The No.1 Website for Pro Audio
 Search This Thread  Search This Forum  Search Reviews  Search Gear Database  Search Gear for sale  Search Gearslutz Go Advanced
Peak Headroom Dynamics Plugins
Old 26th October 2018
  #1
Lives for gear
 
Trakworx's Avatar
Peak Headroom

An AE friend of mine recently attended a mastering session with Ted Jensen. The music was Hard Rock. The mastered tracks are pretty loud, generally hovering around the -6.5 LUFS (short term) area. The peak headroom of the masters is -0.1dBfs. There were no alternate versions delivered.

Now I've read a bazillion GS ME's posts saying they leave 0.3, 0.5 or 1.0dB peak headroom to prevent audible clipping in codecs downstream.

And I continue to see major MEs like Mr. Jensen blatantly ignoring that concern.

So I'm questioning whether I should continue my practice of leaving 0.5dB peak headroom.

If Ted Jensen doesn't worry about clipping in codecs then maybe I shouldn't either.

Is there something I don't know?

Like - Do the various delivery sites' codecs automatically lower the gain before they convert to lossy formats?

Is clipping in codecs really just a non-issue?

It would make it a little easier to please loudness-hungry clients if I had that extra 0.4dB to play with...

Thoughts?
Old 26th October 2018
  #2
Lives for gear
 
Justin P.'s Avatar
 

Yes. For various reasons I've been able to see some master WAV files that were done by some of the top names from Sterling and other major mastering facilities.

They don't seem to care about the peak-headroom as much as you see talked about here.

It's probably a case of "if it ain't broke, don't fix it".

Is this release available on the streaming services? If so, maybe check to see how it came out on the other end on Apple Music/Spotify/TIDAL etc.

I would not think the streaming services are lowing the level before the codec as a courtesy because that would create floating point audio and open the can of dither worms and they could also change their practices at any time. Also, Spotify recommends that 1dB of peak headroom which to me indicates that they aren't going to create it for you if you don't do that.

So for these reasons, I still like to leave up to 1dB true peak headroom on my master WAVs.

For me, that practice is also a case of "if it ain't broke, don't fix it".
Old 26th October 2018
  #3
Lives for gear
 
Trakworx's Avatar
Thanks. Good point about Spotify. This album isn't out yet. I'll stream it after it drops and listen for clipping distortion...
Old 26th October 2018
  #4
Lives for gear
 
SmoothTone's Avatar
 

I've seen this too and come to similar conclusions. They're either still in the CD mindset or trading on the probability that encoder overs or ISPs aren't that audible in the grand scheme of things. They're delivering MFiT files so they're surely aware of the issue. Spotify even recommends -2 for super squashed masters.

I've been a -1 guy for the last year or two. To me, it's just good engineering practice these days.
Old 26th October 2018
  #5
Lives for gear
 
Trakworx's Avatar
Yeah, I try to practice good engineering too, but when I see the most respected MEs in the world routinely ignoring such practices it really makes me question why I bother.
Old 27th October 2018
  #6
Lives for gear
 
Trakworx's Avatar
So I have a 16-44 PT session where a few years ago I ripped several reference tracks from CDs. I went in and checked the peak headroom.

They ALL peak at either -0.1dBfs or at 0dBfs.

-0.1dB: Wolfmother, The Black Crowes, Alice In Chains.

0dB: AC/DC, Soundgarden, Led Zeppelin, Queens of the Stone Age.

I just checked a few of the loudest of those tracks streaming on Napster using my iPhone and Sony MDR-V6 headphones (for their brightness so I could really hear any HF distortion).

I do not hear audible clipping distortion. At least no more than I hear in the ripped waves.

The ripped wave of "No One Knows" by Queens of the Stone Age peaks at 0dB and has a short term LUFS hitting up to -4.2. That's REALLY hot with zero headroom and I don't hear additional audible clipping distortion streaming it on Napster. I can hear data compression artifacts but not additional clipping.

Is the whole thing about downstream clipping much ado about nothing? Just well intentioned internet chatter?

I'm really needing someone to convince me not to follow the lead of all these major MEs and just let it fly full scale... Help me...
Old 28th October 2018
  #7
Lives for gear
 
SmoothTone's Avatar
 

Quote:
Originally Posted by Trakworx View Post
Is the whole thing about downstream clipping much ado about nothing? Just well intentioned internet chatter?
It's not internet chatter in the way I think you mean - well, it's probably BECOME that - but it originates from a lot of the experimental work done by Thomas Lund which formed the basis of the ITU/EBU/ATSC standards that inform the current recommendations.

