• From: Adelaide - Australia
  • Registered: 2007-05-01
  • Posts: 455

Topic: Sample Rate Conversion and levels in Digicheck

Here is an extract of a post I made on the Apple Logic forum and I thought I would pose the same question here, as I am using DigiCheck on my PC to measure all of this and have come up with some other findings...

"I had a voice over session that was accidently recorded in 44.1kHz and when bouncing out the final broadcast mix I used sample rate conversion to take to 48k for the editor, who just drops the audio mix (LR) into his Final Cut session.... Unfortunately, the peak levels are messed with using sample rate conversion... We're talking about 1 to 1.5 db too hot, which is pretty important when talking broadcast.

It meters just perfectly on the realtime bounce, but if you open the resulting file in QT it plays out hotter.
I have also tried bouncing out at 44.1 and just using the realtime conversion in FCP, but the result is the same... too hot by about 1 to 1.5db.

Should SRC retain the levels correctly, or is this normal behavior?"


... my other findings suggest that 'all' SRC seems to mess with levels and I guess it is reasonable to understand that this would be normal as there are new samples being introduced etc.

Can anyone shed any light on this? I am wondering if there is a way to do SRC without effecting level?

Regards,
Matt

----------------
Matt McKenzie-Smith (UFXII, UFX, Babyface) MacStudioUltra OS13.2.1
----------------

mattrixx's Website

2 Reply by MC

  • MC
  • MC
  • Administrator
  • Offline
  • From: Germany/Thailand
  • Registered: 2007-04-06
  • Posts: 25,582

Re: Sample Rate Conversion and levels in Digicheck

SRC should not change levels so drastically. But your post misses some information:

- Did you check the levels with OVS active?

- What IS your highest peak level actually, before and after?

Regards
Matthias Carstens
RME

MC's Website

  • From: Adelaide - Australia
  • Registered: 2007-05-01
  • Posts: 455

Re: Sample Rate Conversion and levels in Digicheck

Hi Mathias,
Firstly, thanks for getting back so quick.
I would also like to say that this is definitely NOT an RME issue, it appears that the sample rate conversions in both QT (and all software progs that use its framework) and Pro Tools are causing this... I am just using Digicheck to measure things.

So anyway, my 'desired' peak level is -9 dbFS and when played at 44.1kHz out of LogicPro or ProTools, it is brick walled at -9dbFS and displays this correctly both on my output metering in Logic and PTs and also across on my PC running Digicheck.... all good there.


It is when I apply sample rate conversion after or during the bounce (even on the highest quality setting in ProTools). It is also present when I play a 48k wave via QT and it applies the realtime conversion via software.
As I said, it's got nothing to do with my FF800 or Multiface, as the editor is seeing the same thing at his place. It would appear that QT sample rate conversion is a bit screwy... but this would also mean that ProTools is screwy with it too.

Playing back the bounce via QT with the system set to 44.1k the files play 1 to 1.2db hotter.
As far as OVS, I can't select it in Digicheck for some reason... I am using v5 release 1.

Got me stumped! :-)

----------------
Matt McKenzie-Smith (UFXII, UFX, Babyface) MacStudioUltra OS13.2.1
----------------

mattrixx's Website

  • From: Adelaide - Australia
  • Registered: 2007-05-01
  • Posts: 455

Re: Sample Rate Conversion and levels in Digicheck

ooh... just noticed I can't turn on the correlator anymore (in the 4 bar meter) with this version.

----------------
Matt McKenzie-Smith (UFXII, UFX, Babyface) MacStudioUltra OS13.2.1
----------------

mattrixx's Website

  • From: Adelaide - Australia
  • Registered: 2007-05-01
  • Posts: 455

Re: Sample Rate Conversion and levels in Digicheck

Just to add....

When I play via QT a 48k audio file that definitely has a peak limit of -9dbFS and play it out of the FF800 @ 48kHz --> lightpipe ---> Multiface (on PC) and meter it it peaks at -9dbFS.

If I play the same file with the system set to 44.1kHz under the same circumstances (note the Multiface (pc) is clocked to the FF800 (Mac)), using QT player soft conversion in realtime, the peaks get to -8.2 dbFS (which seems different again to my previous results).

