Topic: peak lowering in audacity

[DisturbedPyro]

hey gang. i want to lower the peaks in a recording. is there a way to set what you want the new peak amplitude to be over the duration of the entire recording but ONLY reduce the peaks that are higher than the designated new peak amplitude? to explain further, say i want to amplify my recording by 10 db, there are five peaks during the entire recording that surpass allowing me to amplify the file by 10 db without clipping. so is there a way i can select the entire recording, set the new peak amplitute to 10 db, and have it only reduce the above mentioned 5 points while leaving the rest of the recoding untouched? obviously if its only 5 peaks, doing it manually isnt a big deal, but i was just using that as an example. i know the normal amplify function allows you to set the new peak amplitude but that will apply the effect to the entire file and i dont want that. i only want it to affect the peaks. as i said in the subject, i am using audacity. is there a way to do this? thank you

[Roving Sign]

well - you can do it manually - select and de-amplify (especially if its a just few hard spikes)...and then Amplify to 0.

If there are more than a few - you can use a limiter, followed by Amplify to 0

should have the effect you describe...

But - its sort of a two step process...

[Brian Skalinder]

Compression.  Check the stickied Readme1st post for links to some discussions on compression and/or limiting.

[rsimms3]

Hard Limit is what it is called in Audition.  You increase the volume while limiting the volume of any sound that would exceed your preset db i.e. -.01.  Audition will give you an estimate of the amount of distortion you get or would get from increasing the volume by that amount, depending on where you set your limit.

[DisturbedPyro]

ok thanks guys. it looks like hard limit is what im looking for. the readme1st posts said it is "quick and dirty" but i tried using compression but thats not quite what im looking for. that seemed to alter the entire wav, i just want the peaks to be affected. and ive used the envelope tool before and i know that alters the entire area also, not just the peaks. so even if it is "quick and dirty" it does what i want, it lowers the peaks alone and does not mess with the rest of the wav. thanks guys

[Brian Skalinder]

Compression and limiting are very, very similar.  I think of limiting as just a severe form of compression.  As with limiting, you need not apply compression to the entire recording.  Compression may yield better sounding results, depending on the content of the edited portion.  Play around with different thresholds and ratios and you'll find a result you like, I bet.

[rsimms3]

Sometimes compression is the way to go like Brian said.  If you have something specific, like heavy bass, that is the main culprit of loud volume, you might consider reducing the bass which will allow you to increase the overall volume.  I have done that a few times as the bass was just outrageous and overpowered my speakers (I have a cheap setup and listen to 90% of my music in the car).  Hard Limit is a harsher form of compression because you are cutting off and limiting the sound for those peaks while increasing the volume of the remainder of the WAV.  The more you do this type of mastering, the less dynamic and crappy the music will sound.  Very well known example - Metallica's Death Magnetic.  Below are pictures of Pearl Jam's Ten - Original CD, Remaster, and Remix.  Note that the last one looks like "sticks of butter" meaning they compressed the overall sound greatly to be able to increase the overall volume.  I did these shots myself.  It's a real shame the mastering of this new Remix fell into the "loudness war".

[DisturbedPyro]

ok, i see what you mean. hard limiter doesnt do exactly what i wanted. it simply cuts off the peaks at a designated point, thus loosing part of the frequency.

what im looking for is a way to lower the peaks over the entire recording without messing with the rest of the wav, i.e., the non-peaking parts of the wav. and i mean reducing the peaks, like you would by deamplifying them, not hacking them to pieces like hard limiter does

i looked into compression but it still seems applying compression affects the entire selected area, not just the peaks within that selected area. is there a way to reduce only the peaks within the selected part of the wav while leaving the rest of the selected area (i.e., the non-peaking parts) alone?

[Brian Skalinder]

i looked into compression but it still seems applying compression affects the entire selected area, not just the peaks within that selected area. is there a way to reduce only the peaks within the selected part of the wav while leaving the rest of the selected area (i.e., the non-peaking parts) alone?

You need to set the compressor's threshold and ratio properly based on your needs.  The compressor will reduce those portions of the waveform above the threshold and leave alone the portions beneath the threshold.  So set the threshold value to the same level as the highest level of audio you don't want to compress.  Then play around with the ratio to find what value will compress the portions above the threshold down to the level you wish.  The higher the ratio, the more compression is applied to portions of the waveform above the threshold.

