ROM replaced by EPROM

[gertk]

Hi all,

Since last Saturday I have got myself a QL (again). It came without powersupply, leads and the keyboard membrane was broken. Having restored quite a lot of computers in the last years it was time to tackle this one. It also makes a nice addition to my Sinclair collection.

So far I have created a new internal powersupply which is powered by an external single 12-15 Volt AC adapter. It is based on a LM2575 switch mode converter and some linear stuff for the rest of the supply voltages. The membrane has been replaced by a new one and works great. System is nice and stable, only composite video output has a 'hum bar' with 12V AC input (the 78L12 has just too little input voltage). Monochrome is OK, haven't tried RGB but it is probably also OK.

Now for my next challenge:

I have replaced both internal roms by a single EPROM, made a 27C512 with both JS roms and the TKII (ver. 2.20) and connected pin 20 of the EPROM to GND and pin 1 of the EPROM to pin 20 of the socket. The QL starts up with the toolkit logo and F1/F2.
On first attempt I used a AMD 27C512-200 and it seemed to be working OK but randomly the screen froze, developed artifacts etc.
I suspected the EPROM might be too slow (200 ns) and burned another one but this time in an Intel D27512-170. The QL does not even start with this one, just a white screen. Put the AMD back in: works but unreliable. Tried another EPROM, this time a TI 27C512 and it also results in a white screen. In the end I programmed another AMD 27C512-200 and it works! The QL has been chugging a simple basic program all night and this morning it was still running fine.

Any ideas on why the other brands do not work ? Checksums are the same. Is the timing that critical ? From the service manual it seems the only signal that enables the (EP)ROM is ROMOEH (and with the original roms A14/A15) and with all the address lines connected, a single 27C512 should be mapped into the first 64k memory space.

The plan is to use the second socket for a RAM upgrade by inserting a 512k SRAM with some logic to map it into the correct region.

[gertk]

Some pictures:

The EPROM (pin 1 is routed to pin 20 of the other socket for convenience/testing)

Single 27C512

The power supply:

The 'new' power supply

The 78S09 is used to lower the unregulated input voltage for the microdrives and its motors.

[vanpeebles]

Impressive work! Do you think the new power supply works better than if it used the old sinclair one?

[gertk]

Impressive work! Do you think the new power supply works better than if it used the old sinclair one?

After a night running a Basic program the powersupply board was about 30 degrees Centigrade, the adapter in the power socket about the same so it is definitely more efficient.
The LM2575 delivers the main 5 Volt and can handle about 1 Amp. For more power hungry systems you could use the LM2576 which can do 3 Amps. These switch mode regulators can handle input voltages up to 40 Volts although at that kind of input voltage the additional linear 78S09 will run very hot... Maybe a new design with multiple switch mode regulators can be made.

Replacing the original roms by the cmos Eprom also lowered the power consumption already quite a bit and maybe enough to handle the extra 'load' of the planned 512k SRAM

Speaking of SRAM: if I put the SRAM into the ROM socket will it be possible to get access times comparable to that of an external ram card ?

[Dave]

I suggest shortening the flying lead to pin 1 of your EPROM. It's VERY noisy inside the case, and that wire is long enough to pick up all kinds of interference. You may find otherwise marginal EPROMs will work better if you just shorten that lead.

[gertk]

I suggest shortening the flying lead to pin 1 of your EPROM. It's VERY noisy inside the case, and that wire is long enough to pick up all kinds of interference. You may find otherwise marginal EPROMs will work better if you just shorten that lead.

It is indeed quite noisy inside the QL

This evening while trying to get a microdrive cartridge working again the machine crashed randomly again. After putting back the original roms it was rock solid again.

So I have shortened the lead and tucked it away underneath the EPROM so it is attached to pin 20 of the same socket now and will test if it is better now.

On the oscilloscope I could/can not see anything weird going on on any of the addresslines apart from the usual ' madness' on those lines

There is also little noise (no more than with the original roms) on the power pins of the rom sockets.

[vanpeebles]

Oh and welcome to the forum too

[QLObi]

Hi gertk,
the better solution is an extra EPROm board. If you use the standard logic it's also possible to free the upper external 16k bank with a switch at the back of the QL!.
The drawing is published somewhere here in the board, but i put open EAGLE files in the appendix. It can be opend and modified with the freeware version from CADSOFT. The pcb is easy enought to do it yourself, just single side. The only problem is the height of the board. You need to salvage a socket to get the pins and you need to drill a 1.4mm hole for the them. That's the reason why the tracks are so thick on the layout.

QLOBI

[gertk]

Hi QLobi,

Looks like a nice board but I fail to see how this is going to improve my issues with different brands of eproms unless the problem is timing is too fast without the TTL chip in the select line.

Since all rom (internal and external) is selected by the same line ROMOEH, and all of the address lines A0-A15 are connected to the EPROM the only thing that could go wrong is that ROMOEH is generated differently for the 00000-0BFFF space and 0C000-0FFFF which seems unlikely. What type of EPROM (brand/speed/size) do you use on your EPROM board ?

Gert

[QLObi]

Hi gertk,
as far as i see on your picture you forgot to invert the ROMEH ! The EPROMs normaly need an active low signal, but ROMEH is aktiv high ! On all drawings i saw, there is a minimum of this inverter. Even on the orginal ISSSUE 5 drawing from SINCLAIR you will find this option (IC17). In the orginal design only LS TTL chips are used, these a quite slow but work anyway.
If you put PIN20 of the EPROM to permanent GND, you are not able to free the adress range from C000 - FFFF for some extensions like ICE or the DONGLE of some compilers. If you don't need this, its fine to enable it permanent.
I am normaly using EPROM with 200ns access time, but i also tested 250ns and they were ok.
There is no issue with the manufacturer, i am using what's available.
One advantage of the board is the short and noise reduced wiring, but i tested the circuit also with a open wiring you had. The only excpetion is, i soldered a 7400 on top of an existing IC to get the inverter.
If you frequendly change the EPROM the board is easier to handle.

QLObi