Skip to main content ↑ ↓

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: Waveform Plots Considered Harmful (Read 24798 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Waveform Plots Considered Harmful

Reply #25 –
Quote from: Axon on
OTOH, clipped transients - in the limiting case, something like a single clipped sinus - are relatively inaudible, because the transient content is so smeared in the frequency domain to begin with.

The extremely brief moment of actual clipping isn't the main problem. The major part of the audio vandalism is the extreme levels of compression that were applied before the mastering engineer finally had to resort to allowing a bit of clipping to squeeze that last little bit of loudness out.

If that 1000-sample long rim-shot has been pulled down by 9dB, the fact that the initial 10 or so samples have also been clipped is just a pimple on a boil.

Waveform Plots Considered Harmful

Reply #26 –
Quote from: Lyx on
Here's a different example. We are compressing the night sky. Yes, we are. By outputting vast amounts of light into the sky, we in most urbanized countries have made almost the entire sky disappear. Only a low amount of handful of stars are still visible. Dozens of sky objects, which in the past were something normal and well known, are now unknown. This is something, which you CAN directly show another person. It's an issue which - similiar to in this community - is decried by lots of enthusiasts. Does your average joe care? Does he even consciously look at above anymore? Do you think the kids born in the last 15 years even have a sufficient comparision anymore? If you tell them about various sky objects, they react as if you're talking about aliens. However, asume the media would create a new nightsky-hype, thus suddenly making average joe be interested in it again. So all those buy googles and stuff - and one of the first enemies they will get to know - the biggest problem nowadays for skygazers - will be light pollution. And suddenly, there would be popular interest in fixing that (there are simple ways to fix it, and it even saves cash). But do you think anything like that will happen out of average joe by himself? Why would someone, who in the past never cared about the topic, suddenly be interested in it, by pure coincidence at the same time as the media report about it?


That is a fantastic example.

Edit: I'm looking at your blog post Axon. In that big clipped peak that you show the screenshot of, it seems to me as though the clipping in the vinyl master ends earlier than in the CD version. Do you have an explanation for that? The master looks similar to me, but not quite identical...

Waveform Plots Considered Harmful

Reply #27 –
The waveform on the vinyl trace has been differentiated by DC blocking capacitors or high-pass filtering at some stage. This will have definitely happened in the magnetic preamp on playback, and I guess that the same will have applied during the cutting process due to the application of a high-pass filter somewhere in the signal chain. The output of a magnetic cartridge is also proportional to rate of change(ish). In short, vinyl doesn't go down to DC.

The effect should be negligible within the normal audio passband with the possible exception of very low frequencies approximating a square-wave unless some bleeding eejit has allowed the waveform to clip, at which point the signal is temporarily stuck at one extreme or the other and becomes a short-term DC level which will be reproduced as an exponential decay towards zero.

The relatively rapid rate of change at the leading edge of this decay adds harmonic distortion several orders of magnitude higher than would have normally occurred had the clipping not taken place. You will also see the same differentiation on your headphone or loudspeaker outputs regardless of whether the source is vinyl or digital unless the entire system is DC coupled from source to output which is probably unlikely.

This is yet another fine example of why a recorded waveform should never be allowed to clip if it's in any way avoidable.

Cheers, Slipstreem. 

Waveform Plots Considered Harmful

Reply #28 –
Quote from: Canar on
Edit: I'm looking at your blog post Axon. In that big clipped peak that you show the screenshot of, it seems to me as though the clipping in the vinyl master ends earlier than in the CD version. Do you have an explanation for that? The master looks similar to me, but not quite identical...


Bah. I was just wondering about the same thing.

I'm playing around with a new technique for clip detection (well, it's new to me anyway - I hope to post the basic idea soon), and I'm pretty consistently seeing that, for this album, clipping starts occurring about half a millisecond later on the vinyl than on the CD, and it stops about half a millisecond earlier.

The only explanation I can offer is that clipping is occurring on the vinyl at a different, higher level than on the CD. Mistracking, or resonances in the vinyl playback system, would certainly not cause behavior like this. I do have some like HF resonances with my cart, but that doesn't explain why the clipping both starts later and stops sooner.

That, if true, would considerably weaken my point that Mirrored is sourced from exactly the same master as the CD. But the more fundamental point - that vinyl isn't necessarily free of clipping - is still upheld.

Waveform Plots Considered Harmful

Reply #29 –
Quote from: Axon on
I'm playing around with a new technique for clip detection (well, it's new to me anyway - I hope to post the basic idea soon),



Cliveb's Wave Repair has a very good user-configurable clip detector.  It can be set to detect clips of arbitrary consecutive sample length, at an arbitrary level at or below 0dB (and the level can be a range)

http://www.waverepair.com

Waveform Plots Considered Harmful

Reply #31 –
Quote from: Axon on
... also, you know, for a psychoacoustic result that is so well known, it's damn hard to find information on it online. I think I spent half an hour looking for documentation on this sort of thing, and the only reason I didn't search for a longer time was that I've done this search before and came up relatively empty-handed.
       
