I know that when "mastering" our own material, it is helpful to hand edit some of the peskier peaks in order to get a higher overall level when limiting.  My question is HOW do you do it?  Do you do it in Sonar, and if so, what is the process for finding those peaks, and then lowering them?
 
Also, if not in Sonar, do you guys do this in Soundforge or Wavelab or something?  Thanks.

Hi Sven,
 
I wish I could do this in Sonar, but it's a little more difficult. Programs like Adobe, iZotope RX and Wavelab do a nice job controlling this.
 
I like to manually edit peaks as my second line of defense in my pre-master. However, the important thing is to eliminate them before they get on the track you are going to master. Finding where they are at the source is super important, this way you can eliminate them there. That's where you should start at all times.
 
For example, in most mixes, your peaks will just about always come from snare drums, cymbal hits, vocal plosives, guitar sounds with high gain that may not be high passed enough and lack of proper compression.
 
Now that last one, lack of proper compression doesn't mean compress to remove peaks. Your best bet is to not allow a peak to be recorded if you can help it. Sometimes we can't. In the event that is the case, automate the peak first THEN add the right amount of compression and you should be ok.
 
There may be times you may feel a limiter is the way to go. As long as you don't hear any artifacts, use it. The biggest offender for me in peaks is always the snare. I love a snare that cracks so hard it sounds like a shotgun blast minus the decay. When you have a snappy snare like that, peaks will definitely enter the picture. A limiter can totally destroy your "crack" element and make it too consistent and watered down. It will control the peak but it will take away from the impact the snare has.
 
So anyway....try eliminating the peaks before you record them (the best recordings are the ones that have the best recording prints right out of the gate) and if you wind up gaining some, try to control them individually from the tracks in your mix BEFORE you attempt to control them from the mastering stage.
 
This way, if you still have a few left when you go to master, it won't take you an hours worth of hand editing to control them all manually. Good luck, hope this helps. :)
 
-Danny

I use an external editing program. (Cool Edit Pro-Adobe Audition) I still feel they do certain operations faster than most DAW programs. I use the hard limit function which leaves the bottom part of the waveform intact. It lowers the peak to any value you determine. It is very good for bringing anything slightly out of context loud back to match other peaks well. 
 
In fact I open up every track from a multitrack session in the editor before a mix. Lots of stuff can be sorted out early. In mastering you can raise the rms of the track by limiting peaks and adding gain. A lot of mixes can handle peaks being limited to say -4 and then have 3 db of gain added. You have just made the mix 3 dB louder without changing the transients much. If at all to be concerned. It tidies things up before it goes into the mastering session. (The louder or healthier things are prior to mastering, the less each mastering stage needs to raise the rms level and hence sound better)
 
It does not distort the tops of the waveform either. It looks flat on the waveform but up close it can still be rounded. You can redraw the shape of the top part of the waveform in Cool Edit. It is nice to tame some wild transients in premastered mix. A mix can benefit from having peaks leveled out in my opinion. It can sound better for it. But not too much is the key.
 
 

be careful that you do not edit all of the life out of a performance.
 
 
recording with high headroom, and allowing those peaks to exist.....
is really where the excitement lives.
 

Good stuff here, with lots to go on. Last question:  Are there commands in those external editors to find the offending peaks, or is it just eyeballing tracks?  And again, when you say "hand edit" peaks, are you simply using a volume envelope?  I understand Jeff's method is an automated thing (hard limiting), but for hand fixing them, what is the actual process?  Thanks again.

had some fun with an audio editor, and damn if that technique doesn't work great.  Really opened things up on a couple songs that were suffering from some unruly snare hits.  Of all the various things I tried, I found that editing those peaks to bring everything a bit more inline, then normalizing, then limiting a tad bring it up to an acceptable volume really sounded best.  I tried without normalizing, and the limiter had to do too much and it sounded worse. 
 
Is normalizing a thing I should avoid?  I know it is a touchy subject.  I've read plenty of heated debates, but today it really seemed to best to get me where I wanted to go....

I don't ever edit peaks on the master. It's tricky, time-consuming and doesn't always come out as a transparent edit.
 
If you're allowing a decent crest factor to begin with, those stray peaks can always be tracked back to a percussive hit. Automate it out in the track, not on a bus, and nobody'll notice the edit. If it's a soft synth, you can often go into the PRV and lower the velocity of the note, which is even more invisible. IOW, find the source and treat it as far back in the chain as you can.
 
But if maximum loudness is your goal, using two compressors in front of the master bus limiter will do the trick. Try the TDR Feedback Compressor, which has two thresholds built in, one for peaks and the other for RMS. The idea is to get the loudness up before hitting the limiter, so that you're not depending entirely on the limiter to do it all.

Yea, you can go in and edit out peaks in the mix before mastering.  You can step back to the mix (as bit sez) to edit peaks there.  As far as I know, you have to manually do it.  It isn't hard (but it is time consuming if you have a lot of peaks).  If you export the mix and plug it back into the open project synced, it is usually easy to figure out which track has the offending peak.  If you have a fairly level output on a track or buss or mix, you can even normalize to get the rms up.
 
As others have said, the earlier you tame the peaks the easier the job is overall.  When recording is optimum, while editing the master is a last resort.  Automation and Track and buss compression are the 2nd steps.  And using little adjustments along the way means you don't have to go extreme w/ any one step to raise your rms.  That way leads to Bats warning - you can get a flat, harsh product with none of the depth and dynamics you started with.
 
@

Thanks all, and Bit:  man, that compressor you mentioned, the TDR Feedback compressor is super cool!  I messed with it for a while on several different sources and it is really nice.  I can't believe it is free.  It is downright invisible.  It this age of emulations and fake tape hiss modeled vintage blah blah it was quite refreshing. 

On my drum tracks I use a combination of compression and limiting on every track.  When I send my tracks out I don't want the drums to be an issue for the person mixing or pre-mastering.
 

And again, when you say "hand edit" peaks, are you simply using a volume envelope?

 
The way I understand it, yes.  Drawing a volume envelope around a peak or two (i.e., hand editing) can be useful but certainly not practical for an entire track (especially if you are talking about a drum track) ... I guess that would be the equivalent of riding the faders.  I do find this useful for getting an occasional single cymbal cash or two that I was a bit zealous on to sit in the drum mix better.
 

If you have a fairly level output on a track or buss or mix, you can even normalize to get the rms up.

 
Maybe I am misunderstanding but normalizing (I thought) raises everything equally bringing the peak up to some pre-defined value.  Raising the RMS is accomplished through compression and limiting and not normalizing. 
 
EDIT - ... not sure if that statement actually reflects what I am trying to convey (I really do know how rms is calculated) ... I am thinking more in terms of how rms is affected by normalizing (keeping the ratio of loud to soft peaks the same) vs compressing (changing the ratio of loud to soft peaks).