Computers Overview
Commodore PET
Sinclair ZX80
Sinclair ZX81
BBC Micro
Sinclair ZXSpectrum
Memotech MTX
        Disk Drives
        Video Wall
    User Groups
    Video Wall
Memotech CP/M
Atari ST
Commodore Amiga
DEC 3000 AXP
Raspberry Pi



The Memotech MTX Series


Memotech Photos - Smörgåsbord



A Smörgåsbord is really a Swedish style buffet, but in English, the word has been "borrowed" and is used as a way to describe a diverse group of objects, this page is where I will put photos of miscellaneous Memotech related items that don't really fit anywhere else.

If you have photos of any unusual Memotech related items, please let me know and I will post them here.


A Memotech MTX Super-Computer ?

Jim Wills came across this beast in 2013, it was shown at Z-Fest 2003, a German Z80 \ CP/M focused computer fair.


This "slightly" customised FDX was built by Dr. Holger Goebel


Photo by Joachim Schwanter



The MTX main board has been moved into the silver tower, the keyboard just contains the keyboard PCB.



Photo by Joachim Schwanter

Peter Kretzschmar recalled some details of this machine on the Facebook group.


"[It has] ECB (Eurocardbus), board with NEC 7220 VDP, Harddisk, some Memory Expansion [and an 8MHz CPU]. Did you notice the two Monitors ? The right one is a Philips with a long glowing phosphor coating. Some members of the [German] MTX club used these Monitors to reduce the flickering, if the graphics board is set to interlace modes"

Photo by Dieter Hucke

A close up of the internals - the computer board at the bottom is not from an MTX - Dieter has confirmed that the PCB is from a ZX Spectrum 48 issue 2 as can be seen by the variable capacitors near the ULA and positioning of the full 48k of RAM on the main board. (Thanks Dieter)


As the machine definitely contains parts from an FDX and the report by Helmut (below) describes the system as incorporating an MTX512, I assume that a MTX main board is hidden somewhere else in the case?

Photo by Dieter Hucke





A wider angle view of the internals





Photo by Dieter Hucke

Photo by Helmut Jungkunz
Helmut Jungkunz wrote a report about the show in English which gives some brief details of this machine and its use.

"Probably one of the most anxiously awaited persons this day was Dr. Holger Goebel, self-announced dBase madman, who not only had disassembled and improved dBase 2.41 to be less faulty, but had also installed new routines ....

[Download PowerPoint presentation - in German)

I had time to sit down with Holger Goebel and to examine his MTX tower, which incorporates an MTX 512 fully souped up and running with a special BDOS, that allows this CP/M 2.2 based machine to not only run native ZCPR, but also makes use of "Quickstart" technique from RAM by cleverly switching banks. This was the computer, Holger does all of his developments on. "

Photo by Helmut Jungkunz webmaster at

Dr.Goebel's system was still going strong in 2005, you can just see it peeking out from under the desk in this photo.

Photo by Arndt Övermann from Z-Fest 2005


Some kind of Industrial I/O Interface ?
This unusual item was advertised on in 2012.

Manufactured by Arno Eichmann Elektroniksysteme CHE located in Rabenau/Hessen, Germany. (Website obsolete)

The card frame was connected to a Memotech FDX via a 60 way IDC connector on the rear of the FDX that was piggy backed onto the FDX SM1 bus interface card with some cores wired directly to the SM1 board.

The case contains a power supply and 11 Eurocard slots

Close up of the PSU and backplane
Each card contained 4 Mostek Z80 PIO chips, these are a dual 8-bit I/O port devices designed to provide direct interfaces between a Z80 processor and I/O devices

The purpose of this system is a mystery, but may have been used for some kind of data acquisition or control system.

The FDX SM1 Interface card showing the additional cable "piggy-backed" onto the interface connector. Most of ribbon cable with the blue stripe are connected to the interface socket on the rear of the FDX.

A number of cores are connected directly to the solder side of the SM1 interface as shown below.

Solder side of the SM1 interface card, showing the additional connections to the rear panel interface connector for this external electronics module.


Trojan Horses ?

This section has photos of two FDX disk systems where the FDX components have been reassembled in alternative cases.

Photos courtesy of Adrian Graham of binary dinosaurs

This photo shows two 5.25" disk drives sat on top of an enclosure originally intended to house 8" disk drives for use with an FDX.

