Normalizing woth one very high peak

[econoblackcoffee]

Hello! Just when I thought I had a handle on normalizing, I got thrown a curveball. I taped a band last night in Chicago with my Edirol R-09HR. Got it home and dumped it onto my computer. I have both Sound Forge 5 and Wavelab 3 (yes, I know they're out of date). Anyway, so I EQd the show as one file and got it dialed in where I wanted it. The levels were still too quiet, though, so I was going to normalize it and then burn it. The thing is, there is one spot on the show where there is a weird peak that stands out from the rest (like, if most other peaks are riding around -6db, this is at like -.5db). Consequently, when I normalize, with either peak or RMS, it doesn't boost the levels sufficiently, as it is using this almost-0 peak as the referent. I found a reference to this online, and the person recommended normalizing "around the anomalous peak (i.e. select the file up to the peak and normalize and then select the file after the peak and normalize). This seems like a recipe for disaster to me. I am wondering if anyone can give me advice about how to fix this. I am curious if that Wave Hammer may do the trick, but I've never used it and don't want to compress the whole file, only one tiny peak. I tried using the pencil tool in SF to redraw the wave, but each and every time it was audible and distracting. Any advice would be appreciated.

[AAR.oner]

i usually select random spikes and manually decrease their amplitude down to a dB level similar to the rest of the show...be sure to zoom in when you do this, & listen closely afterward to make sure the change isn't audibly noticeable

then do yer peak normalization -- do NOT use RMS normalization unless you really know what yer doing...it can completely kill the recording's dynamic range

[econoblackcoffee]

Thanks! Can you be a little more specific about how to do this? This seems like what I tried with the pencil tool and it was totally audible. Is there another, easier way to do it that ensures greater accuracy? Thanks again.

[dorrcoq]

Leave it the way it is, use your volume knob during playback

[dcbullet]

I use audacity which as an envelope tool that you can make changes to volume as gradual or as quick as you want - it smooths the change instead of it being abrupt. I've used it to decrease sounds like someone shouting in my mics or a cannon going off.

Do your editing software have something like that?

[econoblackcoffee]

Thanks, DCBullet. I'm not too sure; I'll have to check it out. I honestly have rudimentary skills with Sound Forge and just slightly better skills with Wavelab. But there are a myriad of features with which I'm not yet familiar.

As far as using the volume knob, I've heard that argument before and am well acquainted with the "normalizing is bad" argument and whatnot. And if there were more than one peak hitting at just below 0, I would agree. But it's just the opposite. Why should the integrity of the whole show suffer for one peak? There's no reason to leave the extra headroom in there if there aren't any peaks to utilize it.

[rspencer]

i usually select random spikes and manually decrease their amplitude down to a dB level similar to the rest of the show...be sure to zoom in when you do this, & listen closely afterward to make sure the change isn't audibly noticeable

then do yer peak normalization -- do NOT use RMS normalization unless you really know what yer doing...it can completely kill the recording's dynamic range

I use SF6, so you should be able to do it similarly to the way I do.

Highlight a large section, not including the aberrant peak. Select Process > Normalize > Scan Levels. Make a note of the peak.

Do the same again, only highlighting the extreme peak. Subtract your extreme peak from your "normal" peak. This will be the amount of gain you'll want to apply (e.g., -5.6dB).

Highlight just the extreme peak. Process > Volume > adjust the slider to your desired gain adjustment (-5.6dB). Then hit OK to apply.

That should place the extreme peak back into range with your top peak otherwise. Then just normalize the entire file (peak normalization).

There are other ways (compression, limiting, even using RMS), but this is the easiest, most straight-forward way I know. The other methods often end up trial & error in an attempt to find the optimum settings.