The Sinclair QL Forum

The only obvious answer is that the MT3 is paging out it's ROM at $C0000 as per the Trumpcard known boot sequence and running at $10000, but on a soft reset doesn't page it back in, but does on a hard reset (reset switch), which is why the reset code had to copy it back into ram but there's no ram there obviously

I wonder if a change to the GAL code could be made to prevent the ROM being paged out in the first place, will have a play

Am turning my attention to the differences between JS & Minerva ROM's on a JMP $184 instead

Given the reset xxx works with Minerva, maybe there's a clue to how to force a reset so MT3 shows up on the reset with JS

t0nyt wrote: Fri Aug 08, 2025 2:38 pm Am turning my attention to the differences between JS & Minerva ROM's on a JMP $184 instead

Given the reset xxx works with Minerva, maybe there's a clue to how to force a reset so MT3 shows up on the reset with JS

Well that's a failure. Turns out Minerva finds the Trump ROM at $10000, not $C0000 (ignores if found there) so that's why reset xxx works on Minerva but not JS

Have managed to reduce the number of games that won't run with MT3 (configured for 512K, 640K total), as they need the floppy drive to be accessible with 128K or 256k, to 4

When wanting to use those I'll just plug the extra 256k memory expansion board in

Yes that is the reason, Shall be possible to modify the GAL code to hold the rom alwais in 0xC0000, in this way you can reset hard or soft with any memory and will continue to have floppy.
The problems of this configuration is that you can not have any other interface in 0xC0000, and that if you use minerva you will see the message 2 times in the boot screen, but shall work without problem.

The opossite is what I do with my "minerva only" GAL it hold ROM in 0x10000 alwais, that this is the reason only work in minerva, despite hard or soft reset.

There area two pins that are not used in the MT3, (pins 14,15, CLKCT and MA8) they are used only for control the dynamic ram in the original trump 2 card, so if you change the programing of the GAL, you can use any of this as an aditional jumper to hold the rom in 0xc0000, in this way you not need to conect or disconect the expansion card, and will have the normal operation, or the res_512 with floppy. (probably if you fix the rom in place, RAM will detect only 640Kb on any reset, hard of soft, but this need to be tested).

The problems of this configuration is that you can not have any other interface in 0xC0000, and that if you use minerva you will see the message 2 times in the boot screen, but shall work without problem.

Minerva has code to prevent that, it checks whether it has seen a rom before before it re-initialises it!

The Trump Card uses what was originally designated as a IO expansion area, 256k in size, starting at $C0000, for RAM.
The reason for the 'odd' behaviour at start-up is to work around a bug in the JM ROM, which only looks for ROMs at $C000 (ROM slot) and $C0000 (first expansion peripheral slot). The way the TC gets around this is initially mapping it's ROM at $C0000, then re-mapping it at $10000 (if I remember correctly) outside the 'usual' places where ROMs would be, in a portion of the memory map that is normally not used - except if Minerva is the OS of choice. Minerva (at least used to) check for ROMs at $10000 and $14000, so it will find the TC ROM there and because it checks for aliasing (same ROM detected at multiple addresses) it will immediately start the TC ROM at $10000, ignoring it's alias at $C0000, without the double reset.
What is odd to me is that the TC does not use the ROM slot for start-up, as peripherals on the J1 expansion connector can 'hijack' any address used by the motherboard, which includes the ROM slot. I suspect part of the reason might have been the need to get the ROM slot initialized first (as is the normal order of initializing extra hardware) perhaps for the Miracle hard disc.
That being said, there is no reason why TC ROM would not start at $C000, the first thing it does is jump to it's alias at $10000, removing itself from the ROM slot address, at which point anything in the ROM slot re-establishes itself at $C000, and can then be initialized like any other ROM by a small snippet of code in the TC ROM. Basically same thing accomplished on any ROM... ironically except possibly Minerva, which would have to be handled with a bit of code testing from where the ROM slot init code was jumped to.

IIRC, the start up is as follows:

1. At startup, the full 768K RAM is available so the RAM sizing and check works as expected.
2. The QL ROM accesses the ROM slot at $C000, at which point the Trump Card pages in its own ROM at $C0000. The ROM at $C000 gets initialised normally.
3. The QL ROM accesses the Trump Card ROM (either at $C0000 for classic QDOS or $10000 for Minerva) and the initialisation code in the TC ROM pages out the TC ROM at $C0000, through a read access at $1C000 (after the jump to the $10000 area).

With the Trump Card there was never a 'second reset', unlike the Gold Card.

I’m still trying to understand why changing ram size stops the MT3 from being detected (I know the Rom isn’t getting paged in, but don’t understand why not - it’s not directly related to no ram at c0000)

The TC pages it’s ROM into C0000 (which just jumps to the real ROM at 10000) on a power on/reset then pages it out, but not on a reset IF ram size has been changed (except on an 896k total QL where it does)

We know it doesn’t copy it’s ROM into RAM, else it wouldn’t work at all on a JS QL with 640k total RAM (as there’s obviously no RAM at C0000 for it to copy the ROM into with only 640k) at power on/reset

So I still don’t understand why after a reset with ram size changed (128/256/512 from 640) it doesn’t page it’s Rom in temporarily (but does with 896k)

I’m obviously missing something. Is it just the logic in the GAL?

Any thoughts please?

Many thanks
Tony

Hi,

After disassembling some Trump Card ROMs, it seems the Toolkit 2 ROM is located at 10000H, and is copied to C0000H in thevROM inialisation code. Which verifies what has been said here.

The reason for this maybe the way the QL ROM scans for addon ROMs, AH, JM, JS ROMS only scan for the add on ROM header at: C000H, C0000H, so if the Toolkit 2 ROM is at 10000H those ROMs do not know about it, so the solution was to copy the Toolkit to ROM to C0000H then pass the ROM scan to the QDOS to detect thd Toolkit 2.

The Trump Card initialisation then discards the ROM code at C0000H and use 10000H.

Minerva scans C000H, 10000H, C0000H so does not need the Toolkit 2 ROM copied to C0000H, on early Minerva versions, thevRom scan picked up 2 Toolkit 2 Roms.

Thus does seem a very convoluted way of doing things. But tgd design of tge Trump Card too, all the addon ROM space, leaving 10000H to 17FFFFH