Costings for NEW MTX machine !

Post Reply
lezanderson
Posts: 174
Joined: 14 Aug 2012 15:31

Costings for NEW MTX machine !

Post by lezanderson » 09 Feb 2015 16:00

As part of the MTXplus+ project , I'm doing costings (my tiny contribution) to get costs down and having talked to Dave we've concluded that it may be possible to make a £50 MTXplus machines as a Single board Computer !

This is a DRAFT costing, but the overall cost should be :

PCB £15 , I'm assuming this is a double sided Board ?
ICs about <$38 (roughly £25)
Miscellaneous, switches, knobs, sockets, Diodes, resistors, Caps etc £10
Last edited by lezanderson on 19 Feb 2015 15:40, edited 1 time in total.

Martin A
Posts: 409
Joined: 09 Nov 2013 21:03

Re: costings for NEW MTX machine !

Post by Martin A » 13 Feb 2015 12:31

The boards for the Margrom are 2 layer, so the initial PCB cost projections were done using the same service, but a considerably increased board size.

The prototype MTXplus+ sitting behind me currently uses 40 odd IC's and takes up 2 2/3 of the 160x100mm Eurocard size boards it's built on. That doesn't include the diagnostics board. Dave's testing extra features on his, and so uses all 3 boards in full.

The size is what makes the main PCB the most expensive component. Conveniently 160x300mm is the same size as the original MTX board. So it could possible sit a MTXplus+ board in a real MTX case. If it can be routed....

Obviously, the more the board can be shrunk from the space currently occupied, the cheaper it gets. Although an all in 1 board will save the space currently occupied by the Eurocard connectors, space would have to be allocated to things like the joystick and video sockets that are currently off board. I don't think we can get it down to 160x200mm where the next price break is.

lezanderson
Posts: 174
Joined: 14 Aug 2012 15:31

Re: costings for NEW MTX machine !

Post by lezanderson » 14 Feb 2015 12:18

The only possible way to reduce board size without reducing chip count is possibly switching to using PLCC ICs where possible.

PLCC ICs aren't much liked, as it's more difficult to BreadBoard your design with them, however once the design is complete then using PLCC ICs is an option.

Also the PLCC ICs on ebay and from China seem to be more Modern and less likely to be counterfeit. Plus I believe you can overclock them more easily ?? I believe Martin Overclocked a Z84C0020VEC PLCC Z80 to 40MHz ..Quite an achievement !

Martin A
Posts: 409
Joined: 09 Nov 2013 21:03

Re: costings for NEW MTX machine !

Post by Martin A » 14 Feb 2015 14:09

I got a 10 MHz DIP Z84C0010PEG up to 32mhz on a board with slower ram that the one that hit 40.

I don't think the packaging matters too much the CMOS Z80 is just well over-clockable.

lezanderson
Posts: 174
Joined: 14 Aug 2012 15:31

Re: costings for NEW MTX machine !

Post by lezanderson » 16 Feb 2015 15:37

Are there any heat or stability problems with over-clocking to 32MHZ ?? I'd assume that at 32MHZ a Z84C0010PEC will get warm especially if left on for several hours ??

I'd read that the PLCC versions have less capacitance between pins and thus more easily over-clocked .. but this may not be an issue ?

lezanderson
Posts: 174
Joined: 14 Aug 2012 15:31

Re: costings for NEW MTX machine !

Post by lezanderson » 16 Feb 2015 15:52

SRAM speed ratings

What SPEED rating is needed when running at 32MHz and 40MHz , as legacy static ram (SRAM) are normally 55ns,70ns,85ns or 100ns.

32K SRAM or 128K SRAMs 70ns,85ns,100ns speed ratings
512K SRAMs 55ns or 70ns
32k or 64K fast SRAM 15ns (UM61256 , UM61512 type)


So I'm assuming your using UM61512 or similar FAST SRAM at 32MHz and 40MHz ??

Martin A
Posts: 409
Joined: 09 Nov 2013 21:03

Re: costings for NEW MTX machine !

Post by Martin A » 18 Feb 2015 22:00

The 32mhz board is running 55ns ram, 45ns Flash rom.

