1b anyone ?

lucky_bleeder · 09-06-2010, 12:21 PM

Can't believe no one else spotted this....

v3.68pre1b-test - September 5 2010
+ Experimental test change to sample-rounding code, may fix non-resampled timing issues

beatbybit · 09-06-2010, 12:25 PM

http://forum.cockos.com/showpost.php...&postcount=118

almost an hour ago

rob.manderson · 09-06-2010, 01:07 PM

The directory appeared yesterday - empty until today.

musicbynumbers · 09-06-2010, 02:51 PM

Quote:

Originally Posted by rob.manderson

The directory appeared yesterday - empty until today.

ROB!!

why have you forsaken us!! lol

steadyrev · 09-06-2010, 07:46 PM

It's the generosity of his heart!
Like Usain Bolt is taking a year off to give mere mortals a chance to win, so Rob is giving others a chance.
Rob for president!!

beatbybit · 09-06-2010, 08:23 PM

um, any words on the actual pre?..

Justin · 09-06-2010, 10:29 PM

These builds are testing out some changes to the sample timing/rounding code.

Ignore test1.

Test2: this has a very basic change which should (hopefully) not break anything, and fix rounding when splitting/adjusting splits of media that does not have SRC. This is almost certainly going to be in 3.68, unless we hear bad news from anybody.

Test3: this includes test2, as well as one attempt at fixing some rounding bugs when using SRC. It doesn't always work. Bad.

Test4: this includes test2 and parts of test3, to fix those rounding bugs when using (and when not using) SRC, and might actually be correct. If this goes well it might end up in 3.68, if not then maybe someday later.

If you would like to torture test these, I'd appreciate it.

beatbybit · 09-06-2010, 10:34 PM

haha, i meant "any words" from the users - they seemed a bit too lounge-ish..

anyways, thanks a million for these builds, my quick test results here:

http://forum.cockos.com/showpost.php...2&postcount=44

planetnine · 09-06-2010, 11:22 PM

Justin,
I'm not getting any rounding errors at all. Is this because I'm not using Snap-to-Grid, or Beats and Bars as my timeline?

Is this a problem with mixing a musical measure time system with a sample-based time system?

I've not achieved null-fail with scores of duplicates and thousands of splits on my single-sample rate tests, even using automatic fade ins/outs and trying cross-fades.

I'm beginning to think this issue does not affect my methods of working. I've never found this issue directly.

>

dub3000 · 09-06-2010, 11:56 PM

Quote:

Originally Posted by planetnine

Justin,
I'm not getting any rounding errors at all. Is this because I'm not using Snap-to-Grid, or Beats and Bars as my timeline?

Is this a problem with mixing a musical measure time system with a sample-based time system?

I've not achieved null-fail with scores of duplicates and thousands of splits on my single-sample rate tests, even using automatic fade ins/outs and trying cross-fades.

I'm beginning to think this issue does not affect my methods of working. I've never found this issue directly.

>

it may only kick in at particular BPMs, from what people have posted?

rob.manderson · 09-07-2010, 12:23 AM

Quote:

Originally Posted by steadyrev

Rob for president!!

They'd have to change the constitution. I'm only a naturalised American - born in Australia!

Subz · 09-07-2010, 01:19 AM

lol well i came from i host that would phase when sending to a send bus. This is out of my test skills im sorry to say

musicbynumbers · 09-07-2010, 03:21 AM

Will be able to test later as I'm too deep in "deadline" territory to risk giving my "position" away

lol

But seriously thanks for looking at this, I was a "complainer" too and I owe it to you guys to test so will do as soon as.

Subz · 09-07-2010, 03:30 AM

Quote:

Originally Posted by rob.manderson

They'd have to change the constitution. I'm only a naturalised American - born in Australia!

Rob for Governor then

i take it those who where able to see this phase problem in the first place are testing this beta of a beta?

& i'm all up for tightening up unavoidable rounding errors (external midi clock sync anyone?

)

EDIT: i started this post before musicbynumbers posted

planetnine · 09-07-2010, 04:10 AM

Quote:

Originally Posted by dub3000

it may only kick in at particular BPMs, from what people have posted?

I just leave my projects at 120, Dub, as I don't use beats and bars -you think this is why I'm not seeing any problems? I just use REAPER mostly as a playback and editing tape recorder, ocasionally recording on it. I've not been able to get it to glitch with phase.

I take it you line your stuff up to beats or subdivisions -do you have lock to grid on?