As you might imagine, the broadcast folks don't like things out of sorts.

----------------
Matt McKenzie-Smith (UFXII, UFX, Babyface) MacStudioUltra OS13.2.1
----------------

mattrixx's Website

6 Reply by MC

  • MC
  • MC
  • Administrator
  • Offline
  • From: Germany/Thailand
  • Registered: 2007-04-06
  • Posts: 25,582

Re: Sample Rate Conversion and levels in Digicheck

Ouch...hit F2, change from Hardware level to ASIO, then activate OVS, then check again, see the peaks, read the online help about Levels greater than 0 dBFS, then think about what your brickwall limiter in Protools does at -9 (it clips exactly as at 0), then you heureka understand that the SRC process misses a proper input aliasing filter...something like that. Let's see how many points I get :-)

Regards
Matthias Carstens
RME

MC's Website

7 Reply by mattrixx (edited by mattrixx 2008-11-06 07:28:25)

  • From: Adelaide - Australia
  • Registered: 2007-05-01
  • Posts: 455

Re: Sample Rate Conversion and levels in Digicheck

When I hit F2 I get "Die Hilfe konnte nicht aufgerufen werden" ed: Anyway I can get to it if I use the menu... Cool   Now I get the Correlator back too! Yippee!

Ok... Because I am not hitting 0 dbFS, it makes no difference being able to see beyond.

The point you make about the input aliasing filter makes sense, though I had always been led to believe that the conversion was quite good.
All in all, it makes the most sense to do the job at the right sample rate to begin with, alleviating any issues from the outset.

Are there any good SRCs out there that do take all in to consideration?

I think you deserve lots of points!!!  I like to convert my points to beers! :-)  [or points to PINTS]

----------------
Matt McKenzie-Smith (UFXII, UFX, Babyface) MacStudioUltra OS13.2.1
----------------

mattrixx's Website

8 Reply by MC

  • MC
  • MC
  • Administrator
  • Offline
  • From: Germany/Thailand
  • Registered: 2007-04-06
  • Posts: 25,582

Re: Sample Rate Conversion and levels in Digicheck

1. You are not using the release version, and hit the wrong key (F1).

2. I try to explain. Oversampling is always working, for any signal/level if activated. So switching it on you should see a difference IF the signal has been clipped and thus invalid frequencies have been generated within the digital domain. The peak level will rise. You can use the Infinite Hold feature with your material, one time with, one time wthout OVS to check that there is a difference.

3. Most hardware SRCs (of course our ADI-192 DD) will have no problem to handle this material.

4. Now on to free beer for the whole pub: why not doing the sample rate conversion in Protools and THEN applying the brick limiter?

Regards
Matthias Carstens
RME

MC's Website

  • From: Adelaide - Australia
  • Registered: 2007-05-01
  • Posts: 455

Re: Sample Rate Conversion and levels in Digicheck

First of all... THE WHOLE PUB?????? Scary Mathias!!! ;-)

1. You told me (in the earlier message) to hit F2! neutral

2. That totally makes sense and I thank you for the clear and concise explanation.

3. I also suspected that hardware like yours would do this properly, but also assumed wrongly that some of the software available around the place would too...

4. This is what I did in the end anyway. Except,  was working in Logic... so created a new 48k sesh and converted on import and then made sure there were no overs.

Thanks again for your time Matthias... I am one of the guys eagerly awaiting DC for the Mac... btw. I downloaded DC5 from the download page and assumed that was the release version.

Regards,
Matt

----------------
Matt McKenzie-Smith (UFXII, UFX, Babyface) MacStudioUltra OS13.2.1
----------------

mattrixx's Website

10 Reply by david-p (edited by david-p 2008-11-06 13:47:10)

  • From: Wien
  • Registered: 2007-05-01
  • Posts: 99

Re: Sample Rate Conversion and levels in Digicheck

Hi!

Matthias told you to hit F2, and the message you got shows that you actually must have hit F1.

DC5 is betaware.  I suggest that you also get the latest version RC2.

I also suggest that you do your sample rate change BEFORE you apply brickwall limiting.  IN the present case, why not just reduce the level of the sample rate converted signal back to where you want it?

smile smile  david

PS: Why does it claim that I wrote this 12 hours after I actually did??

