so many MLCC capacitors in parallel why???

[Benjam]

I have a question. As shown in Figure 1, he connected so many MLCC capacitors in parallel. What is the basis? In other words, the capacity of the capacitance value is calculated according to what. What is the name of this circuit? Whether can use inductor combination capacitor to achieve, in order to reduce the number of capacitor, narrow the space.

[mariush]

Probably to meet a minimum capacitance ...

They may have chosen 10uF because it was more easily available in that package (easily sourced from more sellers), or cheaper in volume, maybe they didn't want bigger packages to prevent microphonics or cracking of capacitors...

Also capacitance varies with voltage ... see voltage bias on ceramic capacitors :  TUTORIALS 5527 - Temperature and Voltage Variation Ceramic Capacitor : https://www.maximintegrated.com/en/design/technical-documents/tutorials/5/5527.html

edit: better link added.

[Siwastaja]

Low inductance, i.e., quick availability of energy everywhere. Capacitors are like "caches". If you need fast rising/falling edge current pulses somewhere, you need a supplying capacitor near it. In such use, capacitance value is often secondary.

Looks like a huge amount of capacitance. What is this, a large 3-phase BLDC motor controller? For what voltage and current, approximately?

To avoid ringing and overshoot with such MLCC bank, it's a good idea to add electrolytics in parallel, with at least 2-3 times the capacitance. Standard high ESR parts are typically fine because this much MLCC take most of the ripple current.

[dietert1]

Same reason for the multiple connections of each mosfet electrode. If you count, there is roughly one cap for each connection. The caps are probably sitting close to those connections, not close to each other.

Regards, Dieter

[Benjam]

Yes, I want to design a three-phase all-digital fuzzy motor drives, nominal 48 v dc bus voltage, continuous current 12 a, 40 a peak current, I encountered difficulties when choosing capacitance now, I can't use aluminum electrolytic capacitor, because the volume is too big, can only choose MLCC capacitor, but don't know how to determine how specific capacitance values, This is the root of the problem.

[Benjam]

I agree with you. I think it should also be a 10uF capacitor, because it is the most suitable one in MLCC, but I don't know how many to choose, or how to determine their capacity.

[Benjam]

48V, continuous current 12A, peak current 40A, continuous output power ~350W, how do I determine the capacitance value

[Siwastaja]

Replace with 1uF or maybe 2.2uF caps, that way you can use smaller packages, giving smaller inductance, tighter layout. Capacitance is still fine I guess.

You want the capacitors to be "all around" or "inside" the layout so to speak, but you don't want them to be so bulky they make the layout larger. Small loop area from capacitor positive, through the two MOSFETs, to capacitor negative, is important and here you can see how having multiple small caps nearby helps.

I really try to use 0805 if at all possible but of course in such design 1206 would likely get used.

And add some electrolytics to dampen any ringing. If they get warm, it's a sign you don't have enough MLCCs to carry bulk of the current ripple.

[dietert1]

In a 48 V power system one would probably use film caps to build something durable. Those small MLCCs usually have ESR of 1 Ohm or so. So they are of limited use when it comes to currents or current spikes of several A. In the end the smaller distributed caps are only part of the solution and it depends on low inductance layout (multiple layers with wide traces) and low ESR electrolytics that can withstand the required AC current.

Regards, Dieter

[Siwastaja]

You are off by two orders of magnitude. ESR is around 10-20 mOhm. MLCC is one of the lowest ESR types, supporting highest ripple currents, ripple current very rarely is a problem in DC bus bypassing. Film caps as a generic group, on the other hand, contain both higher and lower ESR types, the best types are roughly equivalent to MLCCs in current handling but you need to make sure you get them.

MLCC is absolutely fine in 48V system. 50V rating works fine if you are mentally fine without derating, but 63V or 80V parts can be obviously used. Then package would be 1206 or 1210.

Biggest challenge is mechanical. It's a very good idea to use automotive grade soft terminated parts even if that costs a bit premium; even then, make sure to avoid board flexing when designing mounting holes etc.

Film caps are getting more practical when clearly exceeding that 48V. Mains inverters tend to use film caps even though specialized ceramic caps are also available at those voltages, but obviously not SMT MLCCs.

[dietert1]

Maybe ESR was the wrong term. I looked up the datasheet of a 10 uF AVX MLCC cap. Loss factors of up to 16 %, tested at 0.5 Vrms at 120 Hz. I wrote down my recommendation for film caps under the impression the MLCC parts won't last at 1 or 2 Arms at 100 KHz.
https://docs.rs-online.com/c821/0900766b815bddac.pdf

Nothing but an invitation for the designer to check and do his math. And this type of loss calculation will finally tell you how many of the caps need to be there to carry the total AC current without smoke.

