17 posts
Page 1 of 1
Hi all,

Anybody else notice that the audio after transform to audio is a few milliseconds delayed compared to the MIDI notes? I've tested this with Maschine and ensured the drum hit sample didn't have any offset. It also doesn't happen in Pro Tools or Cubase.

Thanks!

iMac i9 8-core 3.6 gHz, 64GB RAM
Studio One 5.2
Apollo x6 Thunderbolt 3
User avatar
by Lawrence on Thu Jul 09, 2015 2:05 pm
How are you testing it?
User avatar
by 5Lives on Thu Jul 09, 2015 6:03 pm
1) Drag a sample to the arrange page onto a new audio track
2) Create an instrument track with Impact - drag the sample into impact
3) Create a MIDI note starting at the very beginning of the session, triggering the sample
4) Transform the instrument track to audio and compare with the original sample on the audio track

See the difference - I feel it increases the delay the more samples or instrument outputs you have before you transform:

Image

iMac i9 8-core 3.6 gHz, 64GB RAM
Studio One 5.2
Apollo x6 Thunderbolt 3
User avatar
by Jemusic on Fri Jul 10, 2015 12:02 am
Firstly both of those tracks are audio so not sure what you are showing there.

Second I believe you should never do anything right at the start of a session. It is silly and not sensible. Leave two bars and do this test on bar 3 instead. (Even if I start a piece on bar 3 I go back to bar 2 and start from there to hear it play for example. It will always play perfect that way)

I turn midi into audio a lot and never hear any delays by the way. Also 2 to 3 ms should not be audible either. And even it was so you can always slip an audio track back by the same amount to pull it into line.

If you are working with the metronome it can be a good idea to solo each track against the click and use your ears and listen to how the track sits against the click. There are so many reasons why any track will never be bang on a click.

Specs i5-2500K 3.5 Ghz-8 Gb RAM-Win 7 64 bit - ATI Radeon HD6900 Series - RME HDSP9632 - Midex 8 Midi interface - Faderport 2/8 - Atom Pad/Atom SQ - HP Laptop Win 10 - Studio 24c interface -iMac 2.5Ghz Core i5 - High Sierra 10.13.6 - Focusrite Clarett 2 Pre & Scarlett 18i20. Studio One V5.5 (Mac and V6.5 Win 10 laptop), Notion 6.8, Ableton Live 11 Suite, LaunchPad Pro
User avatar
by chris_b on Fri Jul 10, 2015 4:25 am
5Lives wroteSee the difference


In your picture that difference is a single sample.

If it's a concern, note that any effects on your audio tracks might also incur a similar delay - so the real test is to export stems of your tracks and inspect them at the sample level in an audio editor.. but I don't think it's anything to worry about.
User avatar
by 5Lives on Fri Jul 10, 2015 6:51 am
If you have multiple outs from the instrument, I've seen the delay be longer than a single sample. The other issue is if you have multiple instruments in your project that you then transform, they'll all be off, so you'll start hearing phasing, etc. - it'll sound different than the MIDI version, which is a problem and defeats the purpose of transforming to audio. A few milliseconds can add up when you have 100 tracks - and clearly it is not 100% accurate to your composition.

I'm not sure if this behavior exhibits itself if you export the track to audio, but I'll try it (including with automation to see if that is accurate).

Either way, it seems that transform to audio is not sample accurate, which is really unfortunate. Is there a way to file a bug report with Presonus?

iMac i9 8-core 3.6 gHz, 64GB RAM
Studio One 5.2
Apollo x6 Thunderbolt 3
User avatar
by Funkybot on Fri Jul 10, 2015 7:10 am
5Lives wroteIf you have multiple outs from the instrument, I've seen the delay be longer than a single sample. The other issue is if you have multiple instruments in your project that you then transform, they'll all be off, so you'll start hearing phasing, etc. - it'll sound different than the MIDI version, which is a problem and defeats the purpose of transforming to audio. A few milliseconds can add up when you have 100 tracks - and clearly it is not 100% accurate to your composition.

I'm not sure if this behavior exhibits itself if you export the track to audio, but I'll try it (including with automation to see if that is accurate).

Either way, it seems that transform to audio is not sample accurate, which is really unfortunate. Is there a way to file a bug report with Presonus?


1. Are you using any effects in the project? Or is this an otherwise empty project?

2. What exact settings are you using for Transform to Audio?

3. Does it happen with other VST's too or just Impact?

4. If you repeat the same exact steps and start on measure 2, do you get the same results?

5. Can you show us an example of how it's worse in multi-output audio?

6. What kind of system are you running (specs)?

I've never experienced anything like this, and I use multi-output VSTi's all the time including BFD3 in just about every project.

