What is the right reference volume

[warrandeb]

In the past, when I captured some video to be digitized, I always normalized the audio level to 0 dBFS (with a little headroom to avoid clipping). Only recently I began doubting this is not the best thing to do.
The EBU recommendations seem to suggest that an average of -18 dBFS is a good level for broadcasting. However, the dolby digital encoder manual seems to suggest that a speaking voice should be around -31 dBFS.
Can someone explain to me (in layman terms) what the right reference volume level should be?

On a related topic: when I record audio, I always used the maximum bitrate my sound card has (24 bit). However, I believe 16 bit is enough for distribution and archival purposes. When I dither the audio down to 16 bit, and don't normalize the audio to the maximum level, won't the quantization noise level be higher in the non-gained version than the normalized one (and thus the maximum quality achievable with 16 not reached when not amplifying to the max)?

[Cornucopia]

Reference level depends on the content. There is no SINGLE INDUSTRY STANDARD. When doing ENG/Broadcast, I recorded and edited using -12 to -15dBFS. When doing concert audio recording/mastering, I often used -18 to -21 dBFS. The Dolby suggestions are for referencing SPEAKING, and they are also variable depending on other circumstances (for encoding, I've used -24 to -32).

Part of this debate and confusion is over the difference between PEAK and AVERAGE/RMS values. It's necessary to use PEAK values when dealing with most things digital, because that ultimately determines available headroom. But our EARS listen to the RMS, and that's why Dolby chose the values they suggest.

If you are an audiophile, I'd suggest using ~-20 for all your recording/captures. If those get encoded to AC3, etc, and those captures are primarily SPEAKING, I'd suggest readjusting them to the suggested Dolby levels. What you'd really want to do is have a dBSPL meter and map out for the various types of programs what the general level is and what the mastered level has to be to provide (given a standard gain on your amp) a consistent listening level in dBSPL. Yes, it's complicated.

On your other topic: I would always want to archive at the original/highest bitrate, but I agree that distribution using (optimized) 16bit is fine.
You might be misunderstanding how it works: the slightly higher noise level of dither is NOT quantization noise. In fact it helps REMOVE quantization noise. If you start with 24bit recodings (max dyn range of 144dB) with a headroom of -20dB (leaving 124dB for SNR). Then, when you dither and downrez using normalizing to 16bit, your SNR is just less than 96dB, but it SOUNDS like 124dB. When you dither and downrez to 16bit WITHOUT normalizing, your SNR is more like 76dB but it sounds like 104dB. By normalizing while dithering, the amount of empty or "rounding"/noise bits you add to the word length will get lost again in the downrezzing, and will be successfully masked/smoothed out by the dither.

Scott

[warrandeb]

But if there is no real industry standard, how comes every well-mastered dvd/vcd/blu-ray sounds about the same level? (I rarely need to change the volume)

[Cornucopia]

They follow "recommended practices", such as "monitor at full volume (usually ~89dB SPL)" or "reference against both PEAK and RMS levels" or "leave enough headroom for clean peaks" (which means to lower the record/mastering level).
I think also that Dolby's "DIALNORM" architecture has been a success (unevenly sounding volume levels was exactly what it was designed to combat).

However, I disagree with you. While it has gotten much better in recent years for major Hollywood DVD & BD, international releases, semi-pro/corporate/prosumer and VCD releases are NOT similar. And don't even get me started with AudioCD's & Music clips!

Scott

[warrandeb]

That seems fair.
Thanks for the help!

[warrandeb]

Maybe interesting for people who are reading this thread: an interesting read about the subject: http://tech.ebu.ch/loudness

[budwzr]

Part of the problem is if you respect common standards, you just give others another chance to stand out.

Hot Mastering is the result of under-powered playback devices like iPods and phones. But that's where all the growth is. Low levels equals crappy recordings in the minds of 14 year olds.

So the best rule of thumb is What's normal in a specific genre.

[Cornucopia]

Only if you seriously cater to the whims of 14-year-olds.

Sticking to standards helps your material further stand the test of time.

Scott

[budwzr]

I'm glad Dolby Digital has finally caught up with the MP3 spec, whereas playback volume can be reset by the user. But apparently Dolby was too lofty to do it until now.

This is the problem. The old bulls don't want to go off to pasture in a timely manner.

[lordsmurf]

When it comes to video tapes, and computers for capturing, set your OS volume input levels to about 50%.

For audio editing, I actually work visually, because the number sometimes lie. I'd choose a 50-80% occupied spread (levels), when viewing it in a waveform monitor. The % and output decibels vary, because of the dynamic range variations.

When I follow rigid number guides, people complain. That's because treating all audio as equal is quite frankly stupid -- and it's the reason so much audio sounds fubar. But when I use my eyes and ears, there's not a hint of criticism, and that includes some hard-to-please studio types.

Be sure your speakers are reference grade, too. Otherwise you're just guessing at the audio quality.

[budwzr]

Ditto, Smurf.