what is soft clipping vs hard clipping
8080JP
Thread Starter
#1
5th May 2010
Old 5th May 2010
  #1
Gear Head
 

Thread Starter
what is soft clipping vs hard clipping

This is what I thought.

Clipping is when the music goes into the red.

I listen to a commercial dance track and the right channel is always louder than the left and it is often going about .1 or .2 into the read.

It actualy gives the sound some kind of bite for this commercial track.
Very cleverly done.

Does soft clipping go into the red?

When using a limiter set at say -0.1 and the music is gained quite substantailly the music will have a falt top but not go into the red.

So what is this soft clipping.
Is it a gentle way of going into the red?

And is Hard clipping going into the red?
#2
5th May 2010
Old 5th May 2010
  #2
Gear interested
 

Neither is going into the red, per se - that's the whole idea of clipping.

Think of your waveform as an actual piece of paper - and then take a scissor and chop off the tiny peaks of every waveform that passes a certain level. That would be the equivalent of hard clipping - literally your peaks are cut off so that nothing (optimally) passes your pre-set ceiling, such as 0dbfs.

A soft clipper is like a limiter but with a knee, so the clipping is eased into. I suppose this would give you a clipped signal with slightly rounded edges at the cutt-off points, rather then the right-angle edges of a hard-clipper.

All of this does not take into account intersample peaks . . .

FWIW - the soft-clipper on the TC MD3 starts grabbing up to 6db below your ceiling and then there is still a limiter stage after the soft-clipper.
#3
5th May 2010
Old 5th May 2010
  #3
Gear interested
 

Quote:
Originally Posted by dropmasters View Post
Neither is going into the red, per se - that's the whole idea of clipping.

Think of your waveform as an actual piece of paper - and then take a scissor and chop off the tiny peaks of every waveform that passes a certain level. That would be the equivalent of hard clipping - literally your peaks are cut off so that nothing (optimally) passes your pre-set ceiling, such as 0dbfs.

A soft clipper is like a limiter but with a knee, so the clipping is eased into. I suppose this would give you a clipped signal with slightly rounded edges at the cutt-off points, rather then the right-angle edges of a hard-clipper.

All of this does not take into account intersample peaks . . .

FWIW - the soft-clipper on the TC MD3 starts grabbing up to 6db below your ceiling and then there is still a limiter stage after the soft-clipper.
Are modern commercial tracks going into the red?
Is that what modern clipping is in that context and not what you described - is the word clipping used for 2 diffrent scenarios?
#4
5th May 2010
Old 5th May 2010
  #4
Lives for gear
 



The bottom two are 2x-amplified versions of the top one. [EDIT] The hard one is 2x; the soft one is 4x.

Last edited by Brainchild; 5th May 2010 at 08:54 PM.. Reason: I lied.
adl
#5
5th May 2010
Old 5th May 2010
  #5
adl
Gear addict
 
adl's Avatar
 

And what´s the difference between a regular limiter and a softclipper?

It´s even softer to the peaks, right? Meaning it compresses the peaks instead of chopping em off (what the softclipper still does).

Is that correct?
#6
5th May 2010
Old 5th May 2010
  #6
Gear interested
 

Quote:
Originally Posted by adl View Post
And what´s the difference between a regular limiter and a softclipper?

It´s even softer to the peaks, right? Meaning it compresses the peaks instead of chopping em off (what the softclipper still does).

Is that correct?

Yeah, as per the excellent illustration above, both soft and hard clippers CLIP the peaks above the ceiling. Nothing passes the ceiling.

With a soft/hard clipping limiter, you are controlling the gain level at the output stage to ensure that the file itself does not contain any audio over (typically) 0dbfs. That it, you clip it yourself to ensure that it doesn't clip unpredictably when someone is listening to it. That person could still turn this fileup and clip their D/A however. . .

A softclipper uses "compression" through the transition zone - hence the rounded shape - but still clamps down with infinite gain reduction as the ceiling/threshold is reached.

Commercial music that is "clipping" is music that has been clipped - it should not actually produce overs.
24-96 Mastering
Verified Member
#7
5th May 2010
Old 5th May 2010
  #7
Lives for gear
 
24-96 Mastering's Avatar
 

Verified Member
Quote:
Originally Posted by adl View Post
And what´s the difference between a regular limiter and a softclipper?
A brickwall limiter has a release time constant, i.e. a time-curve/slope going back out of gain reduction.

In contrast, gain reduction in clipping (soft and hard) acts on amplitude, i.e. level alone. No release time.

The differences in sound can hardly be described in few words, since different limiters produce different sounding artifacts and since the 'sound of clipping' as well as that of limiting depends a lot on the nature of the incoming material. Best have a play yourself and come up with your own generalizations.
#8
6th May 2010
Old 6th May 2010
  #8
Lives for gear
 

#9
6th May 2010
Old 6th May 2010
  #9
Lives for gear
 

Because I was bored and thinking maybe the intention of the question hadn't been fully answered...

Let's say a line-level signal source is feeding into a high-power PA amplifier.

Say the line-level output's peaks are 10 volts.

