Hello, I’m doing some research in lossy audio coding in order to find the very best among high bitrate (300+ kbps) encoders, never mind why, I know the good codecs are indistinguishable at these bitrates, but still curious, and I’ve come to interest in piece of old software known as DCC2WAV. I am far not a psychoacoustics expert and my methodics to analyze the quality of a codec is simply to compare the source and the encoded waves “bit-by-bit” and see how much it differs in various frequency bands, also to see how wide is the frequency range a codec can handle.
I eventually decided to take a look at the PASC encoder, brought myself a DCC2WAV executable, honestly not expecting much of it, but I was wrong. Surprisingly, it showed me full 20 kHz response range without large dropouts and, the more important, near-lossless quality of bass and mid ranges, pushing majority of distortion into treble frequencies.
Damn, I was amazed, no other MP1/MP2 codec can closely perform as good as DCC2WAV does. This would make me put PASC in the same row of top-tier audio codecs with Apple QAAC and Musepack with only question about it’s treble range. I personally can’t tell anything bad in my listening tests, so the question is: do you find this distortion frequency trade-off acceptable? I can clearly tell that DCC2WAV is better than any other MP1/MP2 codec in terms of DCC recording, but do you think it can compete with modern audio codecs?
Distortion comparison between PASC, QAAC and Musepack:
I would be interested in which other MP1 encoder(s) you have tried.
In my experience, finding any MP1 encoder is very difficult nowadays because no-one is interested in the format. I’ve seen various open-source encoders based on the “dist10” source code but most of the time the MP1 encoder has been removed. DCC2WAVE uses the algorithm described in the DCC specification but it’s slow.
We settled on the NCH encoder in the Switch and WavePad software for our recent DCC Museum releases. It’s closed source but it gives good results that even Patreon patrons with good ears approve.
I previously modified MP2ENC (also based on dist10) so it encodes in 32 bit instead of 16 bits and although the results sound great to me, our “golden ears” users complained about distortion in the high frequencies. I’m trying to find some time to restore MP1 encoding in open-source projects such as LAME / TwoLame / TooLame.
(In case you don’t know, PASC is essentially identical to MP1 except in 44.1 kHz the padding slots are set to 0 in PASC whereas in MP1 they can contain data).
Thanks for reply; I’ve tested at least QDesign I-Media, SCMPX, MPC2ENC and libtwolame, yes, DCC2WAV beats them all despite being a Layer 1 codec, not even Layer 2. For MPC2ENC I’d say it produces zero distortion for range below 9kHz, but the rest is heavily distorted, no wonder it is audible for golden ears guys.
Yes, it is absolutely cursed but still the best. First hearing of NCH, should take a look at it!
Have just tested NCH WavePad MP1 at 384 kbps and, well, it is good, it’s frequency response is quite similiar to QAAC/fdk-aac, but with lower distortion level. However, there is a serious issue with frequency bandwidth which is cut to 14 kHz + one more kilohertz of junk noise not representing the source band at all. I would still prefer DCC2WAV due to it’s outstanding (for Layer 1) frequency response.
