DAC/amp performance with true peaks over 0 dBFS

[Donuts123]

I have been ripping some CDs recently, and use the True Peak Scanner foobar2000 plugin to check what the true peaks are.

It's very common for discs to have tracks where the true peak is over 0 dBFS. Some discs are only just over, e.g. album true peak might be +0.06 dBFS. I guess the mastering engineer was aiming for 0 dBFS true peak there. But there are also some tracks with pretty egregious higher true peaks. One of the largest so far was +3.63 dBFS.

I was wondering, has anyone tried to test what effect playing tracks like that has on DACs/amps? In particular DAC/headphone amps & dongles.

Some might have more headroom than others, and there might be clipping/distortion only at higher volume settings (if the DAC output has enough headroom, but the amplifier clips above a certain volume setting). Or if the DAC output clips, there could be distortion whatever the output volume.

 

[staticV3]

Hi @Donuts123! Welcome to ASR.

People who routinely test >0dBFS playback include:

@NTTY, who measures various CD players: https://www.audiosciencereview.com/...thread&c[content]=thread&c[users]=NTTY&o=date

GoldenSound, who posts on the Headphone Show YouTube channel, his personal YouTube channel, and on his website

In any case, if you want to be on the safe side, then you can use Replay gain plugins which scan each track for dBTP levels and bring it back down to <0dBFS.

 

[gberchin]

OK, somebody is going to have to define some terms. 0dBFS refers to maximum possible digital value, i.e., 7FFF positive and 8000 negative, in a 16-bit context. So if you're seeing digital peaks above 0dBFS, then 0dBFs has somewhere been redefined.

 

[DVDdoug]

There was a long thread about this.

OK, somebody is going to have to define some terms. 0dBFS refers to maximum possible digital value, i.e., 7FFF positive and 8000 negative, in a 16-bit context. So if you're seeing digital peaks above 0dBFS, then 0dBFs has somewhere been redefined.

The "true peaks" don't exist in the digital data. Only in the reconstructed analog waveform. AKA "inter-sample overs".

 

[kyuu]

OK, somebody is going to have to define some terms. 0dBFS refers to maximum possible digital value, i.e., 7FFF positive and 8000 negative, in a 16-bit context. So if you're seeing digital peaks above 0dBFS, then 0dBFs has somewhere been redefined.

The true peak is not a digital value, it's the maximum value of the analog waveform that the DAC constructs from the digital samples, and occurs between the sampled values. This is why it's commonly referred to as inter-sample overs.

 

[oleg87]

People have tested this, and the answer seems to be that some DACs handle it adequately, and others do not. Unfortunately, Amir has deemed it not an important thing to test so you won't find that info here.

My 2 cent contribution here is that I tested the Apple usb-c to 3.5mm dongle and going off memory it has about 3dB of headroom for overs.

 

[Donuts123]

I just saw a YouTube video related to this. See particularly from 03:00, he shows the distortion which happens when a (synthetic test) sample with over 0 dBFS true peaks gets resampled to 48kHz.

If your PC OS/audio driver resamples all audio like that for playback...

 

[gberchin]

AKA "inter-sample overs".

OK, that I understand. But the OP said he was examining CDs for peaks over 0dBFS, which I took to mean "in the digital data".

 

[kemmler3D]

So I get that this is a technical problem that looks bad, but has anyone actually found a combination of DAC and music where this legitimately sounds bad or is even noticeable in practice?

 

[Ropeburn]

OK, that I understand. But the OP said he was examining CDs for peaks over 0dBFS, which I took to mean "in the digital data".

The data causing the peaks is indeed on the CD, or any other digital format.

Simple illustration:

Red points represent digital data, with a maximum value of 1 (=0dBfs), blue is the output waveform. You can see how the peaks are between samples, that's why they're called "inter"sample overs.

 

[NTTY]

The problem can be easily illustrated in Audacity.

Let's take a 5512.5Hz test and sample it at 44.1kHz. Not many samples (highly zoomed view below) at this frequency, here they are:

You can easily draw with your finger a sine going through each points. We get lucky because the samples taken at the highest level (0dBFS) are indeed exactly at the top of the sine.