John Nicol, a former shop owner and retailer of Memotech equipment, used it to build a "half-size" FDX using some original FDX parts.

You can just make out the backplane of the original FDX card cage inside the drive enclosure - the green PCB.

Adrian Graham bought this from John in 2005 for binary dinosaurs, his on-line museum of old computers.

More photos soon . . . .  



Martin Allcorn's "Ex-Bricked" MTX

Martin has made extensive repairs to one of his machines that had major problems, to such an extent that it was formally referred to as "The Brick". I think that the number and type of repairs deserve a little space on this page.


Martin's tale of resurrecting this MTX can be found here

The next photos give an idea of the major components replaced in the process :

This MTX started life as an MTX500, i.e., it had 32K of RAM. In common with a few of us, Martin had previously tried to replace the on-board RAM without success.

More recently he has been successful - see here for details

A common problem with the MTX is failure of one or more of the Video RAMs.

Quite a few people have replaced a failed VRAM with a like-for-like 4116, Martin has used a more reliable 4164 instead - see here for details

Like many others, the ribbon cable connecting the keyboard was also broken, Martin just replaced it with a standard IDE cable.

Other options are also available - see here for details

A 10MHz CPU was installed when Martin was over-clocking the CPU at 4.9MHz (as opposed to the standard 4 MHz).

The over-clocking has now been removed, but the CPU remains.

The MTX uses a 4-way ribbon cable to connect the audio and video outputs on the rear panel to the video daughter board. This cable is also prone to breakage at the rear panel connection, Martin has also replaced this cable as shown.

You can also see the TIP2955 and smoothing capacitors have been replaced to fix power problems, although not visible, the 5V and 12V regulators were also replaced.

Replacement of the capacitors has been described by David Kimberlin-Wyer  - see here for details

The photo also show new labels on the ROMs, Martin replaced the original mask ROMs with EPROMs.

Martin also fitted the jumper highlighted in the photo.

This jumper position, and the open link position beneath it, are shown on the MTX video board schematic as links "a" and "b". They are used to select either mono or colour composite video output. Martin had to swap the link position to enable colour composite output, but all of my MTXs have colour video output, suggesting that was the default position.

A DIY DART Enabler?
This PCB in this photo is the FDX Interface / RS232 card installed in the MTX/FDX auctioned on in July 2014.

Martin, the seller, recalls that he made the small PCB to enable one or more of the RS232 ports, although, after twenty odd years, the details are a little hazy.

Not all FDX interface boards had the RS232 components installed and the on-board PAL was not configured to enable the DART.

The sketch is Martin's circuit diagram from the time, he cannot recall the exact function of the PCB, but I think that it must have been used to perform the DART enable function that would have been done by the PAL on cards supplied with the RS232 components installed and the ports enabled.

Photos courtesy of Martin G.

A Wasted Video Board ?
A novel use for an MTX video board, it has been used to convert the 6-pin Audio/Visio output from a Texas Instruments TI99/4A to SCART.

These photos were taken from the listing page when the item was sold on ebay UK in May 2015

The MTX Video board has been mounted in the box for the original TI modulator.
Power Crazy  ?
Internals of an MTX PSU from a non-working MTX. The PSU had a two pin Euro mains connector and what looks like a user added fuse installed internally.

As of July 2020, this PSU is under investigation on Memorum

Photo courtesy of Steve G.
MTX Customisation
Manfred Flume generously donated a couple of MTXs and an FDX to me in 2016. Manfred was an active member of the MTX User's Club, Germany back in the day and made quite a few modifications to his MTXs, as well as writing various pieces of software for Tape and Disk.

One of the MTXs that he donated was in an MTX500 case, but the computer board was populated with 64k of original RAMs, i.e., it was an MTX512 board.

Manfred recalls that he had an MTX running at 8MHz using a customised clock circuit :
"It should have a PLL generated clock which powers up with approximately 4Mhz and goes slowly up to 8Mhz as some parts did not work when I used a 8 Mhz crystal. But when I started with 4 Mhz and increased the clock in some seconds it worked. So I designed this clock circuit."
Solder side of the boards inside Manfred's "MTX500"

The full size image shows that a number of modifications have been made with jumper wires between various points.
The 4000-05 computer board inside Manfred's "MTX500"

I think that the "MTX500" may have had the customised clock circuit installed at some point, although it has been removed, there are some artifacts on the computer board that suggests that this was the case.
Solder side of the computer board.
The original Z80A CTC (MHz) has been replaced with a Z80B CTC (6MHz) and the original Z80A CPU (4MHz) has been replaced with a Z80H CPU (8MHz).