Here is a more recent summary which has links to all of the key technical documents in the appendix: http://www.aes.org/technical/documen...04_1_15_10.pdf

And here are a couple of examples of Lund's earlier work:

http://cdn-downloads.tcelectronic.co...3_overload.pdf

https://service.tcgroup.tc/media/Lev...109%281%29.pdf

Quote:
I'm really needing someone to convince me not to follow the lead of all these major MEs and just let it fly full scale... Help me...
I guess it's up to each of us to draw our own conclusions based on what the science says and what we (and our clients' audience) hear. I see the -0.1 guys as pragmatists. They're usually clipping converters and pushing FS because in the grand scheme of things it gets what they hear as the most musical results. And no one - other than a few vocal anti-loudness-war-campaigners - complains about it. Some rules are there to be broken.

For me, while my standard practice is to leave headroom and I tend to do my loudness processing in the digital domain with a combination of limiting and clipping, if there are times when driving the analogue chain hotter and just clipping the ADC gets more musical results for a given mix, I have no qualms in doing that.

And if all bets are off and someone wants super crazy loud, I like the fact that my usual 1dB of headroom gives me a whole dB of "free" gain (no additional distortion or squashing) to play with.

Last edited by SmoothTone; 28th October 2018 at 03:49 AM.. Reason: Typo
Old 28th October 2018
  #8
Lives for gear
 
Trakworx's Avatar
Thanks for the links! Yeah, I know that the science is real and there is clipping happening in the codecs for sure, but the question is - Does it matter?

Seems like it doesn't matter to the big name MEs. Or their clients. And I'm not hearing added distortion in my preliminary listening tests.

I exported some of the reference tracks as 128kbps MP3s and re-imported them. It's clear that as expected the MP3s are clipping more than the original waves but it's not audible, just like in Napster.

Like you, I don't clip my AD. Not usually. Today I experimented on a pop track I was mastering. Set the loudness where I wanted it with my limiter ceiling set to -0.5 as usual. Sounded good. Then I raised the ceiling by .4dB and lowered the amount of limiting by 0.4dB. I could hear an improvement in the punch with no loss in loudness. Not surprising, but it shows the plus side of having that extra .4dB to work with. I did the MP3 test on that track and got the same result as with the ref tracks; no audible distortion.

So I think the -.1 master may just be better. Same loudness, more punch, no audible clipping from the codec.

I know there are different codecs and different DACs out there so this is far from a comprehensive trial, but in light of everything I'm seeing and hearing, I'm definitely rethinking things... Hmmmm...
Old 28th October 2018
  #9
Lives for gear
 
SmoothTone's Avatar
 

Quote:
Originally Posted by Trakworx View Post
in light of everything I'm seeing and hearing, I'm definitely rethinking things... Hmmmm...
Hmmm indeed.

It's times like these I often ask myself, what would Dave Collins do? I respect the quality of his work and get the sense that his practices are based on thorough knowledge of the technical literature and detailed listening tests.

So I pulled up both the 16bit flac and mp3 downloads of Hardwired and sure enough, full scale signal with 1000s of overs. But it sounds great on my monitors and on a bluetooth speaker (except for that click of the kick beater which is a little OTT for my tastes).

I've posted this same query over at DC's forum so we'll see what those guys have to say.

Thanks for the food for thought.
Old 28th October 2018
  #10
Lives for gear
Quote:
Originally Posted by SmoothTone View Post
except for that click of the kick beater which is a little OTT for my tastes
lars clearly hasn't been able to hear 5k since at least 1990.
Old 28th October 2018
  #11
Gear Head
 
moostapha's Avatar
 

It's a much more difficult test, but...

Even when I can't reliably ABX a particular 320-mp3 with the wav, it at least feels like listening fatigue sets in earlier with compressed audio. Intuitively, it seems to make sense that something that's clipping worse would also increase (or accelerate) fatigue even if each clip is not immediately obvious.

There have been a few people writing/talking about that effect over the last decade or so, at least in passing. Tony Andrews (funktion one) comes to mind, as well as some producer guy on YT who's name I can't remember. I know Andrew Shcepps has given some very reasonable-sounding talks about the tradeoff.

I think it was Schepps who told a story about an intern he was working with who complained about being tired after a day of listening to compressed music and compared it to his own experience growing up when he worked all day in a studio and went home to put on more records without ever feeling fatigue.

I can't say I've done any scientific testing, but it seems like I suffer from that effect.

