ad1938+adau1452 noise

I tested five PCBs, and all five PCBs showed the same phenomenon. The only difference is the time it takes for the noise to increase—some boards have a shorter rise time, while others take longer. The image above shows the board with the shorter noise rise time.

Hello,

How many layers is your PCB?

Do you have some ground and or power planes?

Can you take a screenshot of the layout around the part?

Thanks,

Dave T

I am using a 4-layer PCB with a ground plane as the reference layer. Below is a screenshot of the PCB layout around the AD1938 section.

To provide some additional information about the issue: I connected a 2.2 MΩ resistor from the ADC IN N port to ground, and this issue occurred. If I do not connect this resistor, the speaker produces high-frequency noise, and the FFT spectrum from the AP test shows a spike at the location indicated by the red arrow.

Hello,

Thanks for sending the pictures of the PCB layout. There are definitely some things that were done that can cause issues. I NEVER have ground planes on the top or bottom layers that get attached to actual pins on the part. If you need some shielding and want to put a plane then do NOT attach it to any of the pins. Let the plane have its own vias to the actual ground plane on an internal layer. 

Then how the bypass caps are connected to the power pins on the part is also important. I have seen people put them on the bottom of the PCB!!! So going through a via to get to the part is really adding an inductor between the pin and the bypass cap!! Not a good idea. Then if connected to some small plane on the top layer it can ring or oscillate!!

We had one of our big customers with very good layout engineers did mistakes like this. They sent us the product that was having issues with the LRCLK not locking to the MCLK and it also turned out to be a PCB issue. 

So I wrote up a PowerPoint explaining how best to do it. Unfortunately, It will be a major change for you...

I feel confident that if you change the layout and follow these ideas you will see better performance. 

Thanks!

Dave T

AD193x Layout Practices.pdf

Audio inputs should always be connected when testing the circuit for noise, never floating. Even the most expensive and highest quality audio components will pick up horrible noise on inputs that are left unconnected. If you want to test "no input" then short the (+) and (-) pins together and then make the noise measurements.

You're seeing noise because these inputs are floating. I suspect that adding the 2.2 MΩ resistor is stopping the (-) input from floating, but now you're inviting Johnson-Nyquist noise, and high value resistors have more noise.