HDMI output for A1200?

[CarlK]

Hi everyone...

Yesterday I bought me an A1200. Out of pure nostalgia. Yes, I'm one of those who was a teenager during the 80's, and never forgot the best computer ever made! Finally I made the decision to spend some money and see if it really is as good as I remember.

This time aroud I got me the machine I never could afford back then (I used to have a modded-beyond-recognition A500)

The world hasn't stopped eveloving though, and I quickly realized that today I have a problem displaying the output of my dear A1200. Reading the hardware thread, I see that many other face the same issue. There seems to be a remedy on its way from Individual Computers, with the announce of their new scandoubler-flicker-fixer-frame-buffer -thingy...

To the point:
I myself enjoy electronics and hardware as a hobby, and recently investigated a video enhancement for the old M$ Xbox. That project didn't really make sense though, but now I clearly see a use for the knowledge I gathered: To me, it seems possible to add a top-modern HDMI v1.3 interface to the A1200 (And probably any Amiga using the "Lisa" graphics controller chip).

I just read the Lisa specifications, and by having a 24-bit graphic port and being clocked at 28MHz, it could fit within the specifications of a Silicon Image Sil9134 or Analogix ANX8560.

There are quite a number of things to work out, such as:
* Where do the I2C control interface hook up? (clockport?)
* Should the native amiga audio be digitized and fed along the HDMI I/F?
* Will the HDMI TX accept the Amiga resolutions?
and many other things as well, but all in all it does seem doable...

Any ideas and comments, specifically from people with experince from amiga hardware are most welcome!

/Carl

[meega]

Hello CarlK, welcome to A.org.

Not wanting to put a dampener on your ambitions, you might well succeed in your project, but have you actually tried using Workbench in SuperHiRes-Interlaced with 256 colours on an A1200 with no graphics card?

Time for a coffee (while the screen redraws).

All the best.

[sigmason]

It's probably do-able.  Just consider however that while this eliminates the 'interface' quantizing data through A/D - D/A conversion, it means you need to not only keep up with the video bus of the Lisa (not impossible) but also feed the HDMI display and deal with those issues in a predictable manner.

Remember that the inputs to the Lisa you plan on tapping while digital and RGB are meant to drive the logic of the Lisa.  The software creating those signals expects things like palettes and 'copper' (not the metal!) and you are getting in before all that.  So you'll have to replicate that function on your end.  I suspect todays programmable logic and MCUs could keep up.  How cost effective it would be, that's another question.

You should also be aware there are companies that make SCART to HDMI converters like Cyp and while they do loose some quality in the conversion, they likely work without you spending a good deal of time and effort.

http://www.cypress.com.tw/

Sigmason

[Eriond]

Hi Carl!

Glad to hear from You again. So, you too realized that the Xbox projct was hopeless  :roll:
Well, I've given up on that...

As for your ideas to apply the SiI9132 on the Amiga; its rather interesting! I'm still waiting for the real datasheet from Silicon Image, so maybe that alternative you found (Analogix?) could be something?

@sigmason: I have not been able to find a good block diagram over the A1200, so I can't really argue with you there, but from xperience with other systems (PC, Xbox and the like) the only thing you should find after the graphics controller is a video encoder of some sort. I don't beleive that the graphic bus is fed "back" into the amiga custom chips (like the copper).

Still, to pull this off, I need to do a lot more investigation and then there is this thing about expanding Amigas... Does anyone have a link to a working PLCC socket that would clip over Lisa (and not fall off as soon as anyone sneezes)? Would be greately appreciated.
So would a timing diagram of Lisas output be. The one in the specs was rather simplified. I would like to see a dump from an logic analyzer containing all 24 RGB lines, the C28OUT and BLANK.

But the more I think about it, the nicer it seems...
Hope we can avoid frame buffering, that would keep the design simple.

Damn you, Carl! Now you got me hooked on another mad project of yours :-P

Cheers,
Eriond

[alexh]

You could perhaps ask Jens schoenfeld to add a digital interface to his new Scandoubler with the ability to get the data before and after the scandoubler?

Dont forget that HSync and VSync are not available on the Lisa.

Jens has come up with a technique where he doesn't need them, but he's keeping the details close to his chest.

[sigmason]

Considering the Amiga must create all the signals it needs to produce the video output, I have no doubt that an analysis of the schematic after the Lisa chip should reveal all the logic that went into the production of the video signal.

Whether you replicate the logic or do it some other way is of course up to the person putting in the effort.