Regards, Dieter

[TimFox]

A 10 uF capacitor at 120 Hz has a reactance of -245 .  If the DF is as high as 16% at that frequency, the ESR would be 21 .
However, at 100 kHz, that capacitor (neglecting self-inductance) is only 0.16 reactance.  For such a large capacitance, the manufacturer does not quote a DF at 100 kHz.

[Benjam]

Thank you for your answer, but my doubts have not been answered until now. Let me reiterate my question again. How many MLCC capacitors should I use in parallel to meet my needs? The parameters of the motor I want to drive are as follows: nominal 48 v dc bus voltage, continuous current 12 a, 40 a peak current.

Please give me an answer directly and tell me why? thank you very much.

[thinkfat]

Maybe to counter the inevitable capacity derating you need to apply for MLCCs at high voltages.

[radiolistener]

Most of all, these capacitors are distributed all around PCB. It is needed to reduce path between power consumer and capacitor. There are at least two reason for that: to reduce resistance and to reduce inductance.

[T3sl4co1l]

What switching frequency and source/supply inductance?  How much ripple is tolerable, or conducted emissions?

Tim

[radiolistener]

If all these capacitors are grouped in the same location on PCB, most of all this is needed to get capacitor high higher capacitance and low ESR.

Also, it is possible that some of them are used to reduce wire path between capacitor and power consumer (to reduce resistance and inductance) and some of them are used to get low ESR of the combined capacitor.

You can't see these PCB layout design details on schematic. On schematic all these cases looks like a bunch of capacitors are just connected in parallel.

[TimFox]

Back in graduate school, I hand-drew a diagram for a board combining analog and digital devices (before CAD).  For clarity, I showed the individual bypass capacitors for the ICs in the lower left corner (with a rectangle for each device showing the power pin numbers). Explicitly showing each capacitor at the IC on the drawing would obscure the signal connections.   An undergraduate technician interpreted that as putting the capacitors directly to each other in a corner of the board, rather than bypassing near the sockets.  He asked me why I didn't use a single 2 uF capacitor for the 20 0.1 uF bypass units.

[Siwastaja]

Thank you for your answer, but my doubts have not been answered until now. Let me reiterate my question again. How many MLCC capacitors should I use in parallel to meet my needs? The parameters of the motor I want to drive are as follows: nominal 48 v dc bus voltage, continuous current 12 a, 40 a peak current.

Please give me an answer directly and tell me why? thank you very much.

As many as it takes to "surround" or "cover" the layout. I have done roughly similar BLDC controllers and use around 10 to 15 caps, X7R 1206 2.2uF 63V would be a likely choice.

And no, they don't heat up.

Then slap in two 100-220µF 63V electrolytics. It's a good idea to check if these heat up.

A TVS that holds off at 50V but starts to clamp beyond 63V would be a good idea.

Motor control algorithms are then the biggest task. Don't forget to monitor the DC bus voltage and prevent regen if exceeding ratings.

[Benjam]

Thank you for your answer, but my doubts have not been answered until now. Let me reiterate my question again. How many MLCC capacitors should I use in parallel to meet my needs? The parameters of the motor I want to drive are as follows: nominal 48 v dc bus voltage, continuous current 12 a, 40 a peak current.

Please give me an answer directly and tell me why? thank you very much.

As many as it takes to "surround" or "cover" the layout. I have done roughly similar BLDC controllers and use around 10 to 15 caps, X7R 1206 2.2uF 63V would be a likely choice.

And no, they don't heat up.

Then slap in two 100-220µF 63V electrolytics. It's a good idea to check if these heat up.

A TVS that holds off at 50V but starts to clamp beyond 63V would be a good idea.

Motor control algorithms are then the biggest task. Don't forget to monitor the DC bus voltage and prevent regen if exceeding ratings.

If I don't want to use an aluminum electrolytic capacitor, can I? How to do?

[Siwastaja]

Electrolytic cap was recommended for damping oscillation from parasitic LC circuit forming with the input wire inductance and that MLCC bank. It's likely a problem because the ESR of the MLCCs is so low. Adding higher-ESR capacitance in parallel solves that nicely.

If you don't want to use elcaps, you can use MLCCs with explicit series resistors. But this will be an expensive solution. Elcap is just the right component because it offers more capacitance (you need at very least 2-3 times the MLCC bank capacitance), and the ESR is high enough internally (even in "low-ESR" types) for that damping purpose. So it's the smallest and cheapest solution.

TVS clamping is also an option which prevents damage caused by ringing and overshoot but can't remove the overshoot completely. Maybe this is the best for you if you really want to avoid elcaps, since the "3 times more MLCCs and as many series resistors" solution gets annoying, big and expensive pretty fast.