- Joined
- Mar 17, 2024
- Messages
- 19
- Likes
- 23
I recently ran some tests to establish the correct replay level for various streaming services, to ensure faithfulness to the original mix. This is because I have supplied a number of reference mix rooms as well as a number of QC rooms for the studios themselves and in the latter case, these use high-end domestic gear as well as Dolby CP850 and 950 cinema processors in their replay chains. The tests were done with an all-digital replay chain, including AES/EBU or AES67 outputs from the AVPs so absolute digital level was measured.
One item that became clear is that various services have some normalisation algorithms that affect peak-v.s.-average level. Netflix delivery spec is for -27dBLKFs, and this means that various test tones (e.g. reference Dolby pink noise - the .wav file I have is the one Dolby engineers themselves use, and was recorded straight out of a DMU at -20dBFs) are subject to some peak level reduction due to recorded level and crest factor. In the case of Netflix, this resulted in a reduction in replay level of -4dB, so lining up to this would have meant an SPL after EQ of 75dBc rather than 79dBc on this test tone - any content recorded to the -27dBLKFs spec passes un-changed, so this was a red-herring!
Further to this, various streaming devices have also included (and more recently standardised) upon a -6dB reduction in output level from the embedded HDMI source, and this holds whether it is PCM, Dolby bitstream for 5.1 or Dolby Atmos. This means for correct replay level, a +6dB offset must be applied in the pre-fader gain structure. Source devices were Apple TV 4k, ROKU Ultra 4K, LG OLED eARC, Sony OLED eARC.
Finally, I noted that the internal test tones of various sources were different from the Dolby DMU reference pink noise fed in at 0dB. Not to say there's anything wrong with this - the pink noise sources likely have a different crest factor, but if you are referencing the same standard that is used to line up the dubbing stage and the level of the original mix this is important.
The Lyngdorf MP60 2.1 'Speaker Verify' pink noise was 0.9dB above the Dolby DMU pink noise, and the Trinnov AL32 units were around 1dB above, so lining up your room with the internal generator on each of these can result in a replay level that is actually slightly too low. The QSYS internal pink noise generator (whether pink noise, or cinema pink noise) needs to be set to a gain of -21.3dB to match the reference DMU noise. Loading the DMU pink noise into the file replay on QSYS does result in correct replay level in the digital domain, so comparison is straightforward.
The Dolby CP850/950 and RMU all have a -10dB gain reduction to allow for boost in their internal EQ - this is published by Dolby Labs. Therefore an offset it required to match this to the Home Entertainment devices, and indeed the mix-room Protools output to ensure the gain structure matches for all sources.
I noted that as I always apply SPL trim in the analogue domain right before the power amplifier to maintain unity-gain through the digital domain, I set the LFE gain in the Lyngdorf MP60 2.1 to 0dB (and verified that this passes through at unity) rather than +10dB - I'd add this gain in the amplifier, but noted that the LFE noise generator was +20dB above in-band pink noise of the other channels on 'Speaker Verify' - something to watch out for! I am sure this is likely correct for the analogue XLR outputs - using AES67 output option on the Lyngdorf is not that common in most installations with most installers choosing the internal DACs to feed the amplifiers directly - although I haven't tested this.
Finally, many Home Entertainment mixes are done in Theatrical rooms - these often have a roll-off at around 40-50Hz on the screen channels. If you have high-end or bass managed screen channels that extend an octave or so below this, then the extra energy will increase the overall SPL reading by a few dB. Its therefore worth, post-EQ, setting the centre speaker to have a 40Hz high pass rolloff, set the SPL to 79dBc (for Home Entertainment) or 85dBc (if you are listening to a true theatrical mix as contained on many BluRay discs), and then setting all channels to match this in-band on a real-time analyser rather than focussing on the SPL reading. Once done you can remove the high pass filter (the SPL reading willl increase slightly). Same applies to surrounds, that often have a low frequency rolloff at a higher frequency.
Another area to be aware of is that Theatrical LFE generally rolls of below 30-35Hz, due to the port tuning of common (e.g. JBL 4645C) cinema subwoofers. For common sealed cabinets that extend lower and have shallower rolloff (like the Alcons CRMS-LFE18) I put a 'Theatrical Match' filter in that performs this function in dubbing studios so equipped. It's switchable so you can enjoy the full low frequency extension of the system, but where a mix has been brought in from elsewhere as is common for dubbing and localisation, it avoids a translation issue.
