For fun, I thought I'd run an old NOS DAC I had lying around to see how it measures...
This is the ancient Philips TDA1543 DAC (data sheet says 1991), 4 chips in parallel - supposed to improve noise measurements. This particular model is the MUSE Audio one you can find on eBay for about $60 with case & wallwart... For this test, I actually ran it with a better AC adapter (Samsung 12V, 2A unit).
The DIR9001 digital receiver in this unit allows up to 24/96 but the TDA1543's are 16-bit chips which I presume will truncate 24-bit data.
Setup: Touch (WiFi) --> proper coaxial cable --> TDA1543x4 DAC --> shielded RCA --> E-MU 0404USB.
Behold... Stair-stepped NOS output (this is 16/44 1kHz sine wave recorded at 24/192 from the analogue RCA output):
July 2013 update:
As per my recent testing methodology, let's start with the oscilloscope graph of a 0dBFS square wave at 24/44 (note this is only a 16-bit DAC so cannot benefit from the extra bit depth).
Impulse response (16/44):
Summary RightMark results:
Frequency response drops off at the upper end on NOS DAC's at 44kHz sampling rate and you can see that quite clearly above - this improves significantly with 96kHz. Looks like the performance overall improves with 96kHz sample rate except for the stereo crosstalk which remains quite poor (even worse than the SONY CFD-S05 boombox). Many folks who use NOS will upsample the data before feeding into the DAC to take advantage of this fact.
THD graph (a bit nasty):
Jitter (16/44 Dunn J-Test) - Due to the stair-stepped aliased nature of the sine waveform, some of this is distortion but most of it is jitter sidebands.
Well what can one say about NOS DAC's like this... They measure poorly. Yet, many swear by them. Sure, this is an inexpensive NOS and there are much more expensive ones out there like the AudioNote products (the AudioNote DAC 0.1x looks like it only has a single 1543 and upper models all seem to use the AD1865 - don't you love their term "1x oversampling"?!).
Personally, I find the sound isn't bad. It does have the high frequency roll-off which works OK for fatiguing music but that also takes away some of the "sparkle" in good studio productions. Goes to show just how amazingly tolerant hearing can be!
Looking at all those sidebands in the jitter plot, I'd estimate it to be >5ns; but yet it still sounds quite good! IMO, more evidence to say there's no point to quibble about the odd sidebands and whatever picosecond jitter found in well engineered gear...