Monday, February 24, 2020

Class D Amplifiers do not Have Horrible Phase shift

A poster on Audiogon keeps making two claims about Class D amplifiers which are not true. The first, that the ultrasonic frequencies matter to a human listener, and the second that they have horrible phase shift.I'll quote him here:


You just don't get it tweak, this is why many find the upper mids/highs objectionable on Class-D, why would I want to listen to an amp with this much phase shift already 25 degrees! at 2khz, and it get's progressively worse all the way up to 70 degrees!!! at 10khz.

And to support his claim he uses a link to this graph:


Unfortunately for this individual, he's built his reputation on this chart, and has ignored the heading:  "Output impedance vs. frequency."  Notice the scale on the left. That's milliohms. One thousandth of an ohm.  The output impedance starts at 5 mOhms and goes up around 500 Hz. Until then it has a damping factor ( 8 Ohms / output impedance) of around 1,600.  Truly outstanding.  The phase plot is the phase angle of the output filter.Clearly, 5 milliohms cannot be the speaker load.


What is the output filter? 

To understand what this means, we have to understand a little about amplifier circuits. Class A, AB and D typically have an output filter of some sort. Almost all linear amplifiers do as well if they have feedback, it prevents oscillation. I've put together a little diagram of what this looks like for a linear amp:


as well as for a Class D amp:

The component values are not real, they are just what XSim uses for default. For the purposes of this discussion what matters is the filter circuit and how it is not the speaker.  I'm using the wrong tool for the diagram, so you'll have to use your imagination. The components in the middle are actually inside the amplifier, and they contribute a great deal to the amp's output impedance and a tiny tiny bit to the phase angle at the speaker. These are either L1,R1 or L1,C1.

Here is the mistake the "high phase shift" proponents continue to make.  They claim that the phase angle at the output filter (L1,R1 on top, L1, C1 beneath) is the same as at S1. It is not.  Of course, Class AB and Class D amps use an output filter for very different reasons. For the linear amplifier (AB) it is used to prevent ultrasonic oscillations due to the feedback loop.  For Class D it is to attenuate the ultrasonic switching frequency, but you get the ide: an output filter is quite normal and not unique to Class D.

When Numbers are Complex

Complex numbers have both a magnitude (i.e. Ohms or dB) and angle, but the angle does not mean the same thing. In the case of a filter network, the angle refers to the relationship of the current to voltage. In the case of an amplifier's output (db) the phase refers to the relative place in time.

So, in essence, we are dealing with 2 complex numbers (impedance and outupt) and none of the 4 data points are interchangeable. They are not on the same scale (Ohms, dB) so therefore you can't just substitute, or add any of these. You CAN do complex AC circuit analysis however, but the output would be the dB chart.

What is the actual output phase angle?

Unfortunately B&O / ICEpower has not been consistent in measuring phase angles in the output, so to see what a typical phase angle chart looks like we have to turn to the 250 ASP instead of the 1200 AS2, plotted above.  You may download the full document from here:

https://icepower.dk/support/download-page/

I present the amplifier's' frequency response and measured phase angle across different output loads, below. Notice now we are talking 4 and 8 Ohms, not 5 milliohms.



The bottom set of lines is the actual phase angle of the amplifier as seen from the speaker. Unlike this poster's claim, the phase angle at 2 kHz is not 20 degrees but nearly zero. At 20 kHz the phase shift changes to about 20 to 30 degrees. It is difficult to judge how far from typical this is for solid state amplifiers because honestly neither this nor the plot for the output filter nor the amp itself is often available.

Anyone who uses the first chart to claim Class D amplifiers have horrible phase shift is misinforming you. 

Further, several experts agree that humans are simply not very sensitive to phase shift. They make this point collectively here:

https://audioxpress.com/article/zero-phase-in-studio-monitors

So even if the poster and his chart were accurate, it's hard to see how this would matter.

I'm still not sure ...

OK, fine. Here are plots from XSim of the SNR-1.  The top chart is of the output, the bottom is the impedance of the circuit.

They both have phase angles.

They are not the same.

The phase angle on the top refers to timing changes with frequency. The phase angle on the bottom refers to whether the impedance is resistive, capacitive or inductive. 

You can't use the phase angle at the bottom as a substitute for the phase angle at the top.

2 comments:

  1. Now it's clear that class D AMP phase shifting problem is BS, great help! it bother me a little though I can't hear any unusual sound through amp with IcePower 1200AS2, but quite annoying.

    ReplyDelete
  2. Good to know the phase shiting is BS, it's annoying although I can't hear any difference between IcePower 1200AS2 and AB AMP.

    ReplyDelete