I have started using the EBU loudness standard (LUFS etc) in my productions lately, so the loudness metering in Cubase comes in handy. However, how reliable is Cubase’s implementation of it?
I am mastering to -1 dBTP, but I get different readings for the true peak level on Cubase (-1.0) and TBProAudio’s dpMeter2 (-0.7). Not a big deal, perhaps, but I would like to know which meter is correct, since true peak ought to be measured, as far as I am aware, to EBU specifications.
I have dpMeter2 inserted on my mixbus in the Control room.
The final insert on the Master Bus (main out) is Cubase’s Brickwall limiter (which follows another limiter, which is good, but not capable of true peak limiting), set to - 1.0 dB and enabled for true peak detection.
From what you wrote, I’d guess the -1.0 you’re seeing is because of the Brick Wall limiter keeping it from going any higher. Maybe if you bypassed it, and “another limiter, which is good, but not capable of true peak limiting”, you’d see -0.7?
The Brickwall limiter is keeping it at -1.0, but ideally the reading should be identical on the dpMeter2 - the LUFS readings are. The other limiter is UAD’s Precision Limiter, which is not capable of true peak limiting. Bypassing the Brickwall limiter and only using the Precision Limiter results in true peak overs around 2 dBTP.
No, not for broadcast. But I want to start “mastering” according to the latest loudness recommendations for streamed music. So I am very interested in knowing how reliable Cubase’s loudness and true peak metering is.
Loading them into Cubase, both meters showed the same dBTP for the loudest file. On the one with the lowest intersample peak value, dpMeter2 was closer than Cubase, although not by much.
Inserting the Brickwall limiter with a setting of -1 dBFS resulted in a -1 dBTP reading on both meters for the loudest file. On the quieter, however, Cubase still showed -1 dBTP, while dpMeter2 indicated -0.8 dBTP.
It seems that dpMeter2 is slightly more accurate than Cubase’s onboard metering. It would be interesting to hear someone from Steinberg comment on this.
Yeah, I see your point, Alexis. What I meant is that dpMeter2 was closer to the calculated true peak for the quieter file when I ran the files without any limiting. Since dpMeter2 indicates a true peak level of -0.8 dBFS for the quieter file after inserting the Brickwall limiter, this suggests that the Brickwall limiter does not stop true peaks at the set threshold (-1 dBFS in this case).
The test files have calculated values for true peak.
“TEST#1. File with … 0 0 0 +1 +1 0 0 0 … impulse. It has calculated inter-sample peak value of +2.098 dB.
TEST#2. File with … 0 0 0 +1 -1 0 0 0 … impulse. It has calculated inter-sample peak value of +0.765 dB.”
dpMeter2 was closer to these calculated inter-sample peak values, i.e. more accurate than Cubase.
I attach four jpg files that should answer your first two questions. Regarding the third one, I haven’t tried it yet.
The first jpg 0799TP shows the two meters’ readings for the 0.799 dBTP file.
The second jpg 2098TP shows the same for the 2.098 dBTP file.
The third and fourth jpgs show the two meters’ readings for the test files after the Brickwall limiter.
I am still interested in a comment from Steinberg regarding the true peak metering in Cubase, if anyone from the company sees this post.
“All posts are seen by Steinberg”, they have said multiple times. Beyond that …?
Thanks for posting all your hard work!
After looking at it all, I wonder if the displayed dBTP is a +/- approximated sort of value, which depends on things like the shape of the waveform and maybe even on one specific algorithm or another.
It’s interesting to note neither the Steinberg or dpMeter2 were were spot on displaying the true peak of the louder test signal (“2.1” dBTP, correctly rounded, wasn’t displayed by either one).
Similarly, on the softer test file, neither displayed the true dBTP (correctly rounded to 0.8 dBTP), with the dpMeter2 displayed a little closer as you pointed out (0.7 dBTP vs. O.6 dBTP).
I wonder if this were done with 10 different test files whether one or the other would consistently be closer, or it just depends on the waveform shape and some other things …
And how weird that though Cubase’s BWL displays that it limits the louder file to the programmed -1.0 dBTP on the dpMeter2 and Cubase’s Control Room, as well as on the softer file as far as Cubase’s Control Room is concerned - but not for dpMeter2 on the softer file (where the dBTP displayed for the -0.8 dBTP file is … -0.8 dBTP!??).
Very interesting work you’ve done and displayed, Magnus_N!
Thanks for your appreciative and encouraging words! The questions you raise can probably only be answered by Steinberg, so a response from them would indeed be welcome. I have been looking for dBTP test files and these were the only ones I found.
Arguably, a variation on the meters of 0.1 dB is no big deal. The behaviour of the Brickwall limiter makes me more curious.