Classical ambisonic surround recording: UFX, FF400, Alesis HD24, Edirol R-44/88, Samplitude ProX 3.

david-p's Website

11 Reply by MC

  • MC
  • MC
  • Administrator
  • Offline
  • From: Germany/Thailand
  • Registered: 2007-04-06
  • Posts: 25,582

Re: Sample Rate Conversion and levels in Digicheck

I and others have the release version installed. Only F1 will call the help, the help error message bug existed only in the pre-release, and F2 never did and does something else than calling the Level meter Setup dialog.

Regards
Matthias Carstens
RME

MC's Website

12 Reply by MC

  • MC
  • MC
  • Administrator
  • Offline
  • From: Germany/Thailand
  • Registered: 2007-04-06
  • Posts: 25,582

Re: Sample Rate Conversion and levels in Digicheck

Don't know, but RC2 is history...

Regards
Matthias Carstens
RME

MC's Website

13 Reply by mattrixx (edited by mattrixx 2008-11-07 02:55:36)

  • From: Adelaide - Australia
  • Registered: 2007-05-01
  • Posts: 455

Re: Sample Rate Conversion and levels in Digicheck

Mmm, I had downloaded and installed the latest version according to the downloads section... I am still getting the German error message. (note: I ran the un-install part of the installer too). Seems it is not the latest version

About to try again...

[sorted... un-installed and re-installed again]

----------------
Matt McKenzie-Smith (UFXII, UFX, Babyface) MacStudioUltra OS13.2.1
----------------

mattrixx's Website

  • From: Wien
  • Registered: 2007-05-01
  • Posts: 99

Re: Sample Rate Conversion and levels in Digicheck

MC wrote:

Don't know, but RC2 is history...

MC: I thought that RC2 was the same as 5.0, and hadnt even got around to looking for the latest version!  wink

mattrixx:  I recommend that you uninstall DC and then remove all references to Digicheck in the Registry, before installing 5.0. ed::-)

Best wishes,

David

Classical ambisonic surround recording: UFX, FF400, Alesis HD24, Edirol R-44/88, Samplitude ProX 3.

david-p's Website

  • exdc
  • exdc
  • New member
  • Offline
  • Registered: 2008-11-07
  • Posts: 6

Re: Sample Rate Conversion and levels in Digicheck

Um, well-informed compatriots,
exuberant newbie asks, eyes wide:
*is* a version of DC really planned for OS X?
That would be a huge boon for us Mac owners.
Seems a shame to keep a PC in the studio just to run DC,
which is way too much great code for free not to take advantage. of. it.
exdc

MacPro5,1/HDSPeAES/Aurora 16
API/Millenia/Empirical/Great River
200 Delta/Jim Williams
Revox/doodads
  • From: Adelaide - Australia
  • Registered: 2007-05-01
  • Posts: 455

Re: Sample Rate Conversion and levels in Digicheck

Hi exdc,
See here.... http://www.rme-audio.de/forum/viewtopic.php?id=49

Thanks david-p.. I did exactly that and all is well. (I am a bit of a novice when it comes to the mess that appears to be windows).

Cheers

----------------
Matt McKenzie-Smith (UFXII, UFX, Babyface) MacStudioUltra OS13.2.1
----------------

mattrixx's Website

  • From: France
  • Registered: 2007-09-14
  • Posts: 113

Re: Sample Rate Conversion and levels in Digicheck

mattrixx wrote:

Should SRC retain the levels correctly, or is this normal behavior?"
... my other findings suggest that 'all' SRC seems to mess with levels and I guess it is reasonable to understand that this would be normal as there are new samples being introduced etc.
Can anyone shed any light on this? I am wondering if there is a way to do SRC without effecting level?
Regards,
Matt

Have you checked out if the peak increase happened anywhere in the audio, or mainly / only along strong rising edges such as transients?
As you mentioned, an accurate interpolation of missing data could lead in peaks higher than original audio, not a surprise for my understanding
I am using Voxengo R8brain Pro for my personal SRC instances, in which you can easaly set your expected peak to meet any normalized value, and it's an excellent product from a musical and user interface extend

HDSP9632, HDSP9652, ADI-8QS, FFUC, Quadmic II