A bit more on the Sgt. Pepper's Lonely Hearts Club Band 50th Anniversary mixes. Yep - just can't let it go. :)
I decided to run two of the tracks (Sgt. Pepper's Lonely Hearts Club Band and Good Morning Good Morning) through Izotope's Insight meter to see the overall Integrated and Short-term LUFS values, and to see what the overall True Peak levels were. I also turned Apple's Sound Check off for this test so there would be no automatic gain adjustments going on - this would be essentially what was delivered to the streaming services. The screencap at the top of this post is of the opening title track taken about 2 seconds before the track ended. Both songs were played all the way through in real time for this test to let Insight analyze the entire song. The two songs I chose for comparison had identical readings when running through Dynameter - both showed an 8 PLR (Long-term, Macro Dynamics) with an accompanying 5 PSR (Short-term, Micro Dynamics).
See that big red -9.6? That's the overall LUFS value from beginning to end, and although it's not in Death Magnetic territory, it's still way over where we should be (-16 would be ideal here, and if you can see the fine print at center bottom is where my Target value is - and why the right-hand meter bars are jacked to the top). The streaming services are going to reduce the level of Sgt. Pepper's Lonely Hearts Club Band by anywhere from 4 to 6 dB to compensate for its loudness. Argh.
The real issue though is not the overall LUFS value, but the (True) Peak value. Take a look at the meters on the left - the peak levels are dancing around -0.5 to -0.8 dBFS (dB Full Scale) during playback. The potential for artifacts when converting to compressed formats (like AAC and mp3) requires us to use at least -1.0 dBFS as a True Peak limit so we can have that extra dB of headroom for any conversion errors with Intersamples when converting to compressed and streaming formats. The fact that both songs tested basically halve that 'safety zone' means that the probability for conversion errors is effectively doubled, and those higher overall Integrated values will almost certainly contribute to that happening as well.
Is this the reason for some of the anomalies I heard in the last test? Possibly. It still could be some of the other reasons I mentioned, but this is certainly another big question mark...
Moving on, here's a screenshot of the second song tested (Good Morning Good Morning), again about 2 seconds from the ending. Can you spot the problems?
Again, just so much wasted potential in these mixes - they're stuck in the 'traditional' digital mixing/mastering paradigm from the last few decades.
That's it for now - I will be focusing a bit more on this in the future with a comparison of the original 1967 Stereo Mix from the actual vinyl. A big thanks to Mike for suggesting this and offering his copy as a testing case. :)