Phase shift

[maty]

Only some words about phase shift, some designers do care, just now: https://www.diyaudio.com/forums/solid-state/344540-alpha-nirvana-39w-8ohm-class-amp.html

Hugh [Hugh Dean, aka AKSA] has been doodling on LTSpice and he came up with something really extraordinary: a 43% efficient SE Class A amp that can drive 39w into an 8ohm load with the usual Aksa-approved harmonic profile and low phase shift...

Updated with image:

- End off topic -

 

[Julf]

So what audible effect does the phase shift have?

 

[NTK]

So what audible effect does the phase shift have?

To answer this question a few months back I mathematically created a few WAV files based on the 2 kHz square wave. File 'square1.wav' is the 2 kHz square wave band limited to include only up to its 9th harmonic (18 kHz). I phase shifted the harmonics to my 'taste' to create 'square2.wav' and 'square3.wav' so the waveforms all look very different from each other (see the Audacity screenshot). If you perform FFT on these files, the magnitude plots will look identical, but the phase plots will (obviously) be different.

I listened to these 3 files with headphones by putting them into a playlist and loop repeatedly. I couldn't tell of any transitions, and the sound appeared constant and continuous, at least to me. This gave me some indication on how well my hearing can detect phases.

 

[Killingbeans]

- End off topic -

Okay... So, what is the rest of the thread supposed to be about?

 

[splattened]

Okay... So, what is the rest of the thread supposed to be about?

Maybe he meant "end, off topic" signaling that we can discard the pretense of discussing phase shift and start the off topic conversations now instead of page 2 or 3. He knows his audience.

 

[Cosmik]

So what audible effect does the phase shift have?

Bruel & Kjaer paper on phase response of speakers.

A poor phase response has no influence on the reproduction of pure sines; nor on steady state music, such as a sustained chord from an organ. But it shows up in transients, such as booms from kettle-drums, or bass drums, pizzicato from strings, short blasts from horns, attack on piano and guitar and the clash of snare drums, cymbals and triangles....
...If the boom from the kettledrum is considered under such a situation, it results in a coloration of the reproduced signal — too boomy if the bass arrives first at the ear; and too sharp if the mid frequencies arrive first.

 

[ahofer]

I think the “off topic” notation is a legacy of the thread from which this post was removed (1ET400 amp). It was OT in that thread, but is the main topic here.

1) What would be an audible measured phase shift at different frequencies?

2) since the phase shift of Class D in the old thread was actually a time delay, wouldn’t that be of no consequence unless you had multiple amps in your setup with different time delays?

 

[JohnYang1997]

Phase basically = Fr
if you have twice of the flat fr you want, there is no phase shift to worry about
If you interpret this sentence, it's a non issue for amplifiers even dacs(mostly). But can be causing audible difference in speakers/transducers.

If it's delay, then it's never an issue it's no difference from you play the music 1 sec later or 1 sec earlier.

 

[Julf]

To answer this question a few months back I mathematically created a few WAV files based on the 2 kHz square wave. File 'square1.wav' is the 2 kHz square wave band limited to include only up to its 9th harmonic (18 kHz). I phase shifted the harmonics to my 'taste' to create 'square2.wav' and 'square3.wav' so the waveforms all look very different from each other (see the Audacity screenshot). If you perform FFT on these files, the magnitude plots will look identical, but the phase plots will (obviously) be different.

I listened to these 3 files with headphones by putting them into a playlist and loop repeatedly. I couldn't tell of any transitions, and the sound appeared constant and continuous, at least to me. This gave me some indication on how well my hearing can detect phases.

That matches my own experiences and tests. Phase-shifting harmonics changes the shape of the wave in spectacular ways, but produces no audible difference.

 

[Julf]

Bruel & Kjaer paper on phase response of speakers.

Coincidentally they happened (at least at the time) to be selling phase measuring equipment.

I was disappointed to see that the paper didn't address verifying the audibility of phase shifts in any way.

 

[KSTR]

That matches my own experiences and tests. Phase-shifting harmonics changes the shape of the wave in spectacular ways, but produces no audible difference.

Try 100Hz + 200.5Hz, or any other test frequency in the 50...300Hz range plus its H2+0.5Hz. Besides the pitch of harmonic being very slightly off, what else do you hear when you compare that to the version with an exact harmonic?

 

[Julf]

Try 100Hz + 200.5Hz, or any other test frequency in the 50...300Hz range plus its H2+0.5Hz. Besides the pitch of harmonic being very slightly off, what else do you hear when you compare that to the version with an exact harmonic?

I hear a beating caused by the frequency mismatch.

 

[maty]

Why I love my modded KEF Q100 speakers? Thanks to a very good 5.25" coaxial I have time aligned ALL frequencies ->"magical" sound versus traditional 2-ways (KEF 6.5" coaxial has not these "magical"). With acoustic instruments the "magical" is more great.

