I will export a song from Cubase into WLE9 using no dithering or heavy efx on the stereo mix down except for the Limiter in iZotope Ozone 7. That’s all. And my Cubase Stereo Output shows no trace of clipping. Once in WL I use my set of plugins to master things and then render (but not before checking my import on the Global Analysis tool). Typically I will find something like .0014 DC offset, not a big deal and no clipping. So next I render the song using my plugin chain and use the GA tool again. I find 3 clipping points with the identical DC offset numbers. (Apparently my plugin chain has increased the gain a bit but this is easily addressed with the gain tool, not a big deal.) This small amount of DC offset is too small to remove according to WL. Great. I let the GA place the marks on the clip points and then go to them and reduce the gain in as narrow a spread as I can by -1db. I go back to the ‘play’ start on the song and rescan the song again with the GA and see that the clipping issues are resolved - however - now my DC offset is 46% for one channel and 48% for the other. Really?
Why does this happen? Is this tool accurate regarding the DC offset?
PG, have you looked at this? This behavior doesn’t make any sense to me. Following my description/process path listed above, editing 2ms of clipping from each of three problem areas in a stereo track should not create any DC offset.
Editing clips will change the dc for sure. But to a little amount if the clips are small. The change depends on the file size too, as the dc offset is computed on the whole file.
I am not in the office now and cant check more closely.
If you have a time selection it’ll only analyze the selection. And if you have a tiny selection (say 3-4ms) the DC numbers will indicate much larger (for reasons I certainly don’t know), whether you’ve done a fix or not. You probably still have a small selection at your fix point. It’s the same if you make a tiny selection anywhere in an unmodified file and analyze, you get larger DC numbers.
What do you get if you analyze the whole file after your fixes?
Hm, I’m not sure if your questions are to me Bob, but let me answer PG’s question and your question as well. In my example I am loading and then scanning an entire stereo mix of a 5+ minute song using the GA tool (results: no clipping points and .0014 DC offset). After rendering with my plugins and scanning the entire song again, I find 3 clipping points and the SAME .0014 DC offset. I let the GA tool place markers at the clip points and then edit them minimally with the Gain Reduction tool, -1db. I then remove the markers (or not, I’ve done it both ways), return the play curser to the beginning of the track (which removes the time event window from the clipping points) and scan again (the entire stereo mix) with the GA tool. The clipping points are gone - perfect - but suddenly the DC offset tab is saying that my track is 46% and 48% out. At this point I use the DC offset tool and the result is .00000 for both tracks.
I should point out that the two tracks DO NOT look 46% and 48% out, and this is what bothers me. Does this tool actually do what it is supposed to do, recognize and then remove DC offset?
PG, would it make sense to develop a new DC offset removal tool?
Imagine a DC offset removal tool that analyzed the audio over time - maybe analyzing the entire WAV in 1ms blocks or something. This way, it could analyze and calculate a DC offset “curve” for the entire WAV (DC offset is going to be different depending on exactly where in the WAV it is measured). If the tool then used this “curve” to perform DC offset correction, maybe the correction could be even more useful.
toader, ha, well from what I know about DC offset correction, this would be a micro adjustment that otherwise should be inclusive of the entire wave. In other words it is not likely that I would create a song track that had a small section who’s DC offset was in question. DC offset correction is best served by - from what I understand - a macro tool that addresses the entire wave file that may have been compromised by (more likely) an analog processor than a plugin.
Besides - provided the tool is actually working properly - you can isolate a section and check it of DC offset. And I can see where this might be valid if you have inserted something questionable maybe, but this is not something I have done or see myself doing.
On another note - OMG! - your idea would be an incredibly difficult tool to create I think. Haha, PG would sooner produce WL10.
If you were to adjust the DC offset in 1ms blocks separately, you would be adding a 1kHz square wave to your file. Not very nice!
