Gimli Givin'me Grief.

[rayc]

I posted this prob in the MP3 Clinic but then thought it'd make more sense to drop it here as it may be a recording rather than a mixing issue.
The track in qustion - Gimli's Lament - sounds fine on CD & as a wave file. However, after being decoonstructed to become a mere MP3 the latter 1/3 of the track is massively distorted.
I remixed to lower all instruments etc in that section & the problem persists.
I imported the wave file to a wave editor (waverepair) & saw something that suggests a REAL prob. The picture is attached - I've only shown the problem section (highlighted (?) in black & a small section of the normal section - in white.
I've processed the file to remove DC offset & no real change happened.
Any ideas?

 

[SouthSIDE Glen]

A few Qs, Ray,

1. What kind of time frame does the highlighted section cover?

2. Is that the WAV before MP3 encoding or after?

Just shooting in the dark here, but I'd wonder whether there's some harmonic series creeping in at the end there that strict DC offset isn't enough to supress. You might want to watch a quality spectrum analysis while the song is playing and seee if there are any notable changes in the overall response when you get to the problem area, especially if a series of bumps start appearing at regular harmonic intervals in the spectrum. If so, a nice paragraphic EQ, possibly with a harmonic filter setting to make things easier, to notch out that series might help you.

G.

 

[rayc]

Glen,
The black section covers about 2min 20 secs. The image is just of the left of a stereo pair but the right is, to all intents & prposes, the same.
It's the wave prior to compression to MP3.
The last 55 sec section of the song was recorded after the rest, basically to take advantage of the 'cellist I had on hand I decided to record a coda beginning with some overlap & fading it in.
The vocals & flute/recorder were imported from files sent to me as MP3's & converted to wave but they are across the entire song.
Interestingly the distortion doesn't seem to be a problem with the soundclick compressed MP3 I have online (I'll come back and edit with a link - http://www.soundclick.com/util/getplayer.m3u?id=5014534&q=hi ). I only noticed when I tried three different mixes on a portable MP3 player - including a remix yesterday.
WEIRD?

 

[rayc]

Glen,
I paid attention. I ran the track whilst checking with Pro Live 2007 Spec An.
There was massive action in the 25, 37 & 51 zones though decreasing across those bands.
As a quick check I threw a stereo para across the mix & with narrow Q hit those three zone - the wave file result was pretty much the same.
I went back in & put a slope across the three in a shelf EQ in cakewalk & the wave is represented below.
I THINK you were spot on - I'll have to play around with some fine filtering as opposed to my gross attack.
I'll have to take a close look at my set up & mics from those sessions. Something was amiss.
THANKS heaps.

 

[SouthSIDE Glen]

I ran the track whilst checking with Pro Live 2007 Spec An.
There was massive action in the 25, 37 & 51 zones though decreasing across those bands.

Yeah, those numbers fall very close to a spacing of about 12.5Hz with a possible fundamental at ~12.5Hz, which may not show up in your spectrogram if it only goes down to 20.

I'm not sure of the cause; I'd be tempted to try and blame some room modes affecting the cello mic, but that's just another shot in the dark. As for the difference in the MP3s, again this is just another guess, but maybe the Soundclick algorithm includes highpass filtering as part of it's compression, and therefore removes the LF offset that shows in the WAV.

It at least looks (FWTW) from your timelines that you may have tamed things well, but of course the proof is in the sound. If you want to fine tune more, you might want to try a sharp low shelf or high pass to take out the theoretical fundamental at 12.5 (which might otherwise be just high enough to escape a DC offset filter). It's up to you as to if/how much you want to shelve at 25 or whether you want to notch it instead; you might want to leave something there for the cello.

Hope it works out for you. Sometimes we get lucky .

G.

 

[rayc]

Glen,
I went back in & sussed out each instrument with the spec an. I found the vox & recorder the - MP3/converted to wav - imports semed to have the most extra baggage in the range up to 51. That's interessting as I had BIG problems with imported MP3 tracks from the same source on the last project I completed. MP3 conversion or further back? I know the soundcard being used is a stock/generic - the mic is a goodie though can't remember which specifically but it's a sennheiser rectangular prism type from the mid 80's (it's not mine so I don't have it listed).
I removed the previous filter ran a para across the tracks individiually dialing in 13 as the high pass shelf (wouldn't wear 12.5 - built in cakewalk stuff) & that seems to have dealt with the offending stuff. I had thought I'd need to do more up to 51 but as you can see the results, at least graphically, are much better & being narrower, less destructive to the SOUND - but you knew that already.
This seems to have worked and, even if it hasn't, you've walked me through a very, very good tutorial.
Ta teach!