More Taps for digital Bliss?

[Purité Audio]

does having more 'taps' improve sound quality ?
Chord appear to believe so.
http://www.chordelectronics.co.uk/chord-dac-technology.asp

Keith.

 

[Thomas savage]

I know the more I ' tap dat ass' the better life is, I guess music sounds better too. So yes the more taps the better. Mind you I am not much for taking a interest in tapping Dave or Hugo.

 

[Mivera]

does having more 'taps' improve sound quality ?
Chord appear to believe so.
http://www.chordelectronics.co.uk/chord-dac-technology.asp

Keith.

If you compare a modern Intel Quad core Skylake processor to the FPGA they use in that DAC, it's like a Bugatti Veyron engine vs a Vespa scooter engine.

See this thread if you want to know how the forward thinkers are doing DSP circa 2016:

http://audiosciencereview.com/forum...t-gen-cutting-edge-digital-audio-systems.254/

 

[bibo01]

It depends on the filter type.
In response to Chord, in its closed-form filter HQPlayer has 16k taps, input-relative. You multiply that with the conversion factor and you get total number. For example: 44.1 kHz to 11.3 MHz output it means 4M taps...

 

[Mivera]

It depends on the filter type.
In response to Chord, in its closed-form filter HQPlayer has 16k taps, input-relative. You multiply that with the conversion factor and you get total number. For example: 44.1 kHz to 11.3 MHz output it means 4M taps...

Yes this is because Hqplayer uses the same technology Interval Zero uses for it's DSP engine.

 

[amirm]

The number of taps is a figure of merit for FIR filters. The more taps, the lesser the ripple, and sharper the response. The drawback is increased memory, computational horsepower and latency through the system.

 

[amirm]

Here is an example showing what happens when you change the filter order:

Increasing that allow the transition from what is passed to what needs to be stopped shorter/more steep.

 

[Mivera]

The number of taps is a figure of merit for FIR filters. The more taps, the lesser the ripple, and sharper the response. The drawback is increased memory, computational horsepower and latency through the system.

This is the exact reason Hqplayer requires so much horsepower to preform it's best algorithms.

 

[amirm]

This is the exact reason Hqplayer requires so much horsepower to preform it's best algorithms.

Indeed. With so much fast memory and lightning fast speed of CPU cores, you can go crazy with filter lengths and such.

 

[Opus111]

Aren't Intel CPUs using 'BruteFIR' rather than direct convolution? Certainly its much faster for long filters to use FFTs.

 

[Ken Newton]

does having more 'taps' improve sound quality ?
Chord appear to believe so.
http://www.chordelectronics.co.uk/chord-dac-technology.asp

Keith.

Chord's, so called, Watts-Transient-Aligned (WTA) reconstruction filter - as I understand it - is directed at an rather interesting subject. Which is the temporal resolution with which continuous-time signals can be reconstructed from discrete-time samples. The more perfect the brickwall band-limiting during reconstruction, the greater the temporal resolution of the reconstructed signal. Which, of course, potentially has significant implications for stereo signals. How many taps are featured by an FIR filter is sort of a secondary issue. The real issue is the length of the filter's impulse response, which is set by the number of coefficients in the filter's kernel. So, a longer impulse response implies more filter taps and a sharper filter slope. In short, the longer the filter impulse response, the greater the number of taps utilized in the FIR convolution, and the sharper then is the band-limiting and the finer the reconstructed signal's temporal resolution. So, how much temporal resolution, and thereby, filter taps are needed? Chord is suggesting that it's much more than are generally utilized.

Also interesting here is that many NOS fans would argure that NOS produces the better sense of temporal resolution and stereo illusion because it has no filter taps.

 

[Opus111]

I put the question to Rob Watts himself over at Head-Fi that an analog reconstruction filter would have the longest impulse response (essentially infinite). His reply was that such a filter wouldn't be linear phase and hence introduce phase distortion. But it seemed to me at the time that he was assuming the ADC anti-aliasing filter was linear phase. In any case, ADC anti-aliasing filters in ADC chips have fairly short impulse responses as the designers care about latency.

 

[Ken Newton]

Yeah, the WTA would seem to be nothing more than a FIR reconstruction utilizing as many taps as possible, except for their cryptic suggestion that the WTA somehow improves signal temporal accuracy over other FIR filters even should both feature an identical number of taps. I can't imagine how they might do that unless it's by incorporating some proprietary windowing of the SINC function. So many technical roads, not all of which lead to a musical Rome.