phaeton

I'm using a Wireworld Starlight 7 , nothing "esoteric" but from what I can hear, very good. I've heard no difference in A-B comparison with Wireworld Silver Starlight and JPS Superconductor3 many have claimed that Vega usb port is among the best usb implementation, so I think an ultra-expensive-super-esoteric usb cable is not really needed..... just IMHO of course....... and, again of course, I'm not saying that in general usb cables are not relevant........ (instead, with other dacs I've owned, they were very relevant)

maybe, but I do suspect that the real problem is that your power amp is gaining way too much. is there any way to reduce the gain? I'm currently running the Vega through a Spectral 30slg2 set at max gain (30db if I remember correctly) and driving a Pass X150.5 set again at max gain (30db). in this configuration the Spectral pot is at the same level using the Vega and the Marantz SA-11. when some time ago I tested the pre-less config, my usual level was around 60-70, anything below 20 was non-audible. even considering that my loudspeakers are quite low in efficiency (Amphion Xenon, 85db), I really do suspect that your power amp is amplifying really too much................. in any case, a cheaper solution than a Taurus could be an attenuator like these Attenuators hth mike

seems very strange... can you adjust your amplifier gain?

ok, thanks

excellent, very good!!! btw, the pads you found seem to be very interesting, are they made of soft plastic, rubber or what? I already have a couple of sets of vibrapod (The Vibrapod Company) but I think that given the peculiar chassis of the Vega/Taurus a more "rigid" coupling is needed. have you heard any improvement in SQ with the pads?

Hi ejn, well, nothing really esoteric or high-end :-) just a floor doorstop made with soft plastic :-O easy, very cheap, very effective.... :-)

very good finding, thanks for sharing mike

myabe because these crystal are used mainly in many other areas than audio dacs? mike

I must admit I don't have much experience in engineering audio DAC. I only have daily experience in engineering DAC for not-audio apps, digital tx/rx PRN crypted systems, mainly in the 10-15GHz range, and target locking digital systems. Both relies on very stable, very precise clock, particularly the tx/rx systems, since we have to packet a lot of stream using the same carrier, and every assigned slot must be very precise and stable before going to the QPSK. We typically use a variant of the Discroll implementation for the crystal circuit, and we never went under 10Hz in calculating the jitter to use in the cross-correlation integral calculations (also leaving a very wide error margin), so I tought that for much lower freq apps like the audio ones, the threshold couldn't be lower (0.1Hz, as said in some posts). but, given you all seem to converge on the point that ultralow freq are important for audio apps, then there's something that I'm missing, maybe related to the peculiar s-d conversion typically used. I think I'll start with studying the doc junker linked some post below, seem to be a very good starting point. mike

what is something like a "jitter phase noise" ???? jitter is a time domain property, and it's an integral (call it RMS if you want) quantity PN is a freq domain property, and it's a spectral density of power measured at a certain offset of a fixed clock

didn't know that an XO "sounds". I was used to think that a well designed circuit "sounds". I said that at -100db PN starts to be irrelevant, and if every XO would have the same PN above 100Hz, and a resonably similar PN below 100Hz, yes, they would "sound" the same. at least until you will present me a human being been able to hear 1mHz differences in a 1kHz pure tone.

-180dB is actually irrelevant. the 100MHz offset was just an example to show the effect, and in fact, I said in one of my earlier post that you should consider the higher offset down to some -100dB, going below that start to be irrelevant. but this has to do with the probability to have that shift in freq. what about the effect of the shift? I do not want to repeat, but a -10dB in the 0.1Hz-10Hz range in the PN means that you have an high probability that your clock will shift from (let's say) 100,000,000.1Hz to 100,000,010Hz , i.e. a shift in the 0.0000001% to 0.00001% range in the output (well, I'll follow Miska's hypotesis, and assume this is only 50% of the effect). instrumentally relevant for sure. audible? IMHO not. I'll leave it to you to calculate what 0.0000001% means @ 1kHz mike

real circuit can be VERY different from simulations.... ;-) you can simulate perfectly what happens in the digital domain (that don't care in any way about PN or other funny things :-) ), and that's very easy to do using any high level language (I started with asm, and that's not easy at all, but now you can use c, c++, even eiffel if you want :-) and that's much simpler) same as you can simulate very well what happens in an hypotetical perfect world, even using very complex component models.... SPICE is here to help ;-) but when you have to deal with real world, parasitic impedance, non-zero trasmission delay, EMI, etc.. you have to make your hands dirty and start soldering :-) , and that's fundamental if you want to implement a real femto clock ;-) but I agree with you that to model the clock-output relation some c-programming is enough, but you know, I'm an old-school engineer, I prefer to build circuit and measure ;-)

yes, but this is only from the input point of view. if you assume that the system is linear, or near/locally-linear, you have 2 inputs, the stream and the clock, and one output. remember the reachability/observability property of a linear system? you are "looking" at the output, what if the clock is not fully reachable from the input?? what could be interesting to do is to arrange a board with a variable master clock (I've done something similar for a demo of a PRN-correlator crypto system) driving an s-d dac, and feeding the dac with a fixed stream of 3 sin (20, 1k, 20k). this way you can really measure the transfer function of the clock (i.e. fixed input, variable clock), maybe repeating the measure using a stream at 44.1k, 96k and 352.8k. once you got your TF, should be easy to build an equivalent circuit, and then clearly understand how PN influences the output. mmmmhhh..... responding to myself (yes, I know it's schizofrenic :-) ), way too complex, I'll live with the doubt and enjoy my femto-clock Vega ;-)

interesting point of view. Yes, you can think of the cross-correlation between the PN and the output of a converter this way, but I do suspect that the relation is not so simple as the "demodulator" one. Remember you are not feeding the converter with your clock (i.e. the clock is not the input of the dac, in that case the relation would have been direct and your analysis spot on), instead you are feeding the converter with a stream and "driving" the converter with the clock. The relation you have to analyze is not the fc-fm, but the cross-correlation between the theoretical clock of your input stream and the actual clock with which you feed your dac, something very different. But, repeated, your point is really interesting, very intriguing, is there any docs or collaterals I can read more about this?