        Any additional information you can provide on the audibility of clipping/limiting would be greatly appreciated.
I don't have that. I do have a reference for the audibility of gaps in noise, which (at the limit) is what clipping of over compressed pop music comes pretty close to...

Penner, M. J. (1977).
Detection of temporal gaps in noise as a measure of the decay of auditory sensation.
Journal of the Acoustical Society of America, vol. 61, no. 2, Feb., pp. 552-557.

...I mentioned that because the threshold is quite high - it's 2-3ms. If you look at the gap in an audio editor, you would never believe it was inaudible, yet it is.

That experiment is far simpler than clipping though - because the surrounding audio is spectrally white and completely random. Most CDs aren't like that!  With clipping, there are audible effects that make much shorter durations audible. TBH, I think discussing the duration as if it has some kind of correlation with audibility is silly - clearly you need a full psychoacoustic model to judge whether the clipping is audible or not, as you have to take account of gap detection, harmonic distortion, and level change - all wrt known psychoacoustic thresholds. This isn't a trivial task, but it's not impossible.

Cheers,
David.

Waveform Plots Considered Harmful

Reply #32 –
I think this is a great thread. I don't have time to reply properly, but...

I think lyx's description of the world is spot on...

http://www.hydrogenaudio.org/forums/index....st&p=587368
and
http://www.hydrogenaudio.org/forums/index....st&p=587425

...and of course the commercial world will always be driven this way. The tragedy* is that much of the music that's damaged in the process will have value beyond it's time, and beyond it's target audience - this value is dramatically diminished if all we have are clipressed versions.

Of course clipression is just part of today's sound. Just like, in the 1960s, 4th generation dubs played back on a Dansette record player were part of that era's sound. However, lots of music from the 1960s has transcended that, and is still listened to today. However, the music that only exists in that form fairs worse than the music where a 4-track master tape still exists, and the recording can sound decent on CD.

I visited a friend the other day, and he'd just spent some serious money on a new stereo. He was complaining that he couldn't listen to half of his CDs any more! He loved lots of indie music, but to his surprise, even CDs from "rich" bands (e.g. Oasis) sounded terrible. I think, as the clipression generation grows up, and starts to listen to "their" music more on decent stereos than on their mobile phone's built-in "speaker" there will be a commercial imperative to re-master these recordings.

However, just as with the 1960s, some of the masters won't be there to re-master, and some of the artists won't allow it to happen.


I think any clipping should be avoided (unless it's for artist purposes - and then not normally on the whole mix!) because, as someone else has said, it's an undesirable process.

How bad the clipping and overcompression sounds is purely subjective - as we rarely have a "good" and clipressed version to compare, it's not like we can ABX or use a standard psychoacoustic model to detect the difference - we just have to listen to what we have, and say if it sounds pleasing to us.

Axon is correct in that the pictures don't correlate perfectly with what we hear - but it would be nice for both the pictures and my ears to tell me that there are no problems of this kind at all!


Cheers,
David.

P.S. *(the almost criminal thing is that many public institutions, which are supposedly there to guard against the almost all consuming power of what lyx describes, are now complicit with it, or are even encouraging it. Will those who care make them change direction, or simply tear down such institutions because they believe they have become worthless? That's a bit deep for HA!).

Waveform Plots Considered Harmful

Reply #33 –
I think that's why ReplayGain (or something like it) should be present in every consumer playback device (or as near to it as makes no odds), and turned on by default.  And when it's turned off, it should have a fixed 20dB attenuator, to stop "ReplayGain off" meaning "louder" to Joe Sixpack.

Then we'll have an end to the Loudness War, and instead we might get a return to quality metrics.

Apple's already half-way there, but I'm not sure they fully understand the problem, and they're not "every consumer device".

Waveform Plots Considered Harmful

Reply #34 –
Apple is indeed - intentionally or not - currently the most powerful force in making the loudness war pointless. However, it is IMO indeed  only halfway there - because in addition to something like RG (soundcheck) it also needs to offer *dynamics compression*. It needs to put the compressor-choice completely into the hands of the consumer.

WTF, why's that, isn't compression == bad, and noncompression == good? Well, tell that anyone who wants to listen to dynamic content in locations with really high noisefloors. Those are situations which are intuitively associated with the iPod, and therefore - from a music producer POV - neutralize the soundcheck effect. Right now, there is still a "point" in compressing music regarding the iPod, which is the use in high-noisefloor environments. If you build a compressor right into the player, that point vanishes as well and there is no argument left anymore for compressing music for ipods (from a music production perspective).

Also, i think a good default amping for no-RG mode would be -10dB.

Quote
Will those who care make them change direction, or simply tear down such institutions because they believe they have become worthless? That's a bit deep for HA!