Let's oversample that by 8x in Audacity, and as all digital filters do in our DACs:

More samples and we see the sine appearing. Everything is awesome, we are living the dream.

Now, let's record the 5512.5Hz test tone with a 67.5° phase shift, still at full scale:

Try to draw again the sine with your finger, and yeah you need to go beyond the frame. Nothing is clipped here, else Audacity would show red samples to indicate an over, but you start to feel the problem that @Ropeburn illustrated before.

But let's upsample that by 8x:

Oh, nasty! Many samples are clipped (over) and show red, because they just can't go over 0dBFS. This 5512.50Hz sine @0dBFS with 67.5° phase shift requires +0.69dBFS headroom to correctly recompose the sine. If there's no headroom in the interpolator like here in Audacity, then several consecutive samples are at 0dBFS, and the signal is clipped at the output of the DACs.

The interpolators in our DACs have variable headroom and are mathematically protected quite differently when processing these "overs".

Let's have a look at real life examples. This is a Myryad Z210 processing a 5512.50Hz sine @0dBFS with 67.5° phase shift:

THD at -104.7dB, no sign of stress, no clip, nothing. The Interpolator of the DAC in this CD player has the required +0.69dB to process the signal.

Now, the same with a Sony CDP-557ESD:

Yep, -30dB THD. We easily see the odd harmonics which show the clipping. The interpolator of this CD player does not have the necessary headroom.

It can be worse, though. This is a Teac VRDS-25x that has an ASRC prior to the DAC (to convert all digital input into one unique sampling rate, probably 48kHz), and that interpolator is not mathematically protected against these "overs", and....

Ouch. Not only the player clips (-30dB THD), but the ASRC interpolator is overloaded, generating tones of distorsion all over the place.

I hope these few examples clarify.

Cheers

 

[gberchin]

Now, let's record the 5512.5Hz test tone with a 67.5° phase shift, still at full scale:

Though the sample peaks are at full scale, the analog sinewave that was sampled in order to produce them had amplitude greater than full scale.

 

[NTTY]

Though the sample peaks are at full scale, the analog sinewave that was sampled in order to produce them had amplitude greater than full scale.

Yes but because of phase shift was not seen as clipping. So that will depend on the capturing tool and the operator. Of course my example was on purpose.

 

[gberchin]

Of course my example was on purpose.

Yes, a highly contrived example pure sinewave at exactly 1/8 the sampling rate. In the real world that never happens. Try it at nearly any other input frequency and you'll see the clipping clearly.

I suspect that the most common source of inter-sample clipping occurs when resampling transients, where the resampling filter has some transient overshoot. Under those circumstances it is possible for a signal that was not clipped when originally sampled to clip when resampled.

 

[krabapple]

What matters most is if you can hear it.

Rare or non-clustered overs in a track are unlikely to be audible in normal listening.

 

[Donuts123]

I went and looked at some CDs in FB2K. True peaks over 0 dBFS are really common, by no means limited to loudness war victim modern stuff.

Here are a few, I have listed the tracks with highest true peak values.

1993

https://www.discogs.com/release/741491-Various-The-Chart-Show-Ultimate-Rock-Album

Disc 1: Highest track true peaks (dBFS) +0.04, -0.29, -0.67
Disc 2: +0.15, -0.18, -0.32

1992

https://www.discogs.com/release/2812847-M%C3%A1ire-Brennan-M%C3%A1ire

+0.27, +0.22, -0.37
Album DR 13.

1989
VSOP 4: Classic Meets Pop
+0.13, +0.12, +0.03, -0.09
Album DR 12.

1996

https://www.discogs.com/release/1995323-Shirley-Bassey-The-Show-Must-Go-On

+1.06, +1.06, +1.05, +1.04 (only two tracks have negative true peak, -0.04 and -0.08)
Album DR 8.

1991
https://www.discogs.com/release/6772706-Various-Things-Are-Swingin
+0.07, -0.37, -0.53
Track DR between 10 and 14.

1989

https://www.discogs.com/release/1024769-Richard-Clayderman-The-Love-Songs-Of-Andrew-Lloyd-Webber

+0.04, -0.64, -0.83
Album DR 14, track DR between 13 and 16. The +0.04 true peak track has DR 16.

