Low-pass related advice

Question

Hi,
Any thoughts on how to get the "best" sound for a live music recording.
I muted, in the spectral frequency display, all frequencies 25k and above.  This caused the recording to have more punch, immediacy, or whatever the right word is.  This seems to be an improvement.  Is this the right way to go for music?  Or, should I be using a low-pass filter, or a shelf-filter, or something else?  Or, should I not do anything like this at all?  Beyond that, what is the best frequency to use as the cutoff.
The digital field recorder that I am using is relatively quiet.  Regardless, I applied noise reduction with a noise profile as my first step.
Thanks,
Steve

Answer 1

RonNovy wrote:
The phase of the higher frequencies can distort the lower ones whether it's digital or analog.  Increase the bandwidth and add a filter to remove frequencies outside the range of human hearing and the distortions won't completely go away, but if you correct the phase relationship of each frequency then you can minimize the affect they will have on each other and get a more transparent sound.
That's a non-sequitur - for several reasons. Firstly you have no idea of what the individual harmonic relationships are supposed to be, and therefore can't 'correct' the phase relationships of anything. Secondly, you can't say specifically that it's 'distortion' as such without being a lot more specific about what the difference is, and how it may be percieved. Thirdly, it's well established that what is normally percieved as 'distortion' in this context is not harmonic-level related as such, but has everything to do with the timing of transient arrivals, and not their relative phases as such. Fourthly, it's been well established by quite extensive research that in terms of harmonics and timbre, anything beyond somewhere between the 5th and 7th harmonic of a fundamental make no difference to its perception - and that means pretty much that anything above 16kHz isn't going to affect the results.
As I suggested before, you might get problems if you get bad intermodulation distortion effects, but basically, if you want your results to sound 'smoother', simply roll of all the HF. But don't be under any illusions; that won't make them less distorted, it will actually make them more distorted - that is, if you use my definition of distortion, which is simply that if a reproduced signal is different when it comes out from the way it went in, it's 'distorted'. Will it be more 'transparent'? No. Transparency relies on nothing you can percieve getting between the original and reproduction. All you will be doing, to use a metaphor, is looking through a window that obscures the edges of everything - you get a nice smooth outline, but you don't see what's actually there. You would see what was actually there if the window had the same resolution as your eyes had, and exactly the same thing applies to an audio recording - if the entire system has a resolution that's the same (or very slightly greater than) your ears, you'll hear what you would have heard if you were there.
What you don't have to do is extend your recording in any way beyond the range of normal human hearing, because if you do you are likely to introduce potential problems, and not solve anything at all. Like a lot of the early stories about how 'digital' works (the ones about audio being 'sliced up', which is absolutely wrong, because it really doesn't work like that at all, by any analysis), the public has been fed a line over higher sampling rates for years, and it's simply not based on any acoustic facts at all. At one stage, higher sampling rates appeared to have an advantage, because they moved the point at which anti-alias filters had to operate completely out of the audio band. Since those filters had extremely poor transient responses, they made a perceptible difference to the results. But the important thing about this was that it was not caused by the digitising process per se, but the filters. Modern methods of oversampling, and much cleaner A-D design have removed these issues completely on newer processors, so the original estimate of what the maximum sample rate needed for audio was is still valid, and most importantly can be achieved without significant error. And it follows that anything at all added to this, in information terms, is going to be noise in the system.
How do I know all this? You have to bear in mind that I have an extensive background in electronic engineering, but most importantly I have a Masters in Acoustics, plus professional acoustic qualifications, and about 40 years of experience, both as a recording engineer and acoustician. And I should point out that I've hardly touched on the issues of Timbre perception - this is a complex subject, and to really get a grip on it you need to study it carefully; it does not work the way most people intuitively seem to think it does.