped him a bit with the VHDL and he did very well.

[neon indicator]

SD card trivial. USB stick needs USB Host stack. Not trivial. The 18F PICs can only be a USB client for instance.

A Spartan3E can be bought cheap. Well able to do this and you could put Soft CPU core, USB host etc, etc.
But for analogue video in you need to add a decent Video ADC.

[ppppenguin]

Originally Posted by neon indicator

A Spartan3E can be bought cheap.

I would recommend a Spartan 3A unless you have a specific reason to use 3E. The 3A emerged as the mainstream series ahead of the 3E which I think of as a bit of a backwater. I know the Aurora uses a 3E but not sure Darryl would recommend it for a new design either. Actually all the Spartan 3 stuff has recently been taken off the main "recommended for new designs" list. Spartan 6 now seems to the main low cost series but I would have no worries about using 3A for quite a while yet.