Conventional wisdom has it that there were distinct versions of the MS-20 over time, the original having a pair of KORG35 chips on the KLM-127 board, but no supporting screw into the heatsink from the VCA panel ^[1]; the screw was apparently added after some time, and then for no very clear reason about a year later the filter was changed, with a daughterboard (KLM-307) being wired in in place of the KORG35s, using LM13600s. ^[α] [2] While online comments sometimes say it’s not so simple, this often seems to be interpreted like ‘there were specific points of change from early no-screw KORG35, to screw KORG35, to late-model LM13600’. If so, it should be possible to use serial numbers to construct a timeline for the changes?

Examples Found

Interim Conclusions

Well, that’s a no. There seem to be some anomalies here, if we believe the story that the heatsink screw was added and then the filter was changed, simple as that.

Heatsink Screw

Clearly, the screw may be present at any point in the sequence. (Current totals: 160 with screws, 135 without.) Why this would vary as much as it does is unclear. Something to do with the weight of the heatsink? And that would perhaps be determined by the heat dissipation requirements of the voltage regulators . . . ? Strength of the boards used at different stages? Or perhaps it’s to do with the mechanical strength of the main attachment of the board to the panel, which is mostly via the potentiometers’ soldered legs; if the pot type differed over time, perhaps some legs might have been thought weaker. But on the basis of pictures I’ve seen (see Board Versions, below) of non-screwed KLM-127 boards’ component sides, there doesn’t seem like that much difference.

Filter Type

The 43s are strange representatives of the block, all found so far being over 439000, and all 100V; perhaps it started at 439001. Which would make sense since the lower 43s are occupied by the EPS-1. Presumably Korg planned to retire the MS-20 before the thousand was up. With those exceptions none of the numbers found are outside the 140001–149999 range, so production probably stopped at around 10,000 (plus perhaps only a hundred 43s, and perhaps minus an unknown number of 14- tuners? ^[6]). So there’s nowhere else for a significant group of “MkII”s to be hiding. As some of the last of the 14- block are certainly KORG35, it seems that only some of the later MS-20s had LM13600 filters? And the earliest LM13600s started in the 3000s? Perhaps it was simply an availability issue rather than a design preference. ^[7]

There may be a reporting bias in these results — as the KORG35 filter has acquired a superior reputation, perhaps people are inclined to say that they have it for sale when they think they do, but people who don’t are inclined not to mention it. Probably several of the unknowns are in fact LM13600 but it’s impossible to say which. Even so, it seems that the LM13600 filter may be substantially the rarer of the two. Currently 9:50, but take that with a large pinch of something. And it should be noted that only a very few of these results are clearly demonstrated; some of them may simply be writers’ opinions, based on conventional wisdom about the relationship of screws and filters. ^[8]

I think it is possible to say that the KORG35 filter is not “old” or “MkI” in production terms, though it is the early design; but while it is a later design, the LM13600 board was only used intermittently. And it seems several KORG35s come later in the sequence than the last LM13600. (Update, 2023-04-07: We now have a confirmed LM13600 in the 439- block.) (However, in recent years Korg themselves have taken to referring to the Rev.1 and Rev.2 filters. Not, as far as I can tell, the whole synth, just the filters.)

Filter/Screw Correlation

So far — but this is a very limited sample size — all LM13600s identified have also had the screw; which might make sense in terms of the slight extra weight of the daughterboard, but it’s only about 40g. And it’s clearly not the determining factor in screw use. However, I can’t rule out the possibility that the first KLM-307s were attached by a different means to early-revision KLM-127s which did not use the screw.

Summary of Filter Distribution

While the identified filters form a small sample size, they are beginning to show some clustering behaviour, indicating that there were distinct periods when the two filter types were in use. As a quick summary of the above, these are the currently observed clusters across the serial number range. (Some clusters have only one member.)

The chart also shows the distribution of serial numbers found. Note that most of the numbers do not have identified filters. The table below lists the clusters.

At this point it appears that there were at least six distinct periods where the KLM-307 boards were used, mostly significantly shorter than the periods when the KORG35 chips were used. As at mid-2024 this number has been stable for about a year, but it could change as more information is acquired.

Board Versions

At least versions C, D, E, F, & G of the KLM-127 board exist. I did think there was no significant distinction between them, but now with close examination — albeit with a small sample set of 1–3 each — it seems that:

1. versions C through F have KORG35 chips and associated components on board, but version G has connection points for the KLM-307 board instead; and
2. versions C, D, & E have no panel screw hole, but F & G do.

Unfortunately there are only three cases where I can directly associate board types and serial numbers — 147891 is version D, and the personally examined 148721 and 149500 are G and F respectively. It is unclear whether there may have been earlier boards with panel screw holes and/or KLM-307 connections, or whether these things may have been done differently. ^[9] But the first known serial numbers are associated with screws, and some LM13600s were used well before the only known KLM-127D. So these relationships remain tentative.

