Does Phase Distortion/Shift Matter in Audio? (no*)

[AmadeusMozart]

To summarize, it seems that there are these primary questions here:
1. How much phase distortion are we talking about?
2. What frequencies are we talking about?
3. Is this specific amount of phase distortion at these specific frequencies audible?

Perhaps some of the more knowledgeable persons are able to provide an answer (or some insight) for me.

Using LtSpice I am modelling a single ended tube amplifier using UL and CFB using only local feedback in the output stage.

All is fine, I get at 400Hz 0.65% distortion at onset of grid current (= maximum drive).

But when running simulation at 50Hz I get 3.25% distortion.

Reducing the signal capacitor between driver (SRPP) and output tube from 1uF to 50nF gives 1% distortion but phaseshift goes from 168 degrees to 141 degrees (amplifier is inverting due to SRPP) at 50Hz (at 400 Hz 179 degrees).

I am not sure but suspect that the inductance of the output transformer introduces some phase shift that then is offset by the phase shift in the smaller capacitance.

But the conundrum is: what will be more audible: the higher distortion or the higher phase shift?

FWIW I've read that 2% second harmonic distortion of 30Hz at 60Hz is actually perceived as being louder than the fundamental 30hz. Don't know how true that is.

When I grew up "HiFi" was defined as a true reproduction of sound between 50Hz and 15kHz.

Looking for your response, many thanks in advance and am now ducking under a chair after showing my ignorance. (wink)

 

[Ingenieur]

Red

Perhaps some of the more knowledgeable persons are able to provide an answer (or some insight) for me.

Using LtSpice I am modelling a single ended tube amplifier using UL and CFB using only local feedback in the output stage.

All is fine, I get at 400Hz 0.65% distortion at onset of grid current (= maximum drive).

But when running simulation at 50Hz I get 3.25% distortion.

Reducing the signal capacitor between driver (SRPP) and output tube from 1uF to 50nF gives 1% distortion but phaseshift goes from 168 degrees to 141 degrees (amplifier is inverting due to SRPP) at 50Hz (at 400 Hz 179 degrees).

I am not sure but suspect that the inductance of the output transformer introduces some phase shift that then is offset by the phase shift in the smaller capacitance.

But the conundrum is: what will be more audible: the higher distortion or the higher phase shift?

FWIW I've read that 2% second harmonic distortion of 30Hz at 60Hz is actually perceived as being louder than the fundamental 30hz. Don't know how true that is.

When I grew up "HiFi" was defined as a true reproduction of sound between 50Hz and 15kHz.

Looking for your response, many thanks in advance and am now ducking under a chair after showing my ignorance. (wink)

Reducing C increases phase ang V to I. That is how power factor is corrected, adding C.


Easier conceptualizing in the +/- 90 deg quadrants and assuming V angle to be 0 deg, we only care about the delta V to I, not absolute. Adjust by adding 180 deg.
1 uF -12 deg
50 Mac -39 deg

Phase angle = arctan(X/R) or tan(ang) = X/R
X = (Xl - Xc)
Xl = 2 Pi f L j = 2 Pi f L / 90 deg
Xc = 1/(2 Pi f C j) = 1/(2 Pi f C) / -90 deg
R = resistance
Z = R + jX

X/R @ -12 deg = -0.21
Basically R ~ 5 x X and lagging inductive

X/R @ -39 deg = -0.81
Basically R ~ 1.2 x X and lagging inductive

X has increased lagging by reducing C which is leading

L : I lags V
C : I leads V

Assume X = 10/90, ie inductive (no R) = Z
Assume V = 100/0
I = 100/0 / 10/90 = 10/-90 deg
ie, 10 A 90 deg behind V (@0 deg)
Or lagging behind

Assume X = 10/-90, ie capacitive (no R) = Z
Assume V = 100/0
I = 100/0 / 10/-90 = 10/+90 deg
ie, 10 A 90 deg ahead of V (@0 deg)
Or leading V

 

[KSTR]

All is fine, I get at 400Hz 0.65% distortion at onset of grid current (= maximum drive).

But when running simulation at 50Hz I get 3.25% distortion.

Reducing the signal capacitor between driver (SRPP) and output tube from 1uF to 50nF gives 1% distortion but phaseshift goes from 168 degrees to 141 degrees (amplifier is inverting due to SRPP) at 50Hz (at 400 Hz 179 degrees).

Something that is often neglected in LTspice distortion simlations: Any RC time constants -- notably from highpass / DC-blocking filters -- must have fully settled before distortion measurerement via .four becomes valid. Ususally this is a problem only when chasing numbers below 0.01% .

