The Sinclair QL Forum

The zx8301 is the more volnurable part. The zx8301 at ebay are so expensive. We need a replacement like the spectrum 128 ula replacement. See the link
https://zxrenew.co.uk/ZX-Spectrum-128k- ... p151287561

The spectrum ULA is relatively simple in comparison to the QL video ULA.

Also - the ULA for the spectrum is documented very well in a book, and there are numerous CPLD/FPGA variants of it available on line.

I have built the Harelequin board variant of the spectrum that only has TTL chips in it -and surprisingly few for a functional machine!

As the output of the ZX8301 is TTL for the video signals to the RGB socket - you could employ TTL buffers to protect the ULA outputs, and use of transorbs /fast zenners or protection devices could also be used to protect the chip.

Hmm. Can't see the 8301 (which I can find on eBay between 20 and 30€) being much more expensive than any replacement would be possible (Spectrum replacement ULAs are much less complex than the 8301, and still they're about the same price as an original QL ULA).

Agree 8301 replacements are becoming increasingly rare, and in the long run, it could make sense to think about replacements. The 8301 is, however, much less well (rather not at all) documented as the Spectrum chips. There's a lot of research to be done before you could actually synthesize a modern replacement (which could also add fixes to compromises being made when the original was designed (like adding 8 more colours with an extra pin & cable instead of the silly flash)

I have a bunch of documentation on this which was produced by Nasta. I will dig it out and share it.

Brane2 wrote:

Dave wrote:I have a bunch of documentation on this which was produced by Nasta. I will dig it out and share it.

DO you have anything about inner makings ULA that should be in Ferratni's doccumentation ?
I have geneal contexxt and inner and I/O cell content ( couple of NPN transistros and resistors), but would like to know their cahracteristics ( resistor vaues, transistor Hfe, etcetc).

I'm thinking about doing SMD PCB with those components and pads that could be used to test actual behaviour and possible implementations on the process of the time.

There was an indepth description of the 8301 by Nasta on the Forum.

I combined all the messages into one file see this link on QL Homepage:

http://www.dilwyn.me.uk/docs/hardware/8301.zip

tofro wrote:Agree 8301 replacements are becoming increasingly rare, and in the long run, it could make sense to think about replacements. The 8301 is, however, much less well (rather not at all) documented as the Spectrum chips.

Documentation is only an issue, if the exact timings have to be replicated. The questions are:

• Do you actually want to slow down RAM access?
• Do you actually want the weird QL video timings?

If the answer to both questions in "no", then the logic it is no rocket science anymore. For example, the Q68 video controller delivers all QL features like flash, second screen etc. Just not the timings. The easiest way would be to abandon the QL DRAM completely and use an FPGA with > 128 KB block RAM like the LFE5U-45, or for cost reduction an LFE5U-12 with external RAM. Zero-wait access, so it would not slow down the machine. For my taste, the output timings would probably be more standard than the QL ones.

Brane2 wrote: BTW: Where is Nasta these days ?

Lurking from the background

Brane2 wrote:So it would IMHO make sense to do something that would also serve as pure ULA replacement and keep the rest of the machine physically close to "real deal".

Sure, but it costs a lot more time than re-using the existing FPGA logic. Not someting I would consider.
And with the disadvantage of requiring an external video converter.
Getting rid of an external converter would be my only practical motivation. To exactly replace the ZX8301, I'd always prefer to buy a real one.

Brane2 wrote:You can replace all the RAM that QL would ever need with single DDR SDRAM chip for $2.

DDR termination would add more overhead than it's worth. The least is DDR2, if one wants to go beyond SDR. But that still comes at the cost of an extra power supply and various complications using the FPGA infrastructure and I/O banks, requiring more costly PCB. It is possible to route two separate (main RAM + VRAM) SDR devices to an FPGA in 0.8mm BGA case on lowest cost 4 layer PCB. But not DDR2, if plenty of other pins (peripherals + video) are needed. While (without cache) nothing is won over SDR performance-wise.

Brane2 wrote:And fit all the circuitry you'd ever think of ( including CPU) within ECP5-25 or even 12. For literally couple $ ( -12 is $6.ish at 1pc qty).

Thats why the Qzero uses these chips. Prices are rising though and like all modern FPGA, delivery times are a pain.

Brane2 wrote:And use FPGAs fast ISP configuration FLASH in x4 mode as QL's ROM.

These are paged devices so one would need to implement cache or get painfully slow even in x4 mode. It's much easier to shadow the ROM in RAM, and in that case, primitive 1x SPI is more than fast enough. You don't even notice.