Line-in Encoding vs. Sonicstage: Different Quality at 64kbps?

[ZosoIV]

Here's an interesting observation I made when playing around with my NH-1 and different encoding modes - Hi-LP, which sounds pretty poor to begin with, sounds clearly worse when encoded via SonicStage. I've heard this mentioned on the board someplace else, but did some testing to verify such observations. On the three test tracks I used (Pink Floyd - Wish You Were Here, Miles Davis - So What, Van Halen - And The Cradle Will Rock), the optically-recorded Hi-LP encodings contain much less stereo flanging, more precise "hits," and less warbling compared to those encoded with SS 3.3. The difference is obvious, i.e. not even worth ABX'ing.

I find this interesting, considering how limited a chip-based DAC is in terms of power and capability. With a DSP, the object is to encode with as little CPU cycles as possible, which decreases energy consumption on a battery-powered device. On a computer, which runs on mains and may have 20-50x the processing power of a ~20-50MHz DSP like found in the NH1, one would think that the encoder would be far better tuned. If anything, you would at least expect the software-based encoder to sound the same, simply being faster! This does not seem to be the case – the software codec is faster at the expense of sound quality.

While the sound quality of realtime-encoded Hi-LP isn't audiophile by any means, it compares favorably IMO with the latest "autov" Ogg Vorbis encoder at -q0 (44.1kHz/64kbps). I have played around a lot with the latest Vorbis encoder and encoded the same samples mentioned above from my FLAC library for testing. The Vorbis files exhibit stereo collapse and sound very rough on instruments like the sax, but do not sound overly metallic like most encoders around this bitrate. The A3+ encoder tends to sound more metallic, but more stable than Vorbis. Either one is an OK choice for cheap earbuds on a train/subway, but not really anything else.

[Top Cat]

Seems like you're right on with the speed concerns. Remember their advertised "40 sec CD to MD transfer"?

Also it might be some differences in algorithms, like joint-stereo in HW and dual-mono in SW or something

[pauljones52]

I recorded a track using sonicstage 3.4 in high quality mode, and i found that it sounded better than my optically recorded track directly from my unit.