You could test the output stage distortion and phase shift in isolation, just drive it directly (with proper DC bias if required).

Same for the driver, just running into a RC highpass and some reasonable test load.

In general, an AC analysis is required to see the frequency response and any phase shifts that come with it. Your value of 40 degree shift at 50Hz indicates that you are already right in the highpass knee (1st order filter at the -3dB point has +45 degrees of phase, half of the final value of +90 degrees), so level has dropped as well by almost 3dB and thus distortion will be accordingly lower.

 

[AmadeusMozart]

Red

Reducing C increases phase ang V to I. That is how power factor is corrected, adding C.
...

Thanks, it explains the why.

It leaves open the question what is more objectionable (audible): The increased distortion or the increased phase shift in the bass region.

...... Your value of 40 degree shift at 50Hz indicates that you are already right in the highpass knee (1st order filter at the -3dB point has +45 degrees of phase, half of the final value of +90 degrees), so level has dropped as well by almost 3dB and thus distortion will be accordingly lower.

Simulation indicated about 1dB drop in output. I'm using UL and cathode feedback and as a result the inductance of the output transformer and its leakage all play a part. One has also to remember that the cathode capacitor is in series with the B+ smoothing capacitors so they play their role as well.

Still some time away from completing the amplifier and then it will be testing with a PicoScope and an old HP 339A distortion meter which will give the measurable results. (chassis has just been completed but still waiting on parts to arrive - all shipping goes with slow detours these days).

 

[KSTR]

It leaves open the question what is more objectionable (audible): The increased distortion or the increased phase shift in the bass region.

The phase shift can be undone easily as it is a linear error. The distortion cannot be undone. Then again, 3% at full blast at 50Hz will be likely inaudible with normal music content, assuming it is low orders only. And your speakers may distort much more at those levels.

As of now it is not clear if/why a higher cut-off frequency in the feed from the driver circuit really produced a lower distortion (check when producing the same output voltage for a fair comparision, so you have to compensate your 1dB loss with increased input level).

 

[AmadeusMozart]

.....
As of now it is not clear if/why a higher cut-off frequency in the feed from the driver circuit really produced a lower distortion (check when producing the same output voltage for a fair comparision, so you have to compensate your 1dB loss with increased input level).

The primary of the output transformer is 18H with a DC resistance of 101 Ohm and a reflected load impedance of 3500 Ohm. When modelled it shows a phase shift between input on g1 / cathode and secondary at 50Hz. But at 400Hz that has largely disappeared.

Due to using UL and cathode feedback the distortion produced is pretty consistent across the output level only rising when the tube gets into grid current or when the output level goes in single digit mW output level to a maximum of 1.47% at 0.01mW.

I suspect the phase shift is due to the cathode feedback loop and the reflected inductance. More concerned about what would be audible.

 

[AmadeusMozart]

The primary of the output transformer is 18H with a DC resistance of 101 Ohm and a reflected load impedance of 3500 Ohm. When modelled it shows a phase shift between input on g1 / cathode and secondary at 50Hz. But at 400Hz that has largely disappeared.

Due to using UL and cathode feedback the distortion produced is pretty consistent across the output level only rising when the tube gets into grid current or when the output level goes in single digit mW output level to a maximum of 1.47% at 0.01mW.

I suspect the phase shift is due to the cathode feedback loop and the reflected inductance. More concerned about what would be audible.

Update:

I've solved the problem by adding a fraction of global negative feedback (-2.5dB). Resorted to this to reduce the phase shift at the lowest frequencies which potentially can result in instability with a reactive load. The value of the coupling capacitor no longer is critical and has less impact on the end result, distortion is lower too and is within expectations.

 

[witwald]

You can do your own, or just use headphones which are linear phase by default (single driver).

Doesn't the single driver mean that they are minimum phase, not linear phase? The headphone driver is just a bandpass system.

 

[René - Acculution.com]

Doesn't the single driver mean that they are minimum phase, not linear phase? The headphone driver is just a bandpass system.

KSTR has already answered this:
"Sorry for the sloppy wording, but that was addressed in a later post.
I should have written heaphones / single drivers don't have excess phase as they are minimum phase. And with a flat FR that means linear phase for most of the passband."

Which is correct.

 

[kongwee]

Electrically yes, mechanically no.

 

[KSTR]

Doesn't the single driver mean that they are minimum phase, not linear phase? The headphone driver is just a bandpass system.

Sorry, and yes you are correct and it is what I meant to say, minimum phase -- with no excess phase. With a flat response in the passband of interest it is almost linear phase, that is. No extra phase contribution other than from the roll-offs.

 

[René - Acculution.com]

Electrically yes, mechanically no.

