How loud is loud, how to measure it? Is THX calibration bad for your health?

[sarumbear]

Thanks for the link, it had an interesting passage in the end, where he points out that if different providers use different normalization, he listens to the new normalized version at the suitable loudness and then decides if a new mix is needed. This answers my question above.

Regarding the k weighting.
Wikipedia defines LUFS or LKFS as k weighted average with respect to dBFS.
K weighting seems somewhat close to the human hearing. Only dbFS is indeed not linked to any analog loudness, which then would mean that the Netflix LUFS specification means nothing with respect to the actual sound. But I think I got it now because this part is handled by the Netflix spec saying that the mixer should mix at 79 to 82 dB, hereby anchoring an analog level. Pretty neat actually.

That Wiki article needs correcting. May I have the link please?

Here is the link to the standard that defined LU and LUFS:

https://www.itu.int/dms_pubrec/itu-r/rec/bs/R-REC-BS.1770-4-201510-I!!PDF-E.pdf

 

[EEE272]

That Wiki article needs correcting. May I have the link please?

Here is the link to the standard that defined LU and LUFS:

https://www.itu.int/dms_pubrec/itu-r/rec/bs/R-REC-BS.1770-4-201510-I!!PDF-E.pdf

LUFS - Wikipedia

en.m.wikipedia.org

Directly the first line:
"Loudness, K-weighted, relative to full scale (LKFS)"
I checked some other languages and they say the same. But I am really no expert on this.

Edit:
Actually, I just checked and it also says it in your document. It is in the appendix A.

 

[sarumbear]

LUFS - Wikipedia

en.m.wikipedia.org

Directly the first line:
"Loudness, K-weighted, relative to full scale (LKFS)"
I checked some other languages and they say the same. But I am really no expert on this.

Edit:
Actually, I just checked and it also says it in your document. It is in the appendix A.

Note the word: LOUDNESS. Not INSTANTANEOUS level that you can measure with a real time meter like a SPL meter or PPM. You need averaging (integrating) over time to calculate the value. Whereas dBFS is the level at any moment.

Meanwhile, if you investigate more you will see that they changed the unit to LUFS.

 

[EEE272]

Note the word: LOUDNESS. Not INSTANTANEOUS level that you can measure with a real time meter like a SPL meter or PPM. You need averaging (integrating) over time. Whereas dBFS is the level at any moment.

Meanwhile, if you investigate more you will see that they changed the unit to LUFS.

I know it is integrated, that is why I said averaged (over time) but that might have been the confusion.

 

[sarumbear]

I know it is integrated, that is why I said averaged (over time) but that might have been the confusion.

Then why/what are you disagreeing?

 

[EEE272]

Then why/what are you disagreeing?

I am not. I was curious how digital values relate to real world sound out of my loudspeakers. Because as indicated by the topic of this thread, I would like to protect my ears. For that reason, we wanted to find out how loud a mix actually is. We discovered that they are often mixed at very different levels - not always reference. It is thus fine to turn down the volume a bit. Nevertheless, if you turn it down too much, the perceived tonality changes. This could be compensated via loudness correction but it requires knowing what sound level originally came out of the speakers.
With the Netflix specs, this is well defined. Now, we also heard and read in your article, that people that create content are not always super strict. Thus, there might be some natural deviations in tonality.
Which seems to imply that we should also not exaggerate the need for reference loudness. Hence, turning down the volume a bit (or even more with some compensation) seems fine. Further, people that always stick to 0dB on their receivers might even listen louder than intended.

For me, these were a lot of interesting points and I am very happy with the information.

 

[MGG]

Mine is Simply Called Decibel X. Maybe the Pro is the subscription model. Also maybe, they were charging a flat fee and now switched to subscription.

Anyway, Decibel X seem to offer a bit more information than the SPLnFFT, especially on the dosimeter, but I rather prefer the simple and elegant UI of the latter. Not to mention that it is also free!

Decibel X is declared as free app that offers "In-App Purchases". "Free" part seems to be an trail and "In-App Purchases" subscription that follows:

https://apps.apple.com/us/app/decibel-x-db-sound-level-meter/id448155923 (click more next to short app description)

* Choose from different subscription options. Our standard subscription options are:
1-month Subscription
1-year Subscription

* Subscription payments will be charged to your iTunes account at confirmation of your purchase and upon commencement of each renewal term. Subscription with a free trial period will automatically renew to a paid subscription. You can cancel your subscription or free trial in the iTunes settings at least 24-hours before the end of the free trial period. The cancellation will take effect the day after the last day of the current subscription period and you will be downgraded to the free service.

Decibel X PRO seems to be a one-time purchase, but description under Notes describes same same subscription model as non Pro version: https://apps.apple.com/us/app/decibel-x-pro-dba-noise-meter/id1257651611

It's all rather confusing, so I might be wrong.

That may be the in-app purchase price. I see no price, nor paid any.
View attachment 173243

As far as I can see @MC_RME is right SPLnFFT is not free, but price is rather low: https://apps.apple.com/us/app/splnfft-noise-meter/id355396114

 

[sarumbear]

I am not.

