Prepping Samples for MD


i’m exploring a Blackstar HT-Metal tube overdrive guitar pedal that is so powerful, it is like an amplifyer head… and offers cabinet emulation output.

the channel A is on ‘clean’ setting, allowing for clean boost of the input by passing it through the hot vaccuum glass tube. Decently sized hot glass. The pedal also has an EQ section and a Bias setting… the signal path is completely analog.

anyway this is working great as an insert effect for the bassdrum.

in Global Settings i assigned Trig track 1 to Output E, taking the signal by lead to the input of the HT-Metal pedal. The output of the pedal then returns to the Machinedrum Input A, and on Trig track 16, i made an Input machine for the returning signal.

As regards finding out whether the two-tier system regarding 44.1kHz and 22.05kHz for automatic encoding of samples by the Machinedrum, the way to verify things is to feed the MD two versions of the same sample, then check what file size they are after the import and MD’s subsequent encoding process (happens at time of import).


That is pretty good to know about. Did you read that or figure it out yourself?

The sample rate thing is interesting too, and does beg the question as to what happens between 22k-44k…


was on a special page at Elektron Users, a collection of tips and also a couple of manual quotes.

sample preparation and choice is a fascinating process.

i found ultimately with going a bit extreme in the search for economical sound files, topping and tailing them, it was a diminishing sum game.

so too for sample rate reduction…

going through many samples at the moment, i’ve realised i need to forget about memory limitations and focus on simply finding the samples that resonate with me in the most enjoyable fashion.

ultimately, the Machinedrum shines as a drum machine. A sound sculpting laboratory also, yes, very much.

so i am linking ableton on a spare laptop via the tm-1, set ableton to external sync, and running pre-composed synth/chord loops on there. two channels only, one is side A, the other, side B.

The Machinedrum controls the individual track volumes, the ableton crossfader, and the bass, mid, treble gain/attenuate parameters with the other three rotaries.

therefore using Trig machine 14 to be a Mid machine,
giving access to six control channel rotaries.

the audio output of Ableton then feeds back into Trig track 15 via an Input Machine…

having immediate access to chords and more spacious sounds, or basslines, relieves the pressure on the drum machine to overperform.


I actually do a similar technique for one very much DAW produced track I have done. I route the two main sources to left and right of stereo bus, bring them into the MD and use the MD just to manipulate these two feeds.


very cool … using the MD almost like a mastering plugin…

so, i’ve actually made the call to go all-hardware now, so it will be a Korg Prophecy feeding into Input A, and a Midi Machine on the Machinedrum playing that.

Input B will receive input from a 5 string bass going through an Octave Multiplexer and a Blackstar preamp/tube-overdrive.

Essentially the Recording machines are Input machines, accepting a mono input and transforming that into a stereo left and right signal… two functions thus are achieved with a Rec Machine, proper L/R signal from a mono signal, and the option to record at any time.


It works well. One other way I’ve used the input was for an acoustic guitar, through a preamp, so I could get tempo sync’d delays with the beat programmed from the MD.

I’ve always wanted to try hooking up a midi foot controller and using that to mute/unmute rec trigs effectively making the MD a looper.


nice one, mute/unmute rec machine trig tracks to achieve hands-free recording trig/untrig… what a brilliant solution.

i’m thinking the EC5 would be perfect for this, as it plugs directly to the Prophecy, then route through the Midi Out to the Machinedrum Midi In …

whilst at the same time, playing the Prophecy with a Mid Machine and sending midi clock tempo for arpeggiation/fx sync.


prepare a sample for Machinedrum at 22050hz and the Machinedrum will keep it at that rate.

any rate above 22050hz gets auto-converted on import to 44100hz.

so, that is a potential 50 percent memory saving for any file that handles 22050hz rate.

right now i am looking at the evidence lol … same file, same length, same everything, different sample rate. On Sample Slot 21 the 22050hz version at 62kb and the 44100hz version at 125kb is on Sample Slot 22.


filesize reduction is a diminishing sum game, however :smiley:

one rule of sample rates is that the rate should be twice as high as the highest sounding frequency in an audio file

in addition to this, something to think about perhaps is that as upper partials contribute a great deal to the sense of timbre in lower spectrums, it is unpredictable as to which way a sample will go when converted to this rate, so the best thing is to try with a range of sounds.

sometimes, not much difference. sometimes, the 50 percent memory saving is also evident in the quality of sound.