Nathan.

>

musicbynumbers · 09-07-2010, 05:55 AM

I think your right that at 120 bpm things are aligned as 120 is tied in mathematically with 60 (seconds) so relates better to samples.

so no matter how many minutes you go into the song at 120 there is no "free-wheeling/drift" but at other tempos the sample amounts and beats grid drift out of sync over time so locking to the beats grid will pull samples out.

That's where I believe the rounding comes in.

I think

Just going to test the new test 4 now.

Quote:

Originally Posted by planetnine

I just leave my projects at 120, Dub, as I don't use beats and bars -you think this is why I'm not seeing any problems? I just use REAPER mostly as a playback and editing tape recorder, ocasionally recording on it. I've not been able to get it to glitch with phase.

I take it you line your stuff up to beats or subdivisions -do you have lock to grid on?

Nathan.

>

Dstruct · 09-07-2010, 06:07 AM

Quote:

Originally Posted by Justin

These builds are testing out some changes to the sample timing/rounding code.

Ignore test1.

Test2: this has a very basic change which should (hopefully) not break anything, and fix rounding when splitting/adjusting splits of media that does not have SRC. This is almost certainly going to be in 3.68, unless we hear bad news from anybody.

Test3: this includes test2, as well as one attempt at fixing some rounding bugs when using SRC. It doesn't always work. Bad.

Test4: this includes test2 and parts of test3, to fix those rounding bugs when using (and when not using) SRC, and might actually be correct. If this goes well it might end up in 3.68, if not then maybe someday later.

If you would like to torture test these, I'd appreciate it.

Will it fix http://forum.cockos.com/project.php?issueid=138 ?

Arnar · 09-07-2010, 06:15 AM

I should be trying this , but I can´t.
Hopefully later tonight or tomorrow.

Justin · 09-07-2010, 08:28 AM

Quote:

Originally Posted by Dstruct

Will it fix http://forum.cockos.com/project.php?issueid=138 ?

It does make it more consistent -- for me, test4 at least, results in a peak of -25dB for all states (i.e. starting at the item start or not).. not sure if that's a win, though. Can you tell me about the underlying media? Was the 44khz version generated from the 48khz version? If so, what process was used?

Dstruct · 09-07-2010, 08:33 AM

Quote:

Originally Posted by Justin

It does make it more consistent -- for me, test4 at least, results in a peak of -25dB for all states (i.e. starting at the item start or not).. not sure if that's a win, though. Can you tell me about the underlying media? Was the 44khz version generated from the 48khz version? If so, what process was used?

Yeah, I think I did the 44.1kHz version from the 48kHz version. Done with REAPER (Extreme 512pt HQ SINC resampling mode + Dither + Noise Shaping).

Mercado_Negro · 09-07-2010, 08:58 AM

Project 48kHz
Media items 44.1kHz

https://stash.reaper.fm/oldsb/996055/test4_1.gif

Files null perfectly when starting at 1.1.00 but if they're both moved to another location (in this case, to 2.1.00) they stop nulling with each other.

Cheers

EDIT: After moving the items to 2.1.00, if I place my edit cursor at 2.2.00, the rest of the splits null again.

Mercado_Negro · 09-07-2010, 09:10 AM

Same project at 48kHz and media items 44.1kHz

https://stash.reaper.fm/oldsb/996069/test4_2.gif

Minimizing/maximizing REAPER doesn't show the green 'info' icon in items. You have to click on one of the items or play the project to see them again.

musicbynumbers · 09-07-2010, 09:55 AM

I tested just now (and had no problem) in a similar way to this bug thread

http://forum.cockos.com/showthread.p...light=rounding

http://forum.cockos.com/project.php?issueid=138

Basically,

1-change tempo to anything but 60bpm/120bpm or simple devisions of that (i used 135)

2-take a few randomly placed short audio drum hits on different tracks,

3-send them to a buss/folder, render the drums to a new track (that's the full length of the drum samples),

4-invert the rendered track's polarity (so that it cancels with the original drum hits.)

Now to test use various ways to copy and paste the original drums and the rendered stem version. I used control & drag to copy, duplicate selection, nudge and copy and paste and EVEN copying the individual hits BUT looping the stemmed polarity inverted item.

ALL worked well in this new pre test 4 version

will try and do more tests later but that's a great result so far!!

This image should explain it

Arnar · 09-07-2010, 10:13 AM

Question .