My comments started with with reminding you what you said below is wrong as that there is no correlation with dBFS and LKFS or LUFS.

Given that it targets -27 LKFS, which is around 78dB K-weighted, if I understood correctly, we kind of have at least a reference level of the content.

If you have now understood that is the case then ASR did it’s magic.

 

[Descartes]

Free app from EA Lab NIOSH Sound level meter

 

[Soundmixer]

I am still curious about how the normalization affects your creation.

Normalization is an encoding function and is therefore not a factor in the creation process.

Is there a lot of variance in the end between your original loudness when mixing and the final version after normalization?

That is determined by the person doing the encoding. Typically, normalization was set in the default position during encoding. Normalization isn't currently used anymore, it died with the introduction of Atmos.

When they normalize for this level do they apply some loudness compensation?

The answer is no.

Or is that not relevant because they did specify that mixing should take place at 79 to 82, so the loudness difference is minimal?

Mixing on the dubbing stage is monitored at 85db, and the content for the home is mastered and monitored at 77-79db average level. Perceptively, that is a pretty big difference.

 

[audio2920]

Normalization isn't currently used anymore, it died with the introduction of Atmos.

At the replay stage with dial norm, yes.

But the normalisation I was referring to earlier was the loudness spec for VOD, which I think is what @EEE272 was asking about.

Normalised in the sense that all mixes (on a given platform) are gain adjusted so that they have the same integrated loudness. (Maybe with a tolerance of between +/-0.5 to +/-2LU depending on a platform's particular spec).

Often a mix can be made to land in spec if the mixer thinks about this as they go, thus no adjustment needs to made.

I did one recently that I found a bit tricky to predict as it was quite dynamic (by my standards, for a home ent mix) and I landed about 1.5dB above spec. Thus "the file" was turned down 1.5dB to hit average loudness before being sent to the distributor. (In doing so it brought peaks down to -3.5dBTP, which means I could have then opened up the limiter a fraction, but since the whole thing was already on my own limit for how wide a dynamic I want, I just left it. I was happy with what I'd done and didn't need the extra headroom)

 

[EEE272]

My comments started with with reminding you what you said below is wrong as that there is no correlation with dBFS and LKFS or LUFS.

If you have now understood that is the case then ASR did it’s magic.

I will have to call it a day but you might be right that I still don't get it.

Maybe an example could help. I want to make a song, in which I play a 1000 Hz tone, followed by a 100 Hz tone at the perceptual same loudness (e.g. 80 dB for the 1000 Hz tone and 90 dB for 100 Hz).
Now, depending on the setting of my amp, the values that my music program generates can be quite different.
If my amp is calibrated and set to 0, the input would need to be -25 and -15.

Now I export it to the standard of Netflix. This impacts the whole song because it needs to fulfill the -27 lufs.
An amp at a different setting might lead to a different perception of the loudness of the two tones.

How do we make sure it sounds as I intended? I could listen to it after normalization, but then my amp could still be at a random value. Or should I assume that the amp is in a specific setting? (Which I thought was the case, based on the spec that it should be mixed at 79dB)

Sorry, I think I am the same question over and over again...

In any case, I call it a day for now. Thanks for all the input so far!

 

[sarumbear]

I will have to call it a day but you might be right that I still don't get it.

Maybe an example could help. I want to make a song, in which I play a 1000 Hz tone, followed by a 100 Hz tone at the perceptual same loudness (e.g. 80 dB for the 1000 Hz tone and 90 dB for 100 Hz).
Now, depending on the setting of my amp, the values that my music program generates can be quite different.
If my amp is calibrated and set to 0, the input would need to be -25 and -15.

Now I export it to the standard of Netflix. This impacts the whole song because it needs to fulfill the -27 lufs.
An amp at a different setting might lead to a different perception of the loudness of the two tones.

How do we make sure it sounds as I intended? I could listen to it after normalization, but then my amp could still be at a random value. Or should I assume that the amp is in a specific setting? (Which I thought was the case, based on the spec that it should be mixed at 79dB)

Sorry, I think I am the same question over and over again...

In any case, I call it a day for now. Thanks for all the input so far!

I have no comment on how to be sure how it sounds as intended. I was only correcting you that LUFS and dBFS have no fix correlation. You can’t interchange the two values. You can’t make calculations between them.

That’s all.

 

[audio2920]

How do we make sure it sounds as I intended? I could listen to it after normalization, but then my amp could still be at a random value. Or should I assume that the amp is in a specific setting?

Sorry, I think I am the same question over and over again...

In any case, I call it a day for now. Thanks for all the input so far!

I understand now. I guess I've not explained the working to "target loudness" process very well.

I've spent too long on ASR already today too, haha!

It sounds like you actually have understood the "problem" though.

Looking at your example and a rough guess rather than proper maths: if you mixed that at a monitor level of -20dBFS=85dBC and by requirement of averaging -27LUFS for distribution your mix got turned down 6dB, the new monitoring level to achieve the same SPL at the consumer end would be -20=91dB.

