[LAD] meterbridge 0.9.2 IEC-scale bug & fix
gheskett at wdtv.com
Wed Sep 26 17:23:48 UTC 2012
On Wednesday 26 September 2012 12:58:35 Fons Adriaensen did opine:
> On Wed, Sep 26, 2012 at 12:08:00PM +0100, Chris Cannam wrote:
> > On 11 June 2012 18:47, Robin Gareus <robin at gareus.org> wrote:
> > > Anyway, some of you who package or copied code from meterbridge may
> > > be interested in this as well.
> > > [...]
> > > // def = (db + 60.0f) * 0.5f + 5.0f; // 5.0 .. 10.0 // bug?
> > > v0.9.2
> > >
> > > def = (db + 60.0f) * 0.5f + 2.5f; // 2.5 .. 7.5 // fix!
> > You're quite right -- I just double-checked the standard and your fix
> > is the right one. I have cargo-culted this code into a couple of other
> > places myself without ever seeing this mistake, so I'll have to fix
> > those.
> There is more to fix.
> The meterbridge website still claims that those meters 'almost' conform
> to some standards while in fact they even don't come close.
> According to the standard, a VU measures the average of the absolute
> value. For a steady input signal around 1 kHz, it must rise to 99% of
> the real value in 300ms and overshoot it by 1 to 1.5% before falling
> back to 100%. The one from meterbridge measures RMS and rises to the
> final value in around 5.3ms, that is more than 50 times too fast (at
> a sample rate of 48 kHz, and worse for higher sample rates).
> According to the standards, a PPM or IEC meter must have a controlled
> rise time of 5 or 10ms (depending on the standard). The one from
> meterbridge indicates peak sample values instead.
You are very correct, Fons. However, as a BC engineer, I have observed
severe clipping of the waveforms because they exceeded the headroom of the
DA's, which in this case was _+28 dbm, while the true vu meter sitting on
that same line was reading -3dbm peaks. For average loudness readings the
legal and pricey vu meter is fairly accurate, but it fails to detect the
transients in todays music, or even in tv's spoken dialog during a soap,
where the human ear as assaulted by the clipping cracklies the 70 year old
analog std meter simply doesn't detect. To that end, the best production
audio boards are also equipt with the much faster LED setups that latch
such a peak condition excursion for at least 100 milliseconds to make it
obvious a peak has been exceeded.
To that end, the operators under my control have always been instructed to
take it down a snudge until the overload LED at +16dbm only blinks at 5
second or more intervals. Listener fatigue can be very apparent in much of
todays audio streams when the old 'vu' meter is peaking at _+4.
And it shows in the ratings books when those instructions are being
ignored, folks might be fans of the program but something they only 'feel'
makes them channel hop looking for a more pleasant experience. When 1
point in the ratings is as important as it is to a facilities cash flow, we
look for every conceivable advantage. Bad audio hurts.
"There are four boxes to be used in defense of liberty:
soap, ballot, jury, and ammo. Please use in that order."
-Ed Howdershelt (Author)
My web page: <http://coyoteden.dyndns-free.com:85/gene> is up!
Gary Hart: living proof that you *can* screw your brains out.
More information about the Linux-audio-dev