Too loud? My experiments with Limiters = square waves...

Discussion in 'Mastering' started by headchem, Jan 2, 2006.

  1. headchem

    headchem Guest

    Here is a simple track I've been using to get familiar with my Limiter. Do you all think it's too loud? My average RMS is between -11 and -10 dB. To my unexperienced ears, I haven't lost too much of the dynamics, but tell me if this is fatiguing to you.


    Here is a screen-shot of one of my bass drum hits (I think it is, at least...)



    Is this a bad thing, or can I get away with quite a few of these square waves (if this is even bad enough to be a square wave)? Nearly every bass drum hit looks like this. Again, to my non-ME ears it sounds fine, but tell me if I should think it is too harsh.

    My auto-volume compensation in iTunes for this track is -1.2 dB, while commercial songs with the same -11 to -10 RMS get -6 to -10 dB reduction in iTunes. Why does that happen? Is there another way of measuring loudness besides RMS that I don't know about?

    Thanks for your input. If you feel so inclined, suggestions on the mix are welcome, too. (I think the bass may be too loud, and the electric guitar needs some EQ or different distortion?)
     
  2. beachhunt

    beachhunt Guest

    If you want to make a song loud but good, don't use a limiter primarily for loudness, use compression. Limiters should catch the occasional spike (or catch nothing at all), but they will deteriorate your sound when used in place of compression. You may still like the result, so as always do what makes you happy. However, if you are truly worried about the song being fatiguing, use compression to get there, and maybe push a dB or two into the limiter if you want to squeeze every last drop.

    If you want, I can give you temporary FTP access on my server so you can upload the non-clipped, non-limited wav and I can try to reach the numbers you're after, and let you know specifically how I got there. I'm no 20-year veteran, but I'm willing to help.
     
  3. Massive Mastering

    Massive Mastering Well-Known Member

    I think there's something wrong with that MP3... Left only...
     
  4. BobYordan

    BobYordan Guest

    It plays in both speakers in my 2.76 Winamp that states that it is an
    approx 159-170 (variating) kbps & 48kHz mp3 file, when streaming it.

    I dont sound so bad in my surf pc speakers, but I guess the
    bass woofer might handle the clippings and makes them sound softer?
     
  5. BobYordan

    BobYordan Guest

    Sorry double post. :oops:
     
  6. audiowkstation

    audiowkstation Active Member

    Never let digital zero be the limiter. Never never. If you want compression..do so with the compression or limiters..not using digital limits as the limiter. I can do a very loud file peaking at -5dB and this way, consumer equipment does not add to the distortion nor will I have translation issues.

    Digital zero was never meant to even be approached..let alone touched. My advice is to seek a sound that is nice and stay away from digital 0.0dBfs. As a RULE...I simply do not allow the limit of digital to ever change the shape of the waveforms. If I want limiting or even clipping..it should look the same at -20dBfs as it does at -1dBfs. The moment I reach anything above -0.5dB peaks..I know that my peaks are too hot for 90% of the consumer gear made.

    Clip it in the mix at lower levels if you must clip..and never do it in mastering.

    If you want clipping, clip the mic pres, clip the line amplifiers and clip the buffer amps..keep the 2-bus low and mastering should never be used as a clipping tool. Never. Those who do are adding to the problem. I know 26 mastering engineers who don't care about your music. They are playing the role of producer when they clip the master. Anyone with any equipment can do that. It does not take a good mastering chain or a good engineer to do this. Any HACK can do it.

    If you desire mastering clipping, never send me any work..because the artists treasures are too sacred for me to screw up. Screw it up yourself and be proud.

    Your -1.2 limit..although itis to be commended..is not digital clipping from going to 0.0dB..it is due to stuffing more level against the new stop (-1.2) than can be processed and the level stays the same and time travels on. To me, this does not sound good. Open in AA and see how it looks reconstructed.
     
  7. dcollins

    dcollins Guest

    But the sound of analog clipping is totally different, and probably less musical than judicious digital clipping.

    A little bit is no problem, it's the use of 6dB or more that starts to suck...

    Waaa wahhhh.....