But you, as a mixer, have no way of communicating that to the consumer that that's whats happened.

Edit: in the real world, if you mixed at -20dBfs=79dBC but then found you were 1dB under average loudness spec, you'd gain the mix up 1dB, and if you wanted to carry on mixing/tweaking you'd adjust your monitoring level to 78dB to retain the same SPLs you gotten used to. This new monitor level becomes the "reference level used for this mix" but as above, no one else will ever know this....

 

[JRS]

I think someone mentioned this before but most of these standards are only meant to make sure that at the end of your work life, you are still able to follow a conversation properly. Your hearing might still be damaged. Seeing that the table stops at 40 hours, it seems the source might be work related.

In any case, the time doubles per 3 dB, which is reflecting the amount of energy you put into the speakers.
If you go for 74 dBA, you can basically listen the entire week...

EDIT: I think I found it. Your document says it is based on: https://www.itu.int/rec/T-REC-H.870-201808-I/en
Page 47 describes the test of workers exposed to noise. At 83 dBA during the work week, they noticed a 10dB dip at 4KHz compared to the average person without additional noise exposure at the end of their worklife and around a 2-3dB dip after only two years.
A few pages before they describe the difference between music and work noise. While no long term study has been done, they observed the temporary shift and found out that music can be worse than work noise - with the funny exception of classical music and they also point out that the person who mentioned to dislike the music had the worst effect, which they do say might indicate that a subjective opinion can influence the outcome... In my opinion, it seems that this part is not well explored yet.

Great point which I planned to add. The OSHA guidelines are designed to prevent you from being stone deaf at retirement, the WHO standards seem considerably stricter.

 

[JRS]

One of the reasons I don't go to the theaters is they are too loud. I have had to leave concerts and many a club. I like my hearing.

Me too, but I will add the observation that the levels we are talking about seem much different in a theater vs ones home. I get sound pressure is a physical measurement, and so one can say objectively they are the same, but there is something to be said for direct sound vs the mishmash of direct and short delay intensity in an average listening room, which is perceived as much more objectionable.

 

[EEE272]

I understand now. I guess I've not explained the working to "target loudness" process very well.

I've spent too long on ASR already today too, haha!

It sounds like you actually have understood the "problem" though.

Looking at your example and a rough guess rather than proper maths: if you mixed that at a monitor level of -20dBFS=85dBC and by requirement of averaging -27LUFS for distribution your mix got turned down 6dB, the new monitoring level to achieve the same SPL at the consumer end would be -20=91dB.

But you, as a mixer, have no way of communicating that to the consumer that that's whats happened.

Edit: in the real world, if you mixed at -20dBfs=79dBC but then found you were 1dB under average loudness spec, you'd gain the mix up 1dB, and if you wanted to carry on mixing/tweaking you'd adjust your monitoring level to 78dB to retain the same SPLs you gotten used to. This new monitor level becomes the "reference level used for this mix" but as above, no one else will ever know this....

Great! I now got it.
I think I also understood why they chose for -27 lufs @ 79dBC=-20.
It seems pink noise measures +1 with k weighting compared to C. I found this on an audio site.
That means -27 lufs of pink noise is -26dB C weighted pink noise over the duration of the whole movie.
Given that it was mastered at -20dB=79, on a reference calibrated system, it would then play at +6dB or -20dB C weighted pink noise.
Of course, it does not change the fact that it would be played too loud on such a reference system (and you should listen to -10dB), but at least the ratio is the same as compared to cinema.
Of course, as you stated yourself, although that is probably to be avoided, many mixers/engineers will lower their amp volume if the mix is too low and was raised by the normalization. Nevertheless, given that, as you stated as well, it is usually in the ballpark of a few decibels, tonality should not change completely.

 

[sarumbear]

But you, as a mixer, have no way of communicating that to the consumer that that's whats happened.

Edit: in the real world, if you mixed at -20dBfs=79dBC but then found you were 1dB under average loudness spec, you'd gain the mix up 1dB, and if you wanted to carry on mixing/tweaking you'd adjust your monitoring level to 78dB to retain the same SPLs you gotten used to. This new monitor level becomes the "reference level used for this mix" but as above, no one else will ever know this....

Correct. It is not the specification reference that everyone can know and refer to, it is your own reference that only you can know. In which case the term reference may be misused.

 

[Soundmixer]

I get your point. It's odd. I think that's why I'm always a bit preachy to anyone who will listen that they need to abandon the idea of a reference level with 105dB peaks - in the home environment.

It simply doesn't exist and it's not how this stuff is mixed.

If people wanna play stuff that loud, I don't mind, but I promise it's unlikely it was mixed to play that loud, haha.

BINGO!!!!!

 

[Chromatischism]

Are we saying there are no 105 dB peaks in disc movies when played at an average loudness of 80 dB? Because there certainly is in the LFE channel, and I'm pretty sure there are some movies that hit that in the front speaker channels, but I'm not sure.