AMD Ryzen 3950X, ASUS Creator x570 Mobo, 32GB HyperX Predator RAM (3600mhz), Radeon™ RX 5500 XT 4GB GDDR6 graphics card, RME Fireface 800, Windows 10 Pro, Studio One 5, Reaper 6, Cubase 10.5, Avid Artist Mix (EuCon please), Behringer X-Touch One, MIDI Fighter Twister, various other MIDI control surfaces and hardware instruments
User avatar
by 5Lives on Fri Jul 10, 2015 8:32 am
Hi,

1. No effects. An otherwise emtpy project.

2. Default settings.

3. Has happened with Maschine as well - will try Battery and others too.

4 & 5. Will try this tonight.

6. 2014 Macbook Pro 15" maxed out

iMac i9 8-core 3.6 gHz, 64GB RAM
Studio One 5.2
Apollo x6 Thunderbolt 3
User avatar
by Lawrence on Fri Jul 10, 2015 8:35 am
...so you'll start hearing phasing,


Not following this one. If you transform a track to audio it's still only one track playing so where does this phasing you hear come from? Afaik, phasing only happens with misaligned identical or like signals, not by just shifting a single track a few milliseconds one way or another.
User avatar
by Lawrence on Fri Jul 10, 2015 9:08 am
Anyway... I tested this and it looks ok to me.

Disclaimer: All of the following is just personal opinion, not meant as food for technical debate.

I'm comparing the Transformed render to a real time recording of the instrument playing because the latter represents what I actually heard before the render and that comparison tells me that the Transform is playing exactly what I heard before I Transformed.

Image

Zooming in to check the alignment. Looks perfect to me...

Image

I think there are generally 2 things going on with these kinds of random tests where some users may be unnecessarily worrying about something....

1. Some aren't comparing the renders to real time prints, which is what you hear, what you need to reliably match, because it's not so much what a thing - looks - like, it more important what it sounds like. .. and ...

2. Many modern daw users tend to greatly over-exaggerate the need for sample accuracy, worrying about it in situations where it has no effect on the result at all. (i.e., if you like how something sounds you want it to sound the same after the bounce and it moving a few samples one way or another won't change that unless you have super human hearing)

For example, when you record audio from live players it's never in the exact same place, for any take or pass, ever, so a track moving a few samples one way or another in a daw after a bounce is generally undetectable and irrelevant, and you'd never even hear it unless it was a big offset. In these cases some tend to freak out over things they see on the screen, not anything they actually heard.

There may also be some other variables in some of these kinds of tests when you start looping things through digital converters, like Record Offset, which I haven't set yet for v3...

Image

... which when setting it to align your system for analog-like audio placement may not reflect exactly where a thing technically hit inside the daw, but more put it where it actually should be, where it hit the input converter. So when visually comparing some things you may be comparing things that have been offset by the host to things that haven't been offset by the host, hard to know sometimes.

As is typical, some of these kinds of tests are fraught with multiple potential potholes and can easily produce flawed conclusions. I typically don't even worry about stuff like this unless I think I hear something really different after a bounce or something.
User avatar
by 5Lives on Fri Jul 10, 2015 11:14 am
Lawrence wroteNot following this one. If you transform a track to audio it's still only one track playing so where does this phasing you hear come from? Afaik, phasing only happens with misaligned identical or like signals, not by just shifting a single track a few milliseconds one way or another.


Say you have multiple kick samples layered together and then bounced to separate tracks. If they are delayed by different amounts, you'll end up getting phasing vs. what you intended. Or more crucially, if you have say a string ostinato from a sample with the sections each playing something slightly different, but rhythmically intertwined, it is very easy to hear the difference if the timing is off even slightly (literally off by milliseconds). They don't call it a "pocket" for nothing.

I'm happy to dive more into what I'm seeing this evening, but please don't passive aggressively assume that it is an unnecessary worry for users or that I don't need sample accuracy. I'm very technically savvy and I was actually trying to bring up a potential bug that Presonus might be able to fix. Instead, it seems this forum of users is more interested in denying that any bugs exist...

iMac i9 8-core 3.6 gHz, 64GB RAM
Studio One 5.2
Apollo x6 Thunderbolt 3
User avatar
by Lawrence on Fri Jul 10, 2015 11:20 am
Say you have multiple kick samples layered together and then bounced to separate tracks. If they are delayed by different amounts, you'll end up getting phasing vs. what you intended.


