Korg PCM Data Name Graph

Ever get the feeling there are too many PCM wav’s in your ROMpler.

I started a bit of a thought experiment by listing all the PCM’s on Korg M3, Krome and Kronos. About 3995 PCM names excluding most of the drums one-shots. Surely you don’t need that many …

Using some simple data reduction like removing numbers and the list is suddenly 1590 unique names. A lot of them are still samples of differing volume and stereo pans.

Once I got the list reduced. I replaced the space between words to be links in a DOT notation file. I used an online Graphvix service to render to SVG and then imported the result into draw.io.

[edit] Added the Synth nodes. Fell over with the Kronos but worked well enough with the Krome and M3. Seems to organise with Krome only PCM on the left and M3 only PCM on the right and PCM on both in the middle.

Getting a bit more harsh on data reduction to only 1 version of each primary unique name in an attempt to get a root set of PCM sampled instruments.

Synth Root PCMs Spec PCM
Korg Krome 175 583
Korg M3 308 1077
Korg Kronos 465 1505+1036 EXs

Drilling into these a bit more. I came up with my Root PCM Data list. Interesting to split out what is really different in the PCM data sets between synths. Answering a question I had … what sounds do you get on each and which are unique and do I even miss them :>

Suppose I should go back and include my M1 PCM data list too. From memory, its only 100 PCM’s … interesting not much has changed in the last 30 years apart from increment expanding of the library of available PCM’s and sample quantity/quality as hardware capability has improved.

1 Like

Great @pac_71 !

I really appreciate all the new tool/software/webapp I’ve discovered with your post :+1:

Except the sound quality progress you’ve mentionned, does a given instrument “sound” the same ?

I only have access to the Krome and iM1 but I have been thinking about a M3 or Kronos. Its hardly sport to compare the almost 4gb of Krome samples to 4mb of M1 samples. I belive the M1 Pianos has appeared on most Korg synths to capture that late 80’s digital piano sound.

From what I can see, the M3 EDS engine is a cut down version version of the HD1 engine in the OASYS and later Kronos. Krome uses EDS-X which I have interpreted as EDS with more sample memory.

It took me a day to get to this point of having some Patches to compare. Going off comparisons I have seen between the M3 and Kronos (Vid1 M3,Kronos mutiple patch comparison, Vid2 OASYS, M3, Krome EX, Kronos Piano comparison) its hard to pick and in a room alone I doubt the average punter would notice.

I think we all know the Kronos is a beast of a synth but interesting to know the better part of the sounds of the beast are in my Krome. So no more excuses for me :>

Its also one of my objectives now to see how much I can get the Zynthian to line up to my list of Root PCM instruments.

You are in the graph database world !!!


I’ve thought about how one would display such relationships. The World of D3 & svg seems an obvious direction. Rendering a context to the screen statically, I’ve found get unwieldy as the parameters increase, and the ability to move around a 3D instance of a field is suddenly something you really feel you want to do!

Im’ sure a very small subset of patches would crop up if such metadata were collected. We used to manage a few sound effects for commercials and the sounds people went for were depressively repetitive.

About 50% of our requests were gun shot and train, aircraft and car types.

Its amazing how many string sounds that don’t sound quite right there are out there! Stick a name like Solina on it ( tm’s and all that…) and it shot up the requests.

Tagging get tedious very quickly, so any analysis that can group, compare and contrast is REALLY useful.

In the zynth world, ideally, we would be able to compare and contrast across engines from a lowest common denominator parameter definition.

Love to see a small sample of the results.

We are heading towards a graphics server ain’t we … ?

I almost threw it into Cypher to upload into Neo 4j but I am not quite sure if a graph or transactional databases will give the best performance. Good old DOT notation and Graphviz was enough to prove there are an underlying set of features that can be extracted. I sort of have the feeling that I can probably get way with a straight DB with a bit of curating of queries and table structure.

Adding people, patch parameters, tags and even patterns/sequences is where I want to go in making a real musical graph. Sort of like Audiobus Statesaving and Patchstorage on steroids.

A machine learning “patch sounds like” function a far off dream, along with matching up the “how to make a song” data with the resulting output Soundcloud audio.

Interesting you mentioned D3 … I think the biggest thing missing from modern music creation is taking advantage of all the available hardware capabilities. Design the UX, the sound, the visual all integrated on the one platform including communications and social connection.

Zenology/ZenCore is not a bad example. Its all the same under the hood, just depends on what form factor you like and how accessible you want to control the sound.

Between a good DSP platform with all synthesis types and IR filters, IO board and your choice of UX … you will never have to buy a synth again! If only such a thing existed! :wink: (Tongue firmly in cheek) The musical graph makes the multitude of choice more tailored and personally accessible rather than information overload induced paralysis.

I drilled into the Korg synth engines a bit more. Very little has changed across the Korg range in 30 years apart from tweaking the basic PCM model and updating samples/hardware. Probably the biggest changes are with the Kronos that allows for wave sequencing as PCM input in the HD1 PCM engine and combi layering with the other 8 synth engines.

Looking at the history. The Krome is essentially a cut down of the last major workstation the M3 with better samples and probability cheaper to produce. So much so the M50 (another hardware copy of M3 without Karma and other stuff) patches can be loaded onto the Krome. Sort of similar to the Nautilus being a cut down/cheaper Kronos.

I would be really disappointed if we don’t see son of Kronos (soon?!?) sporting lots of increased buttons and knobs (a la Roland Fantom) with all the new sound engines like Modwave and lesser extent Opsix/wavestate which are sort of already present in the Kronos.

I might have been a bit harsh on Korg about no innovation in their synth engine. I guess the same can be said for the basic analog model of subtractive synthesis. I did find some an interesting series on analog synthesis and an article on synth block diagrams.