[RDP] Meters vs. dBFS(?)

Dan Mills dmills at exponent.myzen.co.uk
Sat Mar 28 13:36:33 EDT 2009


On Sat, 2009-03-28 at 17:58 +0100, Qtronix wrote:
> Hi all!
> 
> Sorry if this is a stupid self-explainable thing, but I don't understand
> this problem with the meters showing a whole different dB than the
> normalize-dB function in the import dialog, and others.
> 
> My mp3-files on the Windows machine are imported and processed through
> dBpowerAmp with a normalize-value of -3 dB. When I import into Rivendell
> after that, however, the meters yell through the roof and I have to use a
> normalize value in import of -19 dBFS to get a 'peak' of -3 = (-16 - 3) on
> the meters. Why is this?

Metering, here be a can of wyrms...

SMPTE calls for a nominal RMS reference level of +4dbu = -20dbFS as
being equal to 0 VU on a standard VU meter. Rivendell more or less
follows this standard (I don't think the integration times are quite
correct, but it is close). 

Broadcast metering is almost never done on digital peak meters (at least
not in the studio, engineering may have some use for them), and as I say
digital full scale is either 20db (SMPTE) or 14db (EBU) above nominal
working level (EBU defines it as 18db above 0dbm, but almost everyone
works around +4).

This is done because we are working in a production as opposed to
distribution environment and headroom is needed because we cannot know
in advance exactly how high peaks from microphones and the like will be.
Distribution media of course can be pushed right the way to the limit,
so will normally need to be turned down quite a bit on import.

The differences in required headroom probably tie into the fact that the
type II PPM meters used in broadcast in the EU are semi peak reading,
where a VU is something close to average reading. 

Rivendell has (almost) VU meters with the associated integration time so
they will read zero for a steady state -20dbFS tone and general
experience around here is that normalising a rip to around -13dbFS gives
correct results (the normaliser is peak responding).  

With modern converters there is generally no need to get anywhere near
0dbFS as the noise floor is almost a complete non issue. 


Clear as mud?

Regards, Dan.




More information about the Rivendell-prog mailing list