Soundstage

[Cosmik]

My understanding is that the delay is the charge and discharge time of the capacitor through the resistor.

What if it's done with DSP? No charging or discharging, no accumulator, but a simple bit of code that says output = input - prev_input?

Phase shift = delay..?
Lag? Lead? Is a phase lead an example of time travel into the future?

 

[PierreV]

What if it's done with DSP? No charging or discharging, no accumulator, but a simple bit of code that says output = input - prev_input?

Phase shift = delay..?
Lag? Lead? Is a phase lead an example of time travel into the future?

No, because of... previous input...

 

[sergeauckland]

What if it's done with DSP? No charging or discharging, no accumulator, but a simple bit of code that says output = input - prev_input?

Phase shift = delay..?
Lag? Lead? Is a phase lead an example of time travel into the future?

One issue with all digital filtering or pretty much any calculation is latency. It can be short, but it can't be zero. Filters especially can have long latency as every tap involves sending an output back to an input and round again. On DACs, a PLL can be pretty much perfect if it's got a long enough locking time, typically several seconds, but commercially nobody wants to press play then wait a long time for an output, especially when that output has to be locked to video.

No free lunches with analogue or digital.

S

 

[Cosmik]

No, because of... previous input...

Well, differential calculus. I can make the time tend to zero, but I still get a differentiator that gives me the 90 degree phase shift and whose output starts immediately I apply the input.

I can demonstrate a practical device that performs instant phase shifting (I swap two wires). It's a phase shift. No delay was involved. To some people it looks like a delay or travel into the future, but in fact it happened instantaneously.

 

[Cosmik]

One issue with all digital filtering or pretty much any calculation is latency. It can be short, but it can't be zero.

But we are talking about two different time delays here. Yes, a practical DSP differentiator needs a delay, but for thought experiment purposes this can be regarded as tending to zero. This is not the delay that people are ascribing to a phase shift i.e. related to the frequency of the sine wave and what proportion of a cycle it has shifted by.

 

[andreasmaaan]

If you differentiate a sine wave with a C and and a R, say, you will get a phase shift. Where was time involved? Where was there any delay?

Of course, the network will introduce a non-constant group delay dependant on the values of C and R and the load impedance (and FWIW corresponding to the effect on the amplitude response).

Or am I missing your point?

 

[Cosmik]

Of course, the network will introduce a non-constant group delay dependant on the values of C and R and the load impedance (and FWIW corresponding to the effect on the amplitude response).

Or am I missing your point?

My point is that a phase shift is just a specific form of distortion. At the output of the box you get something that looks like the input delayed by some proportion of a cycle, yet when you applied the sine wave to the input, the output started up almost instantaneously, certainly nothing like the delay that (to some people) a phase shift looks like.

If I create a filter that creates a phase lag of 1 degree to an incoming sine wave, is that the same as a delay? Logically, then, if I create a complementary filter that causes a phase lead of 1 degree that must be either an anti-delay i.e. a prediction of the future (which someone above says it can't be), or it's a 359 degree delay. 1 degree's worth of delay and then suddenly 359 degrees-worth of delay? Something does not compute.

I don't see phase shift and delay as interchangeable. A delay might cause what looks like a phase shift on a repeating waveform, but it is really just a delay. A phase shift might look like a delay (or a prediction of the future), but it is really just a specific form of distortion.

 

[andreasmaaan]

To be honest this discussion lost me long ago. The only debate here is about semantics. There is no substantive disagreement nor any useful problem to solve.

I don't see phase shift and delay as interchangeable. A delay might cause what looks like a phase shift on a repeating waveform, but it is really just a delay. A phase shift might look like a delay (or a prediction of the future), but it is really just a specific form of distortion.

But will respond anyway

I don't recall saying a phase shift is delay. What I said was that phase is timing. What perhaps I should have said was that - in the context in which I made my original statement, i.e. differences between two close-to-identical loudspeakers - phase and timing are alternative ways of describing the same phenomenon.

The reason I originally chose the word phase as opposed to timing was to reflect the fact that these differences in this real-world context are likely to be frequency dependent as opposed to constant across all frequencies.

 

[Cosmik]

I don't recall saying a phase shift is delay. What I said was that phase is timing. What perhaps I should have said was that - in the context in which I made my original statement, i.e. differences between two close-to-identical loudspeakers - phase and timing are alternative ways of describing the same phenomenon.

The reason I originally chose the word phase as opposed to timing was to reflect the fact that these differences in this real-world context are likely to be frequency dependent as opposed to constant across all frequencies.

But if I have two channels (could be full spectrum audio or perhaps a narrow bandwidth signal) but one of them is going to a satellite and back and I want to line their timing up, am I going to think in terms of a black box that applies the right phase shifts to the right repeating waveforms at various frequencies, or am I going to want a box that applies a delay to the signal? (A box whose output genuinely does nothing for a finite time after the input is applied.). I think we need to maintain the distinction

 

[Krunok]

phase and timing are alternative ways of describing the same phenomenon.

Exactly. And I strongly believe this "phenomen" shouldn't be called "distortion". Or maybe one could call it like that, but then it should be called "time distortion".

One must admit "time distortion" sounds much more fancy, almost like a phrase from Star Trek, while "phase shift" sounds so common..

