| Electronics > Projects, Designs, and Technical Stuff |
| MLCC derating at RF (HF) frequencies |
| (1/2) > >> |
| bd139:
I'm building a hefty LPF for a 50W PA at the moment operating on 3-15MHz. This is 71V peak potential across the caps according to LTspice. Now being the cheap arse I am I used some old 100V MLCC leaded C0G caps I had lying around. At full whack, one of them caught fire. Whoops. Replaced it thinking it was damaged during soldering and went incendiary as MLCCs do but no it exploded again. And so did another one. Thus, I suspect that this was related to either a parasitic characteristic of the cap or being very close to the line somewhere. Plan is to buy some 630V 1206 MLCC's and try them. I did look at the traditional silver mica caps for this job but they are ridiculously expensive at £2-4 a pop compared to £0.03 a pop for the MLCCs. Any things I need to consider other than working voltage? |
| dmills:
Note that 70V is peak voltage for 50W into 50 ohms, but filters will usually have far higher impedances on internal nodes with correspondingly higher voltages (if there is a loaded Q of say 5, then the voltage will be ~350V not 70V!). Look at the full datasheet for the MLCC series in use (Not the short form one), they will often either specify a maximum RMS current or derate the voltage with frequency (amounts to the same thing if you think about it). Regards, Dan. |
| T3sl4co1l:
Don't forget to model the filter as it actually is, with stray/lead inductance included. Shouldn't be a big deal in that frequency range, assuming nominal (~50 ohm?) impedance, but worth a check. I've seen some fairly surprising things, like running 20VA (at ~500kHz) into an X7R MLCC (leaded). Got quite hot, but not over Curie temperature. C0G of course doesn't have that loss mechanism, but I guess the ones you've got just aren't up to the current or voltage... AFAIK, there's only two limits on C0G, voltage and dissipation. Note that voltages can get whacky around the transition band, where reactive power in the filter gets intense. They're never rated for power, but estimate it as, say, 1/4 to 1/2 the rating of a same sized, generic resistor. Just because on the one hand, film resistors dissipate the power on the surface instead of the bulk; but the bulk is more conductive (alumina is maybe 8 times more conductive than dielectric ceramic?). I don't know about leaded MLCCs, but the power dissipation is sure to be less. Calculate dissipation by ESR, check the datasheet. Tim |
| bd139:
Thanks both for your input. Much appreciated. Will think about these things in detail this evening. This is designed for 50 ohms in and out. I didn’t model the stray inductance and capacitance here. The target of 7MHz meant they were within the tolerance band of the components used anyway. |
| daqq:
What kind of filter is it? Reflective? You could be dealing with VSWR. |
| Navigation |
| Message Index |
| Next page |