Richard made a bunch of good points here. Henry, I definitely think you should make sure to absorb
everything Richard said.
Regarding normalized levels across an archive, I don't even think it's feasible unless one had a
real-time RMS analizer/gain control running as people listened to archive material. Since there may
well be a huge range of program dynamic ranges within a single archive, I think the net-net would be
overall "softer" sound requiring good amplifications for listening stations (something like any of
the new generation of very-loud-capable pro-sumer headphone distribution amps would do the trick).
Something like a college music library, for instance, would the equivilent of a classical, rock and
jazz FM station. I know that some college and NPR FM stations have just that kind of content, but my
experience listening to them is that the compromise processing is disappointing for all music types.
And that's really what we're talking about with an archive-wide "level normalizing system," the
equivilent of FM processing, at least in regard to gain control and dynamic range control.
Regarding postings about levels from Richard and Shai, I'm not about to argue for less-conservative
levels in a 96/24 world, so go with their guidelines. I should probably rethink my own rather "hot"
levels. If I were doing more aggressive DSP or making original wide-dynamics recordings, I'd have to
bring all the gains down a few notches. The hot levels are probably hereditary. One of my father's
favorite sayings was that VU meters are "a guide, not a rule." He would always monitor off the tape
and go by his ears, rarely looking at meters. I always tell people putting up old Fine Recording
tapes on modern solid-state machines to take the "0 VU" tone down -2 so you don't run into headroom
problems. Ampex 350's had a lot of headroom before they'd get fuzzy. By the way, the "hot" levels
were pretty standard practice in US studios back in the day, in order to both eke out a couple more
dB of s/n from the brown-oxide tapes of the day and also to provide an already-hot master to cut
loud singles for AM radio play. In the classical world, it was a way to use maximum headroom to
reduce the hiss floor on the finished record. None of this is necessary in today's 24-bit world, so
I'd say go with the conservative levels and normalize after you've done all the DSP you plan to do.
I stand by the idea that it's a generally good idea to normalize at -1dBfs peak.
-- Tom Fine
----- Original Message -----
From: "Richard L. Hess" <[log in to unmask]>
To: <[log in to unmask]>
Sent: Friday, October 19, 2012 11:57 AM
Subject: Re: [ARSCLIST] Analog to digital dBFS standards
> It really depends on material. With tapes, 0 VU reference levels were set at 185 nWb/m to perhaps
> as high as 500 nWb/m. MRL makes calibration tapes at 200, 250, G320, and 355 nWb/m. G320 refers to
> the German 320 nWb/m standard that is measured differently than the other three. MRL's measurement
> for tape fluxivity is slightly different than Ampex's so the difference between 185 and 200 is not
> what a strict calculation would imply.
> I use 250 nWb/m calibration tapes and I generally set those for -20 dBFS.
> With many converters, one must be aware of what they are looking for for 0 dBFS. For example, the
> RME converters in the "Lo Gain" environment consider +19 dBu as 0dBFS which is 15 dB above +4 dBu.
> Their higher-end Fireface UFX with high-level balanced outputs can go to +24 dBu for 0 dBFS to
> make them compatible with the SMPTE standard of +4 dBu = -20 dBFS.
> However, in a transfer facility one is generally not bound by the same rules a large broadcast
> plant would be bound by, so if the signal on a line is -1 dBu at lineup tone it does not really
> Since I do not even have VU meters on my A80RC machines (there is a very small risk that they add
> a slight amount of distortion), it is easy to set up the machine using the meters in the PC. I
> will admit to adjusting 250 nWb/m to +4 dBu on the Sony APR-5000s, but I will drop that as needed.
> Most of the master tapes are played on the A80RC while the Sonys see more general tapes.
> The Sony APR-16 multitrack (5 audio formats from 4 T half-inch up to 16 T one-inch) is generally
> calibrated to -20 dBFS for 250 nWb/m. There was one series of tapes where I aligned one preset
> about 4 dB lower because the dbx was misbehaving and clipping even with 20 dB of headroom. There
> were no tones on those tapes for calibration.
> Tom's very detailed response is good and I concur with his points, though I tend to record a bit
> lower than he does because I have been surprised by louder passages down a tape and I do not like
> to adjust levels during a transfer.
> I looked at the levels briefly in a symphony concert I recorded last Saturday with a pair of DPA
> omni mics in a reverberant space. I have a preliminary boost of about 3.5 dB above the nominal
> recording level and peaks are coming within 0.5 dB of full scale. The largest peak-to-VU
> difference is on applause where it is almost 20 dB. The music appeared to have about a 10-12 dB
> difference. The Sound Devices 722 recorder I use for this has both VU- and peak-responding
> indication in its LED meters, so it is easy to see how to set the levels.
> I generally peak normalize across an entire file. It seems pretty transparent in Samplitude. I
> know of one person who switched to Sequoia (Samplitude's more-featured big brother) because it
> sounded so much better than what he was previously using even on simple level shifts. I will
> slightly adjust the first and second half of a concert to be closer in level than they were if the
> material changes ("fireworks" in the first half and an early, smaller-ensemble symphony in the
> second, for example).
> I think it is easy to become obsessed with worrying about standards, but once we normalize the
> file, any standardization goes out the window. People listening to music are used to hearing
> normalized CDs for the most part. 16-bit might have been one thing driving that. TV audio was 20
> bits fairly early on which makes it more comfortable to keep everything referenced to -20 dBFS. I
> think this is where much of the cinema world is at, but DialNorm metadata also comes into play
> there. (I did get into a short discussion of this with another industry professional at Neil
> Muncy's memorial service this week--if Neil noticed, I think he would have been happy! It was a
> topic (along with grounding) that I wish I had more of an opportunity to discuss with him.)
> Levels are getting better. The European loudness requirements are complex but seem to be making
> some difference as AGC based on them becomes more available in software.
> In reality, in an archive I doubt that there can be normalized loudness levels across the entire
> archive as loudness is a function of peak and average levels which involve compression. I think
> doing archival transfers we should take the material as it was delivered to us and provide those
> raw files as the preservation copy. I am often asked to increase intelligibility in oral
> histories, and for that I use a variety of manual and automated tools. In order to stay in budget,
> adding compression (rather than adjusting each phrase manually) is often necessary. The access
> copies are different from the raw copies.
> On 2012-10-19 9:51 AM, Henry Borchers wrote:
>> Hello all,
>> I've been hitting a brick wall with my research and I was hoping that with all the experts here,
>> someone could point me in the right direction. Iím currently looking for research done on digital
>> reference levels. I am particularly interested in looking for references related to the amount
>> of headroom standards digital audio archivists and audio digitization technicians use in their
>> digital masters and the digital level dBFS that analog equipment have been calibrated to. Iíve
>> been able to find a lot of references about dBFS standards when it comes to audio for DVD, TV,
>> and cinema (such as SMPTE standards) but not much for the digitization of audio only content. I
>> have been having trouble locating good research regarding this area and I was hoping someone here
>> could point me in the right direction..
>> Henry Borchers
>> Broadcast Media Digitization Librarian
>> University of Maryland
>> B0221D McKeldin Library
>> College Park, MD 20742
>> (301) 405-0725
> Richard L. Hess email: [log in to unmask]
> Aurora, Ontario, Canada (905) 713 6733 1-877-TAPE-FIX
> Quality tape transfers -- even from hard-to-play tapes.