I've identified an additional, consistent trigger for DSD Converter failure: Invalid parameters — namely, waveform behavior at the very start of the audio stream.

Files beginning with immediate high amplitude — no silence or fade-in — often crash the converter. In one case, the waveform hit near-maximum volume at sample 1. After inserting a brief silent lead-in and gradual fade using Audacity, conversion succeeded.

This suggests the SDM may not handle abrupt signal ramps before internal stabilization. Even re-encoding alone (e.g., via FFmpeg) doesn't solve this having tried even manually converting to PCM stream and back; even cleaning up the header with a HEX editor. Unless the waveform itself is softened at the start, it will crash.

Your program is awesome, but more descriptive errors would go a long way in helping users pinpoint the issue, understand where in the chain it starts, and troubleshoot efficiently.

We're glad to play a small part in supporting your work by sharing what we've observed.

Best regards, Petrina Riccardo

P.S. Maybe an extreme DSD like 1024 would help but I did not want to wait the time.