This review page is supported in part by the sponsors whose ad banners are displayed below

Upsampling, DSP and other critters
While agreeing on the importance of separating various circuit sections of his player, Reith prefers one big toroid over two smaller transformers, hence his potted 80VA choice runs multiple secondaries to feed the transport, converter, DSP module, output stage and touch pad discretely. The underlying rationale is well-known and practiced collectively. Various sections should not communicate or 'cross-talk' via a shared power supply and particularly the spinner's servo control should not impinge on the susceptible converter chip.


For spin duties, HIFIAkademie relies on the stout Philips CD Pro2 which also appears in high- and ultra-priced machines. This Philips assembly is contained in a metal frame whose upper cover Reith replaces with a bigger aluminum plate that sits atop rubber-suspended and sprung legs to 'float freely'. Decoupled from the main enclosure, it becomes a sub-chassis transport by definition. For rationale, you'd expect a "chassis resonances on the read-in process are minimized to increase fidelity". Reith concedes that possibility but finds it nearly impossible to prove. His reason was to decouple laser-head movements from the main enclosure to avoid mechanical noise. That's passion gone the extra mile.
Be it CD or one of the digital inputs, the signal path is identical. The data stream first encounters the upsampler, then DSP module, then the converter and finally the output stage. Even though he wouldn't commit to any particular upsampling theories, Herr Reith naturally does have his reasons. The cdPlayer processes all data at a unified 24-bit/96kHz. This doesn't improve the 16-bit/44.1kHz standard per se but the working conditions of the conversion process. Reith argues that D/A conversion filters tend to be optimized for a particular operating frequency. They otherwise work
less well and with possibly audible negative side effects. While acceptable input signal frequencies are variable of course, it simplifies the internal DSP's programming if it has to only work at one rather than variable sampling frequencies. Hence the upsampler precedes the DSP.
Reith is quite skeptical about the "legendary jitter suppression" via upsampling. Poorly implemented, upsampling can in fact end up as a veritable jitter generator. The higher timing precision of data processing which is possible with upsampling is more a byproduct of the applied math and the capacity of the necessary buffer which allows a shallower reconstruction filter. (Those who enjoy technical discussions on this subject might find Herr Doug Rife's Theory of Upsampled Digital Audio PDF worthwhile reading. Rife is the brains behind the Melissa loudspeaker measurement system.)
The DSP processing occurs post upsampling. If not engaged, it simply passes on the data untouched. To have this option is obviously highly unusual for a CD player in this class. Its DSP circuitry and software coding were all developed by Reith himself "to do exactly what I wanted". Such hands-on coding from a small firm is anything but customary but why shouldn't Made in Germany not also apply to immaterial goods? To access the DSP functionality requires nothing more than a laptop or PC connected via
USB. That's after a small Reith-written program has been downloaded and installed from his home page. Once called up, it looks like this:


It offers 8 different filters (high-pass, low-pass, Linkwitz etc) and 16 address frequencies. Per channel. If you wonder why bother, imagine the following simplistic scenario. You're suspicious that your upper bass is somewhat thick due to speakers and/or room while on top, you lack a bit in airiness. An effective correctional contour might look like the middle curve.


Groovy? One part of our audience will condemn it as meddling with the signal and thus sacrificing fidelity. The others remember vividly that "damn, we've sunk hundreds into Cable & Co. and this program can precisely control slope, frequency and knee of frequency response corrections in a CD player? Halleluja!" Both reactions are right and wrong. It is signal manipulation, albeit before conversion to analog. But now take a look at your room's impulse response. Would you consider going homeless just because your room manipulates the signal in such a wholesale fashion?


Just remember how that fancy floor-to-ceiling wall of glass won't get less reflective just because you twiddle with some frequencies; just as loudspeakers which blur the transients will still blur them, just somewhat louder or quieter. DSP is not a panacea but it can be a very useful Swiss army knife. Before you wield this knife, become informed as to its nature, know its bypass settings and consider assembly of a small test rig so you needn't rely on your ears exclusively.


(Sample correction of HIFIAkademie client: The 40-60Hz mode is attenuated, the midrange rise diminished and the treble somewhat boosted.)

Behind the DSP module, BurrBrown PCM1794A converters with 8-times oversampling handle digital filtering but these filter characteristics can be set to flat or soft via the touch panel. Sonic differences are mild. Using just one rather than paralleled DACs seems advantageous to Reith. While S/N ratio suffers a tad, he calls it preferable over remaining distortion that peaks cleanly above a lower noise floor. (This recalls the dither argument where addition of noise during mastering masks small-signal distortion. Wiki explains dither here but Stephen Dawson explains it better.)

To pursue these matters when his chip is already specified at 0.0004% THD and 127dB dynamic range is fascinating. Ditto for his socketed operational amplifiers in the output stage which allow one to "roll op-amps and tweak the sound to personal system and liking. The final 5% performance isn't due to specific parts but appropriate interactions. The ultimate op-amp or circuit is a myth. One will sound better in one installation than another." Those inclined to tweak won't need the soldering iron in this instance.