In such scenarios, there is a third option which is rarely noticed: make the existing institution irrelevant by circumventing it with a self-created alternative.
I am arrogant and I can afford it because I deliver.

Waveform Plots Considered Harmful

Reply #35 –
Switchable compressors are already built into some consumer equipment - I've seen it on a 1990s car radio (as a "loud" switch) and recent home theatre receivers (as a "night mode" switch).  I agree, it would not be difficult for Apple to include one in iTunes and the various iPods.

Actually, I should check to make sure "Sound Check" doesn't already include one.  That's always a possibility.  I'm also not sure what "Sound Enhancer" in iTunes does.

IMO, however, most modern music is naturally not so dynamic as to impede walkabout listening.  There are exceptions, of course, and a switchable compressor would definitely help with classical and other dynamic music.

I don't think -10dB is enough to prevent an ultra-hot master from sounding louder than if it were RG'd.  Perhaps -15dB would work though.  Some ultra-hot masters are showing up near -14dB and certainly over -13dB.  We do need to stop people from thinking RG is a "lame" "quiet" button.  I don't think the preamp boost with RG on is a good idea, but a preamp cut with it off is.

Waveform Plots Considered Harmful

Reply #36 –
And you think that people will switch RG off, because SOME songs sound 4db louder with RG off?

Have you investigated the downsides (cost) of generally reducing the signal significantly? Have you tried to asume the POV of someone who prefers RG off, and make that option feasible for him, or are you only focussed on "making non-RG look bad"? If you asume a totalitarian POV, you can as well force RG, and remove the "off"-option.... no use for a useless option. If you include that option, you need to make it useful and feasible - else there's no point in including it.

P.S.: You might be interested in this:
http://www.hydrogenaudio.org/forums/index....mp;#entry573516
I am arrogant and I can afford it because I deliver.

Waveform Plots Considered Harmful

Reply #37 –
Well, Sound Check isn't a compressor, just a level shifter.  I checked this on both my iPhone and my MBP.  So it's basically ReplayGain with a pretty face - which is no bad thing.

I'm still not sure what Sound Enhancer does, but it isn't compressing either.  I've simply turned it off again.  I think it might have a short-term expander and/or a stereo expander, but I really can't be bothered poking at it.

The above determined using a 17.5 minute version of Blue Room, which has a long-term dynamic range of about 40dB.  It would have been very easy to see compression used on that.

Waveform Plots Considered Harmful

Reply #38 –
Quote from: Lyx on
And you think that people will switch RG off, because SOME songs sound 4db louder with RG off?

Yes.  That's what the Loudness War is all about anyway - they wouldn't do it if it didn't sell.

One of the interesting things about RG, as it happens, is that it allows the mastering engineer to pass on his level setting.  So we might see digitally-distributed tracks coming straight from the shop with RG tags on them.  This is both good and bad:

- Responsible engineers will set a sensible level, allowing plenty of dynamics, and match their RG tag to it.  This would be excellent news for audiophiles.

- Mediocre engineers will keep doing what they usually do, probably without RG - or they'll use a generic RG tagger after the fact.  This is okay.

- Bad engineers will game the system to produce tracks that are still "hot" on an RG-compliant device, by setting the RG to +0.0 (or at least to a higher numeric value than justified).  This would be catastrophic.

So it is clearly necessary to define proper standards for this kind of thing.  Standards *do* work - notice how consistent movie soundtrack recording is, and how DRM-on-CD has mostly failed to catch on, since Philips forbade DRM-CDs from using the CDDA trademark?

The question is, who gets to make the standards and enforce the trademarks?

Waveform Plots Considered Harmful

Reply #39 –
Quote from: Chromatix on
Quote from: Lyx on

And you think that people will switch RG off, because SOME songs sound 4db louder with RG off?

Yes.  That's what the Loudness War is all about anyway - they wouldn't do it if it didn't sell.


Wait, wait - now your are confused. First you were talking about "reducing amping", so that SOME LISTENERS don't turn RG off for the sake of loudness. Now you are talking about "reducing amping", so that RECORD PRODUCERS dont overcompress the source material.

Make up your mind. Which of the two is the DIRECT purpose for reducing amping when RG is set to off? If you asume that both are the same, then think twice.

- Lyx

P.S.: Also, dont simply asume that record labels act logically. Decision makers in the industry are primarily driven by beliefs, greed and fear. Each of those isn't necessarily identical to actual profits. (*cough* DRM *cough*)
I am arrogant and I can afford it because I deliver.

Waveform Plots Considered Harmful

Reply #40 –
Isn't a more accurate analogy that we are reducing the signal to noise ratio of the sky by allowing light pollution?
We have raised the noise floor to the point where much of the signal is lost.

If we were compressing the sky then all the stars would be approaching the same brightness.

The sun, BTW, appears to be clipping.
Creature of habit.