Here's a block diagram of the Amiga 1200:
http://www.amigahistory.co.uk/block.gif

Now consider that by taping the output side you are potentially creating yourself a new problem.  When you tap the input side you have to replicate the functionality to get output.  When you tap the output side, you still eliminate the A/D - D/A but then you have to consider that the Lisa chip will start to output all of the funky variations of video mode that make this so complicated.

As to whether it's better to simply 'replace' the Lisa or deal with the output modes I'll leave up to the person that  tackles this.

There's more information here:
http://www.amiga-hardware.com/showhardware.cgi?HARDID=13
(note the schematics file)

Sigmason

[alexh]

sigmason wrote:
Considering the Amiga must create all the signals it needs to produce the video output, I have no doubt that an analysis of the schematic after the Lisa chip should reveal all the logic that went into the production of the video signal.

Erm, lisa has RGB, alice has H&V. That is why classic scandoublers like DCE FlickerMagic clip over both chips. Like I said earlier, Jens Schoenfeld has worked out how to do without the Alice connections.

sigmason wrote:
Now consider that by taping the output side you are potentially creating yourself a new problem.

I dont think you have to wory about that. Lots of internal scandoublers do it without any problems.

sigmason wrote:
As to whether it's better to simply 'replace' the Lisa or deal with the output modes I'll leave up to the person that  tackles this.

You cannot 'replace' the Lisa.

[Eriond]

Ok, interesting comments. Thanks to all.

I've also requested datasheet for the Analogix ANX8560 (although its "just" a SiI9134 clone). Until the hard facts arrive, I can only do qualified guesswork...

As AlexH wrote, you cannot replace Lisa :-o . Lisa is the chip that actually produces the pixels! My (and Carls) approach is to replace the AnalogDevices ADV101 DAC with a more up-to-date solution, a HDMI transmitter.

What I have read so far in the product briefs, is that a HDMI TX takes a raw garphics bus and repacks it to fit the HDMI standard. Looking at the block diagram of such a chip, there is a block called "Video data capture logic", which to my opinion should sample pixels.

The really interesting question here is at what rate does this sampling occur? Most HDMI chips claim speeds of up to 165MHz, but only a handful specify the lower limit which I guess is more interesting for us who want to connect it to a old/slow graphic chip, like the Lisa. Silicon image say their version begins at 25MHz which would be slow enough.

However, the question is still; is it the graphic chip that dictates when each sample occur or is it the internal clock of the HDMI TX? The difference is very important, because we want to have a resulting image data stream that is compatible with todays TVs and monitors. An aqward propriatary Amiga resolution won't do us any good, even if it is re-packed in a HDMI stream.

Hopefully it works something like this (and now I'm guessing): we tell the HDMI TX which resolution we want (480, 720 or 1080) and it will sample the pixels neccessary to acheive our requested output. But that is probably being optimistic :-(

If the number of scan lines sent from Lisa would fit into a HDMI standard resolution, then this would probably work. I'll have to do some forum searching to see if there are some specs somwhere :inquisitive:

Otherwise, I will have to go for a full-blown framebuffer, which means and advanced FPGA solution. In such case I could probably get hold of a HDMI TX as IP, and do everything in one chip... oh well, it looks I have to do like AlexH suggested and send a question off to Mr. Jens Schoenfeld. :bow:

The quest continues... 8-)

[leirbag28]

@CarlK


 I believe itis possible, but in a much easier way than you think.



@meega

You dont have to run Workbench in SuperHiRes Mode. You can just run it in Hires Interlaced and upscale it somewhat.

The easiest way Ive seen to get HDMI on Amiga is via and Upscale DVD player with S-Video inputs...........My DVD Player upscales my Amiga Output and outputs it to my FLickerFixing Projector via HDMI.........it looks the same as S-video but blurry somewhat because I guess the upscaler expects pixelization and overcompensates.

Therefore HDMI is possible on the Amiga via a simple scandoubler type looking device that attaches to the RGB port and does the Upscaling, yet supports SuperHires interlaced mode as well if someone chose to use that.

Personally I use SuperHiresMode for Text credits and scrolling credits on Video.........looks freakin awesome.


Sometimes getting too technical about things makes it seem impossible.  but it is easily achievable (and by that I dont mean that it is easy creating the hardware) I just mean the concept is easy.

The Amiga itself doesnt have to have a native HDMI output, because then it would need modifications and slow down the system...........

[alexh]

Eriond wrote:
The really interesting question here is at what rate does this sampling occur?

~28MHz (same as a TV) for all Amiga modes except Euro72, Super72 and Multiscan. (Non of which you really need to support)

Eriond wrote:
However, the question is still; is it the graphic chip that dictates when each sample occur or is it the internal clock of the HDMI TX?

I would imagine the HDMI TX would slave to the gfx input.

If HDMI supports 576i then it should be fine.

