Voxengo SPAN 3.0 FFT spectrum analyzer plugin released

[Harry_HH]

But at the final mixing and mastering stage I use also the iZotope Insight for e.g. checking the LUFS. Burocracy or not, the fact is that the LUFS
is standard which is widely used.

OK, how much in your practice LUFS differs from RMS reading?

The core is not how much the LUFS vs. RMS differ but I need to check the LUFS value. Whether the target is -23 LUFS (Integrated) for the broadcast or around -11 LUFS in some indie production, or something between, I wan't to control it, over the time, too (the Insight graph).
As said, this is more my own work flow routine, not any "best global practice".

Thank you for the great plugin, anyway!

BTW I found this presentation for commenting.
http://productionadvice.co.uk/lufs-dbfs-rms/

[.jon]

OK, how much in your practice LUFS differs from RMS reading?

Loudness meters apply human perception-based filtering on RMS data, which means that the difference depends on the material. In other words, they don't even try to express the same thing, even if numerical results may be similar in some (rare) situations. Raw voltage average vs a standard approximation of how we humans hear sound.

For most material in my tests, integrated loudness (LUFS) seems to be between 3-4 unit higher than integrated RMS (which makes sense since our hearing is definitely biased), but there's a lot of variation depending on the actual signal level, frequency content and even duration of sounds- from very similar results to up 8 units difference. Try with different material and see for yourself!

That said, there's plenty of full-fledged loudness meters available even for free, I don't see it necessary to add their features to a frequency analyzer. They all have useful features that would just bloat SPAN significantly if taken to a level where it could replace them.

[Compyfox]

Is there going to be EBU R128 / LUFS metering in SPAN? Seeing as the old RMS-based metering is becoming obsolete.

Sorry, RMS is not "becoming obsolete".

RMS and VU, along with digital max, is still essential for gain staging and plain mixing.

The K-System (v1 or v2) or the ITU-R BS.1770-x specs are not the holy grail. Especially not for pure mixing purposes (mastering is a different topic).

Isn't LUFS is basically the same as DBFS except for some additional filtering applied during loudness estimation? I'm not really convinced it's necessary at all, don't want to promote something I'm not really into. I may be wrong, but EBU R128 sounds like a thing pushed by bureaucrats to somehow reduce the problem of loudness variation in program material. I personally think it will make the overcompression problem even more severe: if somebody limits your loudness you'll want to overcome this by reducing dynamics, using more piercing sounds, ways to cheat the LUFS scheme.

IIRC it's 3 values (integrated/momentary/shorterm) + some gating that take into account of sudden drop for integrated value.

Before even more nonsense and "heck I don't know" spreads (which annoys me to no end).

LUFS measurement is based upon the k-weighting filter (hence the small k post the abbreviated Short Terms Loudness aka SLk and Momentary Loudness aka MLk), and uses ballistics of 3s (Short Term) and 400ms (Momentary). The Integrated Value is an "average" value calculated every once in a while form all readouts taken over a specific course of time (it's the evolution of the Leq(m) measurement after all, while L stands for loudness, eq for "filtered" and (m) for "time").

It's defined in the ITU-R BS.1770-x specs

True Peak should be fairly known by now. It's (very basic spoken) an oversampled digital meter.


Compared to "RMS Mastering" (or as some call "peak mastering", the idea behind the ITU-R BS.1770-x specs is to go with a more reasonable average signal strength (what we call "loudness") and try to hover around a specific value (e.g. -16LUFS SLk, short term, k-weighted). Even if the signal hovers +/-3LU (relative) from the -16LUFS (absolute, so -13LUFS max!), the maximum signal strength in dBTP barely exceeds -1dBTP. Therefore you don't need drastic compression/limiting.

See it like a modern day CD with a loudness of -5dB RMS (with a max peak of theoretically up to +9dBFS, if positive full scale values would exist! that's a lot of work for the limiter) vs a vinyl that has an upper limit of roughly -14dB RMS (else - the grooves would be too wide).