1990
Sony Techno Fair Anniversary Album YDDS-1220
+0.13 (track DR 20), +0.02 (DR 16), +0.02 (DR 18), -0.18 (DR 10)
Album DR 13.

 

[NTTY]

Yes, a highly contrived example pure sinewave at exactly 1/8 the sampling rate.

@DVDdoug provided you with the link where to contribute.

 

[Donuts123]

At least for foobar2000, a good solution is Amplifier DSP (foo_dsp_amp). That allows bit-perfect ~6dB volume reduction (shifts all bits right by 1 bit). No loss of resolution since I think FB2K output can be set to 24- or 32-bit if supported by the DAC.

6dB should be more than enough for the "worst" CDs and allows headroom for DAC-internal oversampling.

 

[Robbo99999]

I tend to run -2dB Negative Preamp extra on any EQ I do to account for intersample overs. When I analysed my music library then this would make almost all the tracks 100% safe, I think -3dB covered all of my tracks, but because I use EQ then most of EQ curve is well well below -2dBFS anyway so figured I didn't need to go beyond -2dBFS to account for the intersample over factor.

 

[Donuts123]

A very significant proportion of the CDs I checked so far have over-0 true peak. Maybe half of them? Here are some more examples if anyone is interested. Many of these are from the 1980s and not very compressed. There are a couple of outliers: Billie Jean on Thriller +3.19 dBFS, Valotte on Music Life Virgin Version +3.04 dBFS.

Heartbeat City (The Cars) https://www.discogs.com/release/13924732-The-Cars-Heartbeat-City
It's Not The Night +0.57 dBFS (track DR 14)

Thriller (Michael Jackson) https://www.discogs.com/release/1782800-Michael-Jackson-Thriller
Billie Jean +3.19 dBFS (track DR 14)

Storm Front (Billy Joel) https://www.discogs.com/release/6304490-Billy-Joel-Storm-Front
Leningrad +0.03 dBFS (track DR 14)

In The Digital Mood (The Glenn Miller Orchestra) https://www.discogs.com/release/34371247-The-Glenn-Miller-Orchestra-In-The-Digital-Mood
+0.89 dBFS (track DR 14)

Legend (Clannad) https://www.discogs.com/release/5129078-Clannad-Legend
Battles +1.57 dBFS (track DR 14)

Faith (George Michael) https://www.discogs.com/release/28682467-George-Michael-Faith
+0.91 dBFS (track DR 13)

Whenever You Need Somebody (Rick Astley) https://www.discogs.com/release/1513322-Rick-Astley-Whenever-You-Need-Somebody
+0.11 dBFS (track DR 14)

Music Life - Virgin Version https://www.discogs.com/release/1546667-Various-Music-Life-Virgin-Version
Valotte +3.04 dBFS (track DR 14)

Revenge (Eurythmics)
The Last Time +0.28 dBFS (track DR 14)

...But Seriously (Phil Collins) https://www.discogs.com/release/13712405-Phil-Collins-But-Seriously
Heat On The Street +0.59 dBFS (track DR 14)

No Jacket Required (Phil Collins) https://www.discogs.com/release/13748012-Phil-Collins-No-Jacket-Required
Sussudio +0.47 dBFS (track DR 16)

Face Value (Phil Collins) 16029-2 / 299 143 https://www.discogs.com/release/2049760-Phil-Collins-Face-Value
Highest true peak Behind The Lines -0.45 dBFS (track DR 15)
(So no tracks over 0 dBFS.)

Face Value (Phil Collins) CDV 2185 https://www.discogs.com/release/10058743-Phil-Collins-Face-Value
I'm Not Moving +0.46 dBFS (track DR 15)
Total 7 tracks over 0 dBFS, though four only very slightly: +0.02 to +0.08.

Face Value (Phil Collins) 20P2-2074 https://www.discogs.com/release/4030076-Phil-Collins-Face-Value
Behind The Lines +0.35 dBFS (track DR 15)
Total 5 tracks have true peak over 0 dBFS.