TPS62561 - Oscillations on output voltage

The TPS62560 will attenuate some noise that is present on its input bus.  The higher the noise frequency, the lower the attenuation.  At 5MHz, the noise most likely passes "around" the power supply, not "through" it.  You must be very careful about how you route the traces and where parasitic capacitive coupling exists between the input and output of your board.  Adding some smaller sized, high quality, capacitors  like 0603, 0401 will help filter some of the higher frequencies.  Noise problems are often difficult to solve.  I suggest that you attack the problem at the source by filtering the power bus where it enters the board.  Another option is to use a cleaner power source.  I hope this helps.

Michael,  Thanks for  the feedback. I will look at trace coupling, but I don't normally see this at 200mV. When I have seen that type of coupling, it only transfers 10 to 20mv.

There are 2 bigger issues. 1) The 20Khz, this is well below the 2Mhz that this part runs at, and should easly be rejected. 2) It's very related to load. As the load incresses the ability to reject noise goes down. (or that is the way it seems)

Again both of these issue are "levels of noise". I don't mind seeing 50mV come though, but 200mV (on a 1.1V output) is just not acceptable.

Got any suggestions on the input filter, I really hate putting ELE's on boards, but if this is a load issue then I do need a few hundered uF.

Thanks for the help.

Michael, BTW: I am not convinced that the noise is really coming from the input. When the load is up it apperes on the input and output of the 1.1V regulator but does not show up on the 1.8V output (same regulator type, and same input). I mean I could just litter the board with 22uF caps but I would really like to understand the problem better. Any other suggestions?

Can you post a schematic, including all bypass capacitors on the load.  What is the total amount of capacitance on the output of the TPS62560?  One potential cause of the ripple you see may be instability caused by too much external capacitance.  This is different from noise.  If the power supply is unstable, it will generate a sine wave that is superimposed on the output voltage.  You can test this by adding additional capacitance to the output of the power supply.  If the frequency of the oscillations is inversely proportional to the capacitance value, instability is the most likely cause of the problem.  Can you also post a scope picture of the output voltage ripple with different capacitor values?

I had (2) 22uf caps on on the output and the freq was 29Khz. I removed one of the caps and the freq went to 43Khz. So less capacitance resulted in a larger freq. However, the voltage remained the same 100mV for this case. It is interesting though that once the board boots, I the CPU must be going to sleep and then waking up at specific intervals, because the ripple will be gone, show up to 5 to 10 cycles, then disappear again.  As for total caps on this output (1.1V core CPU voltage) (3) 22uf, (1) 10uF, (6) .22uF, and (5) .01uF. The total area where this voltage is used is less than 2 Sq In. As as second experiment, I removed all of the main bulk caps expect the (1) 22uf cap directly on the output. Leaving all of the .22 and .01 caps at the cpu pins.  The ripple when down to 40mV at 62Khz. So the question now, is at high load how do I eliminate this instability completely. And should I just remove all of my bulk by-pass caps? It would make the board cheaper, but this seems counter intuitive in a power analysis.

Thank you very much for this help.

Michael,

These parts seem very responsive and seem to work best with little to no bulk caps. I removed all of the 22 and 10 uF caps from the board on all supplies except for a single 10 at the regulator and the .22 and .01 caps at the sink pins. Power ripple on all rails even at high load is less than 20mV.  Please confirm that this is TI recomended configuration for a power network.

Thanks for helping with this problem.

CC :)

From two posts above, I assume that when you say "bulk" capacitance, you are referring to the output capacitors that are to the right of the output inductor.  Yes, TI designed these ICs to work with a minimum of output capacitance to help maximize your circuit's performance with minimal bill of material costs by allowing a single ceramic output capacitor.  I'm glad I could help.  Would you mind sharing your application and volumes for this project?  I'm always interested in understanding how our customers are using our ICs.

Michael, No, "bulk" is different from "output". The output cap is now 10uF, the "Bulk" caps are scattered around the board near larger IC's that need the power, and each power pin on any IC has it's own private .22 or .01 uF cap.  Not sure if this changes your recomendations or not , please let me know.

This application is for a system appliance using a Freescale iMX 515 processor. Unit volume is as yet un-known, because several of our larger customers will end up building there own HW based on our "recomended" design. But we are working with a few vendors that could sell 100K to 200K a year.

For a well designed board with fairly small trace resistance, the power supply does not distinguish between output capacitance at the power supply and bulk capacitance at the load.  All capacitance must be added up and considered as part of the output capacitance for the power supply.  Please note that I recommend changing to a  1.5uH inductor to put your Fp closer to the middle of the recommended range.  This allows you to add some more bulk capacitance at your load if necessary and will also help improve load transient reponse.

Michael,

I think that explains everything and I can fix it up from here.

Thank you for helping.

CC :)