Wires appear to have been soldered to CPU pins 6 (CLK), 24 (/WAIT) and possibly 27 (/M1).

As shown in the photo below, the original 4MHz oscillator has been replaced by a 6MHz one.
An "orphan" wire is soldered to pin 11 of the LS04 quad inverter in board position D9. The circuit diagram for the MTX clock shows this was the input to the inverters for the final stages of the clock signal to the CPU, CTC and bus.

The original input to this pin was from pin 6 (disconnected), driven by the oscillator circuit. Manfred's PLL circuit must have been placed between these pins, with this circuit removed, this MTX does not have a working clock!
With the CPU running significantly faster than design, timing dependent functions in the ROM, such as the clock, cursor flash, keyboard repeat rate, etc. would run too fast.

Perhaps having modified the ROMs to cope with the faster speeds, Manfred appears to have replaced the original ROMs.
The original mask ROMs on the 4000-05 computer board were 24 pin TMS4764 (8K) devices.

The replacement EPROMs installed by Manfred are 28 pin devices (Nec D2764), the solder side of the PCB shows that how the ROM sockets have been modified to suit.
This MTX has had the majority of its power regulation components removed. Manfred recalls that "the MTX 500 had, as other MTX boards, a thermal problem with the regulator circuit. So I replaced this with an other Power supply as I did not manage it to get the regulator cooler. "

(Memotech did the same when they used a MTX computer board and a "Skynet" PSU to build the Reflex Controller in my Video Wall system.)
I modified the computer board to accept the plug from a Cisco ADP-30RB router PSU. These PSUs provide +12v, -12v and -5VDC and are readily available on eBay, making a good DC only PSU for the MTX.

Read about how I converted the MTX on this page.
With a working PSU, the next job was to replace the butchered 74LS04 used to generate the MTX clock.

Here you can see a replacement 74LS04 fitted in a new socket that replaced the old chip.
With the 74LS04 replaced, the new DC power supply connected, and the MHz oscillator still fitted, somewhat to my surprise, the system booted first time.

The system appears stable under MTX BASIC, although clock dependent functions such as the keyboard repeat rate, cursor flash and the clock all run fast. A 30 "second" count on the system clock takes 20 seconds, consistent with the CPU running at 6/4 times normal speed.
The system boots MAGROM and displays the Game menu correctly so ROM access seems to work OK at the higher speed.

Unfortunately, none of the games can be run and the system crashes when any are invoked. I guess that the RAS/CAS signals and/or the RAM are just too slow for 6MHz so it looks like I will have to replace the oscillator with a 4MHz one as per design.
There is a MTX Memory Expansion board fitted, but no RAM installed on it.

The original Memotech PAL has been replaced with a TBP24S10 PROM, one socket has been replaced, but there is no obvious signs of any major modification.
However, the solder side of the Memory Expansion board has quite a few patch wires fitted.

I will need to study the patch wires to try and work out the modifications that were made to this board.
The FDX Interface board, with the original Memotech PAL labeled I(nter)F(ace)ONLY. The DART and other RS232 components have been retro-fitted.

A 74LS32 (Quad OR) has been soldered to the board and some patch wires connected - probably to enable the DART without modifying the PAL (PAL pin 15 =  DARTEN).
Solder side of the FDX Interface Board
The grey wire is connected to PAL pin 11 (A432)

The DART was enabled in a Memotech PAL by :
    DARTEN = A7 * A6 * A5 * A432 * IORQL * M1

The custom wiring should have replicated this logic with OR gates - how good is your Boolean algebra? - Prove :-)
Inspection of the board shows that IORQL and M1 have not been used in the logic. Tony has pointed out that the DART has its own IORQ and M1 pins, so does not need them to be included in the custom enabling logic.
FDX Customisation
Wolfgang Joerger generously donated an MTX500 with an 32K RAM board and the FDX interface board, along with what appeared to be a Single Drive FDX, that had an after market upgrade to add a second Epson SD521 floppy drive.

