Best Modplug output setting

[dBlues]

Lets find out about:
 * Which output anti-aliasing filter setting sounds best? Lets just use our subjective judgement - rate the detail, warmth, easiness for the ear (aching ears is a clear sign of something wrong).
 * Does it sound best when rendering to wave also?
 * Say something about how the sound changes between filter settings (if you can, after all it is pretty subjective)

No signal analyzing skillz/filter knowledge is required, but is considered valuable (SimonV:s tests results would still be nice to have ). Lets still say we use a samplerate of 48kHz and bit depth of 16/32.

E.g. Lpchip rated Polyphase with 78% cutoff the best, due to best balance of improved dynamics and lost detail.

[dBlues]

This test is made by listening same pattern with high-detail pads and rock-style, but very clean, high quality drums, and lead synth + bass. The whole frequency range is covered more or less.

Linear, sample ramping 0: Damped, a bit rough and hardest on the ears (causes ear ache). CPU: 100 (scaled for easier comparison).
Cubic spline: A bit smoother. CPU: 110
Polyphase: The sound has more "air", and sense of space gets through better. The cutoff factor can be used to make the song sound softer and warmer without losing too much detail. For maximum detail, use high cutoff factor and low sample ramping. CPU:110
XMMS/Hamming, cutoff 97%: The bass drum seems markedly stronger and punchier with this setting (sample ramping 0). The airy sound is retainded. Bass line sounds warm and nice. CPU:95.
XMMS/Hann, cutoff 97%: Mid frequencies go more "in the back" with this setting. Also not as punchy as Hamming. CPU:100
XMMS/Blackmann exact, same cutoff still: Mid-highs feel smoother and more distinct. There is a change of character. Bass line is less "in front"
XMMS/Blackmann 4 tap 74: The sound is not much different from other blackmann-filtering. CPU:100
XMMS/Kaiser 4 tap: Nothing much to say about this, probably my ears are getting tired Sounds like Blackmann filters. CPU:100

Overall: I think the winner was XMMS/hamming filtering, 97% cutoff, 0 ramping. The feeling of spaciousness and depth/feeling of bass were excellent, also the highs/mids were detailed but non- obtrusive. Strangely, seems this was the most CPU-economic setting (although not by far).

The differences were alltogether pretty subtle, but they were there. Perhaps the most surprising thing was that all settings consumed CPU almost equally (with all instruments and filtering, around 30%). I am running AMD athlon 64 3400+.



As subjective as this test was, maybe you end up with similar results?

[KrazyKatz]

This is an excellent effort dBlues. I think a test like this has been due for some time. When time permits I'm going to do one myself.

[rewbs]

Thanks! This would be even more useful if you could provide a link to your test pattern. Watch out for clicks if you use 0 ramping. The difference between the resampling algorithms is most noticeable in situations where aliasing is very obvious (take a sine wave sample and play it at a wide range of different frequencies).