The 40mhz capable board has 45ns flash and 25ns Cache ram off an old 486 board.

32k ram 32k rom is also very easy on the chip select logic, and I'm using bipolar F chips for that so the delay for that is under 4ns

The diagrams in the Z80 manual leave a little room for interpretation, the drawings for the M1 instruction cycle, show MREQ and RD going low after the falling edge of the clock in the middle of T1 and rising shortly after the rising edge of the clock at the end of T2. That gives 1.5 cycles or so for the logic generating the read signal. Which even at 25ns per cycle easily meets the 30ns that 55ns alliance ram requires from OE going low to the data being presented.

The address bus changes earlier which means using A15 or *A15 for the select means almost the full 2 cycles are available for chip selects 62ns for the ram 66ns for the rom at 32mhz, (46 and 50 at 40mhz) Again just in spec for 55ns ram at 32hz, and easily in spec for the cache ram.

It might even be possible to push it a little further possibly to 44, but I've not been able to source any oscillators they seem to jump from 40 to 48.

I've done continuous tests of over an hour on with the 10mhz part running at 32, and absolutely no detectable warming.

lezanderson
Posts: 174
Joined: 14 Aug 2012 15:31

Re: costings for NEW MTX machine !

Post by lezanderson » 19 Feb 2015 15:38

There's also the problem of I/O read / writing at high speeds. The need to insert WAIT states may counteract any further speed increases above 32MHz? As the I/O bus will be needed to be slowed down to 4MHz or so to communicate with peripheral chips?? It may prove 32MHz is the optimum speed ??

Also 32MHz can easily be divided down to 16MHz/8MHz/4MHz ! Whereas 40MHz would take more logic and be somewhat more messy to divide down to 4MHz for compatibility?

It's pretty amazing you've manage to push the prototype to 32MHz never mind 40MHz !!

Martin A
Posts: 409
Joined: 09 Nov 2013 21:03

Re: Costings for NEW MTX machine !

Post by Martin A » 19 Feb 2015 20:36

The high speed boards aren't really, "the prototype", they are replacement CPU boards with a MUCH lower level of complexity.

The extremely flexible (ie. messy) Memotech memory paging system needs a lot more logic to implement, and that slows things down.

The GAL equations for chip selects reference MREQ, IORQ and RD as well as the address and page port logic which slows things down by 1/2 a cycle, and then add in the delays in the GAL (4x the delay of a 74F04 inverter) and there no way it can match the speeds of the simpler board with it's fixed 32k rom 32k ram layout.

The CTC is another reason the main board will always be slower, the fastest part is "only" rated at 10mhz and neither of 2 I have will clock faster than 20mhz.

Having said that, the main board will run at 16, however the MTX software won't. At the moment the only way to meet the I/O timing for the VDP etc is with software delays, which is fine for any purpose written software. The MTX roms however will need some form of hardware wait generator.

So while it's possible to make 40mhz MTX500 ish system, it wouldn't be compatible with anything other than MTX basic, as the rom's would have to be re-written to introduce the software based I/O delays, and the interrupt system would have to be totally re-designed to run in mode 1, as there would be no CTC.

The 16mhz Prototype board however, running at 4mhz for compatibility, is proving to be both stable and capable of running the vast majority of the software thrown at it so far.

(hopefully having written that I wont Jinx it)

lezanderson
Posts: 174
Joined: 14 Aug 2012 15:31

Re: Costings for NEW MTX machine !

Post by lezanderson » 20 Feb 2015 14:27

Using Fast CPLDs instead of GALs..:

One solution regarding speed is to possibly use fast CPLDs :EPM7064SLC44-7 (7.5ns 44 pin) rather than GALs , you can get some quite large GALs but anything bigger than a GAL22V10 will need an expensive programmer ?

Though large CPLD are either in PLCC for surface mount packages, some are BGA etc..

If you can think of a useful, large CPLD that can do the job then let me know ?

Also :

You could possibly consider putting a lot of the external logic into a large, fast CPLD.. Including the CTC and SIO circuitry??

Post Reply