11
\$\begingroup\$

Capacitors can be wired in parallel for higher total capacitance and lower total ESR.

What are some practical constraints/pointers for using many parallel capacitors (for example, 10, 25, 50, or 100)?

Application example: using 30 330 µF aluminum polymer capacitors to replace two 5000 µF wet electrolytic ones inside devices intended for long-life at low-variance operating temperatures (for example, illuminated ocean buoy).

\$\endgroup\$
4
  • 1
    \$\begingroup\$ voltage balance resistors so if a series cap fails it doesn't take the others with it \$\endgroup\$
    – DKNguyen
    Commented Oct 22, 2019 at 1:21
  • \$\begingroup\$ inrush current? \$\endgroup\$
    – vicatcu
    Commented Oct 22, 2019 at 2:44
  • 5
    \$\begingroup\$ capacitors and PCB traces (or wires) will resonate. 100uF and 10nH (~1cm) will resonate at 160KHz. To dampen (Q about 2), you only need 10 milliOhms [ compute as sqrt(L / C) ] and 20 squares of standard copper foil provides the 10 milliOhms. Thus lots of these potential power-rail issues are hidden, the resonances fortuitously prevented. \$\endgroup\$ Commented Oct 22, 2019 at 10:53
  • \$\begingroup\$ Related : ESR and parallel capacitors \$\endgroup\$
    – J...
    Commented Oct 22, 2019 at 14:34

3 Answers 3

17
\$\begingroup\$

As DKNguyen pointed out it is important to "account for" or "manage" failures. We put capacitor banks on the 28 VDC power bus of spacecraft all the time. EVERY cap had its own fuse, so if the cap shorted out - which was the predominant failure mode - the fuse would blow and take the cap out of the circuit. We were also concerned about partial shorts of a cap, so we broke the entire bank into 2, 3, or 4 sections and each section was fed with a relay so that we could disable that section of the cap bank if there was a partial short. If you can , use a cap that has detailed specs on internal resistance and inductance and make a model on Spice (I use LT Spice) and check out the response across a wide frequency range - maybe up to 10 MHz.

Thanks !!

\$\endgroup\$
5
  • \$\begingroup\$ Would you use NO, NC, or latching relays for something like that? \$\endgroup\$
    – DKNguyen
    Commented Oct 22, 2019 at 3:05
  • \$\begingroup\$ @DKNguyen For a safety application like that, I wouldn't be surprised if safety relays with forced contactors were used. Nowadays, that is. Who knows what was used back in the day. \$\endgroup\$
    – Mast
    Commented Oct 22, 2019 at 9:46
  • \$\begingroup\$ It would be useful to know how you calculate what fuse is suitable for a given size of cap bank / of the caps in that bank. \$\endgroup\$ Commented Oct 23, 2019 at 10:23
  • 1
    \$\begingroup\$ Hi DKNguyen,Since we were obsessed with low power consumption/low energy usage, we used latching. So we manually detected "unusual" current drain "somewhere" and turned off cap banks individually to see if one of them was the culprit. In practise, the relays and fuses were a good precaution, but we had very Hi-rel parts, conservatively derated, and I do not recall ever having a failed cap in the Cap Bank. \$\endgroup\$
    – xstack
    Commented Oct 23, 2019 at 15:39
  • 1
    \$\begingroup\$ Hi leftaroundabout, Every new spacecraft design had a full fledged prototype. We powered "the bird" from a solar array simulator and exercised the heck out of it, while measuring bus voltage and cap bank current. We used wideband current probe (Tecktronix) and looked carefully at current sharing. After gathering that data we "stared" at the fuse "blow" characteristics and came up with a sizing. Generally we set the fuses 10x higher than the highest sustained average current. Example there might be high avg. current for 5 secs, due to a noisy load. So we would use that. \$\endgroup\$
    – xstack
    Commented Oct 23, 2019 at 15:46
8
\$\begingroup\$

Fusing is good, as stated by xstack. Ripple current sharing can be an issue when AC currents are high and tracks are long. Equalising trace impedances is good practise here. Thermal stability should be considered, so the capacitors should be at a relatively even temperature part of the product enclosure.

If, say, one capacitor out of a parallel group of eight was near some hot choke and its ESR fell then it would hog the ripple current and die young.

\$\endgroup\$
1
\$\begingroup\$

Any capacitor turning into a dead short for any reason will have access to all the short circuit current of all the other capacitors combined. Which can lead to something that could be described as either a self repair mechanism (the shorted capacitor will be demolished and neutralized quickly) or self destruct mechanism (so will anything near it). Cascade failure is theoretically possible (if another capacitor is shorted due to the damage).

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.