Oh, when I used Audacity years ago, the compressor was broken.  Not sure if it's fixed now, so maybe that's why you're running into issues with the compressor?

[DisturbedPyro]

i dont know if im doing something wrong or maybe the compressor is broken like you said...but when i set the threshold to -3.0 thinking it would reduce the peaks that exceeded that amount, nothing happened. it had no affect on the wav. so i bumped up the ratio from one end to the other. still, no affect. so i changed the threshold to -20 with a ratio of 10:1, just to see what would happen. i dont understand the results. take a look:



the first is the original smaple, the second is after compression. what i dont get is why some of the high peaks were reduced only a little where other high peaks were reduced a lot more. for example, look at section A and B. A was reduced hardly at all, but B was reduced more. furthermore, on the far left of the original wav, those peaks were reduced a lot after compression. but the peaks in A and B wernt. also, note that the entire wav was altered, not just the peaks. finally, amplifying the wav after compression resulted in a 1.9 db increase. when i set the threshold to -20, how is the amplitute only at 1.9? am i totally not doing this right or what?

[Brian Skalinder]

DP  --  Thanks for the screenshots.  Looks like the compressor works, now.  How well, I don't know, but it didn't work at all when I last tried.  Glad to see they've fixed it.

I suspect the reason the entire waveform was altered is because the threshold was too low.  The compressor shouldn't alter content that doesn't reach the specified threshold.  To identify a threshold that impacts peaks but not much else, you may find it useful to view the waveforms in dB scale, rather than Audacity's arbitray 0.0 - 1.0 scale.  Simply click the project track's upside-down triangle and select Waveform (db).

The reason some peaks aren't compressing much, and resulting in only a 1.9 dB increase, may be the compressor settings.  It may not be impacting some peaks because the attack is too long and/or the release is too short.  Take a look at those two settings in the compressor:  attack and release.  They control how swiftly to engage (attack) and disengage (release) the compressor.  As a starting point, use a short attack (10 ms) and a longer release (500 ms).  Adjust each setting to shorter or longer values to see how the results change.  For kicks, and as a learning exercise, set them to extreme values (both very long and very short) and listen to the results.  Severe compression settings may result in "pumping" and "breathing".  You'll know it when you hear it, and you'll want to avoid it in the final edit.  Lengthier, but good, piece on compression here.

Let us know how it goes.  If you can't get Audacity's compressor to get the job done, perhaps we can find an alternative way to achieve the results you desire.

[DisturbedPyro]

ok i have it in waveforms now

audacity has an attack time range of 0.1 secs - 1.0 secs so i can only change it by increments of 0.1. also, i dont see a release. just threshold, ratio, and attack time. there is a check box option for normalizing to 0 db after compression also, but i know thats not what im looking to do at this point so that is not checked

more screens: the first is the original sample. the second is compressed at threshold -20, ratio 10:1, and attack time 0.1 sec. this one amplifies to 5.8 db. the third screen is compressed the same except the attack is at 1.0 secs. and this is the one that amplifies only to 1.9

but it still looks like no matter what i do, it alters the entire selected area. not just the peaks. any ideas? thank you so much for your patience with helping me through this

[Brian Skalinder]

I guess Audacity's compressor doesn't allow you to change the release.  But it looks like changing the attack made a difference in knocking down the peaks you mentioned earlier, as expected.  Don't sweat the release too much, it looks like the compressor's working just fine.

The reason the compressor is modifying the entire selection:  the entire selection is above the threshold (-20 dBFS).  Draw a straight, horizontal line across the waveform at the -20 dBFS marker.  (see attached)  The compressor will act on portions of the waveform rising above the line, but not portions of the waveform falling below the line.  As the attached image shows, with the threshold at -20 dBFS, the entire selection is at or above the threshold.  Hence, the compressor impacts the entire selection.

Honestly, the levels look just fine in the original, but my perception may be skewed by the duration.  What's the duration of the selected waveform in the images?

Also, maybe taking a step back will help:  what is the audible problem you're trying to solve, and what are the problem's audible symptoms?