The ITU specs force mastering engineers to work more "dynamic based"; or rather force them to "retain the transients" rather than destroy/remove them in favor for "impact" and consistent streams in loud environments. This is not handled with "loudness normalization" (which can be done automatically during playback).


This has nothing to do with bureaucrats. "Signal Abuse" (due to peak limiting) is not encouraged anymore.

+1 it would be great to have LUFS metering as I favour it over RMS readouts now. I have a target range (-10 to -13 LUFS integrated) that I aim for with my masters. The best way for me though is to still use RX5 standalone application and it analyses the whole track or section within seconds.

Plenty of fish in the sea.
In various flavors as well. Offline analysis, realtime analysis, with or without loudness normalization, etc.

The EBU 128 standard however is pretty useless to anyone outside the TV/film industry. -23 LUFS is restrictively quiet. Though music I/we make does end up in film and TV... they can pull down the volume and normalise it themselves, and they do. EBU128 is more for normalising volumes between adverts/shows etc without annoying the viewers.

EBU 128 is but a preset of the ITU-R BS.1770-x specifications.

There are SEVERAL presets at this point.
EBU R-128 (-23LUFS)
ATSC-85 (-24LUFS)
and now as new "Broadcast"/streaming recommendation: AESTD1005.1.16-09
which funny enough, is also a "recommended value" for "Mastered for iTunes" (-16LUFS)
Spotify (non reatime loudness normalization) and Youtube (realtime normalization) still hover around -12LUFS to -14LUFS

Personally, I do recommend a good compromise for Vinyl/CD/Blu-Ray Audio/Tape/MP3/Radio (studio, not transmission plant) of -16LUFS SLk to -14LUFS SLk on average.

But at the final mixing and mastering stage I use also the iZotope Insight for e.g. checking the LUFS. Burocracy or not, the fact is that the LUFS
is standard which is widely used.

OK, how much in your practice LUFS differs from RMS reading?

Depends on the program material, the ballistics and the weighting filter...

RMS/AES-17 (300ms) might differ up to 6dB compared to MLk (LUFS, ITU-R BS.1770-x or "k-weighting")

Even the K-System (v1) meter (or in other words, the Dorrough 40A with a shifted reference point) might differ +/-several dB compared to MLk (LUFS, ITU-R BS.1770-x or "k-weighting")

That said, there's plenty of full-fledged loudness meters available even for free, I don't see it necessary to add their features to a frequency analyzer. They all have useful features that would just bloat SPAN significantly if taken to a level where it could replace them.

Valid argumentation.

Also, plenty of multi-measurement suites out on the market already.




More in my KVR Marks

[camsr]

Thanks for another update!

Why did you remove the groups selection from the edit menu? I think this new way will be slower to operate.

You can enable this bottom group selector, it is switchable via the settings window.

I did enable it, and I noticed the groups are now missing from the spectrum edit window (the gear icon)

[Aleksey Vaneev]

Thanks for another update!

Why did you remove the groups selection from the edit menu? I think this new way will be slower to operate.

You can enable this bottom group selector, it is switchable via the settings window.

I did enable it, and I noticed the groups are now missing from the spectrum edit window (the gear icon)

You can either have the group selector on bottom, or in the line of buttons on top, they can't be displayed together.

[dreamvoid]

As far as I understand it, it is all about bringing (over)compressed and dynamic material to an equal loudness situation. It is not about limiting loudness but LESS limiting and compression of dynamic audio. Overcompressed material gets 'punished'. The higher you compress the program material the lower the perceived loudness in teh end with a R128 regulated playback.

I do not see how less dynamic material will be penalized. LUFS measures loudness, it does not measure dynamic range. In the end it's no different from physical limitation imposed by PCM encoding, just at a different reference level. It can be worse in fact as you can cheat LUFS by using short but loud sonic bursts, can be more annoying than overcompression.