 

[andreasmaaan]

But if I have two channels (could be full spectrum audio or perhaps a narrow bandwidth signal) but one of them is going to a satellite and back and I want to line their timing up, am I going to think in terms of a black box that applies the right phase shifts to the right repeating waveforms at various frequencies, or am I going to want a box that applies a delay to the signal? (A box whose output genuinely does nothing for a finite time after the input is applied.). I think we need to maintain the distinction

Sure! But this is not the context we were talking about, which was how well speakers image.

In this context, it is presumed that both speakers receive the input signal at the same time. What is important, then, is when one speaker outputs the signal relative to the other. If there's a difference between the two speakers, it is likely to be due to mechanical or electrical differences between the drivers (or crossover). Irregularities, therefore, are likely to be highly frequency dependent, despite the fact that broadly speaking the signal will output from the two speakers simultaneously.

 

[andreasmaaan]

Exactly. And I strongly believe this "phenomen" shouldn't be called "distortion". Or maybe one could call it like that, but then it should be called "time distortion".

One must admit "time distortion" sounds much more fancy, almost like a phrase from Star Trek, while "phase shift" sounds so common..

Ha I think you have to admit that anything other than linear phase is a form of (yes, time) distortion. Then we are back to standard distortion questions like "What are the thresholds of audibility?".

 

[Krunok]

Ha I think you have to admit that anything other than linear phase is a form of (yes, time) distortion.

Yep, I do admit.

Then we are back to standard distortion questions like "What are the thresholds of audibility?".

I'm slowly loosing hope such questions will be answered during our lifetime..

 

[Dogan]

My point is that a phase shift is just a specific form of distortion. At the output of the box you get something that looks like the input delayed by some proportion of a cycle, yet when you applied the sine wave to the input, the output started up almost instantaneously, certainly nothing like the delay that (to some people) a phase shift looks like.

If I create a filter that creates a phase lag of 1 degree to an incoming sine wave, is that the same as a delay? Logically, then, if I create a complementary filter that causes a phase lead of 1 degree that must be either an anti-delay i.e. a prediction of the future (which someone above says it can't be), or it's a 359 degree delay. 1 degree's worth of delay and then suddenly 359 degrees-worth of delay? Something does not compute.

I don't see phase shift and delay as interchangeable. A delay might cause what looks like a phase shift on a repeating waveform, but it is really just a delay. A phase shift might look like a delay (or a prediction of the future), but it is really just a specific form of distortion.

 

[wiggum]

I've not seen any evidence that both sides of a headphone are any better matched than loudspeakers, in some cases I've seen, they are clearly not.
As to loudspeaker pair matching, some are matched much better than others, whether that's done by driver selection or tighter quality control over all drivers I don't know.

S

Plenty of examples in the headphone world.

Here is one I found quickly.
https://www.rtings.com/headphones/reviews/kz/as10

L & R FR are very well matched. Phase response are also well matched.

If you spend more time, you can find more examples on rtings.com

 

[Theo]

Coming back to the "phase" question: we should speak of phase spectrum, as phase is related to frequency. The question is then of what happens if a phase shift is created by a system at some frequencies, and more importantly, if the response of each channel (as measured in the room) is different in its phase spectrum (i. e. the channels are not perfectly matched). I would guess that the effect would be a changing image depending on the note played. Instruments would then be appearing to move around or loose focus as the harmonic content would be spread all around the place.
I would think that equalizing only is not helping to improve the imaging provided by a system (it may even degrade it if the algorithm/hardware is not 0 phase at all frequencies). Aren't some phase adjustments needed when trying to compensate for the room response?

 

[andreasmaaan]

I would think that equalizing only is not helping to improve the imaging provided by a system (it may even degrade it if the algorithm/hardware is not 0 phase at all frequencies).

I think this statement would be correct for interchannel differences that are not minimum phase in nature, but incorrect for differences that are, i.e. for differences that are minimum phase, simple minimum phase EQ is the correct solution.

Aren't some phase adjustments needed when trying to compensate for the room response?

Definitely for non-minimum phase discrepancies IMO.

 

[Theo]

Definitely for non-minimum phase discrepancies IMO.

Aren't room induced discrepancies non minimum phase by nature, as they are a convolution of the original signal by a non minimum phase reflection pattern?

 

[Biblob]

 

[andreasmaaan]

Aren't room induced discrepancies non minimum phase by nature, as they are a convolution of the original signal by a non minimum phase reflection pattern?

There's a discussion of this in Dr Toole's book. I can't link to that here, but there is some discussion of it in this older article by Dr Toole, where he says:

"Room resonances at low frequencies behave as “minimum phase” phenomena, and so, if the amplitude vs. frequency characteristic is corrected, so also will the phase vs. frequency characteristic."

I think that is a slight over-simplification, but it is true in the case of at least some room modal behaviour.

There's also a discussion in John Mulcahy's article on the REW website, where he says:

"Room responses are mixed phase, meaning there are some minimum phase regions and some regions that are not minimum phase. The minimum phase regions tend to be at lower frequencies, but we cannot simply say a response is minimum phase below some specific cutoff."

I know that's not a complete response to your question, but suffice it to say that room modal behaviour tends to be minimum phase, but the specific characteristics need to be measured in a given room to determine what behaviour is minimum phase and what is not.