Digital clipping on squeezebox

I notice that there is digital clippling in the squeezebox Transporter when I put volume control on fixed, dsp engine completely disabled and no volume leveling on the track material. In my perception these setting result a digital gain of 0 dB which should not lead to digital clipping. The clipping can be best heard on voice material that is close to 0 dB (pop material).
The digital clipping is only moving away when I enable i.e. the dsp engine and reduce the volume by -3 dB or when the volume control is enabled and set on 96/100 or below.
(This clipping is not heard other zones with the same 0 dB gain setting)

Is there a calculation offset that leads to this clipping although all changes to volume are off?
Have others noticed this problem and is it more general?

Hotly mastered material often has inter-sample clipping, which is exposed during upsampling processes–including those which occur in the DAC chip in the Transporter.

Creating a little bit of digital domain headroom is the proper solution. Some DACs reduce digital volume this for all content as a design choice, in order to counter this issue. You might find this article interesting.

Does the Transporter have a digital clipping indicator? If not, if this is just by ear, the cause might not be digital clipping in the signal path. It could be digital clipping upstream in the source material, it could be analog clipping downstream in one of the amp stages, etc.


Brian, I think you have a good suggestion by stating that the clipping comes from inter-sample addition that is a result of calculations and brings the signal above the 0 dB level.
From Roon/Raat I handle the signal in a Non OverSampling way. Internally in the Transporter, when I remember right, all material is upsampled to 96 kHz before it is fed to the AKM AK4396. Here the signal must go over the 0 dB which is not posible in the binairy 16 bits digital domain.
When I reduce a loud signal that will be close to 0 dB in Roon by -3 dB then I can hear that the clipping is gone. In the positions in between this amount of reduction the clipping is less in corresponding with the amount of reduction.

Wiwavelength, there is no clipping indicator. My ears register when there is some clipping and I can hear this at best at voices. Those are at mid frequency and the clipping is non harmonic in high frequencies.

Thank you both, for me the explanation is very satisfying. I will put the headroom adjustment to some -3 dB at which I hear no clipping.