[rsimms3]

I was just thinking the same thing Brian.  I am not familiar with audacity, is there a function in audacity to assess the amount of volume increase you can use coupled with the amount of distortion you will get with said increase in volume?  As mentioned, the first question is what you want to accomplish in the end.  From your first post it looks like you want to increase the entire show by 10db but you will be clipping in spots with that much gain.  To me, some overage or clipping is acceptable, depending on how much and what it is.  If it is a sharp snare drum snap, it is less annoying that constant bass rumbling throughout the recording putting almost the entire recording over 0db.  In the end, you do what works for you, it's your recording.     

[DisturbedPyro]

The reason the compressor is modifying the entire selection:  the entire selection is above the threshold (-20 dBFS).  Draw a straight, horizontal line across the waveform at the -20 dBFS marker.  (see attached)  The compressor will act on portions of the waveform rising above the line, but not portions of the waveform falling below the line.  As the attached image shows, with the threshold at -20 dBFS, the entire selection is at or above the threshold.  Hence, the compressor impacts the entire selection.

Honestly, the levels look just fine in the original, but my perception may be skewed by the duration.  What's the duration of the selected waveform in the images?

Also, maybe taking a step back will help:  what is the audible problem you're trying to solve, and what are the problem's audible symptoms?

i guess i dont get how the threshold works. i read the links and what you said and it sounded like it would do what i wanted. but...its not. so i guess i just dont understand

this is just a minute long sample of the whole show. i was just using this sample to play aorund with instead of working on the entire show. theres no exact "problem" i have with my recording. i just want to amplify it some becasue it is a bit quiet. heres my issue: the original wav would not amplify due to peaks. so i wanted to reduce the peaks (just the peaks, not the entire wav), so i could amplify the recording. i thought (and obviously i was wrong) that whatever i set the threshold to, it would reduce the peaks throughout the recording to that level. so...since i set the threshold to -20, i thought it would reduce all the peaks throughout the entire wav that were restricting my from amplifying by 20 db. then after compression i thought i would then be able to amplify the recording to 20 db. but as you saw, i cant. depending on the attack range, its either a mere 1.9 or a 5.8 range that i can amplify. note that i do not want to amplify it by 20 (that would be way too much), i was just using that as an example becasue as i was messing around with the compression feature in audacity, and setting it to anything low had little to no results, regardless of what i set the ratio and attack to. so i went with 20 just to use as an example

it seems i am either severely messing this up or compression is not the tool to do what i want

[Brian Skalinder]

Don't expect the compressor to reduce signal all the way down to the threshold -- that's basically a limiter's function (and it generally doesn't sound very good).  Theoretically, a ratio of 10:1 should knock down a signal 10 dB above the threshold so it's only 1 dB above the threshold.  You're running into an issue I discovered long ago:  the compressor doesn't compress the signal as far as I expect.  This is true in all the compressors I've used (Audacity, Audition, WaveLab, Samplitude).  They knock down the signal, but not all the way to where I expect based on the settings.  I've not taken the time to explore the issue more deeply, and have simply assumed there's some limit in the implementation, whether related to the attack, release, or internal workings, that makes it so.  That said, I still find compression very useful sometimes.

You might try applying compression only to the problematic peaks.  This will allow you to use a more aggressive threshold and ratio without impacting the rest of the recording.  And if compression isn't accomplishing what you need, you might try using a volume envelope.

[DisturbedPyro]

it just seems weird to me that will all the in depth functions these audio editors can do, there isnt one that works like a hard limiter but rather than just hacking off the peaks at whatever designated limit you enter, it instead reduces the peaks to that level. so that it would be as if you had manually lowered the peaks using a negative amplify

what ive always done in the past is: i find a peak, select it, and then lower the peak by entering a –x db amount to bring it down to more of an appropriate level to match the rest of the recording. its a pain to go through and find all the peaks so i figured with all the hundreds of effects you can use on a recording, theres got to be one that allows you to enter a designated point at which any peak above that number would be reduced to that level. is there really no way to do this?

[Brian Skalinder]

A compressor or limiter is the only tool I know of that works the way you mention.  The limiter does so in a bit of a harsh way, and the compressor -- by design -- is more subtle, so it doesn't flatten out the waveform.  But it also doesn't reduce the level quite as efficiently, either.  I stopped searching for alternatives because I've found mild (or mild or heavy) compression works just fine for those instances I need it.  I like dynamic range and don't want to squash it too much, so I've never searched for an alternative tool.