The problem with the delay is if they are different in woofer versus tweeter, like 2-ways.

https://en.wikipedia.org/wiki/Loudspeaker_time_alignment

[ Loudspeaker time-alignment usually simply referred to as "time-alignment or Time-Align" is a term applied to loudspeaker systems which use multiple drivers (like woofer, mid-range and tweeter) to cover the entire audio range. It is the technique of delaying the sound emanating from one or more drivers (greater than 2-way) to correct the transient response, improve accuracy and, in non-coaxial drivers, improve the directivity or lobe tilting at the crossover frequencies. ]

In class A and AB, the phase is not exceeded by +/- 6º in the audio band, although it is rarely publicized. Why should class D have another criteria?

https://www.audiosciencereview.com/...-of-purifi-1et400a-amplifier.7984/post-254975

Hypex NC400, measured by amirm at my request. The graph shows how the phase varies with frequency, progressively, but varies.

Of course, loudspeakers with a good DSP you can solve the problem -> Kii and many others. But traditional loudspeakers without DSP...

 

[KSTR]

I hear a beating caused by the frequency mismatch.

Almost right, but the beating is NOT from the frequency mismatch. There is no beating seen on the spectrum.
The root cause is that at low frequencies we are sensitive to exact waveshape, and hence moving harmonics around can be audible.
You now may repeat the test, with exact harmonics shifted around in 45° steps, so you have 8 files, group them in sets of two, phase 180° apart for the second file. With one specific set of those, depending on the phase response of your speakers/system, the contrast in "body" or "fatness" is largest, while with the ortogonal set, 90° off from the first, the difference is minimal.

Some research seem to show that our hearing mechanism physically differentiates and squares the incoming SPL, which explaines a great amount of hearing phenomena in the simplest way, including some that can't be explained with other theories of hearing. Look up Heerens & de Ru, "Applying physics makes auditory sense : a new paradigm in hearing"

 

[Julf]

In class A and AB, the phase is not exceeded by +/- 6º, although it is rarely publicized.

Sorry, but what is that claim based on?

 

[Julf]

Almost right, but the beating is NOT from the frequency mismatch. There is no beating seen on the spectrum.

There is no beating on the spectrum, because there is no intermodulation at that stage. There is intermodulation (and beating) in your ear.

The root cause is that at low frequencies we are sensitive to exact waveshape, and hence moving harmonics around can be audible.

If you say so, but your example is not evidence of that.

You now may repeat the test, with exact harmonics shifted around in 45° steps, so you have 8 files, group them in sets of two, phase 180° apart for the second file. With one specific set of those, depending on the phase response of your speakers/system, the contrast in "body" or "fatness" is largest, while with the ortogonal set, 90° off from the first, the difference is minimal.

That test would only be valid for exact harmonics, not by ones off by 0.5 Hz.

Some research seem to show that our hearing mechanism physically differentiates and squares the incoming SPL, which explaines a great amount of hearing phenomena in the simplest way, including some that can't be explained with other theories of hearing. Look up Heerens & de Ru, "Applying physics makes auditory sense : a new paradigm in hearing"

How about a direct pointer to the study?

 

[Cosmik]

I was disappointed to see that the paper didn't address verifying the audibility of phase shifts in any way.

How many liberties with the signal are acceptable? For sure, phase shift and/or delay can be inaudible when you're listening for it. And many other variations would be too, such as blurring of the stereo image. And it would be even less obvious if you were comparing two systems with different types of blur rather than blurred vs. non-blurred.

Of course the idea behind lossy compression is that you modify the signal in a way that can't be heard when you're listening for it, although people can, apparently become attuned to it. Maybe you can do the same with phase shifts, or maybe it's never anything but a subconscious sensation.

It always seems odd to me that the default is assumed to be 1930s technology with a waveform that emerges from the speaker that doesn't look like the signal that goes in. Anyone who tries to make the emerging signal a time domain duplicate of the recorded signal is considered a crank who has to justify what they're doing, rather than the other way round!

 

[Julf]

How many liberties with the signal are acceptable?

Basically anything that isn't audible.

It always seems odd to me that the default is assumed to be 1930s technology with a waveform that emerges from the speaker that doesn't look like the signal that goes in.

Why does it matter how it looks? What matters is how it sounds. Take a 10 kHz square wave. Put it through a 15 kHz low pass filter. Notice how it doesn't look anything like the original square wave? Also notice it sounds exactly the same.

Anyone who tries to make the emerging signal a time domain duplicate of the recorded signal is considered a crank who has to justify what they're doing, rather than the other way round!

The burden of proof is on the one making claims that contradict established, evidence-based studies.

Nobody considers you a crank for trying to make the emerging signal a time domain duplicate of the recorded signal, but if you make claims about audibility of stuff, you'd better line up your evidence.

 

[maty]

Hybrid preamplifier:

https://www.diyaudio.com/forums/sol...er-single-smps-hybrid-preamp-post5653264.html

Almost 0º at 100 kHz.

 

[Cosmik]

Basically anything that isn't audible.



Why does it matter how it looks? What matters is how it sounds. Take a 10 kHz square wave. Put it through a 15 kHz low pass filter. Notice how it doesn't look anything like the original square wave? Also notice it sounds exactly the same.



The burden of proof is on the one making claims that contradict established, evidence-based studies.

Nobody considers you a crank for trying to make the emerging signal a time domain duplicate of the recorded signal, but if you make claims about audibility of stuff, you'd better line up your evidence.

Aren't you, in effect, saying that if a viewer doesn't spot a man in a gorilla suit wandering past, then the man in the gorilla suit must 'not be visible'?

The person producing the 'blameless' system doesn't need to prove anything. The burden of proof lies with the person who is knowingly (or neglectfully) modifying the signal.