LKFS QC hits with different levels even though True Peak is the same?

[Inglesworth]

Recently, one of our shows that will be airing on Amazon has gotten a couple of QC hits for LKFS being out of range when the file is surely within spec. It is only a couple of episodes, and the client is going to push them through, since we have verified we are all within spec on our end. I just wanted to see if anyone else has encountered something similar and see if there is any insight on this.

We use Izotope Insight meters on our shows to verify LKFS and True Peak levels before shipping out stems. The show in question had an LKFS at -23.7 and a True Peak on the hottest channel at -3.4 (5.1 mix). Our client sends their final product for delivery to Amazon through QC at Deluxe, and they are apparently using something to measure it called Baton, which I had never heard of before. But it is apparently an automated level checking program that they use to measure all of their audio.

The last 2 episodes, they are calling our 5.1 mix with an LKFS that is approximately 2-3dB lower than what we are delivering, but the True Peak values are essentially the same (some channels are off by 0.1dB, but that is negligible, and certainly not associated to a 2-3dB drop). All of our other episodes have been fine, and other shows we mix for the same client are running through QC just fine. It is just the one show, and only some episodes, and the 2.0 (LtRt) is testing just fine in both cases. It is rather confusing.

We have gone as far as requesting the file that the client is delivering to Deluxe, and testing the audio from that file directly, and receiving the same measurements as the stems we send out, again using the Izotope Insight meters. We just wanted to be sure that there wasn't a change happening between our delivery and what goes to QC, and it appears no change is occurring.

We were trying to get on the phone with someone in the Deluxe QC department, but according to our show's post supervisor, they say that their system is automated, and they don't really have any info that isn't on their QC report, and getting on the phone won't provide any new answers.

I am worried that the show is somehow messed up somewhere, and that it will eventually get to air that way. Has anyone experienced anything like this before? Any ideas on if the show is or isn't actually out of spec? Any idea on how to further test it?

[BScout]

Maybe get a trial version of BATON
http://www.interrasystems.com/file-based-qc.php

And see what it is doing by running the files yourself. Might solve the mystery faster.

[its2loud]

If you’re not making the final file being delivered and QC’d then there’s a high probability that whoever is making that delivered file is messing things up on output. This has happened to me before, although with a different streaming company.

Get that file delivered to Amazon and check those levels compared to your delivered mix.

Also, I’m sure you’re measuring with the same algorithm that Deluxe is for the LKFS.

[Norad155]

I had a similar problem... different vendor and different automated QC... but it was measuring ALL channels on the "Master" File at once... so it was I think the 5.1 and LtRt and maybe an M&E... so the peak was right, but Loudness measurement was off. Maybe something worth checking...?

[philper]

Refusal to discuss the issue (by the QC house) is a little suspicious. It kind of says "we never make mistakes, our gear is always perfect, you must be the problem". As we know, that is often not the case....

[seanmccoy]

Not to hijack, but I'm curious about this because I just finished mixing an Amazon Prime series (LoRo only) and couldn't find any references to audio specs for Amazon anywhere (I opted for standard CALM Act). Given how bad the audio was on the first season of this show before they handed it off to me, I'm surprised to hear Amazon has any sort of QC. Were you privy to some guidance from them?

[Leverson]

Talk to somebody over at the QC house and send screenshots of your measurements and what software you are using to measure. Also check to make sure everybody is on the same page whether the measurements are from act-to-act or for whole-program length. Also make sure nobody is accidentally measuring bars and tone too.

I had something like what you are describing happen a little while back when delivering to QC for a Netflix Original movie. The QC house failed our mix with readings that were all really wacky. I always double check my mixes both with Dolby Media Meter and Insight, and I couldn't figure out how they were getting their numbers. I spoke with someone over there and sent over all my measurements and screenshots, and after a short while I heard back that their equipment calibrations were all messed up and now that they fixed it their measurements were reading the same as mine and everything was correct.

So weird things can happen, and I totally understand human error, but it was a frustrating 24 hrs of producers being concerned about failed QC reports that shouldn't have been.