Friends,
Our AC comrade "
Bica" has been so bold as to claim the
Timepiece 2.1 is more "transparent" than the renown
Quad ESL57. Even though he owns both and has had extensive experience in audio as well as in time spent comparing each, that is still quite a declaration. Undoubtedly it borders on the unbelievable for many. Although one can not prove anything with measurements to those that would dismiss them, there are those that understand that high fidelity is not a matter of magic and pixey dust. For those of the later persuasion we off the following:
Note the 35dB span in magnitude of this waterfall graph. John Atkinson of Stereophile cuts his plots off such that there is only a total possible display of 24dB. We're revealing over 10dB more information down into the noise floor/hash/distortion than he does. Above and beyond any of the following graphs, this waterfall response reveals why the Timepiece as well as all the other Millennial Reference Series products reproduce music with such accuracy and realism.
Do you see any little "mountain ranges" of delayed energy eminating from the system within the first 1.5 mS? No? Notice how far the magnitude drops before you see anything resembling such. The first peak of any kind is out past 1.5mS and is centered at 1.5kHz. Yet, it is over 20 dB down in level! Friends, loudspeakers just don't get much better than this - regardless of price or design or manufacturer. Bring on your ribbons and electrostatic panels boys...let's see their waterfall plots!
The graph is sloping upwards because the sweep rate required to reveal such time dependant anomolies does not allow sufficient dwell time for the analyzer input filter to receive the full energy at lower frequencies. This is a function of the measurement process - not the speaker. The more you know about time, the less you know about frequency - the Heisenberg Uncertainty Principle at work here.
http://en.wikipedia.org/wiki/Uncertainty_principleAnd for those that may squawk about the smoothing - if you've ever tested speakers, you'd know why we use it. Without some smoothing of the graph you'd see nothing but "grass" in place of a line for a trace. You can't make out what you're looking at unless you smooth it out of the response.
OK, let's look at Group-Delay. It's easier to understand than Phase.
Moving vertically on the graph above the horizontal axis line that reads "zero," energy is arriveing later in time. We say that "zero-line" represents the point of no propagation delay in the transmission of the signal. Another way of looking at it is to say the zero-line is "now" and anything arriving above that line is behind "now" - in a sense the energy is running late getting to us. The higher on the graph a certain frequency is, the later it will arrive behind those closer or at the zero-line. The amount of delay is designated in milliseconds.
As we can see, the lower frequencies progressively arrive later in time above the high frequencies. Above 2kHz all enegy appears to arrive together at the same moment in time. As we can see, the Timepiece 2.1 does not provide perfectly constant signal delay characteristics. Low frequencies below 2kHz arrive later in time than those above. But the amount of delay is what is significant here. Other than the lack of absolute perfection, this is exemplary performance. At 1kHz the signal delay is approx. 500 microseconds. A 1kHz wavelength is approx. 1 foot in length so therefore, it takes approx. 1 millisecond (1,000 microseconds) to propagate to begin with.
The upshot is that energy at 1kHz is delayed by 1/2 cycle of the waveform. As the frequency decreases below 1kHz, the signal delay does not increase significantly above that at the 1kHz level. Yet, the wavelengths at decreasing frequencies are getting progrssively longer. This means the percentage of delay per wave cycle is decreasing as well. The significance in all this is that the total average delay is only about 500 microseconds and occurs at relaively low frequencies.
Over-all it represents a very small level of impact on the quality of reproduced sound as the frequency of the signal delay "step" is well below the realm of frequencies wherein the wavelengths are relatively short and the ear is most sensitive - i.e., the region between 3kHz and 5kHz. It would be interesting to know just how many systems are designed to cross over in that frequency range. Now you know another reason why so many speaker systems sound more alike than not! Crossing over anywhere near 3kHz is not a good idea if you're concerned about signal delay effects.
Then finally we have the good 'ol frequency response graphs.
Don't pay much mind to how flat this pair is, even though they're exceptionally flat. More important...compare one graph to the other. Sorry, my system can't print out overlays. I can see 'em - just can't print 'em. Still, if you go back and forth between #1 and #2 you're likely to accuse me of just making a photocopy. If you look closely though you can see where they differ by a tad in the size of one ripple compared to the same ripple in the other graph. This is what makes for good imaging/soudstage - matching L/R frequency response - not time-domain/phase response!!! Don't let the "first order network" guys fool ya! It has NOTHING to do with linear phase response. Linear phase (or lack thereof) affects transient response - and rumor has it we kick butt there too. In fact, some say the best they've ever heard. But who can say?
Well, that wraps up this little dissertation. Hope it helps. At least now some of the claims may be more believable for some of you. Does all this mean the Timepiece is a "Quad killer"...maybe...maybe not. But it does suggest you may have good reason to wonder.
Just remember: The more you know, the better we sound!
-Bob