[DisturbedPyro]

so i guess im s.o.l. ok, thanks for the help anyway. id anyone else has another idea, please let me know

[Brian Skalinder]

I don't know if your SOL, yet.  Try multiple passes of the compressor, but with a higher threshold so you're not impacting most of the audio.  You might also try a pass or two of compressor (again with a higher threshold), and then a limiter pass for any remaining pesky peaks.  Or one or both of those + a final pass with a volume envelope on any particularly pesky peaks.

Just for curiousity's sake, and to get a better handle at the peaks we're talking about, would you post a screenshot of the entire performance's waveform, using Audacity's dB scale?

[Scooter123]

In Goldwave, this is stupidly easy.  They even have about 10 presets for this.  I looked at the presets and made my own.

Pick a threshold minimum and maximum and it reduces the volume of those portions across the whole track or within a selected area, like a few seconds down to the minimum.  It does not touch the volumes below of the spread you've selected.  You can raise or lower that spread of db to suit your taste.

Goldwave also has the reverse, which takes lower db sections, again within a range and boosts that spread within a track or a selection within that spread only, so quiet piano parts are boosted but loud cymbal crashes are not.

What I do is a batch file processing, compressing the volume at a very high db level, so it is not very noticeable at all, then the reverse ever so slightly, then compress again etc.  With Goldwave, I can take a whole 2 hour show and program the batch commands, multiple light compression, boosting, compression boosting etc, and it processes the file overnight and even saves it.  So in the morning, I'm ready to look at it one last time for tweaks and file splitting. 

 

[db]

"a" and the pop just after 45 mins looks like the only thing you'd want to alter. can you envelop or pencil tool these away?  if not, i'd let it slide. the waveform looks fine. the thing to consider with compressing a whole mix (as was pointed out above) is that you can lose more than you gain (pardon the expression). it's a taste/opinion thing, but i find compression useful for recording drums (overheads/snare), bass.. etc. individual instruments in a controlled studio setting. applying it to a live recording has more downsides than up as you are killing whatever dynamic range you've collected.

[Brian Skalinder]

In Goldwave, this is stupidly easy.  < snip >

I haven't touched Goldwave in years and years, but sounds like every other implementation of compression:  easy, really.  Just a matter of finding what works for you.  Though I'm not sure Audacity supports batch processing, I know other apps do.

If the waveform db quoted is of the whole show, I agree A and B + the peak a little after 45min might be the only places I'd bother tweaking.  You might apply a volume envelope to reduce the levels of the first 7 minutes or so -- it looks like maybe you tweaked levels to lower them a bit after this portion -- but if this section of the performance was naturally louder, I'd just leave it alone.

Still curious to see the whole waveform in dB scale, if you're willing.  It's a better representation than the arbitrary 0.0-1.0 scale.  Either way, hope we can help you find a resolution that makes you happy. 

[DisturbedPyro]

In Goldwave, this is stupidly easy.  They even have about 10 presets for this.  I looked at the presets and made my own.

Pick a threshold minimum and maximum and it reduces the volume of those portions across the whole track or within a selected area, like a few seconds down to the minimum.  It does not touch the volumes below of the spread you've selected.  You can raise or lower that spread of db to suit your taste.

Goldwave also has the reverse, which takes lower db sections, again within a range and boosts that spread within a track or a selection within that spread only, so quiet piano parts are boosted but loud cymbal crashes are not.

What I do is a batch file processing, compressing the volume at a very high db level, so it is not very noticeable at all, then the reverse ever so slightly, then compress again etc.  With Goldwave, I can take a whole 2 hour show and program the batch commands, multiple light compression, boosting, compression boosting etc, and it processes the file overnight and even saves it.  So in the morning, I'm ready to look at it one last time for tweaks and file splitting.

yes! i like goldwave's compressor much better. i used the "reduce peak" preset and it gave much more of the effect i was expecting. it also did more of what you all said a compressor should do...im not convinced audacity's compressor works as its supposed to. anyway, i like goldwave's much better, so i can just use the preset as a starting point and alter it from there to get the results i want. as always, the experts at tapers section know their stuff. thanks scooter for the goldwave suggestion. and thanks to brian for sticking with me this entire time! you guys rock!

[Brian Skalinder]

Perhaps Audacity's compressor still isn't fully functional.  Glad you tried another option and it worked for you!