Memotech MTX - Various Failures

System Description : MTX500 Serial No.(missing), 4000-04 computer board with 32k DRAM chips.

Problem Description : System dead, black screen with no sound at power on

Resolution Summary : Replaced failed 74S04, 74LS374 and upgraded on board RAM to 64k

Darren sent me two faulty MTX computers, his original "MTX500" (which has a MTX512 keyboard) and an MTX512 that he had obtained later, and made the attractive offer that, if I fixed his original machine, then I could keep the other one for the collection.

Darren reported that the "512 boots to a ready prompt, but the keyboard doesn't work, the 500 simply shows a black screen."   When they got to me, neither of the machines booted and both displayed a black screen on power on. Although Darren wasn't in any hurry to get his "MTX500" back, (he was about to move house), I decided to start with the "MTX500" to get that repair out of the way then I could work on the "MTX512" as my time allows (or doesn't!).

Initial inspection of the computer board revealed that someone had previously worked on the machine; as well as a number of passive components having been replaced, the 16k ROM had been removed from its socket and accidentally reinserted with pin 16 (D4) not engaged in the socket! Had there been no other faults, this would obviously have stopped the system from working, but inserting the chip correctly made no difference to the fault. 

Darren didn't send me his power supplies, so I used one of my known good ones for the testing. The first checks were to put a DVM on the DC voltages - the VRAMs are a convenient point to measure all of the DC voltages; Pin 1 should be -5VDC, Pin 16 should be 0V, Pin 9 should be +5VDC and Pin 8 should be +12VDC.

Worryingly, although the other voltages looked OK, the voltage on Pin 1 was -6.0 VDC which I thought might have damaged the VRAMs. With the video daughter board removed to get better access to ZD3, the 5.1V zener diode that controls the -5VDC line, I found some debris on the -5V side of the diode. I don't know what the debris was and why it should have affected the voltage, but cleaning it off returned the voltage to -5VDC. However, it had no effect on the fault, but did leave me with concerns about the integrity of the VRAMs.

As is common, the cable between the video daughter board and the back panel connectors had broken, so my initial testing was done using the TV output from the modulator. One of the most frequent "fatal" fault symptoms on the MTX is a black screen accompanied by a constant tone from the sound chip. Since I had the black screen, but no audio tone, I wondered whether the RF modulator was faulty, so I repaired the AV connections so that I could check the sound and video outputs straight off the video board. Again, this made no difference to the fault.

Since I have the luxury of having quite a few MTX spares, I swapped out the majority of the socketed chips, i.e., the CPU, CTC, VDP and ROMs with known good replacements, all to no effect. When I was doing this, I noticed that someone had made an odd modification in the region of the clock circuit on the PCB - including adding a transistor where one was not originally installed. The 74S04 (hex inverter) in board position 9D, used in the clock circuit, was running VERY hot - an indication of possible failure.

