Best ISP Limiters?

[Ploki]

We all love our saturators and clippers. But, try something... try a mix with no saturation or clipping tools. Clean analogue emulations are ok but avoid any heavily harmonically distorted tools. Or open an existing mix and turn off all the saturation and clipping and tweek the mix. See what it sounds like and A/B with the last mix you approved with all those tools engaged. Just try it. You might find it interesting.

i've been mixing more with FabFilter lately so saturation is mostly done deliberately. It's a more conscious decision and sometimes it ends up... sweeter

[mystran]

LoudMax with ISP off doesn't have to hard clip. Of course it all depends on material and amount of limiting. I've just tested it on a drum loop with 4 dB of reduction and didn't see any chopped off peaks.

have you checked with a true-peak meter tho

Yes, I always do that. It makes no sense to me, to use a meter that doesn't show where those nasty peaks are hiding. Truth be told, I have LoudMax's Thresh and Out set at -1.1 to achieve a True Peak readout of -1 dBTP. I just accepted it as a fact. I've seen some paid limiters having similar issues.

When talking about "true peak" it is important to keep in mind that every such measurement is an approximation, because the theoretically ideal reconstruction kernel would be infinite.. and in practice every meter, every limiter and every audio interface actually playing back the audio is going to use slightly different finite reconstruction kernels.. which means they will all have slightly different ideas of what the "true" peaks are.

This ISP business is simply not exact science and anyone telling you otherwise is lying.

That said.. any good reconstruction kernels should come fairly close to each other... so rather than having the 6dB (pathological, theoretical worst-case I guess might be unbounded) errors you get from not measuring ISPs at all, you might end up with .1dB difference in meters and then some random audio interface might add another .1dB (or maybe more)... but that's normal, it's just the nature of things.

[Hermetech Mastering]

We all love our saturators and clippers.

Not me, former I do in hardware, latter is a necessary evil.

[plexuss]

This ISP business is simply not exact science and anyone telling you otherwise is lying.

I'm not a DSP expert so I may have this wrong: ISPs are created when the final render is further processed. For example, audio format compression and conversion in a DAC are the most common processes that can cause ISPs. The ISP context of a limiter is more about the limiter trying to determine if ISPs could be a problem in down-stream processing. A digital limiter should not cause clipping (true-peak). An ISP mode is just over-sampling to try and mitigate ISPs from occuring in down-stream processing. Since we have no control over down-stream processing, the guideline is to allow sufficient dynamic headroom to prevent them down-stream. But that's it. That is to say, the ISP feature of a limiter is over-sampling and any ISP indication is a guess as to when they might occur. But there is no way to completely prevent ISPs in down-stream processing because any kind of processing can be used and we dont have control over that.

The up-shot of all that is, if you want to avoid bad clipping then use a true-peak limiter and if you want to get a sense of how your track might cause ISPs get a limiter that will show ISP warnings. But they are just warnings of what might occur.

[mystran]

This ISP business is simply not exact science and anyone telling you otherwise is lying.

I'm not a DSP expert so I may have this wrong: ISPs are created when the final render is further processed. For example, audio format compression and conversion in a DAC are the most common processes that can cause ISPs.

Right. Pretty much any processing whatsoever is likely to change the peaks somewhat (in ways that are more or less impossible to predict in practice) and this definitely includes both sample rate conversion and lossy compression.

The ISP context of a limiter is more about the limiter trying to determine if ISPs could be a problem in down-stream processing. A digital limiter should not cause clipping (true-peak). An ISP mode is just over-sampling to try and mitigate ISPs from occuring in down-stream processing.

Right. The thing here is that if you aggressively limit by only looking at the sample values, you are almost guaranteed to actively push the audio up to the point where you have serious ISPs (even if the source audio didn't have serious problems originally), where as if you try to estimate (approximately) what the intersample levels are, you can (mostly) avoid this.

The tricky part for limiters actually is that after you've measured the original true peaks and you adjust the gain (over time) to bring them into range, that very gain modulation can cause additional ISPs.. and then you might do a second pass to try to clip them, but that can further create additional ISPs again. The only truly reliable way to avoid them completely is to have enough (read: lots of) headroom, but by at least taking the (estimated) true peaks into account in the limiter, you avoid producing something that's really bad(tm).

[bmanic]

Personally I think true-peak phobia is a bit silly in today's audio climate. It doesn't really matter in the grand scheme of things as long as you don't go stupidly high over 0dBFS with the peaks. If something ends up +0.1dBFS it's hardly a problem at all. Then just turn down the overall gain a few tenths of a dB instead. That's what most digital media platforms end up doing anyhow.. and if you need to distribute CDs, not even the cheapest crap players today will suffer from a few inter sample overs.. even the most basic converters today will play that just fine.

Even the various lossy conversions today will take these things into account, right?

In short: ISP is really not a problem. No consumers will notice nor care at all in the real world. If something is smashed to bits, all relevant platforms will auto-gain the output by a significant margin anyhow, in which case the ISP doesn't even begin to become a problem for the DA converter of whatever device you are using.

[plexuss]

The audible implications of ISPs is wide-band intermodulation distortion. Everytime a peak clips, it creates a shot of wide band distortion that sounds like noise or hiss. This robs the audio of dynamic range and can give the audio a "hazy" kind of sound. A clip here or there is not a problem because they are so short the resulting artifacts are usually not audible. But with sections of high density ISPs (more ISPs per unit time), the more ISPs become an audible issue. The louder the track is the more chance there will be a high density of ISPs. I've found, for my music, that targetting -16 LUFS does a good job of balancing loudness, dynamic range and ISP risks with a bias to sound quality over loudness. But for those working with loud tracks, like >12 LUFS, high density ISPs become more of an issue. The issue is only a problem if you care about the resulting comprimised sound quality or not.