DC offset is inherently a matter of the whole file, and looking at it in a narrow window is misconceived. If you have a DC offset of zero for the file, but many overs on one side only, you may simply have an assymetrical waveform. In severe cases, this can often be made less troublesome by adding a phase rotation - see this article for an illustrated discussion.
(Although that article is written by a Paul, it is not me!)
I’ve come across asymmetrical waveforms many times before over the years, and I usually just ignore them. As far as my idea, it wouldn’t just be a square wave. More likely, it would produce an irregular wave at an extremely low volume - probably totally inaudible, with the waveshape determined by the DC offset in the WAV… Still, maybe a 100ms interval would be a more appropriate suggestion - I don’t know.
I was just guessing here that maybe specialized DC offset could be a fix… but maybe not though… I am definitely not a DSP engineer. I’m just an audio engineer that likes to think up crazy ideas outside the box sometimes… It occasionally results in good things… and usually results in learning. Anyway, thanks for the link Paul - I’m always glad to be learning something new…
The only way I can reproduce what you’re saying is if there’s still a time selection somewhere in the file when I run the Global Analysis tool after the fixes. Then it’s only analyzing the selection and not the whole file, and the results are wildly different.
If you do the fixes, then click anywhere in the file to make sure any time selection goes away, then run the Global Analysis, the DC results should be roughly, if not exactly what they were before the fixes, according to my tests.
Either that, or save and close the file just after doing the fixes, then reopen it, click anywhere in the file to make sure there is no time selection still (there might be, even after re-open), and then run the Global Analysis.
PG, would it make sense to develop a new DC offset removal tool?
You already have one… use a High Pass Filter. Eg.,use the Steinberg PostFilter plugin.
The DC offset is, technically speaking, a 0 Hz frequency. Therefore, if you use eg. a 20 Hz HPF, then any frequency below it, is removed, including the 0 Hz frequency, aka DC offset.
This technique is useful if the file has DC offset in irregular places, eg. because only a part of it has been processed.
This being said, a filter is always more “invasive” than the traditional DC removal, which computes first the DC offset, then use simple “additions” on samples. That is, the traditional technique keeps 100% of the frequency content, but 0 Hz. However, a good HPF will do a good job.
This being said, I will consider adding the HPF option in the WaveLab DC removal tool.
Bob, well, I assure you that I thought I was clearing the time window before, and I thought I did this by hitting the ‘Home’ key on my keyboard? Apparently not. But now I have done what you suggest and clicked on an unrelated spot on the track BEFORE hitting the Home key. And, as you suggest, my DC offset is .0000 for Left or Right channels. So thank you for pursuing this with me, I appreciate it - yes I do! - and now I have at least a procedure I can follow that seems failsafe. (I’m a bit thoughtful that going to ‘Home’ does not clear the track of the time window actually?? Wouldn’t this be a natural thing to happen?)
All of this said, on the first Global scan of the track I had 3 clipping points. Fixing those and returning to home (after clicking somewhere on the track first) I hit GA again and found another click. I fixed that and repeated the procedure. GA presented another clipping point. I repeated the procedure and finally found no clipping points, all the while seeing no DC offset. Of course finding the 2 additional clipping points (in a different area of the track, btw) confuses me a bit. (This was the same track I edited two days ago and I only found 3 clipping points total. Seems a bit odd but I guess the lesson there is to keep checking the track until the GA gives it a pass.) Eh, I’m easy.
Bob, yes, I agree on the additional passes producing more clipping points.
I have downloaded WL9 Pro and tried it with this, same plugins, same settings. The same thing happens. From what I can tell, if I correct a rendered clip and it exists in a area that has the adjacent wave area almost clipping (but not) - if I reduce the gain on the clipping area then the adjacent area will likely clip. This doesn’t seem right but as a test I reduced my left channel volume (the one typically clipping) and added a bit more compression. We’re talking 0.4 volume drop and a bit more threshold reduction. Which reduced the rendered file clipping to 0 from the gate.