The most interesting thing about the FDX though, is the 1MB self build Silicon Disk that Wolfgng had added to the system. Not having a Memotech Silicon Disk board in my collection, I had thought about building one myself - thanks to Wolfgang - now I don't need to!
Although the FDX card cage was designed to allow additional cards to be added, none of the spare slots had edge connectors preinstalled. Here you can see that Wolgang has fitted an additional edge connector in the second slot and installed a custom interface board to breakout the FDX backplane to a DIN 41612 socket.
Component side of the Eurocard board, showing the 24 x 256 kbit RAMs, support logic, and what appears to be a customised daughter board - perhaps to enable a generic RAM board to be used to create an FDX Silicon Disk?
Solder side of the Eurocard board. The PCB has some identifying marks, including Kayser, RAM-Floppy and V1.5.
Solder side of the daughter board

Wolfgang thinks that the daughter board and patch wiring were configured to allow the RAM board to be I/O mapped for the FDX, rather than operating through DMA as it was designed to do.
A quick "Google" shows that the same card was used in other Z80 systems.  More info here (in German)

"A project of the computer magazine "c't" was the ram-floppy of the company Kayser. It must be integrated into the BIOS and is simply addressed via port addresses. The RAM Floppy is the drive F."
With the FDX floppy disk controller removed, you can see the Silicon Disk board installed in slot 2, with the 80 Column board below it.
After running the FDX silicon disk handler program and configuring the disk as type 43, the system can see the 1MB disk which can be formatted successfully. (The FDX format program only checks the first 64k of the RAM are working.)

Unfortunately, writing to the disk fails, hopefully, the fault may just be one or more bad RAMs - to be confirmed.
Replaced 3 x 258's

Replaced LS244
Replaced 2xHC374 with 2xHCT374
Replaced 74373 with LS373
Replaced all RAM with NEC 41256-150
Replaced LS240 with patch wires
Replaced LS139 with patch wires

Logic Chip testing with TL866A
74LS00 OK
74LS258 OK (x3)
74LS245 OK (Replaced original 'HCT')
74LS244 OK (Replaced original)
74HCT374 (OK) (x2) (Replaced original 'HC')
74HCT393 (OK) (Replaced original 'LS')

FDX Customisation - 2
FDX Advertised on eBay UK in June 2018

Has a user fitted Gotek USB drive running HxC firmware.

A neat idea, but a better solution is to use the Gotek with Keir Fraser's FlashFloppy firmware like I did with one of my FDXs. Read the details on this page.
The French Connection
The next group of photos are of an MTX512 from France.

The machine is one of two obtained from Memotech's French distributor by Gilles Bronchain.
The machines have what appears to be a locally made, custom PCB, that supplaments the MTX Composite Video output with SCART RGB (probably).

The brown plugs connect the MTX Video board to a SCART converter PCB mounted inside the upper half of the keyboard shell.

(Photos courtesy of Gilles Bronchain, France)
The Video PCB, removed from the MTX, showing extra wires soldered to the board for the SCART PCB.

These wires are not taking modified signals from the video PCB, they connect to the video and audio signals on J13 from the VDP and sound chip on the computer board.

The Purple and Green wires are connected across capacitor C41, providing 5VDC power for the SCART PCB.
A switch on the rear of the MTX is used to turn the SCART output ON/OFF. (This switch is normally used as the channel selector on US model MTXs).
Tucked neatly in the space above the keyboard PCB in the upper half of the MTX case, there is a custom SCART PCB. There are no makers marks on the PCB, so it does not appear to have been a Memotech design.
Close up of the left hand side of the SCART PCB, showing the red and black wires connected to the switch on the rear panel and the 7 coloured wires connected to the MTX video board.
The SCART PCB removed from the MTX
A close up of the left hand side of the PCB
Right hand side
A 7-pin DIN connector is soldered onto the SCART PCB and the MTX end plate has had a hole drilled in it to expose the connector.
The boxes for these MTXs are marked as French SECAM as would be expected, but also contain hand-written "PERITEL" (SCART) text - likely added by the distributor.
I would like to reverse engineer the "Memotech" converter, but there appears to be an "off-the-shelf" converter that you could purchase from that JS Technology would do the job.

You would need to add connections to the MTX Video board to pick up the YUV signals in the same way as the "Memotech" converter does.
NB : I have not actually tried this converter.
The machine also has a piggy back language ROM fitted.

The ROM is connected to the GROM line and prevents external ROMs, such as Node or MAGROM, from working with this machine.


mailto: Webmaster

 Terms & Conditions