DRV8301 Buck Converter Noise Possibly Issue with COMP Capacitors

Drew,

I am not sure I am quite following you.

How is noise on the 5V rail causing noise in the motor?

For audible noise to be coming from the motor, it typically means you have some sort of frequency in the winding current that has dropped within the audible range. The 5V would not cause this I would think though as the winding current is being modulated by the power half-bridges.

Do you have a scope capture of the 5V with and without the noise? If you do an FFT does it show the audible range frequency removing itself.

Is there some sort of ground that is missing? Trying to understand why touching the capacitor makes any difference

Is the 5V from the buck converter used in any way to power the voltage supply or references for the current amplifier? If this is the case then a fluctuating reference could influence the amplification of the current shunts and feed back into the control algorithms. I scoped the 5V output and there doesn't seem to be any appreciable noise on the output which is odd. Again all I can tell for certain is that by coupling the COMP pin with skin to ground seems to stabilize whatever signal is causing the feedback. I'm also wondering if there could be some sort of ground loop issue going on. What equations are used to determine the type and value of the capacitors for the COMP pin?

Internally, the buck is a completely separate silicon die from the gate driver. They just share a ground connection, no other shared signals or references.

The buck converter is the TPS54160 (page 17) of the DRV8301 datasheet. You can refer to it's datasheet for detailed information on the compensation network.

Do you derive the 3.3V from the 5V buck output? Your REF for the internal op amps it supplied by the 3.3V rail. It is possible this is adding a modulation into the current shunt amplifiers and thus in turn introducing a modulation on the gate drivers. 

I would examine the REF signal and SO (sense amplifier outputs) also to see if you can see the noise. It is strange that nothing is visible on the 5V rail.

Can you provide a layout file for around the DRV8301, you can also send to me over the E2E messaging system?

I will take a look at those signals and get back to you. In the meantime would you like gerber files or can you read native Altium board files? I can DM either type to you.

Native Altium would be best.

Drew,

Where do the MCU and DRV grounds tie together? I see that there are two separate planes.

I am worried that a very large ground loop has been formed between the buck regulator (power pad of DRV) and MCU ground.

Nick,

We are aware of the issue on the bottom layer regarding the DRV power pad's connection to ground on that layer. The DRV pad is connected to ground however on three separate internal layers. Do you think that  lack of connection on the bottom layer would be enough of an issue to generate the kind of noise we are seeing? Essentially the current from the GNDs on the bottom layer would need to flow thru the vias up to the ground planes on the higher layers. 

Drew,

It is definitely ideal to keep the loop for the buck regulator ground return on the same plane if at all possible. Via inductance will reduce the effectiveness of the bypass capacitor and diode. It is difficult to say whether that is the root cause or not.

I was actually asking about where the two ground planes for the PCB connect  (high current and low current). Between the MCU ground plane and POWER ground plane. Is it the jumper on Mid-3 near the edge of the board?

Nick,

Yes, they are connected there and on the Mid-1 layer on the left side of the board. There is also a connection near the right side of the board as well on those two layers.

Understood.

A follow up question, does the noise only happen when the motor is spinning or all the time? I am trying to understood if some noise on the main power line (and whether is related to switching of the motor or not) is making it's way into the 5 V rail. This might explain why the compensation network effects the noise.

Did you have any luck looking at those signals?

Since your finger help the issue, a theory is that you are effecting the actual cap value in the compensation network, I know you looked a few different cap values but it may be good to examine a few more. You can use the TPS datasheet for a reference on values. You can also use the WEBENCH tool.

webench.ti.com

My personal preference is for a unified ground plane to reduce subtle current loops and then design attention be given to high magnitude and high frequency current paths to reduce noise and coupling. You will hear compelling arguments for both sides though.