Wrong 500 hz peaking filter gain in fenton/lee implementation?

[iver56]

Here's a figure from the Fenton/Lee 2017 paper. We can see the peak at ~500 hz

However, in the implementation in pyloudnorm, the 500 hz peaking filter seems to have gain set to 0 dB, so it doesn't do anything:

pyloudnorm/pyloudnorm/meter.py

Line 126 in 1fb9146

self._filters['peaking'] = IIRfilter(0.0, 1/np.sqrt(2), 500.0, self.rate, 'peaking')

I'm not sure what exact value it should be set to, but from the plot I can see that it should be somewhere close to 3 dB