an application for leveling audio
spent a lot of time this weekend looking through the EBUs (european broadcasting union) website -
there was a discussion paper on ‘loudness’ a problem that seems to perplex them -
the EBU has specifications for everything, but some specifications appear to contradict others, especially when it comes to metering - i downloaded a pro metering plugin that complied to their specifications, it has at least 6 meter types based on EBU specs and a multitude of other controls that give the impression of 'standard what standard ? -
on loudness control i found a rack mount Dolby unit (megabucks) and a free program called LEVELATOR link below -
Levelator is a black box application (no controls) in to which you drag the file you want levelling - it seems to work pretty well - looking at the waveforms of before and after tracks it looks like Levelator is what i would describe as a frequency response device that overcomes the problem where certain frequencies appear to our ears and some recording equipment to be much louder than the actual db shown in the waveform -
i have yet to try it in the studio but it may overcome a problem that i have with a certain pro CD recoder i have where the meters go straight into overs whenever a bass guitar is played, this happens to such an extent that to stop the meters going into overs, the bass guitar has to be lowered to such a level that makes it dissapear into the background - according to Ns meters which i find exceedingly accurate the bass guitar is nowhere near to 0bd on its own or in the mix and stands out correctly -
http://www.gigavox.com/levelator
Dr J
What does the Levelator actually do? On soft quite songs and on loud compressed songs what should we expect? Will it make both the same volume? Does it change the EQ?
It looks like a look ahead limiter at least in part - designed for speech - some of the forum comments (their forum) complained about problems with esses.
I dunno, things without knobs…I dunno… 
When I ran a full mix through it all it seemed like it did was lower the volume about 10 db. I didn’t compare the EQs with Har-Bal. That would show something I suppose.
looking at the waveforms produced it looks like Levelator works around a psychoacoustic algorithm, matching the frequencys within a track to the corresponding non linear frequency range of the human ear -
in the software domain i never use compressors or what is generally classed as EQ, i occasionally use Voxengo’s multiband harmonic equaliser which alters the sound of a track rather like the drawbars on a Hammond tonewheel organ can alter the sound produced when playing a note without actully altering the note itself -
listening to a track before and after processing i would say that the processed track is easier on the ear - the problems with ears is that although in natures terms we (humans) all have the same ears, it is the processing device between them, the age, subjectivness and to a degree the socioecomomic programming that the device has been subjected to that determins the auctal sound that a tracks produces or conversly the type of sound levels the processing device defines as being “music” -
if you (“you” being a generic term and not implying to you personally Phoo) are only interested in blowing the cones out of your speakers or mesmerising lesser mortals with your superhuman powers so bestowed on you by the great DAW, then forget it, the Levelator is not for you -
however if your aim is to make your songs RADIO FRIENDLY, this app could assist you in getting on the air, there is also a possability (yet to be fully investigated) that by applying psychoacoustic correction to a track may maintain its quality when converting to lossy MP3 -
apart from the megabucks Dolby device, the Levelator seems to be the only application that fulfills the requirements for a “Loudness control” device mentioned in many discussion documents (spanning many years) in the EBU archive, a device that would maintain a constant output level of the radio station when the engineer has been replaced by automation -
the download runs in at 4.5meg and the Levelator takes quite a time to accomplish its task - it is a stand alone application not a VST or Direct X plugin and when used a new track is generated the original remaining untouched -
its well worth downloading and giving it a try, i personally would class it as a ‘mastering’ tool but mastering, like music and art is totally subjective -
Dr J
Hi Doc., and guys:
Thank you for that link… I’m downloading it as I reply to this topic… It looks quite fascinating from what I see on the opening page…
Am I “Jumping-the Gun” here?
Without even installing it yet I think I imagine the application as a utility that adjusts “Meter Ballistics”… Is that something that this application can adjust for? Or… Does this application actually re-write/render the file? or something like that?
