Error Detection

WaveLab
#1

Hi,

I have a similar post to one made some time ago regarding getting a load of errors in some stems and none in others. There doesn’t seem to be much info about what these errors are exactly and what causes them.

The original files were not clipped or overloud, I exported all the files separately from Cubase 6.5. The manual doesn’t go into any depth about these errors or the different parameters you can alter in the detection config and I want to know what caused them. They seem to be particularly bad on drum overheads and Hi Hats (thousands of errors) that I bounced from Toontrack SD2 and vocals.

I have read some responses to questions regarding these errors and the response “if it sounds OK to you, then it doesn’t matter”, seems to be quite common, but ultimately useless. If the functionality is there and there are errors, then I’d like some kind of explanation as to what they are, what the possible causes are and how to make sure they don’t happen again.

Thanks in advance.

#2

Hi,

Are they sounding bad where you see errors ?
Zoom in on wave form and have a look if they clip (overs)
if they do then it can be badly recorded, very compressed, EQ clips
or if samples is missed, this is where most mastering tech is using
this Error detection and correction for ex. vinyl or if there is pops
and clicks in transfer or in recorded media.
try the different Click detection 1,2 or digital

Have a look in WL 6 pdf manual or Help in WL 7

regards S-EH

#3

I thought I’d give an update.

The “errors” highlighted by Wavelab where not audible and there seemed to be nothing wrong with the files. As I mentioned in the OP, the files where not clipped nor where they anywhere near verging on being clipped by either EQ or Dynamic processing, dithering or effects.

However, I processed a couple of copies of the overheads and vocals files through the error correction functionality, on the default settings and Wavelab made a complete mess of the audio when they were “corrected” and I just binned the results. For this reason and in the absecence of any other explanation, I won’t be using this “feature” again.

It seems that my post, along with a few others on the subject, will sink below the surface without any kind of explanation from Steinberg as to what they are, how they got there and what one might do to avoid their re-emergence in the future.

Isn’t there anyone else here that thinks the “error correction” part of Wavelab could do with more of a detailed explanation than the brief mention it gets in the manual? It would be nice, for example if there was a detailed description of what the “threshold” function does and to what “threshold” it is referring to. This would be nice, but it would also be good if the manual went into any depth at all with anything in regard to error correction.

Maybe there is some issue with the way Cubase 6.5 exports tracks that introduces “random” mathematical errors into the exported material? Who knows?

#4

I totally second that. In global audio analysis, there is clip detection and error detection. I would be interested to know the nature of these “errors” if it isn’t clipped! Especially when I hear nothing wrong, and the audio appears to be inphase and all that.

#5

I wouldn’t just assume that, also because the Error Detection feature in WL has always been some kind of a blur, as long as I can remember. You might wanna download the WL6 manual, cause there’s a bit more in there about it.

And maybe PG can shed some more light on the feature in version 7?

#6

Here is something I found and wanted to share:

Instruments with strong percussive attacks (such as drums and piano) can easily be mistakenly analyzed as a clip or a glitch…… Things that cause real glitches are things such as bad edits (edits with dead air or unmatched zero crossings) or digital audio clock errors.

-Matthew Loel T. Hepworth author of Wavelab 7 Power

Wavelab always reports the maximum number of “Possible Glitches” in the analysis. I never hear anything wrong! Maybe this feature could be perfected in the next version?

#7

Thank you Bane, some vey useful stuff there.

#8

I have the same problem - numerous, inaudible, ‘errors’ detected. I then ‘correct’ them, and they become extremely audible. Since I record at -20dB I seldom get overs, and anyway these ‘errors’ often seem to be at zero crossings. I’ve asked Steinberg - we’ll see what they say.

#9

The error analysis tools are very robust. I would say that a certain degree of refinement of the settings is required and experimentation with different audio files (ie different content types) will give the user a better understanding of its usefulness. I use it everyday and more often than not, the error is very real and correctable. Hot point markers are always on target, IME.

#10

BrownswoodArchiving, can you share your optimal settings for Error Detection click correction? For example, type of detection, threshold, default method, options, etc.

And do you have to “define as currently selected audio range” every time you highlight a new section? This seems like an extraneous click. Soundtrack Pro, for example, just assumes what you have highlighted is what you want to run error checks on. I found Soundtrack Pro’s click correction much more intuitive and much better at identifying true clicks and at fixing them. But it could very well be a matter of setting up WL’s error correction to optimal settings. Hence the above question.

#11

I second the concern that folks have about the “false-positive” errors.
In a number of occasions where my file had indeed some clipping, I ended up doing the clean-up using manual spectrum editing instead of relying on Wavelab’s error detection and correction.
I also feel that this is an area where improvements are needed.

#12

Do you mean error detection or error correction? Depends on the source material. I usually leave clipping to default (any overs can be quickly evaluated using the “Get Peak Level” in the Normalize Sound Level dialogue). For Glitches, I tend to lower the threshold and increase the maximum number of error points to report. This will depend entirely on the duration of the file.

And do you have to “define as currently selected audio range” every time you highlight a new section? This seems like an extraneous click. Soundtrack Pro, for example, just assumes what you have highlighted is what you want to run error checks on. I found Soundtrack Pro’s click correction much more intuitive and much better at identifying true clicks and at fixing them. But it could very well be a matter of setting up WL’s error correction to optimal settings. Hence the above question.

No, I check the entire file. I check the entire file with Global Analysis, then do a quick scan using the spectrum view.

#13

Both, since they are part of the same process.

That’s not really an option for me, since I’m editing 2 hour spoken word files. I really only want to correct the errors that I hear. I guess I’m a little spoiled on a 6 year-old program that did it much quickly, cleaner, and more intuitive than WaveLab. I’m getting paid on a ratio, and manual editing takes WAY to long to stay within ratio.

#14

I routinely check 3-4 hour files generated from tape using the method I described above. The hot point checking takes 3-4 minutes and the spectrum scan takes about the same.

#15

Thanks. I’ll try playing around with Global Analysis.

#16

@tony

you probably already know this but marker navigation can be controlled using the numeric keypad.

5 = fwd
4 = bkwd

I only ever navigate through the file in spectrum mode.

Ah yes, that reminds me - WL8 request for pre-roll on the markers! :smiley: