Elektron-Users

Again - you are all missing the point. Forget Midi Clocks, and forget USB, shell scrips and Windows screen resolutions and CROs - Self run, the SPS-1 playing 1/4 note steps into a good audio recorder - the recorded audio divisions vary by up to 96 samples over a 16 event period at 120 BPM. In the end its about what's being recorded as music that I am interested in and that I can hear - and 96 samples deviation at 44.1 kHz is 2.18ms. It's not rocket science. It's what the box plays. Period. If I record the MPC as audio (which is the aim after all) the same quarter notes deviate in duration by a maximum of 4 samples over 16 events which is 0.09ms. That is as complicated as it gets. Please!

A great resource page for anyone interested in timing performance and Midi Delays:

http://www.hinton-instruments.co.uk/reference/midi/promidi/pg05.htm#measure

And a quote I think is most relavant from this page clearly states that any timing errors greater than 0.5ms effect rhythmic timing:

"The trace illustrated above shows a MIDI Out delayed by 0.6ms from the MIDI In. As the device does not read the data until the end of the first byte this breaks down into 0.32ms transmission time (which is constant) plus 0.28ms processing time. If the processing time varies there will be a spread of delays on the second trace and if that exceeds 0.5ms this will start to effect critical timings in rhythmic music."

Regards - David.

innerclock wrote:
Again - you are all missing the point. Forget Midi Clocks, and forget USB, shell scrips and Windows screen resolutions and CROs - Self run, the SPS-1 playing 1/4 note steps into a good audio recorder - the recorded audio divisions vary by up to 96 samples over a 16 event period at 120 BPM.


Shouldn't we expect to see your findings in those other tests, as well, though?

{edit, I'm charging the minidisc recorder now, and will duplicate your method in an hour}

To my non-scientific mind yes Niall but I did have a lengthy discussion with a boffin who uses hardware storage scopes regularly earlier today and he felt strongly the raw audio recording test method in a good DAW application was very valid for tempo deviation measurements where nano-second sampling was too fine. He also said that DSO's can misrepresent if they are set to trigger-sync off one of the actual inputs. I'm not saying this is the case here but either way - he is bringing one over next week and I'll be glad to update the results of course. - David.

http://syncretism.net/snd/snares120bpm.mp3

16 snare hits on quarter notes, 120 BPM. Tempo out turned off. Left channel only.

Also, this was recorded with a dedicated recording device, the Sony MZ-RH1, which should be a suitable medium for all.


I'm sorry, I don't have the wherewithal to look at the results, right now, but if you're looking for another instance, here's one.

keep up the good work guys... i for one find this info valuable, although i will state for the record that 2ms interval deviation, which = more like 1ms deviation 'from grid', is pretty darn good in mho.

i am curious to see the resolution of the whole scope vs. sound card debate. if i had to take a wild guess, i think the sound card method is essentially fine, and someone will find a problem with the scope testing procedure. it doesn't really matter whether the pc OS is 'realtime' since the audio card buffers the information, (theoretically) enough to prevent missed data.

i am glad someone suggested the idea of soundcard samples being skipped/interpolated by the software.. i've never heard of that but at least it suggests a potential mechanism to explain soundcard readings being off by more than 1/44100th of a second, one i hadn't thought of.

regarding screen display, many editors can numerically display the data of each given sample: isn't that good enough?

at the very least, if the md soundcard timing measurements are repeatable across users, we can clearly say that for whatever reason, the md _produces recordings_ with more jitter than the mpc3000. and i agree, thats the important thing, isn't it.



...
and thanks for the interesting comments re the 808 discussion. i am hesitant to believe that 'improving' the 808 timing would help the sound... i listen to a lot off stuff these days made 100% in software and that is usually sample-accurate... a lot of it sounds 'too precise' to my ears, ie kind of dead. so for me, there may really be a 'too precise'. of course those software pieces are typically also lacking the analog timbre variation that makes 808's so nice, so , tough to say...

cheers

I run into the same problem..when i cut loops from the machinedrum i cant cut them all the same length..sound engine jitter when running the machinedrum alone is between 1,5 and 2 ms..

interesting is that you get better results when running the MD in slavesync via turbo midi !!! that actually gives the internal clocking of the MD a rather bad name... maybe a software issue?

