Home Digital Drive

Upsamplers, DACs, jitter, shakes and analogue withdrawals, this is it.

[ Follow Ups ] Thread: [ Display  All  Email ] [ Digital Drive ]
[ Alert Moderator ]

Re: Perpetual P1A's Resolution Enhancement Alogrithm

The software algorithm inside the P-1A is NOT an upsampler. In fact, the upsampling is handled by a fixed hardware based solution from Crystal. In practice, the upsampling process will require extra bits of resolution to handle the situation where the output sample is not at the same time as the input sample, which is most of the time. It is not essential to use extra bits of resolution in the upsample process, but it will sound pretty bad if you don't. The Crystal part puts out 24 bits. When this data gets sent out of the P-1A box, it is dithered, according to the programmable wordlength, not using the available dither in the Crystal part but our own dither. We use a dynamic dither that is modified according to spectral content and level.

The onboard DSP in the P-1A is running the resolution enhancement algorithms. These are computationally intensive algorithms that first process the audio signal to look for certain characteristics, using some of the same kinds of preprocessing used in modern advanced codecs (e.g. multiresolution wavelet filtering). Here the goal is to use the signal characteristics as an aid to steer the modification of the upsampled data. In a practical sense, we modify the upsampler from a fixed, static process and make it dynamic. Accordingly, we modify the upsampling processing to best match the audio data at a specific instant. Nothing mystical going on here, but we are not publishing the specifics so as to protect our intellectual property investment. Anyone with a high performance DSP and lots of time could duplicate the processing.

I like to compare the resolution enhancement process to a video line doubler. In the video situation, extra lines are created out of nothing (so to speak). There are some purists that don't like line doubled pictures, but the other 99.999% of viewers are knocked out with the upsampled image and don't worry about where the 'new' lines come from. Even though from a strict mathematical perspective there is no 'new' information in the picture, it is more pleasing to the eye, and subjectively has more detail. Our audio processing is not analogous to line doubling, but the subjective impact is the same. More detailed, subjectively more information, and less harsh. For the 0.001% that object to the processing on principle, it doesn't matter if you publish the source code and schematics. They still won't like it, so we are not going to go down that rat hole. Moreover, there is no one "correct" way to perform line doubling.

With some types of standard test signals, such as single and multiple tones, we actually do measure greater information after P-1A processing. This is not magic. For example, if I told you that the audio data is a 16 bit sampled 1KHz sine wave, you could easily recreate a 24 bit sine wave from the 16 bit signal. I doubt that anyone could create an acoustic guitar that created tones with only 16 bit precision. It is only reasonable to assume the guitar tones have much greater precision than that, even if sampled at 16 bits. The challenge in the P-1A is to identify the data, and apply appropriate enhancement processing. Advanced codec theory has many decades of work in the identification area, usually applied to companding, and we take advantage of that research and apply the ideas to a different category of algorithm.

misterdsp for P-Tech

This post is made possible by the generous support of people like you and our sponsors:
  Signature Sound   [ Signature Sound Lounge ]

Follow Ups Full Thread
Follow Ups
You can not post to an archived thread.

[ Top Top of Page ] [ Contact Us ] [ Support/Wish List ] [ Copyright © 2025, Audio Asylum, all rights reserved ]
[ General ] [ Speakers ] [ Tubes ] [ Vinyl ] [ Digital ] [ Hi-Rez ] [ Video Asylum ] [ Cables ] [ Tweaks/DIY ] [ Music ] [ Films ]