Has anyone here tested that? Is it possible that a tiny bit of headroom for the codecs at least reduces that effect? Or is it something that, if it's going to happen for a given person, happens regardless of what you do?
Old 28th October 2018
  #12
Lives for gear
 
Justin P.'s Avatar
 

Usually when I hear audible distortion from lossy codecs, it's not on the fast transient higher frequency stuff, but rather on things like cello and low piano notes that are pushed too loud.
Old 28th October 2018
  #13
Lives for gear
 
Trakworx's Avatar
Quote:
Originally Posted by Justin P. View Post
Usually when I hear audible distortion from lossy codecs, it's not on the fast transient higher frequency stuff, but rather on things like cello and low piano notes that are pushed too loud.
I'm thinking different peak headroom for different genres might be the way forward for me...

A Metal track can probably withstand a bit of codec clipping a lot better than a Folk track...
Old 28th October 2018
  #14
Lives for gear
 
Trakworx's Avatar
Quote:
Originally Posted by moostapha View Post
Has anyone here tested that? Is it possible that a tiny bit of headroom for the codecs at least reduces that effect? Or is it something that, if it's going to happen for a given person, happens regardless of what you do?
I haven't tested that but it seems plausible.

It could also just be the loudness/limiting and not the codec clipping that causes fatigue.

Or both...
Old 28th October 2018
  #15
Lives for gear
 
Justin P.'s Avatar
 

Quote:
Originally Posted by Trakworx View Post
I'm thinking different peak headroom for different genres might be the way forward for me...

A Metal track can probably withstand a bit of codec clipping a lot better than a Folk track...
Not a bad concept. The real eye opening one for me many years ago was a piece for the production library I do mastering for. It had some really deep low cello and piano notes and was pushed kind of loud overall. The poor 128kbps preview codec on the website at the time made it pretty ugly when streamed from the site.

However, most of the songs which are a various genres handled the 128kbps codec decent enough. Since then we've increased the headroom a touch.
Old 28th October 2018
  #16
Lives for gear
 
Trakworx's Avatar
Quote:
Originally Posted by Justin P. View Post
Not a bad concept. The real eye opening one for me many years ago was a piece for the production library I do mastering for. It had some really deep low cello and piano notes and was pushed kind of loud overall. The poor 128kbps preview codec on the website at the time made it pretty ugly when streamed from the site.

However, most of the songs which are a various genres handled the 128kbps codec decent enough. Since then we've increased the headroom a touch.
In general I've often found low and low mid sustained notes (piano, organ, cello, synth bass, etc) to cause distortion issues in limiters and clippers as well as codecs. Possibly worse when panned or spread in stereo. Sometimes I have to audition several limiters to avoid distortion in music like that...
Old 28th October 2018
  #17
Lives for gear
yep. low mid sustained notes are always the tell.
Old 29th October 2018
  #18
Lives for gear
Recently mastered a track that has a very loud, climatic final section, then the (rock) instruments all go away and the female singer closes the song with a low register solo note. Could hear zero distortion during the loud section, even with a tiny bit of clipping via Standard CLIP, but lo and behold, if one sample of the singer's last (dynamic) note touched the ceiling, the distortion was audible, terrible, and ruined the track. I tend to pick between no headroom and -1dBFS of headroom depending on genre and client expectations. After extensive testing, i've found that if I don't hear distortion while printing the track, I don't hear it when downsampling, even if peaks get pushed over 0db. I don't like afclip telling me I have 1000 overs though. Feels yucky. I've tested many commercial releases (ripped CD WAVs). ISP's are everywhere.
Old 29th October 2018
  #19
Lives for gear
 
Trakworx's Avatar
Quote:
Originally Posted by M Mastering View Post
After extensive testing, i've found that if I don't hear distortion while printing the track, I don't hear it when downsampling, even if peaks get pushed over 0db.
Interesting. Do you SRC down to 44.1 after limiting, or before?
Old 29th October 2018
  #20
Lives for gear
Quote:
Originally Posted by Trakworx View Post
Interesting. Do you SRC down to 44.1 after limiting, or before?
I guess I'm just saying, it's rare that I can hear momentary ISPs. Even though I know they are there. They may be effecting the audio in less perceivable ways to me, but I just don't hear horrible distortion usually. I try to avoid them because the theory is sound, and perhaps there may be a real world situation where they trip me up, so I err on the side of caution. I don't like ISPs as a general rule. But so many records I know and love are full of them. I think we make a far bigger deal about them here on GS than perhaps is needed. But like I said, the theory is sound.