Close up of the clock circuit modifications. A resistor and capacitor were connected to one side of the blue 47pF capacitor below the 74S04. What is not clear in the photo is that that this leg had separated from the capacitor body, i.e., the capacitor was open circuit and likely the cause of the failed clock signal.
As described in this link on the Learning about electronics website, and illustrated here, it is possible to make an inverter out of a transistor, so I theorised that, rather than replace the failed IC, someone had used a transistor to replace a faulty gate in the 74S04. Putting a 'scope of the clock line to the CPU revealed that the expected 4MHz clock was absent so I decided to replace the failed chip and remove the transistor "bodge" to return the system to its factory state.
The clock modification components, consisting of a 2N3905 PNP transistor, resistor and capacitor. When installed, they were connected as shown below.
The modified clock circuit, the changes from the Memotech design are shown in red. (R35, a 330R resistor, is normally installed adjacent to R36. It had been removed and the transistor installed as shown.) A 100uF electrolytic capacitor has also been installed between 5V and 0V adjacent to the 74S04 but I just left this in place.
Ta-da ! - With the 74S04 replaced with a socket, a new chip installed, R35 replaced and the surplus components removed, the machine now booted to the "Ready" prompt - the composite video output is shown here. You can also see that my previous concerns about potential damage to the VRAMs from the out of spec -5VDC line were unfounded - the video output is fine.
However, some problems remained, there was a constant tone from the sound chip and the TV output was unusable - although the "Ready" prompt was present, it was not possible to get a decent picture from the RF output. Having spoken to Darren, we decided that there was no point in spending any time trying to improve the modulator output, Darren, like most people these days, would be using the composite video output and had no need for a working RF output. So, the next issue to be looked at was the audio output.
The MTX generates sound using a TI SN76489A complex sound generator (CSG) and the SN74LS374N octal flip-flop that buffers the data to it. The relevant portion of the circuit diagram from the MTX manual is shown opposite.
The system was generating a constant tone with no sound commands active, but, did respond to a <CTRL><G> ("Bell") command, although the result was a strange siren type sound, rather then the expected "ding", so the CSG was working, at least, to some extent. Although I should have followed a more structured fault finding approach, I started by swapping out the CSG - as I should probably have expected, this didn't make any difference. At that point, I got the 'scope out and probed the sound port control lines, OUT(6) and IN(3). Both of these appeared to be working as expected so I compared the flip-flop outputs in response to a <CTRL><G> with the outputs of a working MTX. On the working MTX, all of the outputs were seen to change state, but on this machine, bit D7 (pin 2 of the '374) was stuck low. As I did not have a 74LS374N to hand, I replaced it with a fully compatible SN74HCT374N and the sound problem was resolved. Note : since this article was written, at least two other MTX computers have been reported as having a constant tone from the sound chip. In both cases, replacement of the SN74LS374 has resolved the issue.
Although he had sent me an MTX500 and an MTX512, as it was his original machine, Darren wanted me to return the MTX500 to him, rather than the MTX512. When he mentioned that he was probably going to look into making a 32k expansion board, I offered to upgrade the onboard RAM to 64k instead. This process is not without risk, when I first tried this a few years ago, I managed to brick the MTX and ended up using one of Andy's memory boards to get the system working. I did eventually manage to recover the MTX and got the RAM upgrade working. These days, my toolkit (and my ability to use it) has improved and, having done quite a number of PCB repairs since, I am relatively confident of doing these upgrades without issue. In the unlikely event that it does go horribly wrong, then a separate memory card would be the fall-back.
As I had done this upgrade before, I didn't take step-by-step photos, but you can see the steps involved here. This photo shows the new RAMs installed in sockets and the PAL replaced with a suitably programmed GAL16V8A using the JED file from Andy's hardware page.
Proving that the RAM upgrade has worked and the full compliment of RAM is visible to MTX BASIC, PRINT PEEK(64122) returns 1, signifying 2, 32k pages of RAM are installed.
That completed the repairs and upgrade to the computer board, all that remained to be fixed was the keyboard. The keyboard on all but one of my MTXs has a pin header soldered onto the keyboard PCB which connects to the computer board with a 20 way ribbon cable. My other MTX, and this one from Darren, had the ribbon cable soldered directly onto the keyboard PCB. On this machine, the ribbon cable had been torn from the keyboard PCB, rendering it useless. I removed the remnants of the ribbon cable cores from the keyboard PCB and used a Dremmel with a 1mm diameter drill to enlarge the holes so that a SIL header could be fitted. When repairing the broken cable, I saw that the keyboard itself was in pretty poor shape. There was a large amount of corrosion on the metal plate that supports the keyboard PCB, possibly due to some corrosive liquid (coffee?) being spilled on it. It didn't come as a surprise to find that, when the keyboard was plugged in, it had more than a few problems! When new, the MTX keyboard worked really well, but, when not used for extended periods, even well preserved keyboards can be problematic. The key-switches can become stiff and/or suffer from varying degrees of "bounce". The majority of key-switch issues can be resolved by thoroughly exercising the affected keys, i.e., giving them a good workout, will usually restore the keys to an acceptable level of operation. If that fails, then the key-switches can be replaced as described on this page. With the keyboard connected, I found that one key-switch ("M") didn't work at all and quite a number had unacceptable bounce - even after vigorously exercising them. This meant that I had quite a few switches to replace.
With the keycaps removed, you can see how badly corroded the metal plate was. (The shiny new cable header can be seen at the left side)
Mounting plate cleaned up and various key-switches replaced
That completes the restoration of Darren's original MTX500 so he can relive the joys of MTX computing once he moves into his new house. Fault finding and repair of the MTX512 that I am keeping was a much simpler affair but is the subject of another article.