0
\$\begingroup\$

I am using a buck converter to drop from 5V to 1.8V to be able to drive both of these voltage levels in a single design. The buck converter works correctly to generate a clean 1.8V supply, but the input side of the buck converter, the 5V side, is being contaminated with switching noise spikes of around 1.5 volts as a result of the buck converter being in the circuit. I have used the FT440A and AP3429 chips.

I have tried playing with the capacitance on the 5V input to the buck converter as well as adding diodes and inductors, but so far I have had no real effect on the spike amplitude. Is there a set of standard approaches to removing this sort of input noise that someone could highlight?

The default circuit is all that I am using with the exception of trying numerous modifications on the Vin side to get rid of the noise being seen:

enter image description here

Here are the input spikes that I see:

enter image description here

As far as layout is concerned, C1 is as close to Vin as possible. I have tried from 4.7 to 22uF capacitors. I also tried adding 3.3uH and 4.7uH inductors between before Vin.

\$\endgroup\$
  • 1
    \$\begingroup\$ Could you please add a schematic or diagram to back up your questions. We need to see the details of the design. SMPS are real finicky to build and have no noise on input or outputs. The board layout has to be near perfect. \$\endgroup\$ – Sparky256 Jan 23 '18 at 7:32
  • \$\begingroup\$ The ESR of C1 is critical and the chief cause of actual spikes. How confident are you that they are not measurement errors from probe ground inductance? Ask if you do not know how to measure spikes on 50 Ohms AC coupled. \$\endgroup\$ – Tony Stewart Sunnyskyguy EE75 Jan 24 '18 at 1:49
  • \$\begingroup\$ Thanks for the response. The issue was with the trace to C1 such that the capacitor was soldered in place but not actually connected between Vin and GND. The buck converter works fine as per the diagram. \$\endgroup\$ – Jason K. Jan 24 '18 at 5:23
2
\$\begingroup\$

As with any o-scope measurements on switching regulator circuits you can get false indictaions of spikes if you do not use best practise when probing: -

enter image description here

The "bad measurement" technique is susceptible to induced voltages from the switching currents - the loop formed by the probe's earth connection in effect acts like a receiver for magnetic fields. The bigger the loop area, the more interference is picked up.

I'm not saying this is always to blame but it can be a very common mistake when probing high current/frequency circuits.

You can prove this by connecting the probe tip to the probe ground wire and, without making a physical connection to your regulator, move the loop around to see what you pick-up.

\$\endgroup\$
  • \$\begingroup\$ Thanks for the suggestion, but the spikes are quite real. I tried your suggestion and there is no difference in the size of the spikes. \$\endgroup\$ – Jason K. Jan 23 '18 at 16:14
  • \$\begingroup\$ @Jason then you need to consider that your PCB layout is problematic. You are using a PCB? If you are using breadbaord then always expect problems like this. BTW what sort of signal size did you pick up with the probe shorted as per my last paragraph? \$\endgroup\$ – Andy aka Jan 23 '18 at 16:19
  • \$\begingroup\$ @Andyaka is correct, there is significant ripple current on the input to a buck converter, so the input capacitors have to be low ESR and the PCB layout has to have very low inductance from the drain of the top FET to the ground of the bottom FET or diode. If all of that is in order adding more capacitance will reduce the input ripple. If it doesn't, then the PCB layout is problematic. \$\endgroup\$ – John D Jan 23 '18 at 16:24
  • 1
    \$\begingroup\$ Have you put a load on the output that is representative? It may be going into over-voltage protection mode. \$\endgroup\$ – Andy aka Jan 23 '18 at 16:40
  • 1
    \$\begingroup\$ An input LC filter for a buck can be problematic. Google "Middlebrook Criterion" for more info. Can you post a picture of the switch node waveform? If you trigger on the switch node are the spikes synchronous or do they move around wrt the switch? \$\endgroup\$ – John D Jan 23 '18 at 19:17
2
\$\begingroup\$

Thanks for the assistance everyone. The problem was that the C1 capacitor was not properly connected due to a defect in the PCB pad or trace that resulted in an almost broken trace. Repeated attempts to change the SMD capacitor for different values had no real effect and it was only when I attempted to use a through hole capacitor soldered directly between pins Vin and GND that the spikes dropped right off. Other ad hoc attempts at adding capacitors near Vin were not very successful likely because they were not connected close enough to the chip.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.