On Sat, Jun 25, 2011 at 01:55:05PM -0500, Gabriel M. Beddingfield wrote: And what if the phase is < -pi/2 or > +pi/2 ? -- FA

2011/6/25 Fons Adriaensen &lt;fons(a)linuxaudio.org&gt;rg>: you mean the maximal amplitude (-MAX <= x <= +MAX) , I guess ? since x <= A (always), that result is not possible -- E.R.

On Sun, 26 Jun 2011 00:22:58 +0200 Jörn Nettingsmeier &lt;nettings(a)folkwang-hochschule.de&gt; wrote: Bingo! Which is why it is quite insane for people to worry about phase shift in amplifiers. As long as all channels are the same so there is no phase *difference* it's quite irrelevant. -- Will J Godfrey http://www.musically.me.uk Say you have a poem and I have a tune. Exchange them and we can both have a poem, a tune, and a song.

I think you understood what I am looking for below. Does anyone have a code example for this type of filter?

Would this result in some kind of reverb type effect as Fons suggested? I'm looking for a way to adjust the phase of a signal rather than the amplitude. Does such a plugin already exist? If not which ladspa plugin would be the most suitable to start from? I've looked at using parts of jamin and librubberband but there may be a faster to implement option available. Cheers -- Patrick Shirkey Boost Hardware Ltd

On Mon, Jun 27, 2011 at 02:35:23PM +0200, pshirkey(a)boosthardware.com wrote: One of them would sound like a resonance, it would 'smear out' some frequency bands. Many of them combined could be made to sound like a reverb. Such filters are commonly used as part of reverb algorithms, e.g. zita-rev1 has 8 of them. Tricky. A sudden change of phase would sound as a click, and a gradual one would be a change of frequency. This is because phase and frequency are related: frequency = derivative of phase, phase = integral of frequency. Are you somehow trying to match phases across a crossfade or something similar ? There's no general solution for that. The way apps like e.g. zita-at1 do this is to estimate the pitch (fundamental frequency) of the signal and either skip or jump ahead an integral number of cycles with a crossfade. That way the phases do match. There's AFAIK no solution for this for an arbitrary signal. Ciao, -- FA

On Sun, Jun 26, 2011 at 10:44:33AM +0200, pshirkey(a)boosthardware.com wrote: Unfortunately, I can only guess what the context of your question was, and I'd probably be wrong :-( The example Erik gave is a perfect one. As said, there's no such thing as 'the phase of a waveform'. For a general waveform, and no matter what interpretation of 'phase' you'd choose, it would be a function of frequency. If the waveform is cyclic (e.g. a triangle wave) you could define some 'phase' value on it, using either the fundamental frequency or some arbitrary point in the waveform as reference. But even for a simple sine wave the term 'phase' can mean different things. Take s(t) = sin (w * t + phi), with w = 2 * pi * f All the following are correct: (1) If you take 'phase' as a property of s(t) as a whole, you could say its phase is phi. (2) If look at absolute phase at time t, it would be w * t + phi. (3) If you use t = 0 as a phase reference point, the phase at time t would be w * t. It all depends on the context which one you use. To add some more ambiguity, compare s1(t) = sin (w * t + phi) s2(t) = cos (w * t + phi) In many cases it doesn't matter which one you use when defining or explaining something. If you have a maths background you'd prefer cos() for real signals, since that's the real part of the complex single frequency signal exp(j * (w * t + phi)). If you are defining e.g. a oscillator opcode in a synthesis system you'd prefer sin(), as this starts at zero for phi = 0. In both cases you could legitimately refer to 'phi' as 'the phase'. But the two waveforms are 90 degrees out of phase w.r.t. each other... Ciao, -- FA

On Sun, Jun 26, 2011 at 11:43:46AM +0200, Jörn Nettingsmeier wrote: Right. And 'linear' here means 'without a constant term' - we don't want our system to be a Hilbert transform for example. Correct. It it the derivative of the phase response w.r.t. angular frequency (minus that value if your convention is that a delay corresponds to positive time). Group delay actually tells us how the 'envelope' of a signal is modified by nonlinear phase response, something we can easily hear on any 'percussive' signals. Let w = 2 * pi * f Suppose you have some filter that has a non-linear phase response, e.g. P(w) = a * w^2 (radians) The corresponding phase delay is D(w) = P(w) / w = a * w (seconds) The group delay is G(w) = dP(w)/dw = 2 * a * w (seconds) Now if you have a relatively narrowband signal centered at some frequency w1, e.g. a 'ping' with a gentle attack, then it would appear to be delayed by 2 * a * w1, not a * w1, because what we hear as delay is the delay on the envelope, not on the 'cycles'. Ciao, -- FA

