A DRSSTC's tank capacitor is chosen twice. Once for how many nanofarads, which is the resonance and takes an afternoon, and once for how many capacitors, which is the current and is where the money and the failures are.
This page is the second choice. The first is the frequency arithmetic, and it is on the primary page, which is where the ratio of L to C gets decided in the first place.
The reason the second choice is not obvious is that it runs backwards. Fitting larger capacitors to be safe gives you a bank that carries less current, not more, and the datasheets say so plainly once you are reading the right column of the right edition.
What it is and why
An MMC is capacitors in series to hold the voltage and in parallel to carry the current, and the array is sized by whichever of those two runs out first. In a DRSSTC it is almost always the current, because the voltage problem is solved by stacking cheap parts and the current problem is not solved by anything except more of them.
What follows is the ripple rating and how to read it, why smaller parts win, and why the strings do not share the load the way the schematic suggests.
What you decide
Averaged over time, not over the bang
Both the tank and the bus bank get called a bank, both have a ripple rating, and the two ratings are not interchangeable. Nanofarads of film in the primary is this page; thousands of microfarads behind the bridge is the other one.
I_rms(average) = I_rms(during the bang) · sqrt(duty)
duty = bang length × bangs per secondAnd the shape of the envelope inside the bang moves the answer by a factor of the square root of five:
- a flat top gives
I_pk / sqrt(2), so 0.707 of the peak; - a linear rise gives
I_pk / sqrt(6), 0.408; - a quadratic rise, which is what a QCW does, gives
I_pk / sqrt(10), 0.316.
Dissipation is I_rms^2 · ESR. The catalogue current is quoted for a stated temperature rise, and the figure to look for is ten degrees on most MKP film parts, which is davekni's reading of the catalogues and which he adds is often not printed on the sheet at all. Where it is printed it is not always the ten: of the two TDK sheets cited further down, the B3265* one states 20 °C and the B3264*B one 15, and the 2015 edition of that second sheet stated 10. Whatever the stated rise turns out to be, it gives you the thermal resistance for free and lets you work out the real rise rather than guessing at it. This page used to give twenty degrees as the general figure with nothing behind it; the twenty that survives further down is one sheet's own footnote and holds for that sheet's parts, not as a rule.
Check which edition of the datasheet you are reading
TDK's B32642B0333J is not the part it was on paper. EPCOS's edition of May 2015, mirrored by DigiKey, lists the part at VR = 1000 V DC against VRMS (f ≤1 kHz) = 600 V AC and draws its permissible-voltage curves for "Self-heating TA ≤10 °C". The June 2018 edition lists the same part at 500 V AC and redraws those curves for "ΔT ≤15 °C": a larger rise allowed, and a lower rating anyway. Read the 33 nF curve at 400 kHz off each and it goes from about 68 V to about 34. Half, on the same part number.
The public capacitor tables everybody links to were built on the older numbers.
Small capacitors carry more current per nanofarad
This is the counterintuitive one, and it is why "I will fit bigger ones to be safe" is backwards. At a fixed bank capacitance, smaller parts give you more current. From TDK's B3265* sheet of June 2026, the B32652 rows at 15 mm lead spacing and VR, DC = 1000 V DC, the amperes being that sheet's IRMS at 85 °C and 100 kHz "for a T ≤ 20 °C":
- 10 nF: 1.3 A each, 130 mA per nF. A 12.2 nF bank, the one on the machine this site works from, as 9 series by 11 parallel is 99 parts and 14.3 A.
- 22 nF: 1.9 A, 86 mA/nF. 9 by 5, 45 parts, 9.5 A.
- 100 nF: 5.4 A, 54 mA/nF. 9 by 1, nine parts, 11.1 nF, 5.4 A.
Nearly a factor of three in current for the same capacitance, and the price is count. Across the two ends of that list it is eleven times the parts for 2.6 times the current [derived, 99 against 9 and 14.3 against 5.4], because every extra parallel string needs its own series stack. That is faster than the square of the current gain, which is what this page used to say it was: 2.6 squared is 7.0 against the 11 the rows actually give.
And on a QCW, small in the other sense too
A QCW at the size most people build wants a small bank outright. Gao Guangyan summarises the machine as needing a high impedance primary, "coupling of >=0.3, many turns for primary and with a small tank cap of around 8 - 15nF", and on the forum Hydron told a builder who was planning fifty that "50nF is very large for a QCW coil - the required MMC normally works out to be under 20nF". The bank on the machine this site works from is 12.2 nF and sits inside that.
