Audibility of "typical" Digital Filters in a Hi-Fi Playback
Reply #691 – 2014-11-26 16:15:05
Sorry, but as written this is very misleading. The response shown is that of a digital FIR filter. In whole or in part, it is a linear transform, down to the arithmetical limits of the DSP. Such limits were 120dB down a couple of decades ago. You talk about hearing this. I am not convinced. I didn't talk about "hearing this." As in "this" being a specific thing. As I keep saying, the nature of ABX tests is not qualitative. And at any rate, the tests I performed were combination of filtering and quantization which is not a linear transform. Distinguishing files or even observing subjective difference doesn't spit out the cause. Stuart's listening tests however does examine "this" to fair extent. They isolated just the filter and tested its audibility and got positive outcome to statistical confidence. And they were not searching down in 120 db SPL. They compared filter lengths between 44 and 48 Khz and hypothesized the difference in audibility to filter length. Clearly none of their subjects was hearing the extended ultrasonic response in frequency domain between 22 and 24 Khz. As I just correct Arny, JJ's slide very clearly confirms the same that the brain is not a DSP machine where it captures a bunch of samples and then decides what it means. Note how JJ wrote that in capital letters. And this was my key point to xnor. That he can't use DSP explanations to rule out audibility. The brain's model of perception does not match the mathematical model of filter transforms. At the risk of stating the obvious, it can't hear the future! It has to judge the samples as they arrive and modified by the filter coefficients. So sure, we can have our doubts but we can't rule it out as impossibility based on combined science of signal processing and psychoacoustics as xnor attempted. Edit: usual typos.