Showing posts with label metering. Show all posts
Showing posts with label metering. Show all posts

Monday, 22 September 2008

How do we hear loudness ? Peak vs. Average Level

I have seen lots of confusion over the last few days about the differences between the CD and "Guitar Hero" versions which feature in the YouTube comparison video. Some think that the Guitar Version (GH) is louder, but in fact this is not the case. The two version are level-matched - the average level is equal - and this is how our ears judge loudness.

If the Guitar Hero version feels louder to you, it is because it is more dynamic. It has more variety in the average level, and more contrast between the loud and soft sounds, so it sounds more lively, more energetic and more exciting than the overly crushed CD version. (It doesn't sound as good, in some other ways, though - see "Point Four" in this post.)

How do I know the two are level-matched ? Partly because I've listened to them and can hear they match - balancing levels is a crucial skill for a mastering engineer. But also because measurements back this up, as can be seen in the graphics of the videos.



The two screengrabs here show the same point in the song from the video. look at the level display, highlighted by the red arrows. First look at the Guitar Hero version (GH). The pale blue bit shows the peak level of the music, and the darker blue shows the average level. So we can see that the peak level at this point is almost up to the maximum, whereas the average level is about 10 dB below that. (Click on the image to see a larger version)

Now look at the CD graphics from the same point in the song. The average level is the same - about 10 dB down. But the peak level is barely any higher - about -7 dB or so, instead of right up at the top.


This fact that the average level measurements are equal confirms that the clips are level-matched as I'm claiming. Our ears judge loudness by the average level, NOT the peak level, so they are matched in loudness at the same point in the song.


(Actually it's quite a bit more complicated than that, but this explanation is close enough for the point I want to make in this post. There's more about measuring loudness in this post.)


So why do some people feel the Guitar Hero version is actually louder ?


Try watching the video again, concentrating on the meters. Probably the first thing you'll notice is that both the pale and dark blue meters jump up lower and higher on the GH version, just as the overall graph goes higher and is more spiky. Doesn't this mean it's louder ? No. The dark blue meter is centred around the same point on both versions - roughly 10 dB down from the maximum - so the average level is the same, and over the course of the whole track or album, the ear judges it's level to be the same.


However it feels louder, because it has more highs and lows, more contrast - it's more dynamic. The contrast between the loudest and softest parts of the sound - in this case the drums - is much greater, so it sounds harder and punchier. And crucially because it's overall level hasn't been pushed so hard, it hasn't been over-compressed and crushed, so it doesn't have the ugly distortion of the CD version.


As I've said before, the Guitar Hero version isn't perfect - far from it. In actual fact, it's a bit too dynamic for this style of music - rock requires a "wall of sound" effect, to some degree, and the Guitar Hero version doesn't go far enough in the direction. But the CD is so crushed and so distorted, that out of the two it comes off by far the worst, for me.


As examples of what a great, dynamic rock recording can sound like, check out the Machine Head album "The Blackening" on iTunes or Amazon, or even this BBC live recording of Metallica themselves from last week:


Radio 1 presents Metallica Night: Metallica Live


(This one has a bit too much compressor-pumping in my opinion, but still beats the CD by a long shot.)


And if you find yourself persuaded by any of this and haven't already done so, please check out TurnMeUp.org, sign the petition, Favourite the YouTube video and tell your friends !

Thursday, 15 May 2008

Mastering Software Part 1 - Metering Levels and Loudness

I mentioned the Orban Loudness Meter on the SOS forums, and thought it would be good to start a series of posts here on useful mastering software. So, first up - metering apps. Several of the best editing packages include decent metering, for example Wavelab, and Audacity, but here are a few personal favourites:
  •  Orban Loudness Meter Great (free!) utility includes a PPM meter, VU meter, ITU BS.1770 meter and CBS Technology Center loudness meter - it's fantastic to have access to these less common metering systems, which offer measurements of perceived loudness, which can be more useful than pure RMS. (Plus it runs well under Parallels on my MacBook Pro ;-) 
  • Pinguin Audio Meter Not free but comes in several flavours, the Pro version includes an incredibly detailed spectrum analyser, audio phase correlator (more in a later post), PPM & VU monitoring plus a specrogram. Interestingly, Pinguin have done research to suggest that RMS is the most reliable way of measuring perceived loudness.
  • Speaking of Spectrograms, there's Spectrogram - and Audacity has this capability, too.
  • Finally, the fantastic (Mac-only) AudioLeak - an Leq (Long-term Equivalent Level) Analyzer for audio files and sound input streams. Swiftly discover the average RMS of an audio file, and it's A-weighted counterpart. Just watch out - by default it displays a combined mono value which reads higher than other applications like Wavelab - I usually use the average of the left & right values - they should be almost identical if the L/R balance is right, anyway. Also great for DVD Producers wanting to find the right dialnorm setting for a Dolby Digital encode.
  • Update - here's another one - the SSL X-ISM meter. This is useful because it includes checking for inter-sample peaks. This is well-explained on the page I linked to, but briefly: strange as it may seem, when audio has been heavily limited and is peaking at (or very near to) zero, it is possible that the reconstructed analogue output actually peaks above zero. (In theory, up to 6 dB higher !) In cheaper players, MP3 encoders and other utilities, there is often no headroom allowed for this situation, and unwanted clipping can result. This meter will help you watch for this effect and avoid it. As a rule of thumb, don't allow your audio to peak higher that -0.3 dBFS.
Coming soon - how to choose a mastering studio.