paradoxper
Headphoneus Supremus
Is your wallet jittering in anticipation?
I've been looking forward to the Yggy specifically for quite a while, so just a bit. Good thing Yggy isn't priced to be wallet hemorrhaging.
Is your wallet jittering in anticipation?
I read that. That's cool and all but anyone can claim anything sounds better, and all I have read so far is contrarian talk about SD being bad, ASRC being evil, that people with expensive hifi systems are not the target market, mix in some cable skeptic humor etc. No talk of actual performance benefits over conventional optios. Call me unconvinced.
What irritates me is that this anti SD, anti ASRC idea will probably spread and suddenly everone will decide that these technologies are to blame for whatever they subjectively disliked in any particular DAC without any solid argument of causal link.
TGIF eh, JJ?
On the other hand, a very respectable number of Hi-End companies, especially so States side, but not exclusively, have never adopted 1-Bit topologies. Some have done so briefly, only to return to Good Old Fashioned Multi-bit. A notable example is Naim, that have remained steadfastly on the Multibit side, and have only very recently- relunctantly, I am sure- employed Delta-Sigma chips in their latest Digital gear. There must have been very good reasons for this, and certainly not those of economics...
If I understood any of the DAC talk it would probably be very exciting etc.. But fankly, I don't
But for those of us without a background in DSP design, why is this filter better? Does it have better jitter rejection? After all this I still don't understand what advantage this bit perfect filter represents.
Actually it's reasonably simple to understand at a shallow level (which is the only way I "understand" it, so what could be bad about that, right?):
All digital filters used in DACs these days (or by software players that take the digital filtering out of the DAC and do it in the computer before the file is sent to the DAC), whether sigma-delta, or R2R or some other variant of multibit, use a type of mathematics called Fourier transforms. Fourier transforms have this thing called "conjugate variables." Wotthehellizzat?, you ask? It's two variables that are related in that they move in opposite directions - as one gets smaller the other gets larger, or as one gets more precise the other gets less precise. (An electron's position and its momentum are conjugate variables, which is the basis of the uncertainty principle in quantum mechanics.) For the digital filters in every single DAC or software player now on the market, frequency domain behavior and time domain behavior are conjugate variables.
This means as frequency domain behavior of your filter gets better (less distortion from aliasing, for example), the time domain behavior gets worse (ringing). (If you don't know what aliasing and/or ringing are, not to worry - the basic idea is you'd rather not have either.) So because these are conjugate variables, every single digital filter currently on the market is the filter designer's idea of a good compromise between frequency domain and time domain optimization, since right now you can't have both.
What Mike's done is step outside this whole problem that everyone's been wrestling with for a couple of decades and said "We don't need no steenkin' Frenchy math!" (Well, no, I doubt he said anything like that, but anyway....) He's designed/designing (don't know whether he feels he's done yet) a filter using math that doesn't have this conjugate variable dilemma, so that it can, in his words, be "optimized for...time and frequency domain."
Hope that helps.
Actually it's reasonably simple to understand at a shallow level (which is the only way I "understand" it, so what could be bad about that, right?):
All digital filters used in DACs these days (or by software players that take the digital filtering out of the DAC and do it in the computer before the file is sent to the DAC), whether sigma-delta, or R2R or some other variant of multibit, use a type of mathematics called Fourier transforms. Fourier transforms have this thing called "conjugate variables." Wotthehellizzat?, you ask? It's two variables that are related in that they move in opposite directions - as one gets smaller the other gets larger, or as one gets more precise the other gets less precise. (An electron's position and its momentum are conjugate variables, which is the basis of the uncertainty principle in quantum mechanics.) For the digital filters in every single DAC or software player now on the market, frequency domain behavior and time domain behavior are conjugate variables.
This means as frequency domain behavior of your filter gets better (less distortion from aliasing, for example), the time domain behavior gets worse (ringing). (If you don't know what aliasing and/or ringing are, not to worry - the basic idea is you'd rather not have either.) So because these are conjugate variables, every single digital filter currently on the market is the filter designer's idea of a good compromise between frequency domain and time domain optimization, since right now you can't have both.
What Mike's done is step outside this whole problem that everyone's been wrestling with for a couple of decades and said "We don't need no steenkin' Frenchy math!" (Well, no, I doubt he said anything like that, but anyway....) He's designed/designing (don't know whether he feels he's done yet) a filter using math that doesn't have this conjugate variable dilemma, so that it can, in his words, be "optimized for...time and frequency domain."
Hope that helps.
I don't think this is mathematically accurate. The uncertainty principal holds whether or not you choose to think of a signal in time-domain or frequency-domain. You can't be localized in both time AND frequency --- simply not mathematically possible. The frustrations with using fourier analysis, where signals are broken down into a sum of sine modes (sine modes have infinite extent in time and are perfectly localized in frequency) led to the development of methods such as wavelet analysis, where one can use a different set of test signals to (i.e., wavelets) to decompose a time domain signal. here, Wavelets have some localization in time and some frequency content; however, they too cannot be localized in both time AND frequency.
Digital filters are not usually implemented using any sort of direct application of fourier transforms on the signal. One would need the entire waveform to transform, manipulate, and inverse transform back to the time domain. DSP use finite impulse response filters (which use a finite number of samples to filter) or infinite impulse response filters (which use feedback). Fourier techniques can be implemented on small chunks of the signal at a time. In this case, the windowing of the data introduces spectral artifacts because of the truncated signal. Fourier methods would best be applied as a post-processing techinque where the entire waveform in available all at once. In streaming audio applications it runs into windowing problems because of truncation.
Cheers