I wouldn't call that "phasing" really, personally, technically, mmv on that, and I do get the semantics of all that.

It's certainly true that when mixing different kick drum samples together the summed sound changes a little when you time shift or invert some of them, I do that too, on purpose, to get a certain sound. But the one sample shift (one sample?) in your graphic example certainly shouldn't be producing anything audible.

If it is, you have truly spectacular hearing. :)

But sure, if that offset was larger it would easily be audible, the summed difference. If you Transform something that has a much larger gap than what you showed there, I'd agree with you that it could be easily audible, and potentially problematic, and maybe worth investigating.

The "perception vs. reality" roadblock is to demonstrate that what you heard before you Transformed it wasn't exactly what you see in your graphic result viewtopic.php?p=58266#p58266 ... and not just assume it wasn't just because the sample in Sample One is trimmed correctly. So record it in real time and see how it lines up with the bounce.
User avatar
by Jemusic on Fri Jul 10, 2015 3:28 pm
Also in your OP you said the audio was delayed by several milli seconds. But in your image it shows one sample. One sample at 44.1K = 0.02mS Which is it.

Are the samples in question perfectly trimmed so there is NO dead air before the start of the sound.

Have you tried this on another bar eg Bar 2 or 3 instead of the start of the session.

Specs i5-2500K 3.5 Ghz-8 Gb RAM-Win 7 64 bit - ATI Radeon HD6900 Series - RME HDSP9632 - Midex 8 Midi interface - Faderport 2/8 - Atom Pad/Atom SQ - HP Laptop Win 10 - Studio 24c interface -iMac 2.5Ghz Core i5 - High Sierra 10.13.6 - Focusrite Clarett 2 Pre & Scarlett 18i20. Studio One V5.5 (Mac and V6.5 Win 10 laptop), Notion 6.8, Ableton Live 11 Suite, LaunchPad Pro
User avatar
by 5Lives on Fri Jul 10, 2015 5:47 pm
Ok, did some more testing - made sure I didn't have any weird setting on.

When I moved the clips to the 2nd bar, I actually saw MORE delay than the start:

Image

Having multiple outputs enabled did NOT increase the latency as I previously thought - though there is still a latency. Exporting the stem in real time exhibited the same amount of delay as transform to audio. This makes me think that perhaps the MIDI note UI is slightly early for whatever reason - that's a bug since you can't really trust the UI in that case, but not so severe because all tracks will be time shifted it seems by the same amount and therefore be aligned.

To those that don't think this is a bug, fair enough - but see this example in Pro Tools for OFFLINE bounce. I get the same result in Cubase as well - sample accurate bouncing.

Image

Anyway, I just wanted to raise that because clearly it is a bug - whether or not the community or Presonus agrees that it needs to be fixed is a separate matter.

Keep calm and carry on.

iMac i9 8-core 3.6 gHz, 64GB RAM
Studio One 5.2
Apollo x6 Thunderbolt 3
User avatar
by duncanforbes on Mon Aug 21, 2017 7:57 pm
I have exactly the same problem in 3.5 when using transform to audio. I've had this with Maschine and EZDrummer. The delay for me is audibly noticeable. I have to manually change the timing offset of the channel, to get my transformed audio in sync with the rest of the tracks. 5Lives, did you find a solution?
User avatar
by tedannemann on Wed Aug 23, 2017 3:57 am
I assume this is connected also to this mess explained here:

viewtopic.php?f=151&t=18201&p=139599#p139599

viewtopic.php?f=151&t=22254&p=121698#p121698

Bouncing stretched or melodyned material also causes totally unreliable shifts of the audio material. I cannot understand that this bouncing issue isn't treated as high priority - timing is off very off - audibly off! I constantly get a wishy-washy sound due to this issue. If I set up the same project in Logic the timing is perfect up to a single sample!

Studio One 3 (always the latest version) | Mac OS X (always the latest version) | UAD Apollo | Endless list of plug-ins and samples

Please vote for the integration of the Elastique Tape Algorithm: http://answers.presonus.com/7607/tape-r%20...%20stretching
User avatar
by hobotech on Thu Aug 24, 2017 11:19 pm
I agree that something is definitely not right. I'm documenting my adventures in this thread...

viewtopic.php?f=151&t=26322

...still more testing to do...

17 posts
Page 1 of 1

Who is online

Users browsing this forum: leosutherland and 52 guests