The reason is the tank's characteristic impedance. At a fixed frequency L and C are tied together, and:
Z = sqrt(L/C), primary current goes as C, power goes as Cwhile the voltage across the bank does not depend on C at all. So the fix for too much current is more primary turns and less capacitance, and the fix for too little is the reverse. Landon Kageler built his first QCW on a 12.35 nF primary capacitor, eight series by three parallel of the same B32642B0333J the edition change above is about, was disappointed by what it made, and then "simply increased the current by increasing the primary capacitance and lowering the inductance" to 30 nF, at which point he was reaching one and a half feet.
The rating that binds is not the one that looks binding
A pulse capacitor carries two current ratings and they fail in different directions. Barnkob works one array through both in his MMC chapter: six parallel strings of CDE 942C20P15K-F, each part rated 432 A peak and 13.5 A RMS.
peak 6 x 432 A = 2592 A against 800 A planned comfortable
RMS 6 x 13.5 A = 81 A against about 80 A at the ratingThe rating a builder reaches for is the peak one, because peak current is what a DRSSTC is about, and it is the one with three times the headroom. The one that actually decides is the RMS figure, and in his example the array sits on it.
What turns RMS into a temperature is a spec further down the same sheet: the dissipation factor, in degrees per watt. His part is given at 11 °C/W, and the arithmetic closes on his own numbers. Eighty amps across six strings is 13.33 A a string, which against 5 mΩ of ESR dissipates 0.889 W, and 0.889 W at 11 °C/W is 9.8 °C [derived, all three]. He writes it as almost ten degrees.
Ten degrees of what, though, is the question, and his answer is per second of running rather than per bang. He grades it:
0 to 5 C/s very good
5 to 10 good
10 to 15 not good
15 and up badWhich puts his own worked array at the top of good and one notch from not good, on a design he presents as the careful one. That is the honest shape of this decision: an MMC built properly still runs near a thermal edge, and the margin you have is measured in degrees per second rather than in amperes.
Fourteen parts, and what they do to the two claims above
Barnkob keeps a list of capacitors worth using in an MMC, fully specified on all seven figures his method asks for. Run the two arguments on this page against those fourteen parts rather than against the handful of rows they were written from:
nF I_pk I_rms ESR Rth mA/nF pk/rms
WIMA FKP1R032207F00 220 2420 6 9 33 27 403
WIMA FKP1O131007C00 100 1100 6 10 33 60 183
Panasonic ECWH16473 47 282 4.3 250 40 92 66
TPC CMPPX4K0K0405 4000 5000 80 0.75 6.9 20 63
CDE 942C20P15K-F 150 432 13.5 5 11 90 32
Kemet Arcotronics C4 5000 1600 64 1.1 3.3 13 25
EFD SP 2550-2 3750 2500 152 1 10 41 16Three things fall out, all of them ours and all of them arithmetic on his figures [derived].
Small parts really do carry more per nanofarad. Averaged across his list, parts at 150 nF and under give 81 mA of RMS rating per nanofarad and parts at 2 µF and over give 21. Nearly four times, on a list assembled by somebody else for a different purpose than proving it.
The peak rating almost never binds. Across the fourteen, I_pk over I_rms runs from 7 to 403 with a median of 57. So the rating a builder reaches for is typically fifty times looser than the one that decides, and the section above is not a close call, it is the usual case by a wide margin.
And the thermal figure is the wild one. Rth on this list runs 3.3 to 51 °C/W, a spread of fifteen. That is the term the heating arithmetic multiplies by, so two parts that look interchangeable on capacitance and voltage can differ fifteenfold in how many degrees a watt buys.
And the current does not divide the way you think
Between parallel strings, the current divides in inverse proportion to ESR and path resistance, not to capacitance. At the frequencies a tank runs at, capacitance is not what decides it.
One idea worth disposing of, and the disposal is ours rather than anybody's we can cite: small capacitors do not make a poor bank. A network's ESR does not depend on the size of its unit, because R_bank = R_unit·n/m and tan δ/(2πf·C_bank) are the same number with the unit capacitance cancelling out [derived], and dividing the heat between more parts improves the cooling. This page carried the opposite claim in its first version and it was wrong. What limits it in practice is the number of joints and the symmetry of the paths, not the physics.
One more thing that follows from where the heat goes: the leads are a thermal path and not only a connection. davekni takes his film parts' case rise "at hottest point at center between leads", and notes that in his fixture the "Leads provide heatsinking to copper foil". Cool the leads, and space the parts so air can pass between them.
The rig on the DRSSTC department page draws the bank as a matrix rather than one symbol, because the number of parts is the point. Take the tank capacitor out on the QCW page and you have a ramped SSTC, which is a different machine with a different set of problems.
