markk02474

I'm pretty sure Audio USB is only 1.1 and there are not any USB 2.0 receiver chips - 24bits * 96 kHz = 2.3 Million bits/second/channel. USB1.1 = about 8-10 Mbits/second. Firewire at 400Mbits/sec can do higher rates. The real issue is that the external device is stupid. We'd like it to buffer up the right amount of data from the computer and ask for more when needed. A musician wants to buffer very little so there is less delay between hitting a key and getting the sound. Playback doesn't have this requirement. Buffering is needed because the computer and the link to the DAC or spdif out won't be able to perfectly time when data is sent - they could be busy with other things. The problem with SPDIF is jitter. With minute variation in when the input data is converted to an output voltage, you end up with a different voltage than you would have gotten if conversion happened at the exact time it should have. For music, anyway - if the number was constant, the output voltage would be constant, subject to dac error. The DAC getting the jittery signal can compensate for some of the variation with PLL re-clocking and sample-rate converters. Still, most people report less spdif jitter sounds better. The hot PCI card to use is one from Lynx with under 20 pico-second jitter. I don't know about cardbus or other alternatives for laptop. Noise inside your computer is only one possible problem. The certain problem is not having enough room for independant power supplies to the dac and opamp buffers, and not enough room for good low-pass filters after the dac chip. If you look at data sheets for DAC chips, you'll see that there are at least three power supplies and grounds: Digital, Left Analog, and Right Analog. Frequently there are more, with voltage references for L and R channels, though these are most often implemented as being decoupled off the respective analog channels. The filtering after the dac reduces digital high-frequency artifacts above musical frequencies. These are usually implemented with resistors and capacitors. Both types of components have a wide range of sound quality and cost. Physical constraints dictate use of surface-mount parts, and thus the capacitors will be ceramic types. Within those, NPO types are better for audio because capacitance varies less with voltage. NPO types cost more, are larger, and are hard to distinguish from cheaper ones visually. Guess which kind get used most? Opamps used to drive headphone-levels of output are usually not as good as the best ones that are only driving line-level outputs, so an external box can use an opamp for each need. An excellent sounding stereo opamp costs about $2.50. Most gear use $0.20 opamps. Just about any consumer and most pro audio cards and dacs will benefit from modification. Better parts, power supplies, and minor design changes can help a great deal. The next level would be better clock circuitry. After that comes premium parts and power supply mods, space permitting. If you downsample data, you are throwing it away. Upsampling later does not get it back! You may not lose all that much, however. Most DAC chip specs. indicate they don't resolve much better at 172 than 96k. The difference is going to be the low past filter they can use for each, with the one for 96k imparting more phase distortion into audio frequencies than the 172k one. You should check the datasheet for the dac chip being used to make sure it switches the digital filter being used. If all is good, upsampling back to 172k will use the less distructive filter.

My two guesses are iTunes is adding gain to the digital data such that it is clipped (full numeric value for 16 or 24 bits) too much of the time when you play directly, OR, when you rip the music to CD, software scales down music that is too hot and close to clipping. Perhaps you can test with your mixing application for clipping, or disable volume adjustment when ripping.