This is by no means an exhaustive test, as I have not tested other AVR/AVPs or the analogue outputs, but thought it would be worth a thread to share the findings of others, as well as confirming the +6dB gain reduction in most commonly-used streaming devices.
In summary - if you want what you hear in the home to match the creative intent of the original mix:-
1. Add +6dB to your streaming source
2. Align using a verfied external reference noise source, or apply the corrections to the noise sources above (e.g. -21.3dB in QSYS)
3. Align SPL to 79dBc (Home Entertainment) for screen channels using a 40Hz high-pass filter (that can later be removed)
4. Align SPL for LFE for +10dB in-band gain on an RTA. If the LFE sounds to 'fat' (you may like it that way) then apply a switchable 30Hz high pass filter.
5. Align surrounds for 0dB in-band gain on an RTA to match your correctly aligned centre channel
Devices included in this test - Protools with Dolby Atmos RMU, MTRX (mk1), Apple TV 4K, Roku Ultra 4K, LG OLED 63C5, Sony OLED 55 AG9 (2019 model), Dolby CP850, Dolby CP950, Trinnov AL32, Lyngdorf MP60 2.1, QSYS Core 510i, Sonifex AVN-DIO19 AES/Dante Converters, Alcons Audio Sentinel 3 amplifiers with CRMS-Series loudspeakers. Some earlier results referenced are from another system comprising JBL M2 L/C/R, JSX1801 sealed-cabinet subwoofers, and JBL AC-Series surrounds with Crown DCi-N amplification, Dolby CP850 for DCI replacy and BSS806DA DSP and Protools with Dual MTRX and Dolby RMU.
I haven't covered EQ best-practice, HF rolloff and matching to ISO2969 or SMPTE222M, this is an academic topic in its own right, and is a function of human perception in both the frequency and time domain of loudspeaker dispersion, room dimensions and acoustics of early reflections, method of measurement (steady state pink noise vs. tone sweep+tracking filter), not to mention higher levels of distortion in compression drivers technology (vs. broader dispersion soft-domes found in the home) being something mixers can be found to EQ against. Mixes made for Home Entertainment in Theatrical Rooms can translate differently as a result.
One for another post after I have finished reading Sound Reproduction: The Acoustics and Psychoacoustics of Loudspeakers, Rooms and Headphones (AES Presents) by Floyd E. Toole, Sean Olive and Todd Welti.
One item that became clear is that various services have some normalisation algorithms that affect peak-v.s.-average level. Netflix delivery spec is for -27dBLKFs, and this means that various test tones (e.g. reference Dolby pink noise - the .wav file I have is the one Dolby engineers themselves use, and was recorded straight out of a DMU at -20dBFs) are subject to some peak level reduction due to recorded level and crest factor. In the case of Netflix, this resulted in a reduction in replay level of -4dB, so lining up to this would have meant an SPL after EQ of 75dBc rather than 79dBc on this test tone - any content recorded to the -27dBLKFs spec passes un-changed, so this was a red-herring!
Further to this, various streaming devices have also included (and more recently standardised) upon a -6dB reduction in output level from the embedded HDMI source, and this holds whether it is PCM, Dolby bitstream for 5.1 or Dolby Atmos. This means for correct replay level, a +6dB offset must be applied in the pre-fader gain structure. Source devices were Apple TV 4k, ROKU Ultra 4K, LG OLED eARC, Sony OLED eARC.
Finally, I noted that the internal test tones of various sources were different from the Dolby DMU reference pink noise fed in at 0dB. Not to say there's anything wrong with this - the pink noise sources likely have a different crest factor, but if you are referencing the same standard that is used to line up the dubbing stage and the level of the original mix this is important.
The Lyngdorf MP60 2.1 'Speaker Verify' pink noise was 0.9dB above the Dolby DMU pink noise, and the Trinnov AL32 units were around 1dB above, so lining up your room with the internal generator on each of these can result in a replay level that is actually slightly too low. The QSYS internal pink noise generator (whether pink noise, or cinema pink noise) needs to be set to a gain of -21.3dB to match the reference DMU noise. Loading the DMU pink noise into the file replay on QSYS does result in correct replay level in the digital domain, so comparison is straightforward.
