Audio researchers at Yamaha, Denon, Harman Kardon, Sony and few other companies have developed and deployed technologies that promise to sonically restore the stuff that is removed by “lossy” compression algorithms. I visited this topic a few months ago when I analyzed the DSEE [Digital Sound Enhancement Engine] process that Sony developed and includes on their new high-resolution playback machines (click here to read that post).
The basic concept is that the fidelity of a music selection prior to being encoded as a lossy MP3/AAC can be “restored” by a few fancy signal-processing algorithms. It certainly true that once a file is digitized that all sorts of creative things can be done to change the sound. The power of DSP (digital signal processors) and the clever algorithms that run on them have brought revolutionary possibilities to audio.
I attended the NAMM (National Association of Music Merchants) and happened by a small company that had developed a “repitching”software tool. As the guy played a chord on his small MIDI keyboard the tonality of a Michael Jackson tune was instantly changed. Imagine a completely mixed stereo track switching from major to minor mode at the touch of a few keys. I was very impressed!
So it would not seem out of the realm of possibility that the DSP gurus at Sony or Harman could coax some fidelity back into a lossy compressed music file that had been stripped of some of its high-frequency content. Unfortunately, the marketing claims exceed the realities. I thought I would reopen this subject with the Harman Clari-Fi processing that is built in to the HTC One M8 HKE Smartphone, which features prominently in their advertising.
There’s a lot of complexity here. What compression algorithm was used to convert the original file to the lossy format? What bandwidth was used? How much fidelity was in the original file prior to the conversion?
I took the Laurence Juber “Mosaic” track (a track that has universally been recognized as a great example of real HD-Audio and which won the 2002 CEA “Demmy” award) and converted it to a series of MP3 files. I set the parameters of the conversion to CBR (Constant Bit Rate) 64, 96, 128, 256 and 320 kbps. These levels range from really poor to very good. HD Radio, for example, uses 64 and 96 kbps (makes you wonder why they used the HD initials…or not), the majority of music downloads are done at 128 (iTunes Plus uses 256) and if you purchase an MP3 from iTrax (yes, it happens), the files are encoded at 320 kbps.
I loaded these files onto the HTC phone and captured the stereo analog output back to PCM digital files at 96 kHz/24-bits. I did this with the Clari-Fi algorithm turned off and then with it turned on. Finally, I did the usual spectral analysis on the files. Take a look at the first example below:
Figure 1 – The illustration shows a 96 kbps MP3 file of my HD “Mosaic” track with and without the Clari-Fi processing. Note the additional high frequencies AND the ultrasonic noise that was added. [Click to enlarge]
My initial thought is why bother with the whole Clari-Fi thing? People that choose to listen to MP3 files encoded at 96 kbps deserve what they get…terrible sound. And it’s not just the lack of frequencies above 16 kHz. There’s a lot more stuff that is thrown away at this level. The sonic impact of a the extra high frequencies is minimal at best.
I’ll write tomorrow on the use of Clari-Fi on files that have higher bandwidths…like 256 and 320.