This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

TPS82084: Input switching noise control

Part Number: TPS82084

We need to reduce the switching noise from the the Vin pin on the TPS82084, because it is interfering with sensitive analog components on one of our designs. 

This is for an existing batch of built boards which we would like to continue to use with rework.   The noise is conducted to the sensitive components through the ground plane unfortunately, so the better the high frequency decoupling on the Vin pin, the worse the system noise.   Our best results come from isolating incoming voltage trace with a ferrite, and adding a series resistor to the input cap at the 82084 to slow down the switching transient. This keeps the high frequency noise to a small bit of copper at the Vin pin.   It works great but it leaves a 10ns 8V spikes at Vin, which exceeds the device absolute maximum rating of 7V.

Power source----> [Big decoupler to Gnd] ----> [Low Z Ferrite bead] ----> [Mid sized decoupler with series resistor between cap and ground] --> TPS82084

Questions:

1. By any chance can we leave the spikes there and expect the TPS82084 to function with a reasonable (or at least reduced but predictable)  lifetime.

2. Could I rig up a snubber circuit to one side of the exposed inductor on the top of the package, to try to reduce or shift the noise frequency.

3. Any other ideas.  Re-layout will happen eventually but we may choose another buck converter unless I can prove this one will behave itself.

  • Hi Heather,

    Sure, for a quick check to get your overall system working with lower noise, you can try the input resistor or manually adding a snubber on the (internal) SW node. But we can't recommend either of these as a long term solution.

    I'd be happy to review your layout and test results to see if we can figure out how to address this noise in the layout. Most noise is solved with PCB layout.

    A ferrite bead is a great way to block the switching noise from the input bus.

    Do you know yet if it is the frequency or magnitude of the input ripple that is causing the issue? If a ferrite bead helps, then this indicates it is likely conducted noise and not radiated noise which is the primary problem.
  • Hi Heather,

    After looking at your layout, there is a clear improvement needed. Both the input cap and output cap should be routed directly back to the IC, without using a via to the GND plane. As you indicated, this puts all the switching currents into the GND plane instead of directing them to their return source. By placing the caps closer to the device and using a direct routing, the switching currents are kept on the top copper layer.

    We still need the GND vias under the device for thermal relief. Switching currents should not flow through these as they are not the lowest impedance path back to their source.

    From your existing layout, it looks like there is a GND via and trace by the IC's GND pins that you are able to solder to. You might try adding both an input and output cap there, back to the proper IC pins, and then cutting the trace which connects to the GND via outside the IC. Then, remove the input and output cap from their normal positions on the board. See if the noise is improved.

    You can also try wiring in the EVM, with its good layout.