As for SRC, it's the second to last step, I work at 32bit/96K throughout the process, then downsample to 48 and 44.1, then dither and bit reduction to final format.
Old 29th October 2018
  #21
Lives for gear
 
Trakworx's Avatar
Quote:
Originally Posted by M Mastering View Post
I guess I'm just saying, it's rare that I can hear momentary ISPs. Even though I know they are there. They may be effecting the audio in less perceivable ways to me, but I just don't hear horrible distortion usually. I try to avoid them because the theory is sound, and perhaps there may be a real world situation where they trip me up, so I err on the side of caution. I don't like ISPs as a general rule. But so many records I know and love are full of them. I think we make a far bigger deal about them here on GS than perhaps is needed. But like I said, the theory is sound.
Agreed.

Quote:
Originally Posted by M Mastering View Post
As for SRC, it's the second to last step, I work at 32bit/96K throughout the process, then downsample to 48 and 44.1, then dither and bit reduction to final format.
Cool. Sounds like you limit before SRC then? And still don't hear added distortion after SRC?

That's what I've been doing up til now, and it's been OK with my usual -0.5dB ceiling, but I'm seeing that if I decide to output some stuff at -0.1 then I'm forced to limit after SRC to avoid hitting zero...
Old 29th October 2018
  #22
Lives for gear
 
Apostolos Siopis's Avatar
 

Verified Member
Quote:
Originally Posted by Trakworx View Post

That's what I've been doing up til now, and it's been OK with my usual -0.5dB ceiling, but I'm seeing that if I decide to output some stuff at -0.1 then I'm forced to limit after SRC to avoid hitting zero...
I also limit before SRC but allow a generous peak headroom of 3dB for the SRC (ie my limiter's output or the master output is at -3dBFS)
I use FinalCD which produces overs of around .2 to 1 dB during its process, so the final file is peaking between -2dBFS and -2.8dBFS.
There have been tracks that peaked at -2.95dBFS after SRC but they are rare.
EDIT: just as I was typing, FinalCD outputed a file at -3dBFS. This file is a ballad and loudness was never a concern, so no clipping and the mix was so well crafted that even limiting is on the light side. (ie no square waves to finalCD)

I then import the 44.1Hz file and I either insert a bit perfect (or transparent???...please advice on my terminology) under the threshold limiter, to catch those peaks or I automate a volume envelope to raise the ceiling to -0.1dBFS

I never really lowered my ceiling, as I did the tests you described and decided to do whatever my "heroes"/"mentors" do, until further notice.

Last edited by Apostolos Siopis; 29th October 2018 at 09:30 PM..
Old 29th October 2018
  #23
Gear Maniac
 

maybe it's just my mytek 8x192-- but any album with straight up PCM clipping (non ISP transients that go above 0.0) is highly audible in a bad way over here. many mainstream releases are done this way and it utterly baffles my mind. i would WAY rather have 4 dBs of true peak limiting then 4 dBs of PCM clipping on my DAC. some incredibly notable, grammy award winning albums are mastered this way, and i find them mostly unlistenable.

that's not to say anything for peak headroom-- if it's 0.0, -0.1, or -1, i've never found there to be any audible difference-- perhaps a bit after mp3 conversion, but who cares, people who listen to mp3s are not concerned with barely audible non-ear piercing distortions. actually i see a lot of people here talk about -1 but rarely have i heard albums or singles that actually use such a low ceiling.

it's the PCM clipping on the DAC end that boggles my mind. i don't understand how you can pay someone the biggest bucks to literally just turn up the volume slider past red. why do many big-name ME's do this?? how could it possibly sound good on their super hi-end DACs??
Old 30th October 2018
  #24
Lives for gear
 
Trakworx's Avatar
Quote:
Originally Posted by dogoftears View Post
i don't understand how you can pay someone the biggest bucks to literally just turn up the volume slider past red. why do many big-name ME's do this??
I can't speak for them. I've only ever done that one time - where I just turned up the fader and let it digitally clip in the box instead of using a limiter. It was a Metal album many years ago (2001) and we decided digital clipping sounded better than limiting. It turned out to be a very popular album within it's music scene. Of course today's limiters are much improved over 2001's limiters...
Old 30th October 2018
  #25
Gear Maniac
 

Quote:
Originally Posted by Trakworx View Post
So I have a 16-44 PT session where a few years ago I ripped several reference tracks from CDs. I went in and checked the peak headroom.

They ALL peak at either -0.1dBfs or at 0dBfs.

-0.1dB: Wolfmother, The Black Crowes, Alice In Chains.

0dB: AC/DC, Soundgarden, Led Zeppelin, Queens of the Stone Age.

I just checked a few of the loudest of those tracks streaming on Napster using my iPhone and Sony MDR-V6 headphones (for their brightness so I could really hear any HF distortion).

I do not hear audible clipping distortion. At least no more than I hear in the ripped waves.

