I am hearing clicking at the end of events, after some investigation I think I have found the source of the problem.
In my test recordings it appears that the fmod "fade out" property fades the volume in 20ms steps (possibly occuring on frame boundaries). This is fine for big fades (over 500ms), as the steps are very small. But for fast fades (100ms and less) it’s a problem, as the last step often results in a click. The shorter the fade the more obvious the click.
Has anyone else encountered this and have any solutions? Perhaps this could be fixed in a future update?
- Jogo asked 10 years ago
10 years later and I encountered the same problem with any way I could figure out to fade in Fmod. Funny thing ist that at the end and start of loops regions there is a smooth linear fade out/in in about 1ms, which makes it impossible to have a perfect zero crossing loop. I need to cut every sound clip extrernally before using them in Fmod.
- You must login to post comments
I hear it. on lots of stuff. sound definition crossfades, event fade in/out, volume curves on parameter changes.
It is more noticeable on low frequencies as a kind of high zippering noise. Highs and midrange sounds simply mask this kind of distortion more.
It would be great if volume changes that occur over the course of a mixer frame would ramp more smoothly, even if it meant a frame’s worth of volume change latency. Just so long as it sounds smooth as silk.
To get a picture of zippering in a waveform output, play a ‘silent’ maximum DC offset sample and change the volume in various ways while recording to disk. Then try that same sample in various audio editors and sequencer/synths and see their zippering. Sadly, lots of stuff has it.
fmod’s mixer -does- ramp volume changes. Its possible you have it set to hardware and it is using directsound’s bad quality mixer. Try forcing the wavebank to software or switch the output mode in the settings to winmm to stop it using directsound’s mixer.
I’ve rendered a series of test events to disk from fmod designer running on windows using the wave writer output option.
Analysis of the output reveals fades as a stepped series of volume ramps.
During a step of fade, there is a smooth ramp that lasts 64 samples, but then the volume does not change for about 20 ms until the next step.
For example, a 160ms fade will result in 8 abrupt 64 sample volume ramps with 8 20ms plateaus rather than a smooth ramp over the 160ms fade duration.
Because of the way this step quantization of the various fade curves works, sometimes a 64 sample ramp can have a very steep slope, and when some frequencies play trough such a steep volume change, an audible click can occur.
The shorter the fade, the worse the rate of chunky fade ramp step quantization, though I have seen some big level jumps on longer fades as well.
I will send my fades test project to Firelight, but I don’t see a way to attach it or pictures of the test output to this forum.
Please login first to submit.