Do these changes affect the loop length issue?

musicbynumbers · 09-07-2010, 10:26 AM

Also it seems the crash I reported here..

http://forum.cockos.com/showpost.php...5&postcount=22

where by simply "seeking" to another part of the project/video crashed repear is actually being caused somehow by the sample rounding fixes as it doesn't happen on 3.672.

will let you know if the crash happens again on 3.672 but so far it hasn't.

Justin · 09-07-2010, 12:48 PM

Quote:

Originally Posted by musicbynumbers

Also it seems the crash I reported here..

http://forum.cockos.com/showpost.php...5&postcount=22

where by simply "seeking" to another part of the project/video crashed repear is actually being caused somehow by the sample rounding fixes as it doesn't happen on 3.672.

will let you know if the crash happens again on 3.672 but so far it hasn't.

OK, if you can get this to happen with test2, that would be helpful. beatbybit has seen that crash as well, so hopefully we can duplicate/fix.

...this is why this was a "test" version heh.

musicbynumbers · 09-07-2010, 01:08 PM

Will do Justin

might not be for few hours as I'm trying to now export this edited and stretched dv mov file to dv avi without it crashing half way through.

Think I've read that mov files are a nightmare on PC's due to QuickTime shenanigans

timlloyd · 09-07-2010, 01:16 PM

Sorry for the delay Cockos, I'm having a busy mix day. I'll get testing as soon as I can

DuraMorte · 09-07-2010, 01:23 PM

http://forum.cockos.com/showpost.php...&postcount=113

The test in this post was performed at 120 BPM, with 3.66pre7.
Downloading 3.672 now, to see if I can reproduce it.
Edit: Does not reproduce, oddly enough. :/

musicbynumbers · 09-07-2010, 01:31 PM

Quote:

Originally Posted by DuraMorte

http://forum.cockos.com/showpost.php...&postcount=113

The test in this post was performed at 120 BPM, with 3.66pre7.
Downloading 3.672 now, to see if I can reproduce it.

Just in case you don't know. There is a folder on the lol site containing the test Pres. They're not in the root folder.

DuraMorte · 09-07-2010, 01:38 PM

Quote:

Originally Posted by musicbynumbers

Just in case you don't know. There is a folder on the lol site containing the test Pres. They're not in the root folder.

I'm aware. I just hadn't bothered updating in a while, and when I did, the test I'd done failed. Thus, all of my posts on the subject are null and void.
That's all.

musicbynumbers · 09-07-2010, 02:18 PM

cool, hope you didn't think I was saying you were dumb or anything

just that it's easy to miss that little folder on lol

Justin · 09-07-2010, 02:20 PM

Hmm I can't get test4 to crash the way that you mentioned.. anybody else seen that crash?

Justin · 09-07-2010, 02:22 PM

Quote:

Originally Posted by Mercado_Negro

Project 48kHz
Media items 44.1kHz

https://stash.reaper.fm/oldsb/996055/test4_1.gif

Files null perfectly when starting at 1.1.00 but if they're both moved to another location (in this case, to 2.1.00) they stop nulling with each other.

Cheers

EDIT: After moving the items to 2.1.00, if I place my edit cursor at 2.2.00, the rest of the splits null again.

Yeah, the differing project-media samplerates will require snap-to-project-samplerate, which will have to wait for 4.0, to be really useful.

So I guess now try to make sure that

things work right with project samplerate = media samplerate and no stretching
things don't crash any more than 3.672 :/
there are no significantly worse behaviors in test4 (or test2 for that matter) than 3.672, relating to samplerate conversion etc.

anyway...

Mercado_Negro · 09-07-2010, 02:33 PM

Quote:

Originally Posted by Justin

Yeah, the differing project-media samplerates will require snap-to-project-samplerate, which will have to wait for 4.0, to be really useful.

So I guess now try to make sure that

things work right with project samplerate = media samplerate and no stretching
things don't crash any more than 3.672 :/
there are no significantly worse behaviors in test4 (or test2 for that matter) than 3.672, relating to samplerate conversion etc.

anyway...

roger that!

musicbynumbers · 09-07-2010, 02:36 PM

Quote:

Originally Posted by Justin

Hmm I can't get test4 to crash the way that you mentioned.. anybody else seen that crash?

I'm starting to believe it was the evil final cut exported DV .mov file

not had it again so far with other files.

be good for other video guys to test as well with DV .mov's but it hopefully was just this one file.