I think this is the way to treat my tracks, it is less work for one thing, but the results are just as good. And running this way, I am not repeating my efforts…
I will also add that the mid/side feature is very overwhelming. I can see it will take me some time to actually understand how to monitor things and get the most out of it. Are you stacking the same track and using the M/S feature on either track to make use of this feature? It puzzles me… Is there a simple working video of this somewhere?
It occurred to me that there are settings in Global Analysis for “minimum time between 2 points” (for each tab I think), and I think the defaults are pretty high, like 10 seconds. That would possibly account for why all clip points were not being detected on first pass. If you set that to the minimum (like 1 second) it will probably catch all or most of the points on the first pass. There’s also a setting for total number of points to report, which you might want to increase.
But if you have a different way of working now anyway, that wouldn’t really be so important anymore.
Regarding M/S, I haven’t really used it for anything, just looked in to how it worked. I’m still trying to figure out how to get it in the Spectrum Editor. Maybe someone else could chime in to lead you to something instructional, because I don’t really know of anything, although the manual’s probably a good source of information.
I haven’t used M-S tracks - I just leave the tracks as standard stereo or mono. Here is how I have been using the new Mid-Side features:
Any effect slot can be switched to M-S. This means there is now an invisible M-S encoder before the slot, and an invisible M-S decoder after the slot. Anything you put in the slot will be processed according to how you set the slot… choose “Mid”, and your effect only affects the mid channel. Choose “Side”, and your effect only affects the side channel. Choose standard mid-side, and the left side of your plugin affects the “Mid”, and right side affects the “Side”.
Engage the M-S view in the wav edit, or spectrum edit, and anything you edit on the top lane affects only the “Mid”, and anything you edit on the bottom lane affects only the “Side”. Disengage the M-S view to return to standard editing…
Thanks Bob. Toader, thank you for this, too. My main problem is in the execution of the decoder… I get what you are saying - but in practice I am not sure about what happens when you render the file. You know? What I mean is that as I listen to the playback using an efx set to M/S, I am not convinced that the render will be affected? I can isolate either part, M or S to listen (this can be a dramatic change, yes), but when I combine things for a listen, the track sounds basically untouched by the efx. Well, this would be with my Wave plugins or my oZone7 efx, I haven’t spend too much time with the stock efx yet. At this point I need to realign my thinking to expect something more subtle maybe?
Well, OK, I guess I need to render a few tracks to see what happens.
I know it’s lame - what I seek - but if I could be assured of the correct path and commit this to practice, then I would feel more confident that this M/S thing was actually working.
Load the Wavelab “Studio EQ” plugin into the first effect slot in the master section. Click the “channel processing” button to the far left of the effect slot. Select MID from the list. Now open the EQ, and turn down the output knob all the way in the EQ gui. If you’re listening to a standard stereo file, you should hear the mids just about disappear, and yet the sides remain - things should sound very out of phase. Now render. It will render exactly as you hear it. You don’t have to do anything special or listen differently… Just standard default monitoring.
I’m totally confused by what you mean here… Are you actually switching the effect slots to M or S or Mid-Side? There is a button to the left of each effect slot called “channel processing”. That’s where you change the effects to process only M, or only S, or Mid/side. I tested Ozone here and channel processing works fine… it also works fine with other Wavelab plugins. FWIW, Ozone has it’s own Mid/side encoding/decoding within it, so there is probably no need to use the Wavelab M/S function with Ozone. The Wavelab M/S funtions are great for plugins that may not have M/S fuctions - like the Wavelab Studio EQ, etc.
It sounds to me like you’re just changing the “audio channel monitoring” and expecting it to affect the plugins somehow? Audio channel monitoring only affects what you hear - it’s just a quick way to listen to your mix in mono - or to hear the S channel (the information that is lost when you convert a mix to mono). You have to change the “channel processing” of each slot if you want to actually make an effect slot function as M or S or Mid-Side.
