As I had mentioned in a previous blog entry, iFi released a version of the NEO iDSD called the 'Performance Edition'. The reality of this release is it is the same hardware as the first NEO iDSD to hit market at end of 2020. The differences are a newer firmware and a couple hardware add-ons.
Eventually iFi released the 3.33 firmware to all NEO iDSD owners. The major feature added by this upgrade is a choice of digital filters, similar to the Gryphon. (they are both xDSD platform based, and much of this also overlaps with the work done a few years earlier with the iDSD Pro.
I caught a small bug in the firmware right off the top. When GTO filter is selected, the info readout shows 'DXD 352' when, for example playing back 44.1khz and its multiples. The other filters simply show the actual sample rate. There is nothing wrong here, just a peculiar anomaly. I suspect that every filter option except for GTO are the onboard Burr-Brown chip filters. (the Bit Perfect setting is not really a filter.. there is no oversampling) The GTO filter, however, is part of the FPGA/XMOS digital processing, and behaves just like it does on the iDSD PRO. When playing anything other than Bit Perfect mode on the iDSD Pro, the display shows the oversampled rate, which is always DXD x2 (705.6khz or 768khz). Coming back around to the NEO iDSD, the small bug is once GTO is selected, the indicator display will always show the oversampled rate rather than the input rate, regardless of which filter is selected. It isn't a big deal when you understand what is going on. Everything is functioning properly, but can understandably be confusing.
Hi-Fi News review of the Gryphon pointed out several of the filters were misidentified, and that the minimum phase filter was not a minimum phase filter at all.
I have confirmed in my testing of the 3.33 firmware that all filters EXCEPT minimum phase are true to their name. As Hi-Fi News uncovered with the Gryphon, the Minimum Phase filter on the NEO iDSD is actually a short linear phase filter, the standard filter only being slightly longer.
The impulse response for each filter is below.
As far as overall performance differences? Yes, there were some measurements that showed the fw3.33 version to be slightly more jitter prone, but the results were not repeatable and are of no great a significance to change anything.