On Saturday, June 25, 2011 05:04:43 pm Emanuel Rumpf wrote: *headdesk*. Yes... should have been the half-amplitude of the signal... p = asin( 2*x / A ) I.e. MAX = A/2 and MIN = -A/2 -gabriel

You wouldn't happen to have an example of this code? Perhaps there is something in qtractor along these lines already? I have looked at integrating something with librubberband but first things first right? -- Patrick Shirkey Boost Hardware Ltd

On 06/26/2011 10:50 AM, pshirkey(a)boosthardware.com wrote: so this is about panning? that's actually pretty easily done with just a time delay in addition to level difference. unless you want to spread complex sounds out in space, in which case you replace the delay with a frequency-dependent allpass. careful with the term "haas effect". all the poor guy did was check the time window in which an echo would not be perceived as a separate event. basically, the haas effect says you can get hide a P.A. delay system that's 10dB louder than the main P.A. if it hits between 10 and 30ms later, and still maintain good "on-stage" localisation, with the delay being practically inaudible. sometimes also called the "law of the first wavefront". this has nothing to do with stereo localisation. time delays relevant for left-right localisation are in the 0 - 1ms range, some authors give a bit less, others a bit more. that is precisely what dual-fft tools for p.a. system calibration do, and they're extremely useful. they allow you to constantly monitor the system with _program_ material, without having to use MLS noise or any other specific measurement signal. one channel is used for the direct signal from the mixer, the other is fed by a measurement microphone with delay compensation. you get instant phase and amplitude response. good systems also give you an additional "confidence" curve that tells you how much you can trust which parts of the spectrum. for instance, if your program material is a boy soprano, the confidence of the measurement in the low end is practically zero.

On 06/26/2011 05:47 PM, Arnold Krille wrote: it will do the job nicely in its difference setting, but you have to find (and compensate for) the delay manually, like a real man. modern p.a. tools allow sissies like yours truly to just click on the "find delay" button and be done :-D

On Sun, Jun 26, 2011 at 02:54:48PM -0400, gene heskett wrote: The 'quadrature component' of any waveform can be found by applying a Hilbert transform. Which in this context means just a FIR filter. It will have some delay, but if the only purpose is measurement that's no problem. This is utter nonsense. You could as well claim that the earth is flat disk and ignore all evidence that it isn't. If you want to contribute anything useful, please stop spreading this sort of manifest lies. Ciao, -- FA

On the subject of the word Quadrature: The word Quadrature is shorthand for the imaginary component of a complex-valued variable. Furthermore, it is almost exclusively used in the domain of carrier modulated waveforms. The word for the real component is In-Phase. That is why so many texts on the subject use I and Q to describe these two concepts. This choice of words was due to the fact that some radiomen in the '20s created a 90-degree delayed signal from the carrier (a quatrature relationship with respect to the carrier in the ancient college trigonometry literature) and (I doubt ignorantly) left out the fact that it was simply a practical way to perform complex analysis on a narrow-band waveform. @Gene: It was unfortunate that you used the mathematically reserved word complex to describe a complicated signal. The concept of quadrature in an audio baseband signal assumes a single frequency, say 1000 Hz., wherein that arbitrarily chosen "carrier" is assumed for Quadrature analysis. In fact, the Fourier Transform assumes an infinite number of such carriers to create a continuous, complex-valued transform of the pure real audio channel. The FFT is a Discrete Fourier Transform that periodizes the waveform to create a finite number of discrete carriers, each at the bin related to the length of the transform. Each bin of frequency output by that transform has a real and an imaginary component -- Quadrature. @Gordon, please remember that your SDAR code ASSUMES the carrier frequency so as to be able to analyze Quadrature. It is a more complicated concept in general complex analysis of a broadband signal such as audio. Duncan -- In the limit as productivity approaches infinity all jobs become entertainment or neurosis. On 06/26/2011 05:02 PM, Gordon JC Pearce wrote: