Hi Vinnie,
Quote:
Originally Posted by Vinnie R. /img/forum/go_quote.gif
Actually, I don't know if I didn't have enough coffee this morning or what, but that peak in the top end actually doesn't look that bad (0.5dB is not much, especially at such a high frequency that our headphones or ears might not even be able to hear
).
Like I just said, I must of not carefully looked at this plot (I was extremely busy today and in a big rush in general today). I don't think that the moderate peaking has much to do with my impressions I quoted above. I think it is just the overall sound of this dac + built-in output stage... whoever makes this part!
|
A measurement like the frequency response only shows part of the characteristics of a device. That's why I've also added the phase response, impulse response and step response.
My experience as speaker filter designer and DAC filter designer is that small deviations in a frequency range can be heard. The difference between a rising curve from 20Hz to 20kHz or a decending curve from 20 to 20kHz are clearly audible. A rising curve sounds more "analytical, metallic, zizz", whereas a decending curve sounds more "acoustical, harmonic rich". This experience is shared by more people that develop audio equipment. This is something that I can hear with the iPod Classic as well. Though it sounds precise, tight and crisp, it adds a "mettalic haze" to voices, making them "more electronic", and "less acoustic".
Whether it is the frequency response on itself, or the phase response that is associated with such a response, is for me an unanswered question. There are many indications (and publications) showing that we are very sensitive to timing deviations. Take for instance the difference between a Butherworth or Besselfilter. With both of them you can create curves with a passband ripple within 0.1dB, but their group delay in the passband already shows deviations, and both filters are reported to have their own sonic characteristics.
As one can see is that the iPod Classic has a significant phase (and hence group delay) deviation in the passband. This is also shown by the impulse response and step response. Furthermore, by comparing the minimum phase with the measured phase, one can conclude that the phase deviation is caused by a non-linear effect.
I'm not an expert in the field of DA conversion, but these differences could as well be caused by sigma-delta conversion or noise-shaping techniques, where non-linear operations in the frequency domain are applied.
Nevertheless, to summarize about the sound quality in sonic terms, the 6G sounds precise, crisp, but lacks stereo image and has an electronic haze to the sound. The 5G sounds less precise, but its timbre contains more harmonic information and sounds less electronic. For me, the 5G is closer to how I experience acoustic music in real life, and for me is the better sounding device overall. This is not to say that the 6G is a bad sounding device! It's just less than the 5G. (I've added this to my web pages).
Regarding measuring versus hearing. If there is an audible difference, there must be a measurement that can show the difference. The point is that the interpretation of measurements is not always simple or straightforward. The measurements I've published only show a part the electrical transfer function of the device. To end with a quote that we very oftenly use at work: "Measurements are knowledge, but you have to know what you measure".
Marc