It gets 'punished' because it is automatically turned down in volume if measured loudness is to high. Dynamic material is turned up in comparison. It is NOT about reference level (you can even set an offset flag in metadata) but about a reference model of perceived loudness in human hearing. There are several reseaches and methods molded together. As far as I understand it, the weighting between shortterm (your "bursts") and program (integrated) metering works like this: short burst would have no impact really on regulated playback. Would these occur often they would be taken into the measurments more and more, resulting into an decrease of playback level. R128 is now a requirement for certain work, no matter if we agree or not, or if you think it is useful. Therefore I ask you to please integrate it at least into Span+. I regard the quality of you plug ins as very high and have bought several licenses over the years and use them with great joy and success. Please stay with the professionals. Thank you.

[Compyfox]

- MLk (400ms) is for "bursts", not SLk (3s). Integrated is an overview of the whole stream over a given measured time
- nothing is being pushed or reduced, unless the playback engine has either settings for that (metadata - and that are currently not really existing, "Replay Gain" is neglectable as this doesn't adhere to any standards)
- once more, EBU R-128 is just a preset of the ITU-R BS.1770-x specs

[dreamvoid]

As far as I understand it, it is all about bringing (over)compressed and dynamic material to an equal loudness situation. It is not about limiting loudness but LESS limiting and compression of dynamic audio. Overcompressed material gets 'punished'. The higher you compress the program material the lower the perceived loudness in teh end with a R128 regulated playback.

I do not see how less dynamic material will be penalized. LUFS measures loudness, it does not measure dynamic range. In the end it's no different from physical limitation imposed by PCM encoding, just at a different reference level. It can be worse in fact as you can cheat LUFS by using short but loud sonic bursts, can be more annoying than overcompression.

Concerning overcompressed material getting 'punished'. The EBU Tech 3343 paper even calls it like that:
"Having one single number (−23 LUFS) has great strength in spreading the loudness-levelling concept, as it is easy to understand and act upon. And the active normalisation of the source in a way ‘punishes’ over compressed signals and thus automatically encourages more dynamic and creative ways to make an impact."
You get all explained in this document in detail. Also what happens with your "bursts".
https://tech.ebu.ch/docs/tech/tech3343.pdf

[Compyfox]

This is written in a misleading form.

What the writers of this tech paper wanted to address was, that the "overcompressed" material is being "punished" on the long run.

Example:
Have a track that has an average loudness (average signal strength) of -14LUFS with a maximum peak of -1dBTP. That is a PLR (peak to loudness ratio) of 13dB. The old "Dynamic Range Meter" would have called this "dynamic range" (in reality: crest factor).

Now take a modern production like Metallica (Death Magnetic Era). The track has an average loudness of -5dB RMS (or let's say about -6LUFS) and a maximum peak of +0,73 (the signal is clipping). The PLR is therefore merely(!) 6,7dB. That's half of what the more dynamic track offers. The transients are squashed, the track doesn't "breathe". It's a constant square wave so to speak.


Now what the "punishment" is in this case means, is that the "less dynamic track" won't have enough impact as the "more dynamic one" if you bring both of these tracks down to the same target level of (let's say) -14LUFS. What would happen now:
- Dynamic Track: -14LUFS, -1dBTP, 13dB PLR
- Modern Master: -14LUFS, -7,3dBTP, 6,7dB PLR

The modern track would have the "same loudness", but it wouldn't sound "as loud" due to the squashed dynamics, and missing transients.


So what this paragraph actually means with "...and thus automatically encourages more dynamic and creative ways to make an impact.", is that mastering engineers are encouraged to not squash their material to sh*t, retain the transients, and focus more on "subtle signal limiting" than "being the loudness around the block to be heard".

Nobody wants to listen to a 2012 CD remaster of the Beatles compared to a mid 90ies CD release.


10 year old Youtube video, yet still explaining the whole concept the best:




Regarding "bursts". The "bursts" are measured in MLk (400ms) rather than SLk (3s) as integration time. These "bursts" (which are shorter than 5s usually) don't feel as much "in your face" if done in a suitable way. In case of program stream (TV, etc), it's explosions/impacts, etc. And these bursts can up to +9dB higher than your average signal, for a short time.

