StandardCLIP

[audiotools]

Any chance of adding a numerical display of the eaten dB's above or below that reduction meter?

Good idea, I will think about it!

[widebluesound]

I'm loving this clipper - using it right before my software limiter just to shave some peaks so my limiter doesn't have to react as much.

A question: does the linear-phase filter add pre-ringing artifacts? Is there no chance that the minimal-phase filter sounds better in mastering situations?

And out of curiosity, as you've heard this plugin more than anyone else, what oversampling setting do you use?

Thanks!
Nathan Rightnour

[Hermetech Mastering]

I love it too, and use it the same way you do, just to trim the peaks 1-2dB before final limiting. I usually keep it on Linear Phase with the highest setting (4096 taps, I think) and 32x OS, (64x occasionally), cutoff at 93%, always Hard Clip mode. Best to try the different settings yourself and hear what sounds best on a given track. With these settings (not too much GR/clipping), it's insanely transparent most of the time, and lets the following limiter work better on the meat of the track.

[Hermetech Mastering]

One thing I do wonder about, and it's probably just because I don't understand the over-sampling theory very well, is that Hard Clipping just flatlines the waveform, right? And if that is indeed the case (I could be wrong...), wouldn't OS be unnecessary? I guess I should do some listening and 'peak viewing' tests to see what's going on. Hopefully audiotools can give us an answer.

And again a +1 to having a numerical display of the amount of clipping going on in dB, preferably real time AND with a peak hold to see the max amount of clipping that happens. You can work it out with maths (scanning waveform for peaks and adjusting StandardCLIP to suite etc, and I do this in RX5), but it would be great to have the readout in the plugin itself.

[aciddose]

Hard clipping produces an Nth order discontinuity by suddenly making the derivative zero. This results in ~1/N harmonics up to infinity.

It is possible to very efficiently anti-alias the transition of the derivative at various orders by applying a filter up to Nth order (3 is probably more than enough) which makes the harmonic content fall off so quickly that it won't produce much aliasing anymore and will allow only 2x over-sampling to work fine.

Unfortunately that requires estimating the exact position at which the discontinuity occurred which is expensive and very difficult to get right.

So typically over-sampling is the only option to reduce the level of aliasing resulting from clipping.

[audiotools]

Hi there,

A question: does the linear-phase filter add pre-ringing artifacts?

Like every digital or analog filter, the filters used in the oversampling unit will introduce ringing into the signal. But only if the source signal already contains high frequency elements near the nyquist signal. (The most provocative signal would be a one sample pulse surrounded by silence).

The filter will be applied twice to the signal, the first one to reconstruct the signal in the oversampled domain, and the second after the clipping before downsampling, to remove any high frequency above the source Nyquist frequency.

Important: Because the filter will applied at the oversampled domain, potential pre-ringing is also is oversampled -> shorter (In contrast to a normal FIR-EQ, which is not oversampled)

So if you use x64 oversampling with a filter of 4096 taps, the actual filter (from source-samplerate perspective) is just 64 samples long, therefor the pre-ringing part, is just 32 samples.

32samples at 44.100kHz are just 0,00073 seconds, and because of the reverse bell characteristic of the pre-ringing (very slow decrease in the beginning at a low level), only 4-5 samples before an extreme(!) peak signal (like the one sample pulse) will actual contain noticeable (not audible, only visually with a waveform editor) pre-ringing samples.

So, yes pre-ringing will be introduced, but its impossible to hear.

Is there no chance that the minimal-phase filter sounds better in mastering situations?

I implemented the minimal-phase mode, for situations where somebody does not want to introduce any latency to the signal (live, broadcasting ...). The linear phase filter is more effective to reduce aliasing, and doesn't introduce phase-shift.

One thing I do wonder about, and it's probably just because I don't understand the over-sampling theory very well, is that Hard Clipping just flatlines the waveform, right? And if that is indeed the case (I could be wrong...), wouldn't OS be unnecessary?

The problem is, what you see if you looking at the waveform, is not the "real" signal in the analog domain.

It is just a translation which only works, because the signal which was converted into a sampled domain, has contained no frequency above the Nyquist frequency. This is important, because thats the only way which allows to reconstruct the signal into the analog domain again.
Thats why in every A/D-Converter is a filter.
When you now begin to manipulate the signal (clip) without any oversampling, you introduce "impossible" frequency content (not re-constructable waveforms), which is actually aliasing.

Here is an example of audible aliasing:

[Hermetech Mastering]

Thank you aciddose and audiotools. Makes sense, will continue with my settings!

Another quick question if I may, with the filter cutoff, I usually leave it at 93%, (I understand having it too low may start to introduce the roll off into the audible band), but am wondering whether I should just leave it at 100%. Would this be a good idea, or would it take more CPU/be slower at rendering etc.? On my system in RX5 Advanced, x32 OS is about double real time rendering (not too bad for short tracks), and 64x OS takes about the same as real time (a nightmare for longer tracks). I'm happy with the results thus far, and wouldn't want things to be any slower!

[Subbase]

what would be some good starting points for this on Drum bus,and bass buss on a DnB mix?

[Armagibbon]

what would be some good starting points for this on Drum bus,and bass buss on a DnB mix?

Don't clip your sub unless you want that kinda distortion man. And you prolly don't.
Personally I use a clipper just to trim high freq peaks like claps snares and some disto plucks and whatever. Gotta use EQ and compression instead for anything else.

You're gonna use it on a bass bus? Might be a good idea to split, run a highpass on the clipped freq and lowpass on the clean. If you're gonna do any dynamic work on your lows, use a compressor instead.

[FatGuy]

Hi there,
...

I noticed that Gold Clip doesn’t have sample overshoots when OS is enabled, which is a significant advantage. Could you consider implementing this feature as well?

[Johnnyjohn]

A modernized UI would be pretty pleasant

[ampetrosillo]

what would be some good starting points for this on Drum bus,and bass buss on a DnB mix?

Don't clip your sub unless you want that kinda distortion man. And you prolly don't.
Personally I use a clipper just to trim high freq peaks like claps snares and some disto plucks and whatever. Gotta use EQ and compression instead for anything else.

You're gonna use it on a bass bus? Might be a good idea to split, run a highpass on the clipped freq and lowpass on the clean. If you're gonna do any dynamic work on your lows, use a compressor instead.

Why? The threshold of audibility of distortion in the bass frequencies is much higher than in the midband and treble. As long as you don't overdo it, as intermodulation distortion down below is, on the contrary, nasty.