On Friday 07 March 2014 17:34:15 Jaromír Mikeš did opine: Humm, that peak is 7db below the average vu meter reading here, and we won't even look at a board that will not handle 20-24db peaks, preferable a solid +30dbm. But now, with everything digital, it seems the producers are all living by the old AM motto, the louder the better, and I can quite easily hear digital clipping that is less than 6 db above the average. That makes it pretty darned obvious, particularly when their excrement production gear does a sign inversion in an add/mix stage to go with the peak. How then do they handle the noise floor when they are running rms 27db below a lot of the rest of the planet? That means their broadcast gear has to have a -83db noise floor to hold the mandated noise floor over here of -60db or better. Best I ever saw was -66 and that was with the video modulation removed on a UHF transmitter. We could do that to get thru a Proof of Performance on UHF because the UHF amplifiers in those days were all klystrons, which if the video was on had so much Incidental Carrier Phase Modulation that the true intercarrier measured noise was usually in the high -40's, to very low -50's if the visual wasn't tuned for best efficiency. That was not often done because it had a very noticeable effect on the monthly power bill, which at best was in the $10,000 USD/month range for most stations. Cheers, Gene -- "There are four boxes to be used in defense of liberty: soap, ballot, jury, and ammo. Please use in that order." -Ed Howdershelt (Author) Genes Web page <http://geneslinuxbox.net:6309/gene> NOTICE: Will pay 100 USD for an HP-4815A defective but complete probe assembly.

On Friday 07 March 2014 18:33:36 Fons Adriaensen did opine: Ahh, gotcha Fons, so they are saying -23db below the digital clip level. That makes perfect sense in that context, thanks. Now if we could just get the production houses to adhere to that. Cheers, Gene -- "There are four boxes to be used in defense of liberty: soap, ballot, jury, and ammo. Please use in that order." -Ed Howdershelt (Author) Genes Web page <http://geneslinuxbox.net:6309/gene> NOTICE: Will pay 100 USD for an HP-4815A defective but complete probe assembly.

On Fri, Mar 07, 2014 at 11:19:28PM +0100, Jaromír Mikeš wrote: I think r128gain is more like a scanner that outputs the statistic to another processor (in this case sox/ffmpeg). The method used here ("Applying the gain") does not include a limiter in the signal chain. The TGDB field means the gain needed so that Loudness is same as reference (-23LUFS). It also outputs TPDB (maximum truepeak). Since you wanted max true peak at -3dB, for cases where TPDB+TGDB < -3dB the TGDB should be safe to be applied directly as is (if you want to be sure, scan one more time after applying gain). For cases where TPDB+TGDB > -3dB you would need limiter to reduce the peaks. How much limiting should be applied cannot be determined solely from TGDB or TPDB, so my suggestion is applying limiter at let's say TPDB-3-TGDB and then run the scan again. If TPDB+TGDB is still over -3dB then run the limiter again with lower threshold. Most materials that have gone through mastering process would likely be in the first case. Does the program material should have Loudness exactly at -23LUFS or the maximum Loudness allowed is at -23LUFS (so program loudness < -23LUFS)? If it's okay for the Loudness to be less than -23LUFS then you could skip the limiter and just reduce the gain.