Krassi
100+ Head-Fier
Wow! Thanks for this amazing example!
Lovely!
Seeya !
Lovely!
Seeya !
Hi @manueljenkin,By coverage I meant signals measured along with their phase for starters. You can have a different wave with the same harmonic amplitude spectra as a square wave, with the differences being in the relative phase. So when I want to analyse a signal via fft, I expect a complex fft or having both amplitude and phase info. This will also expose in case there are any artefacts in the windowing method used. And in case of timing differences, capturing the exact timing difference, in this case, something at 10us delay and reconstruct within a 44.1khz sampling if it is possible.
Thank you. I'm now looking at those articles first and deriving the full theory (for the whole set of possible aliased signals). For your text, I'll try to do the same for a few other frequencies and multitones. 2756.25 is a perfect multiple of 44.1khz (16x). I'd want to experiment with a non conventional number to make sure there isn't masking of certain phenomenon.Hi @manueljenkin,
Something to play with... The time column is approximate; the sampling rate is 44.1kHz. Each wave is 16-bit integer, and exactly 2 cycles of a sine wave (n.b. you can use a rectangular window), with f=2756.25Hz, and no dither.
In my sheltered little corner of the world, all Fourier transforms give bivariate results (re-im, amp-phase, cos-sin, etc.), otherwise you couldn't investigate the time delays I've introduced. I chose a small number of points in case you want to do it by hand (DFT), otherwise any FFT should allow you to confirm: all waves have the same amplitude and interesting values for time delays. If you want longer records or a .wav file, let me know.
Extra credit: You'll notice that the values in the different columns differ considerably. What is the minimum possible difference? For sine waves, what minimum ∆t does that correspond to?
LOL. I get it.Thank you. I'm now looking at those articles first and deriving the full theory (for the whole set of possible aliased signals). For your text, I'll try to do the same for a few other frequencies and multitones. 2756.25 is a perfect multiple of 44.1khz (16x). I'd want to experiment with a non conventional number to make sure there isn't masking of certain phenomenon.
Also the bigger interest will be on the signal that we get after interpolation using sinc filter across the missing points to a fair precision get a visualizable output. Comparing just samples is not an easy job
"I thought you were disputing the ability of PCM to represent timing differences less than the sample period" - I am.
Yeah I was meaning bandwidth limited.LOL. I get it.
This is the 6th (and final) version of this post: I keep wanting to give you hints, but realize although we're both "prove it to yourself" types, we still do so in very different ways.
I will mention that the waves I uploaded are also about the 6th versions... I chose every aspect of them to maximize the efficiency for someone like me(!!! - good or bad?) figuring things out. But we are different, so some of my choices weren't appropriate for you.
FYI, when you say "aliased signals", I assume you mean "anti-aliased signals". I usually say "bandwidth limited" (same thing).
Have fun!
EDIT:I hope it's already clear that this is incorrect:
[1] I think you are misunderstanding what I was expecting for a proof.
[2] Hope you are aware of things like windowing artefacts, sub Nyquist artifact's (I'm still learning there), and that's where I'm looking to find the proof.
[3] As of now, nope, I'm not convinced yet, but I highly appreciate your effort.
This is the most entertaining thread I have ever had the (mis)fortune to read on head-fi.
Did you work through the Amidror paper? Did you find anything worth sharing?sub Nyquist artifact's (I'm still learning there), and that's where I'm looking to find the proof.
That is exactly what I am concerned about as well. Scientifically minded and well informed people are a minority compared to the people who are misinformed in varying degrees and the latter are exposed to all sorts of marketing bulshit far more than to correct information. For the first group to distinguish themselves from advocating "just another opinion" they should be correct, precise, consistant and complete.My concern is both audiophiles and others who think: I know that's false (a lie), so the rest of what bigshot says is/might be too, even if it's true and valuable. And worse, if you and gregorio and castle say mostly the same thing, but they catch your lie, they may discard all 3.