Hi! What usually causes intersample peaks? And how can i get rid of it (without using ozone)?

The nature of digital audio, really. The "peaks" you see in a digital signal do not represent the true peaks. Both do not correlate.

The Shannon/Nyquist theorem explains such a signal path:

analogue signal -> band-limiting (low pass) -> digital storage -> band-limiting (low pass) -> analogue signal

You can't cut half of the theorem ("visual inspection of the digital storage") and expect useful results. Audio is what comes out of the DA. The storage form is a highly compressed version of the analogue waveform, visual inspection is almost pointless. Thus, max-peak of the digital signal doesn't give any reliable info about the true max peak.

A. Headroom.

B. Oversampled limiting gives better control over the true waveform (which obviously sounds more musical as well ).

C. Slower (releasing) forms of dynamics control generally work better.

Thanks for the helpful answer Fabien! Is that why a lot of mastering engineers lower the final output to -0.3db to prevent intersample peaks?

A good True Peak limiter will prevent ISPs, and provide precise control of the output level.

NuGen Audio ISL and Flux Elixir are both excellent True Peak limiters.

Leaving a small amount of headroom in the master will prevent clipping and ISPs when resampling or converting formats.

Thank you Simon, i'll check them out!

Voxengo Elephant limiter stops 99.9% of ISPs when set to 4x over-sampling, according to my two most sensitive ISP meters (RME Digicheck and iZotope RX3).

Good news! I'm a fan of Elephant, but it still would be nice if there was a technique get rid of ISPs without using a limiter.

Yes, indeed. I'd also like to know if there are any other techniques besides enough headroom.

I don't think so, you are talking about the level of a track, which is exactly what limiters are designed to work on! You don't have to smash it, 0.0dB input, 0.0dB output (or -0.3dB for the conservative), with conservative levels to begin with, and oversampling switched on, and there you have it! If you're talking about tracking/recording or mixing, then you shouldn't really be anywhere near 0.0dB FS anyway.

Ok i see. Do anyone know how well the L2 stops ISPs?

EDIT: Tested it myself, didn't do well..

work at 192k or higher and never downsample

Just don't go too near zero (tracking, mixing or mastering), and you'll never have to worry about it.

Also, don't be too strict with the matter. Analogue audio has no on/off clipping behaviour, it can perfectly tolerate short overloads. IMHO, the great benefit about "true waveform" dynamics control is not really the ISP outcome, but the better sonic result. All non-oversampled limiters produce incredible amounts of aliasing both directly and most of all in the control path. In fact, they are radically undersampled if they don't run at least at 4-8 times the audio sample rate.

Really, don't worry too much about ISP. Analogue can handle them without much surprises.

You would be surprised how many commercial releases have ISPs.

L2 doesn't stop ISPs. Using Ozone's prevent option seems to add deeper limiting to the signal. My best advice would be to man up and drop the ceiling to a point where there is enough room to prevent them from hitting full scale.

Or use Elephant's EL-4 algorithm with 8 times oversampling. If you like, you can set the ceiling to 0.00 and get no overs and no ISPs, no matter how hard you hit it.

I wouldn't, cos I see them all the time!