Ceramic output caps OK?
I’m designing a 53W TOP269K-based supply running at 132kHz. I have a question about output capacitor types.
The main (regulated) output is 5V, 3.003A. Due to space reasons, I wanted to use something smaller than electrolytic or tantalum. We already have a 100uF 6.3V X5R ceramic cap in a (comparatively) small 1210 package, and I’ve used 4 of them in series for my design. Here are the link to PNs:
http://www.tdk.com/spec_gen.php?
class_code=TJA037&item_roots_key=C3225X5R0J107MT&item_name=C3225X5R0J107MT
http://www.niccomp.com/catalog/nmchc.pdf
Questions:
1. Do you recommend a certain amount of output impedance for stability? Should I add some external, low-value series resistor? If I’m done my math right, the Kemet part’s ESR is 120mR (DF=10%, DF=ESR/Xc, Xc=1/(2*pi*f*c)=1/(2*pi*132k*100u). Perhaps the recommended TOPSwitch compensation is intended to be used with non-ceramic output caps, and a more ideal ceramic capacitor needs different TOPSwitch design techniques.
2. For a ceramic cap, do you worry about ripple current? This page indicates to me that it’s not shown in datasheets because it’s not a relevant parameter for ceramic caps; I assume this is because the ESR is so low that self-heating isn’t a problem like in electrolytic and tantalum caps.
3. 4 of these caps seems like overkill. However, the only information I see about selecting caps is on page 14 of AN47. I’m not sure it pertains to ceramic caps, though.
Comments
Yes, I placed 4x 100uF X5R caps on the board. We haven't gotten any boards yet, so I haven't been able to experiment at see what the design really needs.
I have a 2.2uH + (another) 100uF X5R cap as a post-filter. I can experiment with a higher-ESR cap here.
Thanks for the tips. So do I understand correctly that you see no problems with using just X5R caps, and there's no need to plan for adding any external R? In other words, I can achieve good results with just ceramic output caps?
If the X5R caps are limited in total capacitance so that the 132 kHz ripple on them is >3% of the output voltage, most of the output damping will be done by the post filter capacitor. Thus there will be no need for an extra series R. The post filter capacitor will need to have much larger capacitance than the main output capacitors.
Did you mean you used 4 x 100uF X5R capacitors in parallel?
For stability can you add some other type of capacitor in parallel (e.g. polymer), which has more capacitance but which also more ESR? One good trick is to make this capacitor your post-filter capacitor, after the post-filter inductor, so it also acts as the damping capacitor. If you do this, the post-filter inductor needs to be smaller than usual, perhaps 1 uH.
X5R capacitors tend to be very low ESR and you are correct self-heating tends to not be a problem.
You can start experimenting with output damping by using an external electrolytic capacitor (don't use an ultralow-ESR type) as your output damping capacitor, and reduce its size while observing the load step response. When you are satisfied, you can then experiment with reducing the number of paralleled X5R caps.
As you experiment with reducing the number of X5R capacitors in parallel, observe the load step response and the ripple voltage on them. Don't let the fundamental ripple p-p (not counting spikes), exceed about 5% of the output voltage.