Digitakt 1.30 : Bug reports

Elektron Gear Digitakt
#68

I decided to test it myself.

I recorded the sine wave from the factory samples at ~40Hz and did a slow filter sweep from low to high through the whole range.
Filter gain was set at maximum.
I cut out the 2 cycles with the largest amplitude.
Recording done with Ferrofish Pulse 16 hooked up to RME FF 802.

Distortion confirmed on EQ 1-4, clipping only on the EQ_1 filter, EQ_5 looks clean.
The distortion in 2, 3 and 4 can be explained by the low-frequency sine wave combined with a high-q filter, which is expected.
The clipping looks more like agressive limiting or soft-clipping to me, it’s not a flat cut off the top.
But maybe thats just the output filter in the Digitakt converter or the input filter of the Pulse 16(no DC coupling) and the sine is actually digitally hard-clipped before that.

EQ 1:
sine
EQ 2:
sine 2
EQ 3:
sine 3
EQ 4:
sine 4
EQ 5:
sine 5

4 Likes
#69

I Tried to Downgrade after I accessed and edited some of my projects ,adding some of the new features and when I downgraded I did not have access to those projects.

Also I think Elk Herd is not compatible with 1.30 which is kind of problematic

#70

I’m getting wildly different results with the EQ_1 filter.

Anyone have an explanation for this one?
EQ 1 weird
How the F does that happen to a sine?

#71

I think the clipping has to do with the large amount of boost you get out of the EQ, atleast that is what they said in the beta when it got brought up. If you drop the initial level on the sample you can then take more advantage of the EQ without distortion. I generally drop my levels pre filter in digitakt anyways as I like the resonance on the regular filters more that way, so I didn’t experience the distortion as much as others. I think the EQ is working as intended… but maybe they are also working on some tweaks to it to make it easier to work with.

5 Likes
#72

Of course the clipping comes from the amount of boost, but it doesn’t explain why it only happens on EQ_1 and it definitely doesn’t explain the asymmetrical sine thats only clipped at the bottom.

The amount of boost should be the same on top and bottom of the wave, and therefore should also clip at the same point.
The clipping on the first example makes sense, but the last one is not expected at all.

At least EQ_1 is not working as intended.

#73

In the first example the bottom does look to be clipping a bit more to my eye than the top also. EQ1 boosting the most content it is harder to say if the other EQs aren’t acting under the same principle aswell, them being more subtle based on the range boosted? I would be curious if you crank the sample level to 127 on the sine to see if the others start to clip and if you can start to see a difference in symmetry. Also would be curious if the results might differ on a non wavetable sine? There very well could be something up with EQ1 but it could be the built in sine.

20210619_181159
20210619_1811592776×1808 608 KB

Just took the time to count pixels on the factory sine and the bottom wave has 12 at the lowest point while the top has 10, maybe we are just seeing an exaggeration of this? Pretty sure the factory sine is not mathematically perfect.

#74

It’s on 127 for all the examples I posted.

The factory sine is perfectly symmetric:
pure sine
The top and bottom match up to .01dB in my recording, I can’t measure more precise.

#75

Maybe a spectral analyzer would be more appropriate to show the distorsion.
A good modus operandus would be seeking the gain level for which the distorsion is gone.

#78

Hello, i‘m quite new here, sorry for any obvious lack of knowledge :sweat_smile:

Talking about ‚waiting with the upgrade‘ and making a backup: the ‚new way’ of making a backup sounds awesome with Transfer 1.4. but i need the latest OS for that.

So what would be a proper (and SAFE!) way to backup my whole DT how it is right now without Transfer 1.4?

Thx guys.

#79

Typically these EQ cutting/boosting tools are made by adding or subtracting a bandpass signal onto the original signal.
It could very well be a headroom issue in the bandpass itself, that translates onto the combined signal – “could”, who knows, there are many things going on in digital IIR filters, the distortion could be coming from a lot of places.

#80

Before Digitakt OS 1.21A there wasn’t an easy way to do a backup. You can try ELK Herd, which is a third party tool.

You can also upgrade your digitakt to 1.21A first and use the new transfer 1.4 to backup your digitakt projects before going to 1.30. (Tip: save your projects in each revision of the os, and also make a backup of the projects in each revision of the os).

2 Likes
#82

I think the ‘E’ is the bootstrap (bootloader) version.

3 Likes
#83

Do you see the same results when you record the Digitakt using the digital USB output? (Just to rule out any external factors)

#84

Can anybody please tell me if it’s a feature that LFO still gets sent from MIDI track if it’s muted? Or is it a bug… :slight_smile:

Hmm… or it had to do with global mute mode and pattern mute mode… do they operate in parallel?

#85

Audio from iPad over USB clicks – it’s also in the recording.

#86

I wouldn’t trust the waveform display on the Digitakt to make that conclusion. :slight_smile:

EDIT: (I assume it takes the sample data and squeezes it into the available pixel space)

7 Likes
#87

Thanks a lot, didnt know that 1.4 works with OS 1.21A! Can you explain in short what you mean with the revision-tip!? Is is some sort of option while doin the backup? :thinking:

1 Like
#88

Haha yeah I suppose you’re correct… I loaded in a sine wave from hardcore and it looks exactly the same on the digitakt screen.

1 Like
#89
  • First update to 1.21a and check your projects
  • Save the projects when they sound fine
  • Backup the projects with transfer 1.4
  • Update to 1.30 and check your projects again
  • Save them when they are fine
  • Create another backup with transfer 1.4

So you keep backups of your projects for each revision (version) of the os.

1 Like
#90

Ah ok, now i gotcha! Thanks!!
I guess i will wait with 1.30 though… and see whats discussed here!