Could you elaborate?

 

[witwald]

Electrically yes, mechanically no.

Why would you say that? The driver is an electrodynamic/electromechanical system. As the Thiele and Small papers show, amongst others, the frequency response (magnitude and phase) of an individual loudspeaker driver can be represented by mathematical models that produce analog-filter-like response functions. And these models have been validated by test measurements. Note that a simple mechanical system consisting of a mass, a spring, and a damper, is also minimum phase in nature.

 

[kongwee]

Why would you say that? The driver is an electrodynamic/electromechanical system. As the Thiele and Small papers show, amongst others, the frequency response (magnitude and phase) of an individual loudspeaker driver can be represented by mathematical models that produce analog-filter-like response functions. And these models have been validated by test measurements. Note that a simple mechanical system consisting of a mass, a spring, and a damper, is also minimum phase in nature.

Anything before the cone move the air will have phase shift because it is electrical. You have voltage and current that will not flow in sync. That why you have power factor. That is from you transistor to your last bit if voice coil and magnet/electro. You are using this property in your crossover, of course the source and the amps. That will mechanically show at the surface of speaker driver. After the driver, it is pure sound pressure. As far as I known in science, we can only measure in dB and directional angle. These two doesn't have a phase angle like voltage and current. It is alway summation of direct and indirect sound. At least, there must like Voltage X Current X Power Factor, something like in sound pressure to conclude there is a phase shift.

 

[René - Acculution.com]

Anything before the cone move the air will have phase shift because it is electrical. You have voltage and current that will not flow in sync. That why you have power factor. That is from you transistor to your last bit if voice coil and magnet/electro. You are using this property in your crossover, of course the source and the amps. That will mechanically show at the surface of speaker driver. After the driver, it is pure sound pressure. As far as I known in science, we can only measure in dB and directional angle. These two doesn't have a phase angle like voltage and current. It is alway summation of direct and indirect sound. At least, there must like Voltage X Current X Power Factor, something like in sound pressure to conclude there is a phase shift.

You have got this all wrong. That is why I asked you to elaborate. Just a you can make a transfer function relating the complex pressure somewhere in space to the complex input voltage, you can make a transfer function relating the complex displacement somewhere on the cone to the complex input voltage. Those transfer functions will be minimum phase, so from the magnitude you could calculate the phase.

 

[kongwee]

You have got this all wrong. That is why I asked you to elaborate. Just a you can make a transfer function relating the complex pressure somewhere in space to the complex input voltage, you can make a transfer function relating the complex displacement somewhere on the cone to the complex input voltage. Those transfer functions will be minimum phase, so from the magnitude you could calculate the phase.

I am not wrong at all, it is all in textbook. Even with electrical to mechanical sound wave. there is a coefficient value but I remembered nothing come close like power factor. I mean in theory you can have 5v and 5a in AC output but you power factor is zero. Or 180 degree out of phase. Zero watt output, no work done. With 1 value of resistance and 1 value of inductance or capacitance. You get 90 phase shift. 0.5 power factor. 5V x 5A x 0.5= 12.5 watt while still 25 va, 12.5 reactance power. Of course, you must add the impedance value to calculate the true power. But I just wanna show the effect of power factor. Even more complex adding frequency and RMS value. Amir already explain this in his own way. I just add calculation for you. I don't remember mechanical have such calculation, well I don't study mechanical engineering. That why I say yes in electrical, no in mechanical. You could prove me wrong in mechanical known basic theory.

 

[witwald]

I am not wrong at all, it is all in textbook.

Which textbook, and whereabouts? Do you have it to hand?

 

[Frgirard]

Gentlemen, it is strange that you disagree on technical concepts that have conceptual existence for decades

 

[kongwee]

Which textbook, and whereabouts? Do you have it to hand?

Pick up electrical engineering textbook.

 

[witwald]

Pick up electrical engineering textbook.

From Dickason's Loudspeaker Design Cookbook: "If there is no more phase shift than what is dictated by the magnitude response, the device is called minimum phase, which is what loudspeakers are generally considered to be." Here "loudspeakers" is referring to "loudspeaker drivers".

From Schuck's 1986 JAES paper, titled Design of Optimized Loudspeaker Crossover Networks Using a Personal Computer: "As others have observed, most moving-coil drivers have minimum-phase characteristics in their operating ranges. As a result, the magnitude of their frequency response completely describes their linear behavior with their phase response calculable via the Hilbert transform. Ladder-type filter networks are also minimum phase. Thus the overall transfer function between a ladder-type crossover network input voltage and the sound pressure output of each individual driver (but not the summed acoustic output of two or more drivers) can be assumed to be minimum phase."