Wikipedia:Reference desk/Archives/Science/2023 February 6

= February 6 =

Different way of improving SWIFT's frequency response
Continuation of the first thread about the Goertzel filter ringing infinitely with damping cascaded in series and the related concept sliding windowed infinite Fourier transform and the second thread about the same topic. I made the demo about the different version of alpha-SWIFT in which it is cascaded in series similar to typical IIR filters do instead of subtracting the slow-decaying one with a fast decaying one calculated in parallel. But here's the question, is it simpler to adjust than it is described in the original paper as you only have to adjust the filter order? 2001:448A:3043:762C:545C:7B7D:EA1E:AFAC (talk) 19:40, 6 February 2023 (UTC)
 * I think your question is too specialised for anyone here to just know the answer. It is probably something that has to be determined by experience or experiment, or perhaps researching the literature. I expect that you can do this yourself, as no one here has done it for you. — Preceding unsigned comment added by Graeme Bartlett (talk • contribs) 22:41, 12 February 2023 (UTC)
 * All I can say is 4th order cascaded SWIFT is a bank of complex Gammatone filters obtained by combining first-order lowpass filter thing with complex rotation (which in turns, shifts the frequency response and turns into a bandpass filter) and cascading it in series. BTW, is 4th order cascaded SWIFT better than Hann windowed regular sliding DFT in terms of replicating the frequency response of our auditory spectral analysis? 2001:448A:3041:7E63:D3E:5407:F6DD:3DF5 (talk) 02:31, 13 February 2023 (UTC)
 * Asking which digital process best models human hearing asks one to consider alternatives on the criteria of A) their relative computabilities or B) their result's agreement with human listening tests. Regarding A) the OP has done tests with interesting results on which I won't draw conclusions beyond a comment that computation speed is only critical if one is looking for an analysis in real time. Regarding B) my experiences in seeking a quantitative objective measure of the subjective impact of audio distortions, e.g. noises that interfere with a telephone call, are based on mapping the linear frequency scale of a FFT analysis to a psychoacoustical scale such as Bark scale. In that work there seems no strong reason not to chop the distorted audio into the sample lengths of individual FFT's and to take the average result. Note that the motivation for this work is to replace the need for hours of multi-listener opinion rating, which is very expensive to arrange with statistical rigour, with a digital process that can claim to deliver a credible quality rating (MOS) without human intervention. This analysis may go some way towards saving expensive human listening tests on competing lossy audio codecs. However the OP chooses not to segment the audio and to apply instead a continuous Sliding DFT. The motivation needs clarifying because of the uncertain side effects of applying multiple DFT's with their individual Window functions overlapping. If the intention is to model human hearing with a fidelity good enough to identify musical instruments in an orchestra, individual performers or audio target recognition then these are research areas where Wikipedia can offer references and encourage but not provide or even publish original research. The Gammatone filter that models the "auditory filter" created by the cochlea, the sense organ of hearing within the inner ear has been mentioned. The Mel scale (after the word melody) of subjectively equidistant pitches may also be appropriate in music analysis that probably will need to incorporate actual Auditory masking effects in time and frequency. I reiterate an invitation to the OP to register a Wikipedia account and to look into improving the article Sliding DFT where issues can be raised on the talk page. Philvoids (talk) 20:17, 13 February 2023 (UTC)

Service Ceiling for aircraft
What sort of ref do we need for this and for what sort of performance? This just came up in the F-22 article for unknown reasons and ISTR that the invulnerability of the B-36 was based on USAF jet fighters of the time being able to climb up to its height, but not turn at that height. So when then they tried to attack they'd stall. (Not a problem for USN or USSR of course.) Hcobb (talk) 21:43, 6 February 2023 (UTC)


 * The obvious reference would be Jane's All the World's Aircraft, an annual publication by a leading company specializing in this type of domain information.
 * It is important to add caveats that Jane's probably publishes the best available public information - and they have a blurb to this effect in their book - much is not known, or is intentionally obfuscated, surrounding such military capabilities, especially relating to specific aircraft limitations.
 * Along the same lines: the book is commercially available and you can even find copies on major online book retailers at affordable prices - but if you're just trying to find and cite one factoid, .... it's a bit steep. "Those who need to know" probably don't need to refer to Wikipedia, anyway... Wikipedia is not a reliable source; it is not a data repository; and trying to find a trustworthy value for a specific numeric parameter is in many ways a perfect example of a place where Wikipedia's model works quite poorly - namely, free access, public editing, and the challenges of data curation.
 * A better strategy is to use Wikipedia to learn about the concept of an aircraft service ceiling, and cite, then refer, to other reliable sources to provide the precise data. We don't, (can't, and should not even try to...)... mirror every data entry in every edition of Jane's.
 * Nimur (talk) 21:46, 6 February 2023 (UTC)
 * See Ceiling (aeronautics). The absolute ceiling, as defined by the FAA, is “the altitude above sea level at which a climb is no longer possible.” A plane's service ceiling is arbitrarily defined as the altitude at which the plane can no longer maintain a climb rate of more than 100 feet per minute (propeller) or 500 feet per minute (jet). Philvoids (talk) 00:02, 8 February 2023 (UTC)