The Sinclair QL Forum

This was about a Super Gold Card remake:

skagon wrote: You can always get a newer model and just port the logic code to the newer chip. On the VHDL level, it's a day's work tops and Altera will have guides (if not porting tools too) to tell you how.

Firstly, complex ancient logic code can be hard to port to newer development tools. I remember my massive problems even to port my Q60 logic to the next generation development tool. In the end I gave up, and still use the ancient tool (which won't work under a modern Windows, not even in a virtual machine).

Secondly, you don't just need functional sourcecode, you need a timing model. Ancient development tools had poor support for that, so you'd have to read through databooks from decades ago and manually calculate timings.

Thirdly, modern chips won't let you implement the minimum asynchronous delays of ancient chips, simply because they are way too fast. You'd have to work around that, which can be challenging.

Fourthly, modern chips achieve nanoseconds timings, so they can be affected by high frequency faults on signals, which had no visible effect while chips were still slow enough to ignore them. So you can have logic and timings correct, but fail, because you have to interface to an old generation design with meagre signal quality. This is also a problem with QL-SD by the way.

Fifthly, modern FPGA won't have 5V tolerant IO. You'd need level shifter chips surrounding your FPGA.

Sixthly, modern FPGA are a pain for manual soldering. In the best case, you have no BGA, but a 0.5mm pitch TQFP with lots of pins. This hinders building the very low quantities for QL, which are no comparison to the Spectrum by the way.

Seventhly, even to deal with all the bugs and problems of FPGA development software in a modern design is a challenge (once you leave the path of well prepared demos for evaluation boards and start your specific design for your specific PCB).

Eighthly, all of the above applies to logic you still understand to some degree, because you once wrote it yourself. Porting a complex ancient logic written by someone else can be more time consuming than writing it from scratch.

Ninthly, the one who writes this has successfully designed complex mainboards and graphics chips. You can not expect similar knowledge from a trader, or persons he could easily hire for a design.

I still wait for one of those who find chip design for the QL easy, to actually do it themselves and prove.

Thanks Peter - yes I have to bow to your superior knowledge - designing and building hardware for the ZX81 and Spectrum is a lot simpler.

Add to the points raised by Peter, that so far as I know, the original code for the Alteras is no longer available, and relies on some odd out of spec tolerances in the original Altera chips which meant that some traders in the 90s got their fingers burnt by purchasing a load of Altera chips (with the same model number), and programming them only to find that a minor change had been done by Altera - undocumented and even Aletera do not know when the change was made - and you realise you are better off designing a new interface completely from scratch.

As for my knowledge: Obviously it was not sufficient to foresee the problems QL-SD would suffer from. I was more used to newer designs like the Q68 and Q60, where you have decent signal quality due to adequate routing and ground planes. The QL motherboard was never designed to satisfy >300 MHz logic chips. So my decision to use a modern standard PLD was a bit naive. I had better used an obsolete, slower PLD. Shame on me.

And (*drum roll please*), tenthly, would we want SRAM chips instead of (I presume) fast page DRAM (can't be bothered to go and check at the moment) OR we would need some err, modern DRAM modules (with their own "new to us" access timing)...

Mark

Mark, what do you mean? Why SRAM? Or, what do you mean, "modern" DRAM? As in... DDR perhaps? And use it how?

Personally, I am not interested in a new super-QL. The modern computers are enough and there are other options as well for tinkering (like the Pi).
For me, even a SuperGold is marginally an overkill. I want a stacked QL, not a FrankenQL with "new" processors and new graphics and new... whatevers. It no longer is a QL, then.
So, I'm happy with a Gold card or a plain FDD/RAM expansion and the HxD emulating floppy disks. Ok, one of those microdrive SD emulators would also be nice. But that's where I draw the line.

I just wonder all the time, why nobody complains about the lack of flatscreen monitor support for the QL. Storage is obviously important, but I have no idea how to achieve a decent 512x256 QL display nowadays. Do you all press F2 and use TV mode? Or do you all still own working CRT monitors?

I use an RGB2VGA board GBS-8220

GBS here as well. Through an Extron unit.

I have that too, but isn't it a pain? Characters are hardly legible in 512x256, and single pixels (e.g. at the border) can not be distinguished. Do you use it in 256x256 only?

I have a photo for you, luckily. You can check the quality of the image yourself.
Please note that it looks a lot more fuzzy in the picture, firstly because it's a low-light photo, which means long-ish exposure time and thus, some movement and secondly because it's been resized to a lower resolution, to fit the 256k limit.