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

As to why the filter modes attenuate by 6dB over native but only in USB bridge mode: "When we stream bit accurate (blue light) using the the Concero as USB->SPDIF bridge, we don't touch the data to be bit perfect. When we add our apodizing or IIR filters, we must drop amplitude by some amount to allow for the headroom required by the oversampling process. All D/A chips with internal oversampling do this inside the chip but the user never sees it. Since we provide an optional upsampled output, its gain must be backed off or distortion results. Note that in any product which performs oversampling—be in player software or hardware—one must allow room for the Gibbs phenomenon. This is essentially the overshoot fundamental to any system without infinite bandwidth. We simply provide enough headroom in the bits to not clip the oversampling process. If any software does its own oversampling but matches output levels, there will be distortion in the music but interestingly not visible in the sine wave. Not everyone is aware of this but it's immediately audible.

"We were left with a choice. A/ Do we lower the bit-accurate level to match the filter levels? This would no longer be bit accurate for native mode. B/ Do we ignore the headroom required for oversampling and match the oversampled level with the native level and simply live with the distortion (note that sine waves still measure really well, it's only the transients that misbehave)? C/ Do we remain bit accurate but also provide correct headroom for the filters (this is the best of both world but creates the level-matching issue)? The latter was our choice. Note that this is only the case for USB->SPDIF. For SPDIF->analog and USB->analog, levels are perfectly matched since we took care of the required headroom by setting the full-scale output at 1.2Vrms."

About Sabre's usual DPLL jitter attenuation vs. the Concero: "We are not using a PLL to lock into the input rate like adaptive USB would. The Concero uses asynchronous USB and that same clock drives the Sabre. Hence the dual phase-lock loop in the Sabre is inactive. Its master clock is already 'synchronous' to its bit and word clocks. The Concero uses a 49.152MHz clock for the 48k, 96k and 192kHz inputs being an integer multiple of 1024 x 48kHz, 512 x 96kHz and 256 x 192kHz. For the 44.1k, 88.2k and 176.4kHz inputs we run a 45.1584MHz clock which is an integer multiple of 1024 x 44.1Hz, 512 x 88.2kHz  and 256 x 176.4kHz. The non-PLL action thus refers to the fact that we already have two low-jitter oscillators which create the master clock to not need PLLs to regenerate it. The upshot is that we bypass the usual jitter reduction in the Sabre simply because in USB mode we don’t have any jitter in the input clock to start with. That said, the Sabre jitter reduction is activated when using the S/PDIF input since now the incoming 44.1kHz S/PDIF stream is not phase-aligned to the 1024 x 44.1kHz clock reference. In this case the jitter reduction in the Sabre chip will do its job as usual."

On filters in USB bridge mode. "When we apply our own upsampling filters (IIR or apodizing) to the S/PDIF output, they do this by moving the data to 4 x its rate, i.e. to 176.4kHz or 192kHz. Yes you will still encounter the filter inside whatever DAC appears downstream to the Concero in bridge mode but remember that said DAC is now getting data at 176.4kHz or 192kHz. That means the impulse response of the downstream converter has been shortened by four times. So if the DAC downstream exhibits pre-ringing, that ringing will now be reduced by a factor of 4 when the Concero provides initial upsampling.  By upsampling data before it is sent to the external DAC, we really push that DAC's filter 'out of the way'.  

"The terms upsampling and filtering are sometimes used interchangeably. Filtering implies that you are changing the phase or frequency response of a signal. Upsampling implies some form of interpolation to get the data to a higher rate. The confusing bit is that in order to upsample or interpolate, one must apply some form of filter. This filter's job is to typically remove the images present in the data stream due to its sampled nature. This is called an anti-aliasing filter. The Concerto actually performs its 4 x upsampling by applying a filter on the data at the upsampled rate. How this filter is written greatly affects soundstage depth and the perception of transient attack/decay times.

"There's a tradeoff of course. A very sharp filter does a better job of removing sampling artifacts at the expense of softening the attack times since it requires a filter with a longer time-domain response which essentially smears the data a bit. On the other hand with a really short impulse response and no smearing of the data, the aliasing images will be more abundant. There's no free lunch with 44.1/16 data. The other important thing to note is that the choice of IIR (faster attacks with more artifacts) or apodizing filter (smoother attacks with less artifacts) should be tailored to musical content. There is no best or perfect filter for all music types. Some will prefer the IIR filter for one genre and the apodizing one for another. We believe that the user needs to have options for the ultimate listening experience. The reason we add two upsampling modes is for people without filter options to get a taste of how much better things can get without having to replace their existing DAC."