Cool, thanks for the clarification. I've done several reconstruction filter designs for NOS DACs, they're all freely available schematics on my blog and anyone with questions for clarification can post them on the blog. It was something of a challenge but 'very difficult' it wasn't. Its certainly possible not to impact the HF significantly but becomes more difficult (primarily inductor tolerances get very tricky), in the end I've gone for flatness to within half a dB up to 18kHz or so. I don't see why a custom oversampling filter would need to be designed - off the shelf ones are available. Some glue logic to adapt to I2S might be necessary though - which is a whole lot easier than starting from scratch on a filter. I have developed custom filter code for an ARM (M0, M3) processor myself which if anyone's interested I'd be happy to have a look for on my various disks in differing states of repair... When I did this I found that simply running the DAC faster made the sound worse so I've abandoned oversampling and stuck with NOS. The filter characteristics weren't the cause of the SQ degradation which was a loss of dynamics, a 'greying out' of timbres.