Yes. I did distortion measurements at several TPA-boards including TPA3255EVAkit and if you measure with high resolution you will find that tiny voltage drops across the plane that couple into the audio input path will degrade THD-performance.
You should know there is one dominant source of disturbance:
The current flow from the capacitors through the supply pins of the TPA: Whenever the amp delivers a sine wave into the load, the current flow into the chip is NOT sinoidal, but a rectified sine- i.e. a totally distorted signal containing harmonics H2, H4,H6 etc.
Even worse: In fact it is a rectified sinewave chopped at the pwm clock frequency. With amplitudes of several amps, the current loop is the origin of a magnetic field that induces voltages into any loop in the neighborhood, including the ground plane.
Let us assume 1mV differential voltage in the ground plane coupled into the differential input. With 20dB amplifier gain this results in 10mV output of distorted Signal.
Furthermore let us assume 10V output sine wave signal for 25W into 4 Ohms - then you get -60dB distortion level. That indeed is a poor result compared to the capabilities of TPA in a proper layout.
And that is the reason to decouple AGND from the "poisoned" power GND.
In my designs I consider TI-EVMs not as a target but more as a starting point to begin with -that's just my eccentric hobby. Anyway you may be happy with your design as it is - and to be honest I personally do not believe that anybody on this planet can discern distortions below 0,1%.
So have fun
Last edited:
The TPA GND pins and nearby blocking caps are not on focus here. I am talking about the audio signal path, i.e. the input. You may have some preamp or not - at the end there is an inverting and a corresponding non inverting input. Each one is connected by some impedance to ground.
But not exactly at the same location of the ground plane.
What matters is the differential voltage between both inputs, not the absolute deviation from an ideal gnd potential.
So my approach is to use a separate AGND-plane that is nearly free from induced noise current - i.e. no voltage differential between AGND return points.
hint: you will not find AGND plane on the layout I posted.
But you will find a 14pin header at the edge of the PCB to connect via a ribbon cable with a separate audio front end (AFE, not shown here) containing the AGND plane.
Last edited:
Yes. I did distortion measurements at several TPA-boards including TPA3255EVAkit and if you measure with high resolution you will find that tiny voltage drops across the plane that couple into the audio input path will degrade THD-performance.
You should know there is one dominant source of disturbance:
The current flow from the capacitors through the supply pins of the TPA: Whenever the amp delivers a sine wave into the load, the current flow into the chip is NOT sinoidal, but a rectified sine- i.e. a totally distorted signal containing harmonics H2, H4,H6 etc.
Even worse: In fact it is a rectified sinewave chopped at the pwm clock frequency. With amplitudes of several amps, the current loop is the origin of a magnetic field that induces voltages into any loop in the neighborhood, including the ground plane.
Let us assume 1mV differential voltage in the ground plane coupled into the differential input. With 20dB amplifier gain this results in 10mV output of distorted Signal.
Furthermore let us assume 10V output sine wave signal for 25W into 4 Ohms - then you get -60dB distortion level. That indeed is a poor result compared to the capabilities of TPA in a proper layout.
And that is the reason to decouple AGND from the "poisoned" power GND.
In my designs I consider TI-EVMs not as a target but more as a starting point to begin with -that's just my eccentric hobby. Anyway you may be happy with your design as it is - and to be honest I personally do not believe that anybody on this planet can discern distortions below 0,1%.
So have fun
Don't get me wrong, I am just confused and would like to find out why. I agree with your opinion regarding the PGND shifting. My confusion is about the effectiveness of separating the GND plane on PCB. Say the spike current creates a same 1mV voltage on the PGND and we do have a separated AGND plane. In this case, the internal GND of the chip might already be raised by 100mV. Since the internal ground is connected together, no matter how clean your AGND is, the input signal will see that 100mV noise once it gets in the chip.
Anyway, separating the PGND is a good idea in general, I will adopt it so that the power noise won't be coupled into other sensitive parts.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.