I need to spend even more time up here checking out all these topics…
Thanks again Doc., for that link…
Bill…
Hi Bill
if you havne done it yet - fun a track through Levelator, the import the original file and the output file into N and compare the waveforms - they will look totally different - mute each in turn and see which you prefer -
then - normalize the ORIGINAL track to -8db and see how that compares to the Levelator output track - -8db being the average loudness of a Levelator track -
ideally Levelator should set its loudness at -9bd to comply with EBU recomendations, but being American it ‘sort of’ complies with loudness recomendations for US analogue transmission - in reality 1db either way is nothing to worry about -
talking of meters look at the link below - this meter is a multi face, multi spec meter, (6 different meters all complying to various specs in one package) its from a British company and although not free it is excelent value for its capability - demo is time limited - it reads directly from the track and is unafected by the volume control - so in may case its always in the red ? -
http://www.rawmaterialsoftware.com/ppm.php
if you care to PM me your e-mail address i will send PDF on loudness and reccomended output levels and meter types for digital and analogue audio recording and transmission -
Dr J
Hi Doc:
I am in the middle of my Noontime run… However, I got a printscreen of almost exactly what your post described… I’ll post the printscreen and maybe later this afternoon I’ll post some uneducated detail of what I see on the photo…
here’s the printscreen…
I’ll post some comments later this afternoon…
Bill…
[EDIT]
I can’t stay here any longer… BUT… This render of this Test File on this timeline is incredible… If I didn’t know any better I’d say I have over combinsated for (inthis file) in the “Dynamics” department… This utility does render, and does a lot of other processes, as well… I’ll be back here… O.K.?? I have “drawn a line” around a Question that the utility has raised…
Later…
it looks from your screenshot that at the point you have indicated there is more energy im the lower track of the original than the upper -
here is a piece from the WIKIpedia about masking of audio signals and psychoacoustics - you can apply this to recorded audio as to live audio -
-
In some situations an otherwise clearly audible sound can be masked by another sound. For example, conversation at a bus stop can be completely impossible if a loud bus is driving past. This phenomenon is called masking. A weaker sound is masked if it is made inaudible in the presence of a louder sound. The masking phenomenon occurs because any loud sound will distort the Absolute Threshold of Hearing, making quieter, otherwise perceptable sounds inaudible.
If two sounds occur simultaneously and one is masked by the other, this is referred to as simultaneous masking. Simultaneous masking is also sometimes called frequency masking. The tonality of a sound partially determines its ability to mask other sounds. A sinusoidal masker, for example, requires a higher intensity to mask a noise-like maskee than a loud noise-like masker does to mask a sinusoid. Computer models which calculate the masking caused by sounds must therefore classify their individual spectral peaks according to their tonality.
Similarly, a weak sound emitted soon after the end of a louder sound is masked by the louder sound. Even a weak sound just before a louder sound can be masked by the louder sound. These two effects are called forward and backward temporal masking, respectively -
in psychoacoustics a rise in energy (db) does not always imply that the human ear will detect the rise as louder but as the correct level for that frequency (when compared to the mathamatical model of the ears acoustic response) - does the indicated part sound louder ?, it looks only to be 2db above the surrounding audio and 1/2db above the mean loudness - not a great rise -
i pressed play on the screenshot but it didnt work (must be a bug) - so you will have to tell me what instrument/voice Levalator has found that needed such treatment -
two EXCELENT links -
psychoacoustics -
http://en.wikipedia.org/wiki/Psychoacoustics
digital broadcasting -
http://en.wikipedia.org/wiki/Digital_audio_broadcasting
if you had a graph of the human ears response to frequency and superimposed a graph generated by an audio track on top of it, frequencys on the audio track that are above or below the graph of the ear can be attenuated or boosted till the audio from the track lies in a near comparison to the graph of the human ear - this may in the visual sense appear to show anomolys but sounds correct to the ear -
Dr J
I grabbed the file from the folder that I have this test project, in… It is a Two Track rendered mix. 16-bit @44.1khz file… The Levelator has no buttons to play with… as you know… It is not “Open” to the user to play with… (Which is Good) in this respect… I did the “Drag-and-Drop” and it rendered the file… I imported both files into a timeline and there is the result of the two tracks that appear on the timeline…
I only speak from second-hand experience on this… But…
The CBC here in Canada has and I believe… two protocols regarding this… The audio that goes to their “Master” control studio before the program gets to their transmitters or distribution lines, has an attenuating level attached to the audio program… The two protocols are and it depends on the programming… A +8 level/attenuation or a +4 level/attenuation… format… I am only surmising this… and I stand corrected on this protocol… The Master control then adds it’s processing to the distribution signal… Years ago it was some type of “Hardware” that finalized the signal… Today, I believe it’s done with Software… Who knows?? If it is done with software, this could very well be what is used to condition their final product… Who knows?? It certainly sounds like this could be the process…
I know the Technical Maintenance Supervisor at the CBC here in Halifax… Maybe I could HIT him up for a few Questions-and-Answers… as to how they condition their “End Product”…
This is so fascenating…
I have in mind to “RIP” a file from a CD and import it into a timeline and then “Levelator” the same file and then see what it shows…
Then…
Create a TWO sinewave render of say a 400hz tone and a 700hz. tone mixed and rendered and then do a Levelator of the render to see what appears on the rendered Levelator file…
Anyway, I don’t know what caused the Levelator rendered file on that screen shot to appear the way it did on the timeline… But IT was certainlly audible on the rendered version of that file…
Your explaination of the result is probablly right on… My understanding of what your reply is, leaves me wanting to read it a few times to see if I am able to adjust my thinking as to just what the words say in layman’s terms… or “Bluenose Language”…
Thinking… 
???