[heavymetalmixer]

Since the latest version (1.40) Loudmax became the best free limiter IMO. Note that it's with the ISP button OFF, 'cause once I turn it ON the sound always becomes worse (I'm starting to think that true peak limiting in all limiters sounds worse anyway).

So hard clipping sounds better to you?

Limiters and clippers are a different thing.

[heavymetalmixer]

Personally I think true-peak phobia is a bit silly in today's audio climate. It doesn't really matter in the grand scheme of things as long as you don't go stupidly high over 0dBFS with the peaks. If something ends up +0.1dBFS it's hardly a problem at all. Then just turn down the overall gain a few tenths of a dB instead. That's what most digital media platforms end up doing anyhow.. and if you need to distribute CDs, not even the cheapest crap players today will suffer from a few inter sample overs.. even the most basic converters today will play that just fine.

Even the various lossy conversions today will take these things into account, right?

In short: ISP is really not a problem. No consumers will notice nor care at all in the real world. If something is smashed to bits, all relevant platforms will auto-gain the output by a significant margin anyhow, in which case the ISP doesn't even begin to become a problem for the DA converter of whatever device you are using.

Sadly, lossy conversions ALWAYS rises up the peaks, even if it is by 0.1 dB. Unless a codec with TP prediction technology becomes the industry standard (right now it's MP3 and AAC), we'll need to wrroy about TPs.

Also, just like you say I turn down the track so the TPs don't clip, right now I just don't like the sound of TP limiting.

[pixel85]

Personally I think true-peak phobia is a bit silly in today's audio climate. It doesn't really matter in the grand scheme of things as long as you don't go stupidly high over 0dBFS with the peaks. If something ends up +0.1dBFS it's hardly a problem at all. Then just turn down the overall gain a few tenths of a dB instead. That's what most digital media platforms end up doing anyhow.. and if you need to distribute CDs, not even the cheapest crap players today will suffer from a few inter sample overs.. even the most basic converters today will play that just fine.

Even the various lossy conversions today will take these things into account, right?

In short: ISP is really not a problem. No consumers will notice nor care at all in the real world. If something is smashed to bits, all relevant platforms will auto-gain the output by a significant margin anyhow, in which case the ISP doesn't even begin to become a problem for the DA converter of whatever device you are using.

Sadly, lossy conversions ALWAYS rises up the peaks, even if it is by 0.1 dB. Unless a codec with TP prediction technology becomes the industry standard (right now it's MP3 and AAC), we'll need to wrroy about TPs.

Also, just like you say I turn down the track so the TPs don't clip, right now I just don't like the sound of TP limiting.

iZotope RX can compress to MP3 and OGG with a limiting (or normalization) option. So, far converted files are showing -0.01dB TP.

Also, even Limiter in Ozone seems to sometimes have a negative impact on sound when TP is activated. On loud mixes anyway. So I'm not surprised that you have a similar experience with LoudMax.

[michal.ochedowski]

The audible implications of ISPs is wide-band intermodulation distortion. Everytime a peak clips, it creates a shot of wide band distortion that sounds like noise or hiss. This robs the audio of dynamic range and can give the audio a "hazy" kind of sound. A clip here or there is not a problem because they are so short the resulting artifacts are usually not audible. But with sections of high density ISPs (more ISPs per unit time), the more ISPs become an audible issue. The louder the track is the more chance there will be a high density of ISPs. I've found, for my music, that targetting -16 LUFS does a good job of balancing loudness, dynamic range and ISP risks with a bias to sound quality over loudness. But for those working with loud tracks, like >12 LUFS, high density ISPs become more of an issue. The issue is only a problem if you care about the resulting comprimised sound quality or not.

Not so long ago I heard a well-established mastering engineer mentioning -8 LUFS as current standard. According to him (and he knows his stuff) that is today's normal. He also mentioned that going above that is mental.
Personally I can't imagine mastering something so loud. I don't want to go beyond -13 LUFS, because everything sounds too distorted. Your target of -16 LUFS seems great. That's a good balance indeed. I wish it could stay that way globally.

[plexuss]

Not so long ago I heard a well-established mastering engineer mentioning -8 LUFS as current standard. According to him (and he knows his stuff) that is today's normal. He also mentioned that going above that is mental.
Personally I can't imagine mastering something so loud. I don't want to go beyond -13 LUFS, because everything sounds too distorted. Your target of -16 LUFS seems great. That's a good balance indeed. I wish it could stay that way globally.

Here's what 16 LUFS can sound like.

https://soundcloud.com/musicofplexus/fairy-flight

Youlean Loudness Meter Export.png

[sircuit]

You can gain 5 LU in bass-heavy music just by boosting (and compressing) the 20-50 Hz area. Talks about LU should be correlated to music genre. Also, in the end, the only reason we go loud is to give the song density and to have our song perfectly audible at very low level. Listen your music at very low levels and hear elements dissapearing from the mix. Take a top40, do the same, and the entire mix still stands at very low volume.

OT: ISPs are overrated massively imho. We make music for the 99.9%, not for the 0.01%, remember that.

[plexuss]

OT: ISPs are overrated massively imho. We make music for the 99.9%, not for the 0.01%, remember that.

We who? You mean, we you. I make music for myself. That's 1 of 7.9 billion or the 0.00000001%. I guess that's why I can concern myself with ISPs, LUFS and optimizing DR and LU vs making sounding audio like typical top 40 examples. YMWV

[sircuit]

All these technical subtleties that could be a problem (arguably) only on a very expensive audio system while the music ends up on the platform with the crappiest lossy encoder on the market. YMWV indeed