Take a listen to "Sting - Englishman in New York" or "Phil Collins - in the Air Tonight". The so called "burst" or "impact" there is a breakdown of about 10s which can range up to +9LU on the MLk meter, while the SLk meter goes only up to +4LU. Therefore the track is still within specs.



Commercial "bursts" are a completely different topic. And you have to understand that this tech paper origin is from early 2010s... so a lot has happened since then, including "new broadcast recommendations".

So there is a "bigger picture" than just one white paper and "one standard" that you folks try to grasp on as tight as possible and try to force as an inclusion into SPAN.

[Aleksey Vaneev]

Concerning overcompressed material getting 'punished'. The EBU Tech 3343 paper even calls it like that:
"Having one single number (−23 LUFS) has great strength in spreading the loudness-levelling concept, as it is easy to understand and act upon. And the active normalisation of the source in a way ‘punishes’ over compressed signals and thus automatically encourages more dynamic and creative ways to make an impact."

That's sound in theory, but I doubt that in practice this will lead to more dynamic material. As far as I understand psycho-acoustics, overcompressed sound is perceived as louder in comparison to dynamic sound, even if the average loudness is similar. If you can point me to a research that states otherwise, it would be interesting to know.

[.jon]

That's sound in theory, but I doubt that in practice this will lead to more dynamic material. As far as I understand psycho-acoustics, overcompressed sound is perceived as louder in comparison to dynamic sound, even if the average loudness is similar. If you can point me to a research that states otherwise, it would be interesting to know.

Any research, Aleksey. Again I'm not sure whether you are just using incorrect terminology, or do not understand what you're talking about.

"Loudness" is perceptual. Two signals may have similar average intensity, but different loudness. This is why loudness meters exist. If your loudness meter gives similar results for two signals which you perceive wildly different in loudness, it just means the meter is broken or setup wrong.

[Compyfox]

Does anyone even read what I write? Guess not.

[Aleksey Vaneev]

That's sound in theory, but I doubt that in practice this will lead to more dynamic material. As far as I understand psycho-acoustics, overcompressed sound is perceived as louder in comparison to dynamic sound, even if the average loudness is similar. If you can point me to a research that states otherwise, it would be interesting to know.

Any research, Aleksey. Again I'm not sure whether you are just using incorrect terminology, or do not understand what you're talking about.

"Loudness" is perceptual. Two signals may have similar average intensity, but different loudness. This is why loudness meters exist. If your loudness meter gives similar results for two signals which you perceive wildly different in loudness, it just means the meter is broken or setup wrong.

All loudness meters are wrong, because they either measure short-time or long-time loudness. For short-time loudness overcompression will read lower than dynamic material. For long-time loudness overcompression may read the same, but perceptually it may sound louder because long-time estimation does not consider short-time loudness.

[Harry_HH]

That's sound in theory, but I doubt that in practice this will lead to more dynamic material. As far as I understand psycho-acoustics, overcompressed sound is perceived as louder in comparison to dynamic sound, even if the average loudness is similar. If you can point me to a research that states otherwise, it would be interesting to know.

Any research, Aleksey. Again I'm not sure whether you are just using incorrect terminology, or do not understand what you're talking about.

"Loudness" is perceptual. Two signals may have similar average intensity, but different loudness. This is why loudness meters exist. If your loudness meter gives similar results for two signals which you perceive wildly different in loudness, it just means the meter is broken or setup wrong.

All loudness meters are wrong, because they either measure short-time or long-time loudness. For short-time loudness overcompression will read lower than dynamic material. For long-time loudness overcompression may read the same, but perceptually it may sound louder because long-time estimation does not consider short-time loudness.

That's why e.g. Insight measures 4 different loudness key figures.
I personally use mainly momentary Peak, RMS and LUFS Integrated.
In my opinion Span RMS and Peak are OK, the LUFS Integrated would be an useful addition, though.

[Compyfox]

No, it would NOT - as "integrated" is a measurement of "averaged, randomly selected measurement points" over a course of the whole program stream.

If something would make sense, it's SLk for mastering music, because "this" is what you should shoot for if we talk "perceived loudness". IMO and YMMV and all that.