DAC that re-clocks for under $1000?

Benchmark DAC1 ($995)

You can read about Benchmark's technology here: http://www.soundonsound.com/sos/jul05/articles/benchmark.htm

As far as I remember at 50GHz most of samples are redundant. This frequency is not real (1 million times oversampling) and is based on statistical manipulations(finding exact spot for the output sample).

They also mention updating output DAC at only 110kHz. Most of DAC chips have higher THD at 192kHz than 110kHz.

Asynchronous upsampling never brings exact oversampling ratios. If specification says 4x,8x,128x etc. then it is not asynchronous upsampling converter but rather oversampling one based on PLL (phase lock loop).

Realremo - Benchmark is not on the warm side. It is pretty neutral and resolving/revealing. I have class D amp (Rowland 102) that is also neutral and resolving. Combination of two with my previous Paradigm Studio/60 v2 speakers was a little bright (painful on some records). I changed speakers to Hyperion HPS-938 and now everything is creamy smooth and even bright recordings sound nice and musical.

Before Benchmark I had Cambridge CD4SE. I was really surprised with Benchmarks clarity. Initially it sounded like some instruments were missing (too clean). Some people call it sterile - I call it very clean.

I cannot tell you how it compares to other DACs since I never had other DACs in my setup.

Realremo - Airport Express has only optical out but it doesn't matter if DAC has very strong jitter rejection. My Benchmark has jitter bandwidth of few Hz and at frequencies of interest (kHz) has over 100dB rejection (of a jitter that was -60dB or lower). Stereophile wrote primer on jitter http://www.stereophile.com/reference/1093jitter/ You'll find some info on reclockers there.

Jitter is noise in time domain. In frequency domain it creates sidebands that are not harmonically related to root frequency and therefore audible even at very low levels. With music (complex signal) it is basically noise. Since it appears only when signal is present and it is proportional to signal level it cannot be detected without signal and can be judged only as a lack of clarity.

Typical transport's digital out switches with about 25ns transition time. Better transports do it perhaps faster. Faster transition means more problems with reflections on characteristic impedance boundaries but slower transitions suffer from induced noise or threshold/system noise (case of toslink). Whole thing depends on all three components (transport, cable, DAC) but Toslink in general makes about 2x worse jitter than coax. On the other hand if your DAC has strong jitter rejection then Toslink might be a blessing because it breaks ground loops.