    DC
     
  8. Michael Fossenkemper

    Michael Fossenkemper Distinguished past mastering moderator Well-Known Member

    As you can see, people have opinions on clipping. Only you can tell if it's ok with you or not. If it's the sound you want, then by all means use it. Ignore what the waveform looks like. Just close your eyes and listen. then listen again and again. Once you really hear what each process does and doesn't do to a mix, then you can use that knowledge when you approach a mix in the future.

    As far as rules go, that is an individual decision. Experience will tell you when to break them and when not to.
     
  9. headchem

    headchem Guest

    A few things I'm not understanding:

    About never letting digital zero being the limiter - does this mean I should lower my output volume from the limiter by about .5 dB so that none of those flat-looking peaks actually touch 0 dB? In a different phrasing, should I not normalize 100% (if there's headroom leftover)? I didn't push my level up until it clipped like that. I set the Limiter to catch every peak so that nothing coming out was a digital clip. Excluding the soft clip function on my limiter (which I think tries to emulate analog clipping), is a limiter just a device that helps control the amount of digital clipping? I guess I thought that limiting is better than clipping, although I'm not sure anymore since they both chop off the peaks...

    Did you mean -.5 dB, or is -5 dB a realistic peak for some tracks?

    I took a look at two very loud songs and sure enough, one of them, Sheryl Crow's Good is Good, mastered by Bob Ludwig, had plenty of headroom leftover (-.6 dB). However, Green Day's Boulevard Of Broken Dreams, mastered by Ted Jensen, touched 0 dB every bass drum hit. None of Ted Jensen's peaks looked as flat as mine, though. Both songs sounded much more compressed than my track, too, although I know it probably wasn't the ME's will to do so.

    Here's a quote from the recently created NOT LOUD ENOUGH DAMNIT thread. I'm posting it here because it seems relevant.

    Would you all (or Bob Ludwig for that matter) agree that this is a good starting point for limiting and compression? I assume that the compressor goes before the limiter in the mastering chain, right?

    When Cucco said that high frequency peaks need faster attack settings, it gave me an idea I could program into my mastering chain in Reason. Every time my (high frequency) snare hits, the attack on my master compressor is automatically reduced to a faster attack time. When the snare isn't playing, the attack settings are returned to "bass drum settings." My bass drum attack is 28 ms, and my snare attack bounces down as far as 10 ms but zooms back up to 28 ms as soon as the snare hit has passed. My release doesn't change and is set to 119 ms with adapt release enabled. The only difference between these two files is the difference in attack. Does one sound better to you all?

    Not Another Rap (fixed attack settings)
    Not Another Rap (dynamic attack settings)

    Maybe this technique has been around for ages and I just don't know it. Regardless, "Frequency-dependant attack" sounds like a cool button to put on a compressor... :D

    I read somewhere on these forums while searching for posts on RMS values that really loud songs must be prepared from the tracking stage all the way through mastering. Is clipping it while mixing one of the tricks used to get really loud songs? If so, why not just use a limiter during the tracking?

    I'm not too proud to say those few sentences were over my head... Is the -1.2 limit in reference to the gain reduction applied by iTunes? I don't understand how iTunes decides on this, but I was only stating that iTunes thinks it should reduce my track's level by -1.2 dB in order to be comparable in volume to the other tracks on my computer. Most commercial tracks have maybe 1 or 2 more dB in RMS than my track, but they get -6 to -10 dB gain reduction in iTunes. This could be a totally different subject that would confuse this discussion, though.

    What is AA, reconstruction, and a "stop"?

    If I understand you right, you saying that when my limiter's gain reduction is more than 6 dB for any given peak, this should sound bad? This makes sense to me. I don't think my limiter ever reduced gain by any more than -4 on that track I posted.

    I appreciate the offer, and I'll certainly take you up on it as soon as I understand what's going on. I think I need to better understand the role of a limiter before I start emulating others' use of it. :D
     
  10. Reggie

    Reggie Well-Known Member

    I can't feature why one would leave -5db as the peak amplitude on a mastered digital track. That seems like WAY too much wasted headroom. Sure you don't have to worry about the kids' CD players clipping, but they will now have the problem of turning up the system noise as they increase the volume to match the level of their other CDs that were mastered hot with peaks at -.1 or 0.0 or whatever.
     
  11. headchem

    headchem Guest

    Yeah, I think it was just a type-o.
     
