The Sinclair QL Forum

Hi,

I've been looking at the IPC and its comms protocol with the ZX8302. I'm wondering if the IPC could be replaced with by an RP2040 with 3v3-5v level converters ? This might allow easier connection of a PS2 keyboard and better serial handling.

The 4/8bit IPC protocol looks fairly speedy (say for an ATMEGA) but the PIO of the RP2040 could work well.

Before I start to get my head around the PIO sub-system, has anyone explored using the RP2040 as a replacement for the IPC ?

Thanks.

The protocol itself is basically 1 bit - and given the 8048/9 is not the fastest of Microcontrollers, an ATMega or a PIC could easily do this - so no reason why an RP2040 couldn't handle this.

For me the amazing thing about the IPC is just how much they managed to squeeze into this 2K memory.

It provides keyboard scan (and therefore also joystick)
It provides serial receive function for 2 channels
It provides sound
It's possible that it originally also provided some kind of debug facility using INT7

What ever you replace it with - it will need to do the first 3 of these.

Hi,

Maybe a programmable joystick interface could be implemented.

Pr0f wrote: Tue Oct 08, 2024 8:28 am It provides keyboard scan (and therefore also joystick)
It provides serial receive function for 2 channels
It provides sound
It's possible that it originally also provided some kind of debug facility using INT7

Unfortunately, it does none of this very well. That's the *less amazing thing*. (Well, it
does mostly work most of the time, maybe that was good enough for Sinclair.

If you look more closely, it becomes obvious that the IPC has never been in the original design and was a late afterthought to be able to get rid of yet another custom chip. Keyboard works rather well, normally, serial is a mess, and the decision to put the simple 1-bit-beeper into the IPC behind a slow serial interface makes the sound capabilities even worse than on the ZX Spectrum. That was a totally anachronistic design decision at a time where most home computers offered 3-channel programmable sound chips.

Thanks for your quick replies and the background.

I've captured some example comms between the IPC and 8302 on the trace below (brown and red COMCTL and COMData signals - ignore the top reference clock):



Here is the 8301 is issuing a status update command (0x01) to the IPC and the IPC is responding with status byte (0x45). There appears to be one valid COMDATA bit for each two pulses of COMCTL - and the data bit is sampled/valid on the falling edge of the second pulse. For the initial 4 bit command issued by the 8302, there's roughly 5.5uS between pulses (181KHz) - and for the 8 bit reply by the IPC, there's roughly 11uS between pulses (90KHz).

Given the clock rate of the link, I was thinking it would need a hardware input channel - say SPI on an ATMEGA or PIO on a RP2040. I would need to work out how to sample/send data on the rising edge of the second clock pulse for each bit.

Am I on the right lines here - or is there something obvious I'm missing ?

In case it helps (if you haven't already seen it) there's a partially commented disassembly of the 8049 on my site:

https://www.dilwyn.theqlforum.com/qlrom/ipc8049.zip

May be of some help to you to know what the code expects by way of timings and signals, maybe.

...you'll also find a beautifully comprehensive description of the ULA/IPC 'interface' in the Minerva ROM source, thanks to our friend Mr Reeves

tofro wrote: Tue Oct 08, 2024 9:01 am If you look more closely, it becomes obvious that the IPC has never been in the original design and was a late afterthought to be able to get rid of yet another custom chip. Keyboard works rather well, normally, serial is a mess, and the decision to put the simple 1-bit-beeper into the IPC behind a slow serial interface makes the sound capabilities even worse than on the ZX Spectrum. That was a totally anachronistic design decision at a time where most home computers offered 3-channel programmable sound chips.

It would have been better to omit the IPC completely, attach keyboard and beeper the Spectrum way and save time-to-market, board space and costs. Just one serial port in that case, but a decent one.

andy_edin wrote: Tue Oct 08, 2024 5:51 am This might allow easier connection of a PS2 keyboard and better serial handling.

You are aware of SuperHermes (Lite) which already does that?

Peter wrote: Tue Oct 08, 2024 9:49 pm

andy_edin wrote: Tue Oct 08, 2024 5:51 am This might allow easier connection of a PS2 keyboard and better serial handling.

You are aware of SuperHermes (Lite) which already does that?

Strictly speaking - not all PS2 keyboards - there are some it won't recognize - it has something to do with the scan code sets and initialization of the keyboard - which is why the SuperHermes boards won't work with a lot of the 'inteligent' KVM switches.