Say the amplifier's gain is +26dB. That means it will amplify an incoming voltage by a factor of 20. Also, say its maximum peak voltage output capacity is 112 volts.

Now, 10 volts times 20 equals 200 volts. But the amplifier can only put out 112 volts. The original signal will clip the amplifier's outputs (it will be "in the red").

So, in order to protect the amplifier from being driven into uncontrolled clipping, we limit the line-level signal before it enters the amplifier.
Say we hard-limit the original signal at a threshold of half its original amplitude, or -6dB relative to the original.
Now the peak voltage going into the amplifier is only 5 volts. Five times 20 is 100, so we have 12 volts (or 1 dB) of headrooom.

If the original signal is just fed into the power amp, then its output is inadequate by 88 volts. The output is clipped.

From an audio standpoint, the limited-line-level output and clipped-power-output waveforms will sound basically the same. But the amplifier is super-pissed at you in the clipped scenario. It's basically a matter of definition. They're both "clipping," but one is smart and one is dumb.

Soft clipping is just a variation on the hard limiting scenario. The maximum levels are the same, but the waveform is different (and probably less offensive).



----
Pertaining to the article flatfinger linked to...

A sine wave has a single frequency:



The image below shows the sine wave soft-clipped once with the transfer function:
OUTPUT = INPUT * (1 - ((ABSOLUTE VALUE OF INPUT) * 0.5)) and then re-amplified to its original amplitude. It adds a bunch of odd harmonics of the original signal:



The image below shows the result of several more iterative applications of the same function; the harmonic content increases in strength:



Gradually, the function ends up turning any single-frequency input waveform into a square wave.
8080JP
Thread Starter
#10
6th May 2010
Old 6th May 2010
  #10
Gear Head
 

Thread Starter
Quote:
Originally Posted by dropmasters View Post
With a soft/hard clipping limiter, you are controlling the gain level at the output stage to ensure that the file itself does not contain any audio over (typically) 0dbfs. That it, you clip it yourself to ensure that it doesn't clip unpredictably when someone is listening to it. That person could still turn this fileup and clip their D/A however. . .
You see why its unneccesarily complicated for beginners.

What you just said implies that the word clipping is used for both cases

1) A limiter not letting smashed loud music past a point, say -0.1, but because it is smashed it has flat lines at the top of the waveform.

2) When you say the the words clip the D/A you are implying that the waveform has gone into the red.

Therefore when someone on Gearlslutz says that some modern track sounds awful because it has been smashed and is clipping.

I dont know whether 1) or 2) is what is being described.
#11
6th May 2010
Old 6th May 2010
  #11
Lives for gear
 
Jesse Graffam's Avatar
 

Quote:
Originally Posted by Brainchild View Post
Gradually, the function ends up turning any single-frequency input waveform into a square wave.
add oversampling and filtering multiple bands of clipping, and you're started on your way to a clipper that doesn't suck nearly as hard.

btw, this approach was pretty much perfected as good as it probably ever will be in analog form, by 1985. most DSP designers have a ways to go.
#12
6th May 2010
Old 6th May 2010
  #12
Lives for gear
 

Quote:
Originally Posted by 8080JP View Post
...
What you just said implies that the word clipping is used for both cases
...
Therefore when someone on Gearlslutz says that some modern track sounds awful because it has been smashed and is clipping.

I dont know whether 1) or 2) is what is being described.
It doesn't matter. They sound the same (given the same degree of clipping; and adjusting for noise, distortion and other nonlinearities introduced by instrument abuse).

Quote:
Originally Posted by Jesse Graffam View Post
...
btw, this approach was pretty much perfected as good as it probably ever will be in analog form, by 1985.
Could you point out good soft-clipping analog stuff? I have been doing some obsessive researching and haven't come up with much...and yeah...the digital approximations don't really cut it for me.
#13
7th May 2010
Old 7th May 2010
  #13
Lives for gear
 
Jesse Graffam's Avatar
 

Quote:
Originally Posted by Brainchild View Post
Could you point out good soft-clipping analog stuff? I have been doing some obsessive researching and haven't come up with much...and yeah...the digital approximations don't really cut it for me.
I wouldn't really describe it as soft clipping, because the distortion is easier to filter when it's harder and easier to prevent audible distortion. Anyways, check out the Orban Optimod 8100A1 & XT2 chassis. Or the Optimod 9100A1 some people think is an even better/cleaner design. You should be able to find schematics around the ol' net. It does have some rudimentary circuits to handle prevention of audible distortion based on masking (or rather, the lack there-of).

Also, if digital approximations of this design don't cut it for you (which would still probably be more advanced than anything you've tried), then there are actually much more advanced digital clippers than that too. None that have been marketed specifically to the mastering community... yet. I know of at least 2 that should be available this year tho. One as software which is at least on par with the mid 80s Optimod design, and one in a box that's the most advanced I know of which I've had the extreme pleasure of using over the last 3 years.
#14
7th May 2010
Old 7th May 2010
  #14
Let's Hear It For Jean Baptiste Fourier!

