Automation bug?

Winter_Rat · February 27, 2016, 10:43pm

Hi.
I’ve made an audio file 2 sec long (generated it with Test Generator) and automated Pre-Gain, rendered audio and got very strange result. Take a look at the picture

Audio 02 - is the original audio. (R), (R)01, (R)02 - rendered audio (just three times) with RIP(Channel Settings). Audio 02-Untitled1 - rendered audio with Export Audio Mixdown (Channel Batch Export). The shapes of audio look weird to me. Is a bug or do I miss something?

Win 8.1 x64. Cubase 8.5.1 x64.

Regards.

peakae · February 28, 2016, 7:34am

And you have of course no insert efx on anywhere?

Martin.Jirsak · February 28, 2016, 9:13am

Hi,

Could you share the project, please? I tested very similar automation, and it looks exactly the same as the RIP on my system (Cubase 8.5.10).

Winter_Rat · February 28, 2016, 10:07am

Hi, guys.

Sure I have no inserts or whatever in the project. This was a new clean one.
1.Create Audio Track Mono. Insert TestGenerator to the 1st slot
2.Create Group Track (Mono->Mono Out) for the Selected Audio Track.
3.Create Audio Track (Mono->Mono Out) with Group Track as Input.
4.Create Loop 2 sec. Start at the Left Locator. Stop after Punch-Out. Press record. Done.
5.Edit Pre-Gain Automation. Render in Place 1, wait no changes, Render in Place 2, wait and 3. After that Export Mixdown.

I’m sorry Martin, but I didn’t save the project, but I can create another one, what files do you need? .cpr only?

Martin.Jirsak · February 28, 2016, 10:13am

Hi,

Yes, i need the cpr file. The WAV file, I can create myself. Thanks.

Winter_Rat · February 28, 2016, 10:53am

So, I did some investigation now I know that time accuracy depends on the Buffer size - the lesser - the better.
After that I switched off all C-states for processor in my BIOS. Yes, guys, I had all the C-states On before , but till this very moment I had no headaches with it. So now I can get all my RIP versions in the same wave shape and my Export file looks the same as the RIP one, except Real time Export. Take a look, read Track names.

Now the question is why do I get all those slopes? When I use Volume Automation I get very precise shape. And the most interesting thing is the last part of the Automation - I’ve increased the Gain but in all cases I got volume reduction.

GargoyleStudio · February 28, 2016, 3:26pm

Steps in the volume is what I get at 2048 buffer size, as well as it being out of time!! This is very worrying because I can hear the clicking as the volume steps, see the image… Apart from that its doing the right volume changes though.

The first track is 64 buffer, the 2nd with the steps is 2048.

Mike.

alexis · February 28, 2016, 4:07pm

Can you guys confirm please what I only just recently read for the first time … that rendering projects with automation should be done at the lowest buffer size the interface lists?

peakae · February 28, 2016, 4:29pm

Should ?
For the average fader automation it does not really matter.

If you need it very precise, then yes smaller buffer size equals more precision.
It has been discussed in many “sample accurate automation” threads over the years.

alexis · February 28, 2016, 4:35pm

Thanks for that. Would you also know, if Real Time rendering is not selected for a song that has detailed automation, is the dependence on buffer size removed?

lukasbrooklyn · February 28, 2016, 4:51pm

by the way, this also affects the resolution of MIDI (cc/automation) data. and much more at that, because its resolution is very quantized/limited to begin with, what with having 128 states, most of the parameters anyway. automation then effectively turns into on/off states. i reported that ages ago… like most things, it has been left untouched by steinberg.

J-S-Q · February 28, 2016, 5:03pm

If you use the Export Audio method (not Real Time) the automation inaccuracy is still present and smaller buffer sizes will make it more accurate.