Reply to "Uniti Nova review in HiFi News"

Hi SJB -

I'm no filter expert by any means, but I read the Naim white paper (for the NDAC where they first introduced this filter) and two of the references and I get the impression they mean what they say. Although FIR filters are more generally linear phase, and IIR filters more generally not, it is possible to compute the coefficients for an IIR that preserves phase and has zero (acausal) pre-ringing as shown. Since their emphasis is low-cost / -intensity compute load in the fpga, I think their design was a "best of both worlds" approach rather than a simpler tool-box designed filter (especially since they use 40 bit floating point, which pushes the arithmetic S/N below the original encoding and the analog converter noise floor. This seems to be the polar opposite of the many thousands of taps FIR approach used in Chord DACs, where much higher computing is available and taking a larger pre/post sample dataset (is a wider convolution bandwidth) drives down the noise floor and improves the interpolation precision  ... but I don't have the impression that the Naim filter isn't what it says on the tin  

Regards alan

(edit)

ps - in this context, linear means zero phase shift; I'm not clear on how this might differ from minimum phase

×
×
×
×