  12. Michael Fossenkemper

    Michael Fossenkemper Distinguished past mastering moderator Well-Known Member

    In general, a look ahead limiter when gently used, will try and round the reduction a bit rather than lop it off. The harder you push it, the more drastic it flat tops.

    You could use adaptive attack times for different freq's but you have to be careful that it doesn't pump. but I think a multiband compressor might do this a little better. A few high end compressors use variable release times to keep it more natural sounding. A snare hit will have a faster release time than say a vocal line. This will mask the effect of the compression more.
     
  13. headchem

    headchem Guest

    Now that I think about it, a multiband compressor is very similar to that frequency-dependant attack I was thinking of.

    Off topic: If you had a perfectly mixed song, would you ever consider putting it through a multiband compressor? I thought those were used primarily to fix problems in the mix. I know the answer will probably be something like "if it sounds good, sure, just use your ears, etc..." But what do you look for before deciding to use a MBC?
     
  14. Cucco

    Cucco Distinguished Member

    Personally, I can't ever recall a time in history where I've reached for a multi-band compressor. I've had them available in one incarnation or another for a few years now, but I don't think I've ever seen the interface. Generally, if I need to use it, I think I did something wrong.

    So, I guess my question would be, when do you find it appropriate to reach for the multiband compressor? (Same question, only in reverse...)

    J.
     
  15. TrilliumSound

    TrilliumSound Active Member

    "Perfect mix" might not need compression at all.
     
  16. BobYordan

    BobYordan Guest

    Here is one more stupid question, for ya. :D

    Is there any difference between clipping inside eg an application like Sound forge/Wavelab at 0dB making them square tops, compared to clipping inside eg a VST plug, that limits it's output? Eg if a VST plug is set to -0.1dB max output, then I guess that Sound forge/Wavelab etc dont consider the peaks to be to be clipped? :? But looking at the wave form
    (that is some kind of dynamic representation of all the dynamics in all frequencies bands) it shows clippings anyway. It might sound okay or not, listening to it? But the eye is fooled in some way, IMHO. And the audio editor programs stats might lie?
     
  17. Michael Fossenkemper

    Michael Fossenkemper Distinguished past mastering moderator Well-Known Member

    there is a difference in sound between the two. I find that plugs do a much worse job in handling clips. probably because they aren't written to handle them well.
    Nova (apogee UV22 option) was basically just taking the overs down a hair so they wouldn't look like overs.
     
  18. BobYordan

    BobYordan Guest

    Okay, thanx for the answer, I am gonna ask some more. :D

    1) The audio editor programs, what do they do? take down the overs or
    just cut the output?

    I guess the recommendation is probably to use some built in dynamic plug/algoritm/VST, to do the job? To get the best result.

    2) Are there any VST plugs/built in algoritms that can reverse the process or is a clipp not reversible, sound wise? I guess it could be possible to get some dynamics back from a very squashed mix? If eg the clipp gets diminished in time, it might not 'rattle speakers' so much, that it will be heard? This is a very hypothetical topic. ;)

    3) Also within a software, digital distortion seems like an abstract, IMHO. At least for an offline algoritm, that can analyse the dynamics and handle
    potential clipps causing digital distortion in a better way.
     
  19. Michael Fossenkemper

    Michael Fossenkemper Distinguished past mastering moderator Well-Known Member

    I'm not sure why you are asking, but i'll give it a go.

    Clipping is not a linear process, whenever there is a non linear process, it produces junk, (harmonics). Now there are units out there that do a better job at controlling them then others. Now these frequencies that are produced are beyond the sampling frequency. When they go beyond this point they are aliased back into the audible frequency. this is dependent on what frequency was clipped of coarse but there is a formula for figuring out where that alias would end up, but I don't know it. If you were to use a unit that did this process in a higher sampling rate, then filtered out these frequencies before they could be aliased back into the audible spectrum, then you would have a better chance at reconstructing the waveform. Otherwise you could kind of reconstruct the waveform but you'd still be left with these aliased frequencies that don't sound very good. Now I would imagine that different audio engines in various programs deal with these generated non linear frequencies differently. Some do it very well and other don't. Plugins are about the worst I've heard.
     
  20. Michael Fossenkemper

    Michael Fossenkemper Distinguished past mastering moderator Well-Known Member

    Actually Nika used to hang around here. he wrote some really informative papers on digital that you might find usefull.
     

Share This Page