Yes. Experiment and all should become apparent.
What’s lame? You just seem confused about how it works right now. Go experiment and learn. By the way - no offense intended here if you already know all this, but M-S techniques can sometimes save you if you have to pull off a miracle fix on a master… but M/S is not the end-all-be-all. I RARELY use M/S processing for masters unless absolutely required - but when I need it, it’s an AMAZING tool. 95% of your mastering work will probably never require M-S processing… unless you’re one of those people who likes to widen or narrow things a little… I RARELY ever do that myself, but I know some people like to do it.
Toader, invaluable help here, thanks. I am beginning to see the light I think.
Here’s one thing I just learned but wonder if I have this correct? As you load a stereo track and go to the left of things, below the track itself where it says ‘L’ or ‘R’ (depending on which track) and click on L or R, the track becomes replaced with a Mid track and a Side track. At this point any plugin will become useful - if - you open the M/S dialog and instead of choosing M/S, you choose Mid OR Side. At this point, the plugin performs only on the Mid OR Side track, depending on what you select. Which means that you can open two efx slots, choose a mono (compressor? EQ?) plugin twice and designate one for Mids and one for Sides (at the top of the plugin) and then render the track as a stereo mix of the two independent components from the Master section. Have you tried this? I find it rather fascinating actually, mixing the two parts live. I can’t do this in Cubase or WLE9. Have you done this?
As to mixing in M/S, I agree with you, my non M/S mixes sound tighter and more to the point, even as I A/B the track I am working on. However, I am now thinking that using the M/S ingredient might give me a better drum OH mix. I mean I can do a high pass EQ filter and squash the cymbals a bit without M/S, but now I am thinking I can fly the drums into WL, compress the sides and then bring them back to my mix to blend with back into the kit more quickly. Well, I’m thinking this anyway.
OK, toader, please double back with me on this method I have tripped over. Watching my meter within a Waves plugin and switching between M and S, it seems I have things right.
If you choose “Mid”, or “Side” for an effect slot, that is all that effect will process in that slot. If you choose Mid>Side, the left side or your plugin will process mid, and the right side will process side. You don’t have to engage the MS or LR edit view or anything for this to happen. Remember I told you that the “channel processing” button on each effect slot places an invisible encoder and decoder before and after that effect slot. The effect slot processing has NOTHING to do with the “LR” or “MS” button at the lower left of the screen you’re talking about. That is for doing MS editing.
For MS editing, click on the LR button, and yes, the display changes so you can edit the wav - you can literally use any of the edit functions you want to edit only M or only S just by using this view - VERY COOL. Even more powerful is this: Select the “Spectrum” tab, and then press the LR button so the spectrum ms view is engaged. Now, you can do spectrum editing on just the M, or on just the S… This is AMAZING… Let’s say you’re mastering someone’s track, and they can’t go back to remix, but you’ve got a P-Pop in the vocal. Often, a vocal is mono… so you can engage the spectrum tab, go into M-S mode, and then edit out the P-pop on just the M channel… this leaves the S channel alone so you don’t end up with a huge bass hole in your mix… It allows for more transparent editing. You can do the same thing sometimes with “S” sounds or any other spurious noises that do not belong… MS editing allows for very transparent edits.
Whatever you want I guess. Using M/S techniques on overhead mics has potential to cause phase problems between overheads and your close mics (at least if your overheads were properly aligned to the close mics in the first place). Try it though - if it sounds awesome, go for it! I almost never use M/S stuff in mixes… If I absolutely had to though, I’d probably just load a M/S encoder plugin and a M/S decoder plugin into my inserts, then load the plugins between them that I want to be processed using M/S. (Voxengo had a free ms encoder/decoder plugin available a while back.). If you start using M/S stuff in your mix, you might be cautious to check your mix in mono… If things get too wide, things can disappear in mono - so if your stuff is ever played on a big mono PA in a club, it may not sound right.