Bill…
the specs for North American radio is higher than in Europe, this could be partly to the vast amount and type of transmitters in use some of which may be older analogue types converted for digital use and being analogue in principal have more power output that those over here -
in England all (legal) transmitters are owned by the two governing bodies, the BBC and the Independent Radio Commision - the majority of these will be digital from Mic to transmitter so the specs used are not to far removed from that we use -
European standards for digital systems -
-18db (tone) alignment level -9db max level giving 9bd headroom to allow for ‘hidden errors’ (peaks) not shown due to slow VU meters -
audio presented for radio transmission should be at 48000hz sample rate at a minimum of 20 bits - with BWF header (if available to provider) -
for radio instead of 0db being the max level -9db is to be the max output level - with the main program content being between the (so called loudness level) -18db to -9db - the headroom area above -9db is to compensate (as above) for peaks that are to fast and/or to narrow to be seen by VU meters and may not be trapped by a limiter -
for voiceover/news/traffic/special anouncements the EBU reccomend 0bd for for the voice and -3db for any background music used - this lets the news etc stand out from the general program material -
Dr J
DAB is a bit hit in England how is it going in Canada ?
Hi Doc:
I know those specs would be available somewhere… However, I am unable to quote them. I know that my brother would be able to find them quite easily… He worked for the CBC for some 29 years… However he isn’t with us anymore… I still might ask my CBC Tech. maintenance guy these questions to see how he replies to this…
Coming back to this project I have been playing with…
I ripped a song from a CD using the n-Track “Import” tool… Then, I took the ripped file and “Dropped-and Dragged” the file to the Levelator utility… Then, I imported the two files into n-track and did a “Normalize Check” of the two files using the n-Track “Normalize” tool… The difference of levels of two files with the n-track normalize tool reported a 4.89 db difference… This is at “First Glance”…
I then brought the level of the n-Track “Ripped” file down by the amount of The Normalized n-Track file by the “Difference Amount” of 4.9 db. Or… to match the lower level of the Levelator file… on the timeline…
The two tracks were then solo’d One-at-a-Time… As much as I could hear… the two tracks played at an identical energy/response… At least “Level-Wise”…
That particular file played out at the “tail” of the song with no level change… I didn’t/haven’t done a print-screen of the timeline… but I still have it in order to do that… I’d like to do a “Tone” generated sine wave recording project… Then a Levelator process and compare those findings…
Oh yes… The ripped song normalized out at 0.00 db using the n-Track normalize tool… The test project that I posted in the print screen averages out at slightly less than 0.00 db In fact I’ll normalize the test project right now to see what appears…
My test project and Levelator project averages out at 0.09 and -4.61 db respectively and the ripped project averages out at 0.00 and -4.89 db respectively…
Would it be safe to assume that the Levelator utility allows for a headroom of some 4.5 - 4.8 db as a safe margin-of-error for “Overs”?
Having said that, I’d like to attempt a Mixdown of the Levelator file/files to some 0.05 resolution using the n-Track tool for this procedure and then do a normalize of that file using the n-Track tool as a check for it’s (n-Track), accuracy…
I can for see this leveling application as a great processing tool for adjusting the dynamics of individual tracks here, in the setup I have…
Bill…
yes the -4.5 average allows for overs - it is a pity that it does not have a slider or something to set the out level to -9db -
i have a Tascam CDR750 CD recorder which has very fast meters (and inaccurate ones at that -18bd 1khz tine shouw -12bd) for the price i paid for it it should be more accurate than that ? - with the meters being so fast it picks up peaks that both Ns meters and a liniter/compressor pass through un-noticed -
on the Tascam for overs read (hear) loud didital BLATTTTTTTT - so setting it not to go in to overs is not such a bain it is a inevatable that the CD will be a few db lower than i would like it to be - this week i managed to get my latest project to record for all 77minutes of it without going over -
i ripped it in to N which gave me an automatic wave file - i put this wave file into Levelator and compared it against the original and was very happy with it - (there was a slight loss of top end) then monitored it back through the Tascam (with its meters in operation) the tascam behaved as it should, the meters keeping out of the red and staying pretty stationary around -4db - bought new box of audio CDs today so will soon be able to burn Leveled track to CD and give it extended listening test -
all in all and considering it a free app its very good -
if you want a lower level than the Levelator sets (-8/9db) then lowering the out level with Ns faders should have no adverse effects at all -
Dr J
Hi Doc:
I’ve been playing with this beautiful VST Plug for some number of weeks now… The link was posted by some one up here and I jumped on it… Big Time… Thank you who ever you were that posted the link… I can forget, so easily… Sorry…
If you don’t already have it I would go get it Big Time and see what you are able to do, with it… Go Here… I have used it on individual tracks and rendered/mixdown’ed tracks and I’m having the greatest time with it… as the Case-May-Be…
Maybe, the Levelator People might write a Level/Adjust Knob into their application that will allow the adjustment of the “Output” file to be rendered… It might be worth some mail over to them to see how they might accept the suggestion, regarding this idea… Tell them I sent you on this “Mission”… I’ll back you up till your nose bleeds… 
OoHh… while you’re doing this… Tell them we want a .05 db reselution with a “Punch-in and Enter Dialogue” Entry Box… to address the level adjustment/change… as well…
You see… if I wrote them the suggestion for the update, they’d wantta know who this guy over in Bluenose Land is, and what does he know… anyway?? to suggest changes to the Levelator app… like that??