Quote:
Originally Posted by Brainchild View Post
Because I was bored and thinking maybe the intention of the question hadn't been fully answered...

Let's say a line-level signal source is feeding into a high-power PA amplifier.

Say the line-level output's peaks are 10 volts.

Say the amplifier's gain is +26dB. That means it will amplify an incoming voltage by a factor of 20. Also, say its maximum peak voltage output capacity is 112 volts.

Now, 10 volts times 20 equals 200 volts. But the amplifier can only put out 112 volts. The original signal will clip the amplifier's outputs (it will be "in the red").

So, in order to protect the amplifier from being driven into uncontrolled clipping, we limit the line-level signal before it enters the amplifier.
Say we hard-limit the original signal at a threshold of half its original amplitude, or -6dB relative to the original.
Now the peak voltage going into the amplifier is only 5 volts. Five times 20 is 100, so we have 12 volts (or 1 dB) of headrooom.

If the original signal is just fed into the power amp, then its output is inadequate by 88 volts. The output is clipped.

From an audio standpoint, the limited-line-level output and clipped-power-output waveforms will sound basically the same. But the amplifier is super-pissed at you in the clipped scenario. It's basically a matter of definition. They're both "clipping," but one is smart and one is dumb.

Soft clipping is just a variation on the hard limiting scenario. The maximum levels are the same, but the waveform is different (and probably less offensive).



----
Pertaining to the article flatfinger linked to...

A sine wave has a single frequency:



The image below shows the sine wave soft-clipped once with the transfer function:
OUTPUT = INPUT * (1 - ((ABSOLUTE VALUE OF INPUT) * 0.5)) and then re-amplified to its original amplitude. It adds a bunch of odd harmonics of the original signal:



The image below shows the result of several more iterative applications of the same function; the harmonic content increases in strength:



Gradually, the function ends up turning any single-frequency input waveform into a square wave.
Brainchild: Nice work. It's nice to see someone using Fourier analysis to explain the relationship between the signal in the time and frequency domains, and just how the clipping manisfests itself in terms of harmonics. Have you thought about cooking up some .wav files with these artifacts and posting them here?

Don't get me wrong, I love that you used some FFT techniques and it gets the point across, but if you had some audio samples of these cases, that too would help cement the illustration for a lot of folks.

Jean-Baptiste Fourier...Mathametician, snappy dresser...and father of (among other things) the Fourier series. While we're at it, raise your glasses to Cooley and Tookey, who inspired by Jean-Baptiste Fourier gave us the DFT and then the FFT (circa 1967). Gentlemen!
#15
7th May 2010
Old 7th May 2010
  #15
Lives for gear
 

OK so here's some samples of an original sine wave at 100Hz, and then the algorithm applied progressively from one to seven times.

Attached Files
File Type: wav sine wave soft clipping-1.wav (258.4 KB, 277 views) File Type: wav sine wave soft clipping-2.wav (258.5 KB, 134 views) File Type: wav sine wave soft clipping-3.wav (258.5 KB, 112 views) File Type: wav sine wave soft clipping-4.wav (258.5 KB, 97 views) File Type: wav sine wave soft clipping-5.wav (258.5 KB, 62 views) File Type: wav sine wave soft clipping-6.wav (258.5 KB, 60 views) File Type: wav sine wave soft clipping-7.wav (258.5 KB, 78 views) File Type: wav sine wave soft clipping-8.wav (258.5 KB, 201 views)
#16
10th May 2010
Old 10th May 2010
  #16
Lives for gear
 
polybonk's Avatar
 

#17
28th May 2010
Old 28th May 2010
  #17
Gear interested
 

Quote:
Originally Posted by Jesse Graffam View Post
Also, if digital approximations of this design don't cut it for you (which would still probably be more advanced than anything you've tried), then there are actually much more advanced digital clippers than that too. None that have been marketed specifically to the mastering community... yet. I know of at least 2 that should be available this year tho. One as software which is at least on par with the mid 80s Optimod design, and one in a box that's the most advanced I know of which I've had the extreme pleasure of using over the last 3 years.
What are the names of the software you mention and the one in the box? Thanks
#18
30th May 2010
Old 30th May 2010
  #18
Lives for gear
 
Jesse Graffam's Avatar
 

Quote:
Originally Posted by inhahe View Post
What are the names of the software you mention and the one in the box? Thanks
The software one I was talking about is Slate Digital's up coming clipper (and it also does other things).

The box is something I'm directly involved with, so when the time is right for the marketing people, you guys will hear about it one time from myself, the rest is up to you. I'm not much for spam.
Thread Tools
Search this Thread
Search this Thread:

Advanced Search
Similar Threads
Thread
Thread Starter / Forum
Replies
DP40oz / So much gear, so little time!
29
90Brownie / Rap + Hip Hop engineering & production
5
Outlaw Hans / So much gear, so little time!
0
luctellier / Music Computers
7
luctellier / Music Computers
10

Forum Jump
 
Register FAQ Search Today's Posts Mark Forums Read

SEO by vBSEO ©2011, Crawlability, Inc.