Korg Lambda repair and modification
The Korg Lambdas I have encountered have each exhibited problems:
Fault: A very thin, almost silent, Chorus preset; all other sounds were fine.
Solution: A loose capacitor in the filter bank for that preset, which was a simple solder job.
Fault: A loss of Percussion presets and a thinning of the Ensemble presets in Normal Octave mode; in Up Octave mode everything was fine.
Solution: A faulty CMOS switch (IC1 on KLM-184, a 4066) which failed to pass the divided-by-two clock to the first TOG, thereby effectively muting the first oscillator in that mode.
One case proved more problematic, however, and I have documented the process for reference.
Two main jobs were carried out on this machine:
The main PCB in the base of the Lambda is home to the TOGs, divider/keyer circuits, and associated per-key envelope generators. The lines of diodes/resistors/caps that form the envelopes can be seen quite clearly stretching along the board, and the nine divider/keyer ICs can be seen poking up, mounted vertically on daughter-boards. Each of these smaller boards holds a single divider/keyer IC and some zero-Ohm jumpers.
In this example, two of these daughter-boards had been damaged – someone had gouged away the tracks on the PCBs and bridged the gaps with wire, but solder was everywhere and the PCBs were in less than ideal condition. Rather than simply tidy up the existing mess, I chose to replace one of those boards with a fresh PCB. The manufacture was carried out by Jim Harris, who kindly documented the process in the video below.
The results are very satisfactory. I always install replacement ICs in sockets for ease of future repair. Turned-pin sockets are preferable as I find they assist in seating the IC firmly, but here a flat-pin type is used as it was the only one to hand.
When I initially tested this particular Lambda, it was exhibiting thinner sounds than expected. It became quickly clear that one of the oscillators was not sounding. The beat-frequency indicator LED B was locked in one state, revealing the culprit to be the third oscillator (Miii, as per the schematics).
There are multiple possible causes for a dead oscillator in the Lambda: faulty clock (including its output buffer logic), faulty octave switching, faulty wiring and dry/cracked solder joints/tracks, faulty TOG, badly seated ICs, corroded sockets, dirty switches, or even failed capacitors pulling down the power rails. There are probably more, but these are some options that occurred to me during testing.
In fact, everything seemed OK except the TOG itself, so I ordered a replacement vintage part and installed it. Despite initial success (all three oscillators worked, hurrah!) the new-old TOG failed after a few minutes. A pattern emerged of it working for the first few minutes after switch-on from cold, but then failing – and the machine had to cool down again for the TOG to function once more. It occurred to me there could be another faulty component bringing this TOG down, but isolation of the TOG from the other parts and testing it again showed the TOG itself to be at fault.
These ICs have been obsolete for many years, and are expensive to replace, if they are to be found at all. Stocking up with a batch of S50241s, of unknown provenance and with unknown remaining lifespans, is not a viable option.
How not to build a Top Octave Generator
I initially resolved to build a replacement using CMOS logic to fulfill the same function as the TOG – effectively it is a logic device, taking a clock signal and dividing it by fixed amounts to give outputs such that the pulse trains on them represent the frequencies of an octave scale.
Although the frequencies outlined on the S50240 datasheet are specified to give a certain pitch for a certain frequency clock, the Lambda’s keyboard begins at F, not C. To minimise the circuitry used, the TOGs are clocked so the outputs are pitched to match the octaves on the keyboard; thus, the nominal C output becomes an actual F, etc. This requires a higher speed clock than the recommended 2.00024 MHz, somewhere closer to 2.5 MHz to raise the overall pitch by a few tones.
The need to divide by 3-digit numbers in the low hundreds, some of which have no common factors, suggested the 4040 CMOS counter. My initial sketch is presented here for reference only, and in no way represents a functional circuit:
The clock, buffered by a 40106, pushes the 4040 along on each positive edge. The 4040 is a 12-stage (divide-by-4096) ripple counter, which means it has enough stages to divide by several hundred, but suffers the disadvantage of long propagation delays – each stage only flip-flops after the previous one, so there are several consecutive toggles to go through before a stable and desired output is obtained. The principle of using a clock divider is that when the correct number of clock pulses has been received (as derived by AND-ing together the correct combination of output bits), the counter is reset and an output pulse sent. This triggers a flip-flop to give a 50% duty-cycle output at the correct frequency. In theory, the result is a pulse train with a frequency of clock/n where n is the divide-by for a given pitch.
Experiments were sadly not successful. To illustrate one example, the note A requires division of the clock by 358. The total propagation delay using the 4040 to get through enough cascaded toggles is 65+(30*7) = 275ns. As the clock period is approximately 200ns, it can be easily seen that this is too long. In practice, I found this particular example gave a note around E – the incoming clock pushes the 4040 along before that count’s toggles have settled on the correct combination, erroneously triggering the next division faster than the desired division can be completed; a partial set of toggles is carried out and spills over to the next count – the resulting output is therefore at a higher frequency than desired.
Although there are ways around this, pragmatism demanded a tighter solution.
How to replace an obsolete Top Octave Generator the easy way
Time constraints eventually led to the purchase of a custom-built TOG replacement from FlatKeys, which is a small SMD circuit housed in a compact enclosure, and connected to the original IC position by a ribbon cable. It came configured as a 50240, which is identical to the 50241 except for the output pulse width – not an issue here as the Lambda further divides the TOG outputs at the divider/keyers. The enclosure fits nicely down the side of the main PCB, and the ribbon plugs into the TOG’s socket. It works perfectly, and there is no obvious difference between the original and replacement tones. It responds to pitch bend and modulation as expected. Though it would have been nice to make an entirely DIY circuit myself, simplicity proved the greater benefit.