4lay3 skin hh 22050 mono.aif (93.9 KB)
4lay3 skin hh 44100 mono.aif (187.7 KB)

also to mention: these files are 24 bit. When the MD converts to 12 bit i thought it would be a 50 filesize reduction although in practise appears to be a 33 percent filesize reduction.

anyway the 22050hz rate is holding so that is fantastico


Bro your posts are inspiring me to turn on the MD.


more samplechains…

to load into any trig slot numbered 32 and above (where the start parameter geometrically divides the imported sample as though it were a loop or samplechain) …

i set the Hold parameter to about 5 or 6, and then set the Start parameter at either 0, 32, 64, or 96, to access the four different hihats.

hihat in 44100hz samplerate, and also at 22050hz. Machinedrum won’t change the sample rate for these two levels, so if the 22050hz version sounds cool, that is a fabulous economising of memory space.

as regards filesize, there is a further reduction as the resolution is changed to 12 bit by the Machinedrum on import.

4hh1 44100 mono.aif (187.7 KB)
4hh1 22050 mono.aif (93.9 KB)

4 bassdrums in one geometrically spaced audio file…
4bd1_44100_mono.aif (249.1 KB)

4bd1_22050_mono.aif (124.6 KB)

snare drum 4-part sample chain…

4sd 44100 mono.aif (187.7 KB)

4sd 22050 mono.aif (93.9 KB)

bass chain…

fab_bass_chain4_1m v2 22050.aif (170.0 KB)
bass_chain4_1 v2 44100.aif (679.7 KB)

skin drum 4-sample file…
4skin1 22050 mono.aif (93.9 KB)
4skin1 44100 mono.aif (187.7 KB)

layered bassdrum and snare drum combinations, 4 samples, 1 file…

4chain layer bd sd 22500 mono.aif (93.9 KB)
4chain layer bd sd 44100 mono.aif (187.7 KB)

another layered-sound sample chain…

4lay4 bass skin hh 22050 stereo.aif (187.7 KB)
4lay4 bass skin hh 44100 stereo.aif (375.3 KB)


Cool shit, pretty weird it up and down sample converts.

I have fond memories of Setting up 4 ram recorders and sampling on the fly with mute and unmute from a poly six, Juno 6, spaceecho, mc202, monomachine, and ableton


man that sounds like amazing fun.

yes as regards the two-tier sample storage system, if fed a sample via sysex transfer, the Machinedrum will actually maintain that one lower option of 22050hz. It was a snippet of information found on an Elektron Users page, when searching for other information. I found it surprising, as the manual doesn’t say anything whatsoever about this, although really the manual doesn’t say anything at all about the Record machines except for what each rotary does.

anyway as mentioned, sample rate reduction to achieve memory economy is a reducing sum game … the 44100hz versions sound better usually.


thinking about sample chains for the machinedrum i now realise that what may be most useful is collections of sounds of types that might not be usually accessed by the one drum type.

as in … rather than a string of four bass drums, a combination could be much better of one bassdrum, one snare, one high hat, and one clap or skin drum.

or no bassdrum at all in the chain.


Hey, I’m trying to make sample chains for the machinedrum but the logic behind it is a little hard to wrap your head around.
I’m trying to make chains of however many samples can conveniently fit per chain without losing any more audio quality than the MD inherently does.
I’m happy to pitch every sample up one octave above intended playback so that less memory is used.
However, at what interval should I set the One Shot samples to be, and how many of them should I fit in one chain?
For years I’ve attempted but its always been a little messy and I’m finding it hard to catch the start of a sample within the chain.
I hope someone can help me out here.


I need help with calculating file duration…

2,5 MB of sampling capacity equals to around 77 seconds of mono 12 bit .wav files in 22050 hz sampling rate?
With another formula i got around 12 seconds! :confused:


I tried too, hoped some loop possibilities, to use is as a wavetable without success (even with loop option checked from C6).
For one shots, theorically you need 128 samples with equal lenth.
Start 0, End 1 > sample 1
Start 1, End 2 > sample 2

Considering pitch, IIRC +36 = 1 octave, +57 = 1 octave and 7 semitones. Above it seemed out of tune.
Sold MD recently.


Thanks man.
Do you know what is the length of audio between STRT value 0 and 1 at 44.1kHz?
Or does it depend on the PITCH parameter?
Does anyone know?


ok! i found the answer! total time is: 1 Minute : 15 Seconds . 585 Milliseconds
and for 44100 hz is: 0 Minutes : 37 Seconds . 792 Milliseconds


It depends on total sample’s length, and sample slot position :

The ROM machines ROM-25 to ROM-48 operates slightly differently since they are optimized for loops. How the parameters differ from the other ROM machines is described below.

Not linear otherwise.

Pitch reference is 0.