SOLVED! timing issue

[DeeringAmps]

SOLVED! Link to Build 147 is in Noel's post on Page 2 (I'd link it, but its beyond my tech grade...)

2021.04 is dropping clips late!

See the image below, Track 1 was recorded in 2021.01 and started at Beat 1:01 (0 sample)

Track 2 was recorded in 2021.04 and started at Beat 1:01 (0 sample) but dropped at Sample 2166.

The RME Babyface was set to 2048 and YES Sync and Caching was set to Use ASIO Reported Latency

This will NOT do!

tom

I know this belongs in the Feedback thread...

Edited May 4, 2021 by DeeringAmps

[giovannibuchelli]

Same or similar here. There seems to be a problem with reported ASIO latency and with manual offset. I didn't use the feedback threat because I don't know it this is some new problem of the latest version. I've done loopback tests in the past, but not very recently.

https://discuss.cakewalk.com/index.php?/topic/28875-asio-reported-latency-and-manual-offset-not-working-properly/

 

 

Edited May 2, 2021 by giovannibuchelli

[DeeringAmps]

Rolling back to 2021.01 solves the problem. 

NOTE: I have not as yet done a "loop back" test (I will and report back if there is an issue), but I am confidant that all will be well!

t

Edited May 2, 2021 by DeeringAmps
added content

[John Vere]

Hasn’t there alway been a reason we are told never put data at ground zero?  
I think when you hit play there is always a delay before stuff happens so we are advised to start at at least measure 1 and many use 2. 
A long standing feature request is to have a negative measure at the start because of this. It would be used for count in and pick up notes 

And if you need a good tool to perform the loop back try latency monitor 

https://www.resplendence.com/latencymon

Edited May 2, 2021 by John Vere

[DeeringAmps]

1 hour ago, John Vere said:

Hasn’t there alway been a reason we are told never put data at ground zero?  

Good question @John Vere, I don't know; generally I start projects later on the timeline.
But as you can see both 2101.01 tracks "started" at 0, as I would expect.
Mostly I used 1:01 so I could show the "Inspector" and zoom in enough to illustrate the "lag".

However, I "discovered" the problem down stream in a project, so its an "issue".
I can confirm that dropping the clip late was consistent across the timeline.

Interestingly, the RME Babyface was "latency plus 118 samples late"
The RME UFX+ was "latency plus 35 samples late"

Doing a Loop Back test (started at measure 3 BTW) confirms 2021.04 Build 98 is "spot on" in its timing.
Well as "spot on" as I ever see, moving the "Loop Back" 1 sample forward on the time line (meaning 1 sample early) yields the best "null"
with the source track's phase reversed. Pretty typical to what I see with both the RME Babyface and RME UFX; haven't tested the UFX+ as of yet.

t

Here's a screenshot of the loop back in 2021.04 Build 144

Edited May 2, 2021 by DeeringAmps

[Base 57]

16 hours ago, DeeringAmps said:

2021.04 is dropping clips late!

You are right. This is definitely broken.

 

[John Vere]

Glad I didn't update yet. This is a serious issue. I'll update one of my computers and test later tonight. 

[Base 57]

Record Latency Adjustment in Preferences has no effect at all. In a loopback test I have an offset of 374 regardless of the setting.

If I change the Manuel Offset, it changes where the clip boundary starts but the wave is off by 374. Whether checking the Use Reported ASIO Latency box or not the wave is  offset by 374.

[John Vere]

@DeeringAmps and @Base 57  

So I ran the loopback test using my Scarlett 6i6 ASIO mode  with version 2021.01 build 98 and as expected it was right on as seen in track 3.

Up updated to 2021. 04 build 144 and now it is late by what appears to be 800 samples. see track 4 

So defiantly don't update to the new release folks until they fix this. Its a big one. How the heck did they break that?? 

And I was going to update soon. Thanks for pointing this out most of the people using this software would not even notice. 

 

Edited May 2, 2021 by John Vere

[John Vere]

This was interesting as well. I changed buffer size which in past experiences has never had any bearing on loopback results at all. But as you know it does change the reported RTL in SYnc and Cashing settings. I got different results and different settings. At 64 ms buffer it was late only 500 samples now.

But at 560 ms buffer it was late by over 1,000 samples. So it would seem the calculation is totally broken. 

top track is frozen drums, / 2nd track build 98/ 3rd track 256 buffer / 4th track 64 buffer/ 5th track 560 buffer 

[Base 57]

14 minutes ago, John Vere said:

So defiantly don't update to the new release folks until they fix this. Its a big one. How the heck did they break that?? 

And I was going to update soon. Thanks for pointing this out most of the people using this software would not even notice. 

Instead of rolling back I have a workaround. I set a nudge to my offset amount (374). This accurately lines up new recordings quickly.

There are too many things I like about this update to roll back.

[John Vere]

But I think it might be inconsistent. I'm sure they'll fix it. There's seems to be a few threads reporting issues already. 

Lets draw out some staff and see what they say.

@Jesse Jost   @Noel Borthwick

[Noel Borthwick]

Hi guys, thanks for reporting this. I've identified the change that caused this and will fix it. Its caused because the latency offset is actually being applied twice in this case because of some obfuscated code that was doing this unexpectedly!
I'll post a fix for you to test shortly.

[InstrEd]

Noel even working on Sundays!     You need to chill out and play your guitar or the trumpet a little

?

 

[DeeringAmps]

3 hours ago, John Vere said:

At 64 ms buffer it was late only 500 samples now

What I'm seeing with the Babyface is its the buffer plus 118 samples, so at a buffer of 64 I see 182 samples late.

But Noel will have it fixed and it will be like it never happened...

t

[Noel Borthwick]

I've posted a build that has a fix for this and a handful of other issues. Can you all of you affected please try this and report back? Please also verify that punch in recording and loop recording behaves as expected and that audio is aligned.

@DeeringAmps @Base 57

@giovannibuchelli

[giovannibuchelli]

@Noel Borthwick

ASIO latency working fine here.

Thank you sir.

 

[David Baay]

For me, compensation is now correct without punch enabled, but audio is still late by the ASIO buffer size when punch is enabled.

EDIT: When loop-recording with punch enabled, compensation of all takes is correct, and loop recording alone is also good. Only punch alone still has an issue.

Edited May 3, 2021 by David Baay

[giovannibuchelli]

16 minutes ago, David Baay said:

For me, compensation is now correct without punch enabled, but audio is still late by the ASIO buffer size when punch is enabled.

Yep, same here. I tested normal recording and it's ok, but I've just tested punch recording and it's still off.

In my test, the recorded signal is placed ahead of the original, by the same amount of samples of the reported latency.

 

[Base 57]

Sample accurate again with regular recording. The offset is different now. It used to be 92. Now it is 67.

But the Punch is early. I have not calculated the offset yet.

Gotta go attend to the grill