It seems that every new digital limiter and meter these days is obsessed with measuring and controlling "ISPs".

Consider how many CDs were made in the past with digital limiters and/or clippers simply set to 0dBFS - completely oblivious to this problem called Intersample Peaks. I own MANY commercially produced CDs that measure +1.5dB or more on ISP meters, yet they sound great - even when converted to mp3.

So, where did all this fuss over ISPs begin and why is it such a big problem now?

I'll start by saying it wasn't a problem before there were oversampling DACs, but then, they also didn't clip the audio much. I'll say it wasn't even a problem when they were pushing things -10dBfS RMS or so, despite oversampling DACs being commonplace, because it was largely drum hits and short spurts of clipping is hard to hear. Now, it's routine for not just the odd drum hits to be limited/clipped but the entire duration of the body of every sound to get the full force of multiple stages of ruinitification. Now all of a sudden what was once a few inter-sample clips here and there are in the tens of thousands and more audible. This is especially true since now a lot of people aren't setting the peak level to -.3dBfs but often only one least-significant bit.
That, and they've always known about the problem but processing power wasn't as plentiful as it is now, so it wasn't addressed as much.

ISPs are an issue that should be addressed but they are a very minor one. As been said here, much more harm can be done by clipping, distorting and creating super loud mixes, in which case, ISPs are just a minor part of the big picture.

When working in the hot level department, ISP must be seriously taken in account. The usual -0.3 dB safety margin isn't enough for ISP that may be as great as 3 dB.

+1

The IM distortion introduced by a limiter and the ISPs is meaningful but as Miguel points out, not a huge issue. The issue is amplified if using more than one limiter stage where both are producing meaningful amounts of gain reduction.

Give me a record which has peaks 3dB below 0.

Or better - give me HOT record that is 3dB below 0.

I've never seen one, I've never done one. Most records done by pro's are hitting 0dB or -0,1. I think they don't care about ISPs, and headroom to zero too...

ISPs are a problem in the same way that sweating through your shirt is a problem when you cook a chicken with napalm.

You don't want to sweat through your shirt. That is bad for sure. There are several ways to prevent sweating through your shirt when cooking a chicken with napalm, and those methods could certainly be explored. Here, enjoy this chicken ash.

That is the modern ISP conversation.

I was always led to believe that ISPs could potentially be an audible problem when playing tracks back through inferior DACs (of which there are many examples in consumer playback systems), and so, if trying to preserve the utmost fidelity, it's always better to make sure you don't have any. Having said that, I don't think they are generally audible on a decent system with a track that isn't heavily smashed.

Regarding high levels of ISPs, the original CD of the following track, by Kyari Pamyu Pamyu, which was a huge hit in Japan about 18 months ago, registers ISPs of up to 3.9dB above digital full scale on RME Digicheck's meter:

http://www.youtube.com/watch?feature...&v=pb3IdLdIc6s

When smart people refer to ISP's as "free level", you know it's not a big problem.

Obviously I wasn't clear. Despite ISP can be 3 dB above 0 dBFS, I've never advised to turn down the level by the same amount. I just mean that using limiters managing ISP is mandatory when dealing with hot levels.

Doesn't sound all that professional to me.

-18 to -20RMS for me, no peaks over -4dBFs.

What world are you living in? I want to go there.

Well, I make my records R128 compliant, but I don't think there are many of such records out there nowadays.

ISPs can be a problem when the DAC tract is not designed to handle them - so these peaks can overload the analog circuitry after the DAC.

By the way, I heard that Adobe's Audition is good at representing the reconstructed waveforms visually - is that right?

So, you really saying that highly specialized electronic engineers are too stupid to give DA circuits 6-7dBs of headroom? Do you have any kind of proof for this steep claim? IMHO, it isn't much more than misinformed audiophile paranoia.

The main job of a DA converter is to deliver a proper, distortion free performance over the full scale. This full scale specification includes the expected ~6dB tolerance above full scale. Electronic engineers know about these effects and can easily prevent any problems, even in cheap circuits. Just throw in a 0dB FS sine and measure the output of your converters.

I personally never found any audio DA that didn't act properly under these circumstance.

It's the DAC chip itself that clips and there are many that do it with full-scale signals. You see flat-topping in 0dBFS square waves.

I'll have to dig up the thread, but I've posted on this before. As an example, the latest Benchmark DAC attenuates the digital signal 3.5dB before conversion. This reduces the S/N ratio by the same amount, so they have to parallel several DAC outputs together to get the same equivalent performance as before the attenuation.

DC

I'm not claiming that anyone's stupid - please don't read my post between the lines. I just said that it can be a problem.
It depends on the DAC chip itself, power supply etc. All these desicions are made by the engineers. I don't say they are all stupid, but I don't eliminate the possibility of error, especially in cheaper products.

You made a valid point Orson. Other's have mentioned that. I mean the post with text, not the blank one. Welcome to GS, it's good to hear from people in Russia.

Hey, Mark. Thanks!

I believe I came across an article in the net about a portable CD played not dealing with the ISPs correctly.
If we think of it - some engineers could theoretically choose a wrong type of DAC for a portable CD - for example an ordinary line level one. And we get a double AA battery power supply which is in an ideal situation a 2.7db headroom over line level. Add NiMH battery type here (1.2V instead of 0.5), add P-N voltage drop inside the amplifier, add power supply sag due to low charge, and all this becomes pretty plausible I reckon.

The question here is whether there are a lot of devices that cannot handle the ISP and whether we should worry about the poor bastards at all.

ISPs are a problem when D to A converters have no headroom in their filter and/or analog stages. It's more of a problem because of lossy encoding of files and streams and less of a problem because most newer players have digital volume controls ahead of the D to A stage. The waves L1 limiter was the first to identify and address this problem.

And I'm afraid there has been plenty of poorly designed gear offered for sale.

From this EBU paper ISP are a real issue for downstream processing such as lossy encoding, sample rate conversion and D to A converters as well.

What about my software question? Is there any piece of software that reliably displays the future DAC'd waveform? Is Audition good for it - it seems to be just from the look of it.

Yes, Audition is good. iZotope RX is another option: there you can adjust the order of interpolation in Preferences. Every DAC is different, though, and there's no single software that models the output of every possible DAC. Beware that both Audition and RX display the interpolated waveform only when you are zoomed in enough. When you are zoomed out they usually switch to a non-interpolated (digital) waveform to speed-up the display.

Thanks, Alexey -good points you made there.

I believe there was a shootout thread comparing different digital interfaces where this issue was addressed in more detail. Not sure where did the thread go though.