Audio changes in 1.05 / Restored in 1.06!

mzero · February 11, 2018, 9:23pm

1.06 Update

I have repeated the tests again, comparing 1.05 to 1.06, and the 1.06 results to 1.04:

Overdrive: Restored to exactly what it was in 1.04, sample for sample!

Bitcrush: Same in 1.06 as it is in 1.05, but on the whole boosted by 6db if you have Overdrive at 0.00 (as expected). This makes bitcrush have the same effect as it did in 1.04, but with constant gain across the effect range as implemented in 1.05.

YAY! Thanks, Elektron for the quick fix!

1.05 Investigation

As reported elsewhere, there are changes to the overdrive and bitcrusher aspects of DT in 1.05. I’m generally a “the instrument is what it is” kind of musician, and don’t mind such a change, but I also an engineer, so I wanted to see what really changed.

Overdrive

I carefully recorded several different factory samples on a “init” track, with p-locked overdrive settings of 0, 0.01, 8.00, 24.00, 48.00, 64.00, 96.00, and 127.00. I was careful to record with the exact setup before and after the 1.04 to 1.05 upgrade. (I didn’t even touch the cables, volume knob, or my audio interface!)

Surprise: At any amount of overdrive more than 0, the output is sample-for-sample identical to the 1.04 output, except for the now fixed inversion:

With the limits of my having sampled the analog output - these signals sum to zero!

The difference is that the signal w/o overdrive is quieter in 1.05:

What’s interesting is that how much quieter seems to depend on the sample - I saw between 3db and 6db depending on the sample. I think it is more pronounced on the bass end. I’d do more careful tests… but I’m now on 1.05 and don’t have a second machine… ah well.

This suggests that perhaps in 1.04 the overdrive computation was always engaged but with zero distortion, and perhaps now at zero it is removed from the processing chain.

If you had patterns from 1.04 and you want to preserve the exact balance: For any track that had an overdrive setting of 0, just set it to 0.01. You can see from the image that in 1.04 the volume of 0 and 0.01 is the same, and there is no overdrive in the signal at the 0.01 setting that I was able to hear or measure.

Bitcrush

I did the same style test with bitcrush - using values of 0, 8, 16, 32, 48, 64, 96, adn 127.

At a setting of 127, the output is again exactly the same. At other values, the output volume is now consistent, where as in 1.04 at some values it was significantly louder.

At low values it is very hard to see the effect of bitcrush on the waveform, but as far as I can see and hear, the effect hasn’t changed, just the amplitude (so as to make it consistent across the range). Over all, this should make the effect easier to apply, since you don’t get wild jumps in volume anymore.

DaveMech · February 11, 2018, 9:28pm

Thanks for this! Really great info. I’m going to upgrade now as this is easily fixable. Thanks a million!!

X · February 11, 2018, 9:35pm

Thank you so much for taking the time and effort to examine this!

Chiron · February 11, 2018, 9:43pm

Thanks for the effort! These are useful insights!

t · February 11, 2018, 9:43pm

Awesome. Was planning to do this, but you beat me to it!

Buska · February 11, 2018, 9:43pm

Thanks for bringing the facts @mzero

Flohr · February 11, 2018, 9:58pm

Science ftw

re5et · February 11, 2018, 10:16pm

in summary.

1.05 is fixed , and 1.04 was broken due to inversion and ‘stuff’ ???

@mzero do you imagine it’ll change again in 1.06 ?

mzero · February 11, 2018, 10:38pm

First, let me say, I appreciate the between a rock and a hard place that releasing software puts one in, especially when it comes to subjective creative software: One person’s ugly effect curve is another’s perfect response. Do you change it or not? Live with your initial sub-optimal design forever, or change the way the instrument responds breaking exiting music? Support both with a setting? (This quickly becomes unmanageable for both users and developers.) There is no good answer!

So with that said, if I were in their shoes, I’d consider bitcrusher done-and-done.

But I’d think that that volume difference at the start of overdrive is a bit much: It means that if you have a track with overdrive = 0, and plock overdrive on a trig, then you’re almost certainly going to have to plock the volume on that trig as well, which is less than perfect.

I don’t know if the change was really “take overdrive out of the processing chain” in 1.05 from “overdrive is always in there, just set to no-effect” in 1.04 ---- but if it were, I’d be tempted to put it back the way it was.

effy.st.duvall · February 11, 2018, 10:41pm

Thank you for doing this! Good to know that much can be done to revert back to prior levels. It at least provides some peace of mind with how my current patterns may be handled moving forward with each OS update

Flohr · February 11, 2018, 10:56pm

Can you comment on the reverb only in the headphone out issue that some people are reporting?

mzero · February 11, 2018, 11:09pm

I find no difference at all between headphone out and main outs. If I select a very clean sharp sound (CB Digit is a good choice) and play it both with no reverb, and with reverb, I hear the same thing in both headphone and outs: The clean sound is clean, the 'verb’d sound is 'verb’d. I used a long tail setting on the reverb, so I could time the decay to get an indication of the wet/dry mix: came out identical.

Flohr · February 11, 2018, 11:51pm

Interesting. I wonder how it is possible that other people have been experiencing problems? The only options seem to be either (a) they are incorrect and/or delusional or (b) there are hardware defects present in some units that are being revealed by firmware updates. B seems extremely unlikely but not impossible.

Flohr · February 12, 2018, 2:05am

C is reducible to A.

Ryan · February 12, 2018, 2:13am

Damn! OD was always on on 1.04! That explains the lack of noticeable effect when I messed with the od on that OS!

Good to know! Thanks!

ZORDON · February 12, 2018, 2:35am

as always @mzero proving to be a valuable member of the elektron community. thank you for your time !

man909 · February 12, 2018, 2:39am

1.04 OD sounded better. New one, volume jump is noticeable and hyped sounding. But whatever! Don’t use it much.

provoda · February 12, 2018, 2:39am

Thank you @mzero!

I tend to agree, bit reduction sounds much more useful now and it’s easy adjust your current patterns to sound as before. But overdrive “jump” from 0 to 0.01 seems like a bug to me since there is no mentioning of this in 1.05 changes list.

mzero · February 12, 2018, 2:51am

I think you missed the point of my investigation: OD on 1.04 and 1.05 are identical. The only difference is that when OD is set off (0.00), then in 1.05 there is a notable volume drop - though that’s all.

man909 · February 12, 2018, 3:20am

There were some folks on a diff. thread saying it “sounded so much better”

so it sounds the same you say?