[sigmason]

AlexH:

1. Lisa can't make RGB without a synchronization.  It doesn't matter what makes the original H&VSync.  It must be possible from Lisa alone to determine sync (and obviously someone has.)

2. Scan doublers and flicker fixers yes.  If the intent is to effect that as the limit of the change.  In doing scan doubling you are limiting your ability to get anything better though.  Basically as I said, the limit of the advantage in this mode is that you don't do another extra conversion.  It's not allowing you to bypass any of the clocking issues or the side effects that can create.  Just because you have the original RGB data, that doesn't mean you can't inject problems into the output quality from the syncs being inappropriate for certain displays.  I don't know how much or how little a problem that can create in this particular system.  I do know from working with video equipment that it can create problems.

3. It's much more work, but yes it's possible to replace the Lisa.  The only thing stopping someone from doing that is the engineering and reverse engineering of the existing Lisa to replicate it's functionality.  It's way beyond what your average person should do, but it's hardly impossible.  Think about how many Intel compatible UARTs Asian countries spat out in the early days of the PC industry.  Many of them were basically created in a manner that could be used to reverse engineer any chip in the system.  More over, given the cost of FPGAs and their speed today it's more probable then ever that someone really bent on doing this, could do this.

Sigmason

Let me phrase it like this, I wouldn't waste my time replacing Lisa.  It's a lot of work for little return as the market probably isn't large enough to net a return on that scale (that's a business problem, that doesn't make it impossible at an engineering level.)

It's probably better to say, it would be much more practical to work with the existing Lisa outputs and whatever else you need to replicate the scan doubler function like AlexH is saying.  That said, be prepared that you'll have to make sure the video system your HDMI is connected to will tolerate the resulting output.

[sigmason]

Leirbag28:

I don't think making an HDMI output from the signals in the Amiga would slow it down.

If you truly only take the existing signals and ask nothing more of the Amiga hardware then to output them it would have no effect on the Amiga system itself.

Now if you were to somehow how ask the Amiga to put itself in a particular mode, or do something extra to accommodate said change that's a different story and I don't think that was what someone was suggesting.  In a way, would that not be exactly like installing a video card?

Sigmason

[sigmason]

Eriond:

You'll only get from the Amiga what the Amiga will give you unless you start messing with the drivers and everything else.

If you want a clock other then the 28MHz clock that is on the Lisa, you'll have to create the hardware to insure that the data you want is where you need for the HDMI transmitter's sampling.  So think basically a scan doubler like AlexH is saying, but extend that to creating a digital output instead of an analog one like most monitors and older products made.  The big difference is that everything you are doing is digital.  Which means that you won't loose in the A/D - D/A conversions but again, you must compensate for the timing differences and that's where you have your challenges.

This could actually be harder to do if you want to make the Amiga Lisa chips output conform via conversion to HDMI then to get the output of the Lisa to conform to something that 'will work' with an SVGA monitor.  The SVGA monitor will work with simple scan doubling because of it's wide sync ranges.  The HDMI specs, if you want them correct, will be harder to match and the process you use to match them, while constant for all HDMI inputs, could create issues that will be ever present.  As apposed to creating an output that is close enough for SVGA monitors and might work better on some displays and not as well on others.

Sigmason

[alexh]

Sniff... sniff... what's that smell?

You need to go away and read the HDMI spec again.

[Stedy]

Carl,

Why not use a DVI device like the SiI1162?

Datasheets are available without an NDA for HDMI parts and you also open up the possibility of DVI monitors.

HDMI TVs can take a DVI input take a look here at the HDMI website.

To clock the DVI chip from Lisa use the Sync On Green (SOG) output, this will be a composite (hsync and vsync) combined signal. The SII1162 requires separate HSYNC/VSYNC signals. You can either use a sync stripper like the LMH1980 or if you use programmable logic, use that.

Of course, it would be worth double checking that the SOG output is a standard Video sync output first otherwise it may need conditioning/converting.

All DVI/HDMI transmitters will be destroyed if you connect them directly to the Amiga. You need to convert the +5V logic to +3.3V logic with appropriate converters. Some 74LV244 or equivalent devices run off 3.3V but have 5V tolerant inputs.

I have never run a DVI interface at less than VGA (640x480) resolution so I can not say for sure if it will work with say Hires (640x256) or low res (320x256) amiga displays.

For the I2C interface, connect up a microcontroller with one of these ports. For DVI devices, this is used to read the Extended Display Identification Data (EDID) PROM in the monitor and set up natural screenmodes and refresh rates. You can also use the microcontroller to program the DVI chip via I2C.

I hope this gives you a few ideas.

Ian