Estimating ripple current of 8 x paralel MLCC 500V 2.2nF for a resonant converter

[rxpu]

I am trying to estimate the ripple current of a capacitor bank of 8x MLCC cap 500V 2.2nF connected in paralel

Ctotal = 17.6 nF Irms = 8A

The datasheet have only ESR value: 0.009 Ohm up to 1MHZ, butno data for ripple current


Resonant freq of circuit = 200khz

If I calculate the resistive power dissipation for caps each carrying 1A

The resulting power per cap is 0.009Ohmx1x1 = 0.009W

Main questions:

The caps are CG0 type (+-%5), will the current distrubute evenly between the paralel connected caps?

Assuming an equal distrubution, is calculating the power dissipation through ESR enough?

Are there other effects to consider when estimating the maximum ripple current per cap.

Case is 1206.

Thank you for your input.

 

[KlausST]

Hi,

The caps are CG0 type (+-%5), will the current distrubute evenly between the paralel connected caps?

There will always be tolerances when you have 8 capacitors, even if you use them all from the same production batch.
Thus the answer clearly is: No. There never will be a perfect match...
But the question is: what current distribution tolerance do you call as "equal" --> you need to consider this on your own.

Besides the part tolerance there will be a difference caused by the PCB layout.
A bad layout may cause additional impedance ... causing widely different currents...

Klaus

 

[Easy peasy]

you need to know the dissipation factor of the caps at 200kHz ( D = 1/Q ) the VA each cap is doing, i.e. Vrms ac x Irms ac x D = heat in the caps in watts...

- - - Updated - - -

C0G / NPO is a good choice - make sure the V is 2x peak in ckt ...

 

[FvM]

Obviously capacitive impedance dominates the current sharing, respectively it will be almost equal. 1A rms should be no problem, but 360 Vrms @ 200 kHz is marginal for a 500V rated cap. 1000V rating suggested.

 

[Easy peasy]

FvM is correct, 500V rating is too light ... the peak V will be in excess of 500V for 1A at 200kHz in 2.2nF ( V = Irms x Xc )

 

[rxpu]

Obviously capacitive impedance dominates the current sharing, respectively it will be almost equal. 1A rms should be no problem, but 360 Vrms @ 200 kHz is marginal for a 500V rated cap. 1000V rating suggested.

FvM is correct, 500V rating is too light ... the peak V will be in excess of 500V for 1A at 200kHz in 2.2nF ( V = Irms x Xc )

The topology is the capacitor-clamped half bridge LLC. Each cap is paralel connected with fast diodes.

These diodes act as voltage limiter (the diodes clamps the maximum cap. voltage to the bus voltage, any voltage voltage bigger than the bus voltage will be cut and crippled to the bus voltage )

At the same time this acts as a current limiter for the half bridge. Because limiting the cap voltage results in limiting the current flowing through the cap.

Do you think even under this clamping arrangement , a 1000V cap is needed? (in the simulation I see that for overload condition the maximum voltage clamps to the bus voltage ex: 350VDC)

If I you agree with 500V, which kinds of diodes should be used to be sure that the clamping sequence is fast enough, should these diodes be fast-soft recovery or fast-hard recovery?. I also see that regular 1N4007s may also be used. But I am not sure if 1N4007 has enough forward recovery to clamp the voltage rapidly.

Thnak you for your input.

 

[Easy peasy]

4007 would be a bad choice, 600V with Trr < 35nS a far better choice - TO-220 package, on a heatsink ...

 

[FvM]

Instead of the long winded explanation I would appreciate a schematic. For the time being, if you have 1A@200kHz through 2.2 nF, you get 360 Vrms, if you clamp the voltage somehow, it's not 1A.

 

[rxpu]

Instead of the long winded explanation I would appreciate a schematic. For the time being, if you have 1A@200kHz through 2.2 nF, you get 360 Vrms, if you clamp the voltage somehow, it's not 1A.

I send the schematic. It was a rough estimation. The simulation result is Vcap=287Vpeak over 38nF capacitor @100khz @I=5A. (0.28A per 2.2nF)

1A estimation was a rough exaggeration to calculate power dissipation of the cap with a very large safety margin.




- - - Updated - - -

4007 would be a bad choice, 600V with Trr < 35nS a far better choice - TO-220 package, on a heatsink ...

I think the diode kicks in only when there is an overload, In the normal working condition there should be little or negligible current flow through the diodes.

I used for each clamping diode , 2 smd diodes combination (D2PAK) in paralel with large copper area for cooling : 6cm2 x2 double sided with vias)

I think forward recovery time of the diode may also be as important as the reverse recovery time. I can not find the forward recovery time of the diodes in datasheet. I read that it may be not negligible for non-shottky diodes such as fast or ultrafast reverse recovery pn-diodes.
Any idea how to guess the forward recovery performance of a diode in the datasheet?

 

[FvM]

O.K. this looks much more comfortable. The specifications of post #1 are apparently wrong, including the total capacitance. 500V capacitor is just right for 360V DC bus.

You should know if the clamping diodes are ever forward biased, if this happens periodically, they should surely expose low trr.

 

[Easy peasy]

because volts cannot change instantaneously - the forward recovery is not such an issue ...

 

[rxpu]

because volts cannot change instantaneously - the forward recovery is not such an issue ...

I got it, But what happens when the reverse recovery is not fast?

Because as I understand it is important that the diode kicks in if the volatge rises above a threshold.

But why is it critical that the diode shuts down fast?

My uncertain answer: If diodes shut down slowly , there may be a shoot through of the DC bus when one diode kicks in and the other can not shut down rapidly. Is it true?

- - - Updated - - -

O.K. this looks much more comfortable. The specifications of post #1 are apparently wrong, including the total capacitance. 500V capacitor is just right for 360V DC bus.

You should know if the clamping diodes are ever forward biased, if this happens periodically, they should surely expose low trr.

In fact it will not happen peridocally. It will happen only when the primary or the seconadry of the transformer is shorted. The half bridge will limit its current through limiting the Vcap.

But anyhow, I think low trr may also be mandatory even if the diodes are rarely forward biased.

 

[mtwieg]

I typically use ceramic caps from TDK, as they provide very detailed data on each part number (ESR vs frequency, temperature vs Irms and frequency, capacitance vs DC bias, etc), and they make this info fairly easy to find.

However I've noticed that the Irms ratings (from all manufacturers, not just TDK) are lower than expected. For example, here's their data on a 1uF, 100V, X7S in a 0805 package. They suggest that at 1MHz, the ESR is about 8.2mohm, and a ripple of 2.27A rms will give a temperature rise of 20C. But 2.27A rms into 8.2mohm is just 42mW of dissipation. Compare that to an 0805 resistor which is typically rated for 125mW. And I would expect a MLCC to have much better thermal performance than a resistor due to its interleaved metal electrodes.

Of course 20C isn't very high of a temperature rise. But I'm not sure this data is valid for establishing a maximum permissible ripple current. I generally don't see such a maximum specified for ceramic caps, except maybe for ATC and other specialty vendors.

 

[FvM]

If a manufacturer gives maximum AC current ratings for capacitors (either film or ceramic type), the values are possibly rather arbitrary. I did a research about the AC current rating of PP capacitors that could be used for LLC converters and found large variation between manufacturers without significant differences in ESR and measured temperature rise. At this point we decided to make our own specification of useful current rating based on empirical determined temperature rise.