Memory reliability: The Write Spike problem and how to fix it
Many TR-808s suffer from having their first pattern and/or first song corrupted, despite the batteries being good. (These seem to be the usual victims, because the CPU lines to the memory are usually low when the problem occurs.)
The cause is a spike from the power cable inside the machine – usually when it is turned off. The spike, which may be a thousand volts or so (guess) with a very fast rise time, is generated when the current to the power transformer is turned off.
The two wires leading from the power switch to the power transformer are in a clear plastic sheath, and are generally pressed up against the main PCB just behind the mode switch.
The exact placement of these wires probably accounts for the variability between machines – so if you put it back together with these wires close to the mode switch you will probably have the problem. Maybe you don’t have it now because the wires are not so close.
Never leave the machine in pattern play (manual play) or track play, when you are turning the power on or off. I don’t know how helpful this will be, but it may make a difference.
Just figure out a way to place these wires so they are not pressed up against the main PCB. I don’t do this, and I can’t be sure it will fix the problem, but it might and it is easy to do.
Complex, magic bullet fix
This is what I do as standard procedure on all TR-808s I work on.
You don’t need to do a complete disassembly, but you do need to be an experienced electronic technician and have the TR-808 schematics.
You must take responsibility for everything you do, including your own safety. See the notes above about the dangerous mains voltages on exposed terminals inside the TR-808. The TR-808 should be unplugged from the mains for this entire operation.
Take the proper precautions against static electricity. (If you don’t know what this means, then you shouldn’t be opening any electronic equipment).
Leave the batteries in, but expect some corruption of memory as a result of performing the modification. This will probably be the first pattern and/or track.
The write signal from the CPU to the RAM chips takes a long march to the Mode switch, so that the Mode switch can physically write protect the RAM by disconnecting the /write signal in two modes.
The first mod is to put a pullup on two positions of the Mode switch so the write line to the RAMs is pulled up solidly in the two positions where the /write signal is disconnected from the CPU.
The existing 1 Meg pull-up resistor (R91) to VRAM (VCC of the battery backed up RAM chips) is evidently to weak to counteract negative spikes capacitively coupled from the power leads.
The second mod is to put a low pass filter into the write line near the RAM chips to filter out short spikes, but let the long write pulses from the CPU through.
At the rear of the Mode switch, are two pins which are currently not connected to anything. Solder one end of a 33 k resistor to these and connect the other end to VRAM: IC7 pin 18.
Now cut the track leading to the left of IC7 pin 10 (conveniently labelled on the PCB). This breaks the active low /write signal from the Mode switch. Solder a 100 pF capacitor between pins 9 (ground) and 10 (/WE) of IC 7. Now install a 10 k resistor from IC7 pin 10 to the track coming from the Mode switch – just to the left of the cut you made.
Now you have a 10k -> 100 pF first order, low pass filter on the write signal to the RAM chips.
These mods, as far as I know, completely solve the problem of intermittent corruption of the TR-808 memory due to the coupling of spikes into the /write line.
Thanks to Robin Whittle.