The KLM-127 diagram in the (original) service manual does not give a revision number, but it appears to be earlier than rev.F at least. ^[10] (I would of course expect it to be the original production version rather than anything later, but who knows?) n.b. while the KLM-307 is sometimes described as a filter board, some examination of the service manual indicates that in rev.G, the VCA circuitry is also on the KLM-307 rather than the KLM-127.

(Update: Turns out VSE is now hosting a revised version of the MS-20 service manual with clearer diagrams and additional LM13600-version schematics and drawings. It distinguishes “old production” and “new production” KLM-127s but doesn’t give revision codes.)

There were multiple versions of the KLM-128 (patchbay) board too, though I’ve only seen rev.C, and I’m not clear what the differences were. The service manual for the MS-10 shows a revision of the noise generator circuit, with the revised version being the same as the circuit shown in the MS-20 manual. There is no indication that the circuit changed in the MS-20, but it may have?

An Unlikely Explanation

One possibility that should be acknowledged, though I don’t think it likely, is that the serial numbers are not a chronological sequence, but may have been assigned either randomly or by some other algorithm from within the 14- block. That would explain the above results, but still . . . 

There are two specific points of evidence for Korg using incrementing serial numbers in this period:

1. The service manual for the MS-10 states that a change was made in the noise circuit, so that all MS-10s from 132828 had the new type — strongly implying a chronological sequence.
2. The numberplate on Sigma 5663 appears to have been stamped twice with 5662 and then 5663.

And in general, some models with model-number type serial numbers seem to show (relatively) low production volumes with serial numbers concentrated at the low end of the number space, e.g. at the time of writing all BPX-3 serial numbers found are between 400001–401500 rather than randomly distributed from 400001–409999. As a back-of-a-spreadsheet calculation, the average numberspace proportion of the low-end and high-end numbers found (where >1 numbers known) for each model are at 0·044 and 0·489, which would be a −0·234 bias compared with random or other even distribution. I don’t think anything other than normal ascending incrementation is tenable for this time period.

(More recent manufacturing has been able to calculate and print quasi-random serial numbers with built-in validity-checks, but that would have been an atypical use of computing resources in 1978.)

Other Models

Several models of MS-20 have been produced, but this article is only about the originals, not the 21^st Century models (MS-20ic, MS-20 Mini, MS-20 Kit, MS-20M Kit, MS-20M, MS-20 FS) or any future ones.

Korg did produce two models of large flat wall-hung MS-20s, for demonstration and teaching purposes. Model names are unclear — “MS-20b”, “Blackboard MS-20”, or “Big MS-20” are used (see ref.  ^[θ]). Both were produced in very low volumes and I haven’t seen any serial numbers for the earlier version. The second version appears to have numbers in the normal sequence but with ‘EX’ interpolated.

A Note About Colour

I have read some comments online to the effect that the case colours of MS‑20s varied over time between a blacker and a greyer colour. None of these comments cited any evidence. Photographs I’ve seen could be interpreted as showing a large amount of variation but these shots are taken in widely varying lighting conditions, with hundreds of different cameras, so it would be surprising if there were not some variation even without the possible effects of 40+ years of fading in different climates and state of cleanliness.

All I can say myself is that checking the colours of the clean undersides of all the MS- series devices I have shows that they are nearly indistinguishable from RAL 9017 in Northern daylight, but the rear/top/front panels (which could hypothetically have been affected by exposure) are a little lighter — lighter than RAL 9004, but not as light as RAL 7021 — and are perhaps a fraction bluer but that may not be apparent in warmer light. I expect the original colours were specified in the JPMA system. Instruments checked include an early-ish MS‑10 and a late MS‑20, so while I would be interested to see more evidence, a deliberate change during manufacturing seems unlikely.

A Note About Semi-modularity

I tried to explain the meaning of the terms modular and semimodular, giving the MS‑20 as an unclear example, so I might at well add this here for reference. (I’m adding my basic operating definitions in the Glossary.) Is the MS-20 a semimodular synthesiser? Quick answer: On average, yes.

More detailed: It’s a unique combination of non-modular, modular, and semimodular. Most of the functional units are hardwired (normalised), but some of the signal routes cannot be replaced. e.g. there is no way of taking the output from the filters separately, swapping their order around, or feeding an input directly to the main VCA. So these connections are not even semimodular. On the other hand, the Sample & Hold module is a separate module with no normalised routing, as is the ESP, and so are the momentary switch and modwheel; and while the auxiliary VCA does have a normalised control input it doesn’t have normalised signal i/o; so to varying degrees these are in fact modular. For simplicity — I think it averages out.

(This compares to the original semimodular synth, the ARP 2600, in which all the normalised routings between modules can be overridden, but even that has some modules with no normalised routing. It seems to be unusual for semimodular synthesisers to meet any strict definition anyway.)

Comment or Question about this page? write

Article text ©2022–2025 Electropict .
Click images for individual licences.