So timingwise the MD is not as tight as elektron claims..it is actually one of the worst machine on the market regarding that issue... electribes and all other drummachines i measured are well below 0,3 ms.. when running alone


however..the elektron jitter is pretty relaxed from the charakter..it sounds rather groovy... And the general performance of the machine as master and slave is good...its not so late as a electribe when beeing in slave sync for example
I will do more measurements in the future..i still have to measure the MD?s clock output... hope to find this more stable...
For the time beeing i let my ears judge..and sofar i can life with it..just more fine trimming for loops necessary...

Same jitter you have on old apple systems on OS9 ...and the machinedrum sounds more groovy than that..beats under os9 that was programmed only in the computer usually had a bit tired feeling...



Another thing that is not optimal is that 120 bpm on the MD is not the same as 120 bpm in the real world...

we have the same with electribes and 909?s..but...i actually expected the more refined MD to be more precise there..i like to do free jamming when possible..and this works perfect between a protools on pc and a ableton live on an apple for example...you can run minutes before you hear the drift..

actually there is no drummachine i know that gets the tempo so precise to match that performance... electribes are over 0.1 bpm off for example..

so nothing to blame Elektron here...but in an ideal world 120 bpm just would be exact 120 bpm

and one thing..never use mp3?s to measure timings..the compression is altering the timing and therefore all measurements are irrelevant..

a decend sample clock on 44,1 or 48 khz with a plain wave or aiff file is doing a good job..

44 samples or 48 samples are one ms


The reference machine mpc 3000 has max 2-3 samples jitter...
1 sample is measurements tollerance so we have a max from beeing 0,05 ms timing jitter..

an electribe ES-1 is doing 10-12 samples... so around 0,25ms timing jitter..

The MD is doing 60 samples and more ...so around 1,5 ms timing jitter...

wich is 30 times more than the MPC 3000


An 808 is doing 1sample timing jitter.. equals no jitter..but...every view bars its missing one beat badly... up to 60 samples... but such a thing is no jitter..sound more like groove when sometimes a beat is a bit off while beeing tight most of the time...together with the sound fluctuations of the analog sound engine the 808 sound more alive than a MPC 3000 ...

niall wrote:
Innerclock, there are other threads about this subject, and Daniel Elektron posted in one of them that, if you found evidence of clock jitter with the MD as master, they'd be very interested to hear it - posting a link to this thread in it may be a useful way to link the two discussions.


So...i do it now because i like to know...

midi clock of the MDUW on 123bpm ..

Jitter very good...most of the time sample accurate...max i found 6 samples off..
but as the ears tell..not in a nervous jitter fashion..just ocasionally..

So i can state the MD has a rock solid internal clock...
Sound output jitter must be related to the DSP soundengine..what explains why this appears to the ears rather as groove than nervousity...
when its the soundengine ther is probably a relation between the beats programmed and the charakter of the timing derivation has an internal cycle somehow that gives the thing a more organic feel...


i can live with that...

it would be however good if the One would have priority and be more precise... Than the looplength would be constant what helps when working with DAWs... Or having the one exact every 4 bars..


So... the internal clocking of the MD is rocksolid...the soundengine plays around..

sounds good ..makes more work...

A statement that fits to the MD in general... :-/


Only currious thing left...why i got better results in external sync with the turbo midi interface?...

Maybe a lucky measurement...

I ve to repeat that at one point to verify...because when the MD is better on an external supermidi clock than to its internal rocksolid one than there is a little bug somewhere...

Hi David,

my "moving right along" is at the bottom of every post I make/made here, with the mini keys.

so no insult to you or this thread intended.

I just write lyrics sometimes and mostly they are short 1 line stuff, that happen to be one.

cheers

an electribe ES-1 is doing 10-12 samples... so around 0,25ms timing jitter..

Strange I think my MnM is way tighter than the es1 es1mk2 I have.

Wonder if the Mono is different to the MD?