The ripped wave of "No One Knows" by Queens of the Stone Age peaks at 0dB and has a short term LUFS hitting up to -4.2. That's REALLY hot with zero headroom and I don't hear additional audible clipping distortion streaming it on Napster. I can hear data compression artifacts but not additional clipping.

Is the whole thing about downstream clipping much ado about nothing? Just well intentioned internet chatter?

I'm really needing someone to convince me not to follow the lead of all these major MEs and just let it fly full scale... Help me...

I have run into this issue as well Justin. As a matter of fact last night while mastering a song. Which LUFS and dBfs level is right? I've done extensive listening test on the same types of music and have concluded that most rock music (Foo Fighter/ QOTSA/ ect...) in the last 10+ years is short term -6 LUFS and peaks at -.01 or 0 dBfs. I've personally found it tough to avoid distortion reaching the -6 LUFS short term with anything less -.01 or 0 dBfs. Referencing the latest Chris Stapleton album I found his short term -9 LUFS and peaks -.01 dBfs. though this would fall into a different genre, but a noticeably different approach to the mastering level. All references were made off CD wav files.

Side note - I sent a song off to Sterling Sound to have them master it about a year ago for a rock band I was working with on a project. I wanted to hear what levels they used and compare it to the master I had done. The price was $175.00 (after hours) and the engineer mentored under Ted Jenson, so I thought it would be cool to give it a shot. Personally, I thought it sounded great and I could hear a positive difference. the levels were as we spoke about above for the genre. I dug it, and thought it was better than my master. So, I gave both to the band and didn't tell them which was which. They ended up using my master because it had more bottom end to the bass. The band leader was the bass player... go figure??
The Sterling master did not have clipped wave forms and looked similar to the wave forms ran through my Pendulum PL2 (thought could never get a PL2 to produce that hot of a master). I would assume they were using a software limiter. To be straight up honest... I think they have there own in house software developer that writes some of there software used for mastering.
Old 30th October 2018
  #26
Lives for gear
 
Trakworx's Avatar
Quote:
Originally Posted by scorpiwoman View Post
I would assume they were using a software limiter. To be straight up honest... I think they have there own in house software developer that writes some of there software used for mastering.
My friend who attended the Ted Jensen session told me he used all name brand software. Sequoia, Fabfilter, TC Electronics. And the room was amazing.
Old 30th October 2018
  #27
Gear Maniac
 

Quote:
Originally Posted by Trakworx View Post
My friend who attended the Ted Jensen session told me he used all name brand software. Sequoia, Fabfilter, TC Electronics. And the room was amazing.
Damn... Well I guess that theory is blown. Just a heads up I don't believe anyone has stepped foot on the moon. Just saying... so you know who you're dealing with here.
Old 30th October 2018
  #28
Lives for gear
 
Trakworx's Avatar
Quote:
Originally Posted by Apostolos Siopis View Post
I then import the 44.1Hz file and I either insert a bit perfect (or transparent???...please advice on my terminology) under the threshold limiter, to catch those peaks or I automate a volume envelope to raise the ceiling to -0.1dBFS
Can you elaborate on bit perfect/transparent limiter?

I have several limiters but I have no idea which ones would fit that definition, or maybe they all do...?
Old 31st October 2018
  #29
Quote:
Originally Posted by Trakworx View Post
Can you elaborate on bit perfect/transparent limiter?

I have several limiters but I have no idea which ones would fit that definition, or maybe they all do...?
A bit perfect/transparent limiter is a limiter that doesn't change anything to the PCM stream as long as the signal doesn't reach the threshold.
If there is oversampling for example, it's not bit perfect anymore.
Old 31st October 2018
  #30
Lives for gear
 
bcgood's Avatar
 

Verified Member
I mastered a track for a friend awhile ago and RMS hovers around -14 and peak is -0.4

This is the sweet spot for this track. It sounds so good

Quality over loudness is what I say
Topic:
Post Reply

Welcome to the Gearslutz Pro Audio Community!

Registration benefits include:
  • The ability to reply to and create new discussions
  • Access to members-only giveaways & competitions
  • Interact with VIP industry experts in our guest Q&As
  • Access to members-only sub forum discussions
  • Access to members-only Chat Room
  • Get INSTANT ACCESS to the world's best private pro audio Classifieds for only USD $20/year
  • Promote your eBay auctions and Reverb.com listings for free
  • Remove this message!
You need an account to post a reply. Create a username and password below and an account will be created and your post entered.


 
 
Slide to join now Processing…
Thread Tools
Search this Thread
Search this Thread:

Advanced Search
Forum Jump
Forum Jump