Hi.
From what I understand about aliasing, prevention is better than a cure, using LPF to cut out of band aliasing artifacts.
However I am left with a problem. Recently I have been experimenting with producing sounds in a home made synth (pcm wave table ). The aliasing artifacts it produces present themselves as what I describe as a disharmonic hiss. Not quite white noise as some subharmonics are produced. The aliasing is noticeable in the decay part of the envelope. During use the aliasing does not sound bad per se, and adds some complexity to the sound that I'd like to keep, the decay 'hiss' is the down side.
I am looking for suggestions of noise reduction software which can be 'taught' what noise is 'noise' and which can (with the help of gating) help remove the worst of the disharmonic hiss.
Many thanks for any help in advance.
Greg
From what I understand about aliasing, prevention is better than a cure, using LPF to cut out of band aliasing artifacts.
However I am left with a problem. Recently I have been experimenting with producing sounds in a home made synth (pcm wave table ). The aliasing artifacts it produces present themselves as what I describe as a disharmonic hiss. Not quite white noise as some subharmonics are produced. The aliasing is noticeable in the decay part of the envelope. During use the aliasing does not sound bad per se, and adds some complexity to the sound that I'd like to keep, the decay 'hiss' is the down side.
I am looking for suggestions of noise reduction software which can be 'taught' what noise is 'noise' and which can (with the help of gating) help remove the worst of the disharmonic hiss.
Many thanks for any help in advance.
Greg
Last edited:
I think what you're hearing is quantisation noise rather than aliasing noise, most noticable with low bit width samples or output (i.e 8 bit samples).
Aliasing is the metallicness to the sound that you hear as you play higher and lower notes and it is independent of volume level, and depends on sample rate of the device and what the quality of the samples themselves are, the lower quality the more aliasing there will be.
Aliasing is removed by agressive lowpass filtering, but quantization noise is lot more difficult to deal with as it is level and sample dependent. Some combination of EQ and noise reduction can get you somewhere, in CEP you can create a noise profile from a segment in a recording and apply it in the parts that need fixing but more often than not the sound you try to doctor will get harmed more than made nicer.
Aliasing is the metallicness to the sound that you hear as you play higher and lower notes and it is independent of volume level, and depends on sample rate of the device and what the quality of the samples themselves are, the lower quality the more aliasing there will be.
Aliasing is removed by agressive lowpass filtering, but quantization noise is lot more difficult to deal with as it is level and sample dependent. Some combination of EQ and noise reduction can get you somewhere, in CEP you can create a noise profile from a segment in a recording and apply it in the parts that need fixing but more often than not the sound you try to doctor will get harmed more than made nicer.
If it is quantization distortion, you can change it into a 4.77 dB louder white noise without distortion by means of 2 LSB peak-to-peak triangular probability density function dithering. Audacity includes triangular PDF dither, in GoldWave you have to enter an expression to get it to dither.
(Actually you can get rid of the distortion and make it only 3.01 dB louder by using 1 LSB peak-to-peak uniform probability density function dither, but then you are left with noise modulation.)
(Actually you can get rid of the distortion and make it only 3.01 dB louder by using 1 LSB peak-to-peak uniform probability density function dither, but then you are left with noise modulation.)
Last edited:
Don't know if running the sound thru a computer is an option but DSP will give the best results.
Bad Request
Bad Request
But, isn't the problem with aliasing that you have no way to distinguish between a properly sampled signal and an aliased artifact? I suppose if your signal is able to be predicted, and thus you really didn't need to sample the aliased material in the first place, then you could re-synthesize the stuff that got aliased from some other characteristic of the signal that you have now. But, in the general case, a sampled signal by definition only exists from -Fs/2 to Fs/2, and you have no way to know how the signal got mapped into the passband, or whether anything was the result of aliasing or not.
Agreed. In my case, I'm typically looking for stationary phases at lower-than Nyquist that are coherent between signals through time. With stacking techniques, I just need to make sure that aliased phases don't have more energy than those I care about.
My thought relating to audio signals is that this is an efficient way of reducing spikes at high frequencies. However, I'm certainly not an electrical engineer and only barely a non-rookie in diyaudio.
My thought relating to audio signals is that this is an efficient way of reducing spikes at high frequencies. However, I'm certainly not an electrical engineer and only barely a non-rookie in diyaudio.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.