a DIY home build project...

[Pr0f]

I am using a 68901, which could and probably will.eventually go in the fpga with video, the reason for using the separate chip to start with is ease of testing. This acts as interrupt controller for the QL, totally replacing the zx8302. It has a serial controller on board, 4 timers, and 8 programmable interrupt lines. It's this device that talks with the propeller over a much faster serial interface, moving 8 bits not 4 bits as in the zx8302. Extint can be programmed for level or either edge, and all interrupt sources are maskable. The usb fifo full and empty also use 2 of the lines. So I needed something that would work with existing zx8302 and the 68901, hence the propeller.


Oh, and the DUART is the 2nd generation Philips upgrade of the Motorola DUART, giving baud rates up to 230K, independant for both tx and rx on both channels, plus programmable stop and parity, fifos are 8 bytes, but are closely linked to hanshaking, unlike the 16550. Its a tried and tested chip also perfect for 68k bus.

[Dave]

I'm particularly interested in the DUART - which one is it, and have you a functional driver or schematic for it yet? I need a faster serial on the QL for this internet card, which works through serial and 9,600 doesn't cut it.

[Pr0f]

This is the one:

https://www.google.co.uk/url?sa=t&rct=j ... 8D-llqEBoh

It has some distinct advantage over the 68681 part it is descended from. Deeper FIFO's but also more choices of baud rates and higher rates all from the same 3.6MHz crystal. Another advantage is this chip doesn't suffer the restriction the earlier models had on bus clock cycle limits, which means it will happily interface to faster processors.

[Dave]

I'll email you about this.

[Peter]

The serial communication 8049 -> 8302 is a real bottleneck.

Not, if SER is implemented separately anyway, and all you want is the matrix keyboard.

Propeller or FPGA? Not the most important question. Certainly shouldn't be a show stopper that prevents people developing and moving forward...

My point was overall development time, on the basis that the design will later include an FPGA anyway. If mainly the first step is in view, and Pr0f is very familiar with that Microcontroller, it makes sense to just use it and move forward.

In some way (except the CPU debugging) the Q68 was easier to design than this, because more wiring was done inside a chip, less on PCB. And I know how long the Q68 took...

[tofro]

The serial communication 8049 -> 8302 is a real bottleneck.

Not, if SER is implemented separately anyway, and all you want is the matrix keyboard.

That is still somewhat a state-of-the art concept - modern PS/2 and USB keyboards still do it more or less the same way.

What is a bottleneck is the serial ports and sound, which is clearly a nuisance on the QL - Even the Spectrum's main-CPU-driven sound is nicer to program. Some of the stuff the Spectrum guys do with their single bit is really amazing and beyond all capability of the QL's 8049 (current favourite: http://converter.dev.artweb.ee/music/50 ... 6_secs.mp3 ).

Tobias

[Pr0f]

There is code available for download to the propeller that emulates ay8910, Commodore SID chip and a WAV player, so more interesting sound options are there, a driver would be required to push the data down to the propeller, but using an 8 bit fast serial link running at 1M, you can shift a lot of data. Emulation of the QL sound was just a requirement for compatibility with Basic.

Not quite sure what you meant by 'state of the art concept' - you'd have to explain that.

[Peter]

What is a bottleneck is the serial ports and sound, which is clearly a nuisance on the QL - Even the Spectrum's main-CPU-driven sound is nicer to program. Some of the stuff the Spectrum guys do with their single bit is really amazing and beyond all capability of the QL's 8049

Totally agreed. That's (among other things like not using 16 colours) why I do not find the QL a clever hardware design, compared to Sinclair standards of that time. IMO the 8049 has complicated, enlarged, price-increased and delayed the design with no practical gain.

[Pr0f]

so having left my thread unattended for a little while, I have aquired some useful parts, and will be building a proto board so I can test out the DUART ideas, and USB.

I am also fabricating a small header so I can lift the 8049 and place a propeller board in it's place - giving access to the keboard / sound and IPC comms links. This will give me a way of writing the code for the prop and being able to test it on a real BBQL. I aquired another one on ebay yesterday, so that will need repairing, but should become the standard test unit for my other boards...

I keep honing the design of the DIY box, but getting closer to a 'finished spec'

68008FN10 clocked at 12.5MHz (but other clocks selectable)
4MB CMOS RAM 55ns
512K FLASH ROM - split as 2 x 128K x 2 (Chip is split into 2 by address line A21, and a bit from the 68901 will provide a page switch so two images can be stored). The upper memory area stores ROM extensions, the lower 128K is comprised of 48K + 16K, 32K, a gap of 16K for internal I/O and finally the last 16K of the 128K lower area.
Xilinx CPLD for address decoding and memory shadowing
DUART + 2 x LT1133 transceivers
VNC2-48 chip for 2 USB Host interfaces Initial support for Flash drives, hard disks, HID devices like keyboard / mouse / joystick / gamepad, and RS232 slave devices. I would like to add USB floppy...
USB current limit devices for soft current limit
68901 interrupt controller and IPC driver
Propeller IPC replacement - provides QL sound, PS/2 keyboard and mouse, QL network?
I2C bus using philips bus master chip - will provide clock chip as on the Minerva - the I2C connections are buffered out to extended ROM port
SPI driver chip for upto 5MBs - buffered out to extended ROM port

Still in flux:

The following 2 devices will be based initially on an 8 bit bus access, so they are compatible with the 68008, but an option to drive both using 16 bit data bus will be provided via a 96 way connector in place of the original 64 way offering on standard QL. The idea here is that any board designed for the orginal QL should still be pluggable, but if an extension board offering 16bit or better bus is used, the speed gain of doing IDE and video over 16 bits is there.

8/16 bit IDE interface supporting Flash disk, CF and 2 x SD cards
8/16 bit Video interface with Xilinx FPGA and 256Kx16 CMOS 10ns RAM - initially this will present as 32Kx2 in the original screen area (supporting 2 screens), and as 256K in the top of the 4MB address space - although actually mapped much higher and using an echo

The extended ROM port is a 3 row 48 pin connector - similar in spec to the QL's standard edge connector - the idea it to make that interface 'project proof' as much as possible and provide protection against shorts, and abuse of the i2c or spi ports on it. ROM's could be used to replace any or all of the internal ROM areas in the lower 128K (so the function of the 16K plug in ROM expansion is maintained), and some additional I/O area is decoded for project use - along the same lines as Aurora. The 5V line will be fed from a current limited supply device

[Dave]

An exciting project.

The propeller won't need to do QLNET because that's fully implemented in the 8302.