The Dolby CP850/950 and RMU all have a -10dB gain reduction to allow for boost in their internal EQ - this is published by Dolby Labs. Therefore an offset it required to match this to the Home Entertainment devices, and indeed the mix-room Protools output to ensure the gain structure matches for all sources.
I noted that as I always apply SPL trim in the analogue domain right before the power amplifier to maintain unity-gain through the digital domain, I set the LFE gain in the Lyngdorf MP60 2.1 to 0dB (and verified that this passes through at unity) rather than +10dB - I'd add this gain in the amplifier, but noted that the LFE noise generator was +20dB above in-band pink noise of the other channels on 'Speaker Verify' - something to watch out for! I am sure this is likely correct for the analogue XLR outputs - using AES67 output option on the Lyngdorf is not that common in most installations with most installers choosing the internal DACs to feed the amplifiers directly - although I haven't tested this.
Finally, many Home Entertainment mixes are done in Theatrical rooms - these often have a roll-off at around 40-50Hz on the screen channels. If you have high-end or bass managed screen channels that extend an octave or so below this, then the extra energy will increase the overall SPL reading by a few dB. Its therefore worth, post-EQ, setting the centre speaker to have a 40Hz high pass rolloff, set the SPL to 79dBc (for Home Entertainment) or 85dBc (if you are listening to a true theatrical mix as contained on many BluRay discs), and then setting all channels to match this in-band on a real-time analyser rather than focussing on the SPL reading. Once done you can remove the high pass filter (the SPL reading willl increase slightly). Same applies to surrounds, that often have a low frequency rolloff at a higher frequency.
Another area to be aware of is that Theatrical LFE generally rolls of below 30-35Hz, due to the port tuning of common (e.g. JBL 4645C) cinema subwoofers. For common sealed cabinets that extend lower and have shallower rolloff (like the Alcons CRMS-LFE18) I put a 'Theatrical Match' filter in that performs this function in dubbing studios so equipped. It's switchable so you can enjoy the full low frequency extension of the system, but where a mix has been brought in from elsewhere as is common for dubbing and localisation, it avoids a translation issue.
This is by no means an exhaustive test, as I have not tested other AVR/AVPs or the analogue outputs, but thought it would be worth a thread to share the findings of others, as well as confirming the +6dB gain reduction in most commonly-used streaming devices.
In summary - if you want what you hear in the home to match the creative intent of the original mix:-
1. Add +6dB to your streaming source
2. Align using a verfied external reference noise source, or apply the corrections to the noise sources above (e.g. -21.3dB in QSYS)
3. Align SPL to 79dBc (Home Entertainment) for screen channels using a 40Hz high-pass filter (that can later be removed)
4. Align SPL for LFE for +10dB in-band gain on an RTA. If the LFE sounds to 'fat' (you may like it that way) then apply a switchable 30Hz high pass filter.
5. Align surrounds for 0dB in-band gain on an RTA to match your correctly aligned centre channel
Devices included in this test - Protools with Dolby Atmos RMU, MTRX (mk1), Apple TV 4K, Roku Ultra 4K, LG OLED 63C5, Sony OLED 55 AG9 (2019 model), Dolby CP850, Dolby CP950, Trinnov AL32, Lyngdorf MP60 2.1, QSYS Core 510i, Sonifex AVN-DIO19 AES/Dante Converters, Alcons Audio Sentinel 3 amplifiers with CRMS-Series loudspeakers. Some earlier results referenced are from another system comprising JBL M2 L/C/R, JSX1801 sealed-cabinet subwoofers, and JBL AC-Series surrounds with Crown DCi-N amplification, Dolby CP850 for DCI replacy and BSS806DA DSP and Protools with Dual MTRX and Dolby RMU.
I haven't covered EQ best-practice, HF rolloff and matching to ISO2969 or SMPTE222M, this is an academic topic in its own right, and is a function of human perception in both the frequency and time domain of loudspeaker dispersion, room dimensions and acoustics of early reflections, method of measurement (steady state pink noise vs. tone sweep+tracking filter), not to mention higher levels of distortion in compression drivers technology (vs. broader dispersion soft-domes found in the home) being something mixers can be found to EQ against. Mixes made for Home Entertainment in Theatrical Rooms can translate differently as a result.
One for another post after I have finished reading Sound Reproduction: The Acoustics and Psychoacoustics of Loudspeakers, Rooms and Headphones (AES Presents) by Floyd E. Toole, Sean Olive and Todd Welti.
Last edited:
