Playback issues with 0dBFS+ Levels on Consumer Audio

[gene]

The specifications for the CD format were set out in the 1980 "red book" standard developed by Philips and Sony and was later adopted by the Digital Audio Disc Committee and ratified as IEC908. The red book standard specifies the data format for digital audio and the technical specifications for devices and media. In the red book standard, it clearly defines the maximum permissible signal level of 0dBFS (digital full scale) is allowed. Any levels beyond 0dBFS are defined as an “illegal states” or violation of the format, and not mandatory to be supported by the digital playback devices (IE. CD/DVD, A/V receivers/processors).

We previously discovered in “The Current Trends in the Recording Format Arena” how a great bulk of today’s mainstream / pop music suffers from hyper-compression, and digital clipping, all of which are a result of recording levels set too high during the mastering process.

However, another issue that plagues many of today’s recordings and the hardware that must play them is when output levels enter an illegal state or 0dBFS+. This article details what can and will happen on the majority of consumer digital playback systems when attempting to playback these recordings. It is for this very reason that concerned Audioholics such as ourselve’s have presented the case for NOT exceeding 0dBFS levels to ensure compatibility of consumer electronics and the best possible fidelity, free of distortion and compression.

Read the article

 

[Az B]

The actual signal level on modern mastered CD's can be literally -.01 dB short of full scale, but the resolution of the mastering meter circuit under real-time conditions may not discern between this level and 0dBFS (max/clip). It is quite common for manufacturers to implement a circuit that illuminates the CLIP indication PRIOR to actual clipping, so that the engineer can correct any issues before an audible result is heard.

It's also been my experience that most professional digital recording equipment also use limiters to prevent clipping in any case. Extreme amounts of compression also make it easier to prevent clipping, at the expense of dynamic range.

If the source material was indeed clipping, it would be quite noticeable. This type of clipping is particularly annoying and very hard to cover up.

But since this is a theoretical measurement with test tones and not actual source recordings, I'm not sure what the point of the article is. Is the author suggesting that there are CDs out there that clip? I would think that they would be pretty rare and pretty amatuerish as most good engineers are quite aware of what happens when the signal actually clips.

Don't get me wrong, I've heard plenty of CDs that sound like crap, but I've never heard one clip unless it was intentional or mastered by a complete novice or moron.

Has anyone heard a CD that this clipping was noticeable? I'd be interested to giving it a listen.

 

[Dan Banquer]

CD Clipping

I have one cd that does clip, or maybe I should say I see strong evidence of a digital limiter, a few times on one particular track. I can see it on the scope, but audibility is very tough because the clipping/limiting happens rarely. There probably is a threshold if the clipping/limiting happens at some consistent rate, but I don't know what that is.
d.b.

 

[Mudcat]

I was under the impression that it is almost impossible for digital equipment to exceed the 0 dBFS on a mathmatical basis.
dB being a ratio of things will be calculated (using voltage here) dB = 10*log(V1²/V2²)

This equals

dB = 20*log(V1/V2) where V2 is a reference voltage (usually 0.775 for pro gear and 1.0 for consumer gear).

This version of dB was retained for digital gear based on binary encoded information. I'll use 16 bit for my example. Stuff digitally encoded in 16 bit resolution will have 16 placeholders for information like this 1111 1111 1111 1111. And that 0 dBFS is at the very top of the scale.

dBFS = 20*log (1111 1111 1111 1111/1111 1111 1111 1111)

This is a royal pain, so in base 10 it will look like this

dBFS = 20*log (65535₁₀/65535₁₀)
dBFS = 20*log 1
dBFS = 0

Conversely, the lowest measurable dBFS will be -96

dBFS = 20*log (0000 0000 0000 0001/1111 1111 1111 1111)

Again in base 10

dBFS = 20*log (1₁₀/65535₁₀)
dBFS = -96 (actually it is -96.32 but no one cares about the 0.32 at this level)

For 20 bit encoding, the dynamic range is 120dB, for 24 bit - 144dB.

That is why I find it difficult to imagine exceeding 0 dBFS, it is just that there is no way to do it because there is no information beyond than the systems max resolution.

 

[gene]

Mudcat;

Put the calculator away for a moment

It's the intersample peaks can exceed 0dBFS! Check out this article for more info which is also referenced in ours.

 

[djoxygen]

It's not the digital equipment that's creating the situation, Mudcat, it's the analog filters on the D/A stage of playback that interpolate something that wasn't present in the original signal.

Waves mastering plug-ins have contained (much less) information about this for quite some time, and I belive that ProTools will even show you a possible clip on a channel that may exceed 0dbfs on playback. (I have ripped CDs to SD2 for use as mastering references and in a couple cases, while there was no visible clipping in the waveform (i.e. two full code samples in a row), the channel meters would show clipping.) It is definitely the responsibility of the mastering engineers to have the tools to detect and avoid this situation. Of course that's no guarantee they will accept that responsibility.

 

[Dan Banquer]

Digital Cipping

"It's not the digital equipment that's creating the situation, Mudcat, it's the analog filters on the D/A stage of playback that interpolate something that wasn't present in the original signal."
With voltage rails at +/- 12 Volts to +/- 15 Volts I find that hard to believe. I remember testing just the analog output filter on my unit and got a 10 Vp-p swing without clipping. 0 dbfs is typically set at 2.0 to 2.2 vrms out from the output of the CD player or DAC. Are you sure that's where it's clipping?
d.b.