Then tell them I’m having My People looking into this… Sorry… You know if you don’t tell the “Shirts-and-Suits” “CODERS”… what you want how do they know you want IT?
Bill…
Which plugin on that page are you referring to, Bill - the FreeG?
Hi vanclan:
Those guys over there are pretty clever… They are looking to get their product out and working, in the market place… It’s FreeG plugin that I downloaded and am playing with on my setup here… I sent them some feedback on the Pan resolution adjustment… They replied to me the next day and they are looking into some issues I had with the resolution… I explained to their support how I was using their freeG plug and with what multi-track editor I am using… I mainly use their plug to process individual tracks… As well, I use the plug to attenuate and pan tracks that are positioned on n-Track’s timeline… Not all tracks get this treatment… and some groups get this treatment…
This is the type of plugin that I’d like to see an Adjustable Phase feature added to it… Well…
Bill…
thanks for the link Bill, will look into it -
heres one that is rather interesting its called BITVIEWER - it is a VSTplugin that shows how many bits are actually being used in a track - the display has three regions for 16bit 24bit and 32bit tracks -
you just insert it in to a track, play track and watch the display - if your 16bit track is only running at 12bits, you may have to use dither in the mixdown to regain the quality lost by low bit useage -
Dr J
get BITVIEWER (free) from link below (top VST on page)
http://www.tobybear.de/p_utilbag.html
That FreeG looks very interesting, Bill. How much CPU does it take per usage? How many tracks and groups have you used at any one time?
Paul
i know your question was aimed at Bill and no doubt he will give you a full description of how FG assists him in his tracking -
my answer to your question is - (on P111 932meg PC running Win 2K, 256 meg RAM) -
in order to see what is going on with FG and CPU use, we have to have a constant signal, you really cannot make any test comparisons using a normal wave file -
test track was professionally recorded 1k sinewave calibrated to give 0db output without altering anything in N -
now being a VST plugin the CPU use that this plugin takes up is added to Ns CPU use -
track without FG inserted showed 11% CPU use
track WITH FG inserted showed 35% CPU use
that a massive 24% jump -
how does it work and how does it sound -
well it works perfectly well - cant fault it -
however on sound -
to compare FG in combination with Levelator and an original audio wave file i did this
1…inserted a original wave file that was known to go into the red (clipped) in to N
2…inserted FG into original track, set FC to -4db which is the average that Levelator gives -
mixed down to new wave file -
then passed known wave file (1.) through Levelator which generated new leveled version -
cleared screen - imported original wave file, FC generated wave file and Levelator generated wave file -
comparison test -
original wave file was harsh and crackly due to clipping -
FG generated track stayed level at -4db - but sounded weak -
Levelator track stayed around -4db but had a more full bodied sound to it -
so in my opinion and for my own use i would stick with Levelator -
1… it is a stand alone app so does not add a large CPU hit on N -
2…although there is a slight roll off on the top end, tracks i have put through Levelator (to my ears) sound sweeter -
if i definately had to produce a track that was averaged out to -9db (see other posts in this topic) i would rather trim back a Levelator track by 5db and maintain the full bodied sound than set FG to -9db and have a weak sound -
these are my own personal opinions and are based on the way i produce a finished track (which is through hardware units), N track being used to set out a song and making full use of the pan & volume automation -
also my studio PC which is has 24 optical outputs from the soundcard, there is no master output available to insert FG into (all audio is generated by an outboard mixer) - this is a 2 pass recording method - this may seem messy (but it works) -
for the 1st pas - i have hardware limiters to keep the levels to what is required -
i record a CD taking stereo out from mixer into a multiband limiting compressor, (with limiter set to 0db with minimum compression), take output from limiter/compressor through bass and treble enhanser (this puts back in what the limiting compressor takes out) then to CD recorder -
i then rip CD back in to N which auto generates new wave file -
for the 2nd pass - i put new file through Levelator and then record Leveator track back to new CD (bypassing all effects on mixer and all other hardware) - result, nice smooth sounding, full bodied recording -
so which is best ? - thats UP TO YOU - both are FREE and easy to use. so download them and see which one (or both) suits your way of tracking -
Dr J