The two capacitors that were measured here, the Samsung ceramic and davekni's tank, are linked to the threads they were measured in, and so now are Gao Guangyan's range, Hydron's ceiling and Landon Kageler's rebuild. So are the catalogue rows, which used to carry no source at all: the 1000 V family and the B32652 against B32653 comparison are TDK's B3265* sheet of June 2026, and the edition change is TDK's own two editions of the B3264*B sheet, the older of them reachable only through a web archive. The one thing in that comparison that is not quoted is the pair of voltages at 400 kHz, which are measured off the published curves and are marked as such where they appear. The arithmetic on all of them is ours.
Two letters on the part, and one of them is a failure mode
Builders quote tank capacitors by their type code and this site has been printing those codes without unpacking them. Gao's QCW 2 spec reads "5.875nF MKP 3.2kVAC/8kVDC MMC", and the middle three letters carry more than the voltages beside them.
Both are polypropylene. What differs is how the film is made conductive, and therefore what happens the first time a layer punches through (Barnkob's MMC chapter):
FKP film and foil no self healing punch through fails SHORT
MKP metallized film self healing punch through fails OPENAn MKP heals by destroying itself locally: the little internal explosion burns the metallised layer away from around the hole and isolates it, so the part carries on with marginally less capacitance. An FKP has nothing to burn away, and the fault stays.
Put that against what this page already says about a series string, which is that losing one member hands its share of the voltage to the survivors. With MKP the string keeps standing and quietly loses a little capacitance. With FKP the first punch through is a short across one position, and the rest of that string is over its rating from that instant. The choice of letters decides whether a fault is a drift or a cascade, which is why MKP is what the guides recommend for this job.
The rest of the film comparison is worth one look, because it contains a trade that reads backwards. Across polyester, PEN, PPS and polypropylene, the dissipation factor at 100 kHz is 170 to 300 for polyester and 2 to 25 for polypropylene in the same units, an order of magnitude and more, which is the whole reason the tank uses it. And polypropylene has the lowest maximum temperature of the four, +105 °C against +150 for PEN and PPS. The material you must use to keep the heat down is also the one with the least room for it once it is there [our reading of his table].
And the sharing is not even in the first place
His chapter carries two named failures worth having: Amaury Poulain's MMC overheated on the strings with the shortest path to the terminals, and another builder's part melted apart entirely. Both are the same fault as the layout point above, seen after the fact.
His remedy is the same one, plus a number this page did not carry: derate a parallel array by 40 per cent to allow for the sharing being uneven no matter how carefully you lay it out. That is a large factor, and it is multiplied against the RMS rating rather than the peak, which is the rating that was already the binding one.
Getting the numbers the sheet does not print
Everything above needs an ESR and a thermal figure, and a pulse capacitor's datasheet often prints neither. Barnkob's chapter gives the conversions (the MMC chapter):
ESR = tan(delta) / (2*pi*f*C) from the loss angle, and frequency dependent
Rth = 1 / (W/K) when the sheet gives watts per kelvin
I_pk = C * dV/dt the peak rating from the slew rating
X_c = 1 / (2*pi*f*C)
ESL about 1.6 nH per mm of lead including the capacitor's own leadsThe first is the one that matters most here, because tan delta is what manufacturers print and ESR is what the heating arithmetic runs on. It is frequency dependent, so take it at the frequency your tank actually rings at rather than the 1 kHz or 100 kHz the table is quoted at. The last is a rule of thumb and he says what it is worth: it holds for well designed parts, and this site uses lead inductance in three other places without ever putting a figure on it.
And the shape of the can changes the answer
Two capacitors of the same value and the same rating are not the same part thermally. Work by El-Husseini, Venet, Rojat and Joubert on metallised polypropylene geometry, which he cites, finds that taller capacitors run hotter than shorter ones under the same electrical stress, because the current has further to travel through very thin metallisation and the total I²R is larger for it.
Put that beside the section further up, which says smaller parts carry more current per nanofarad. The two are about size from opposite directions and they do not cancel: the array wants many small parts for the current, and each of those parts wants to be short rather than tall for the heat. A tall capacitor of the right value can lose on both counts.
The cooling advice that comes with it is the part builders get wrong, because the case does not shed heat evenly. Roughly two thirds of it leaves axially and one third radially, so the terminals are where a capacitor is actually cooled, not the middle of the body. And the thermal resistance in the sheet is quoted for still air, which means forced air lets you de-rate it rather than obliging you to accept it.
What will get you
Reading a current rating without its conditions. Every ampere on this page is quoted with the frequency, the temperature and the permitted self heating it was measured at. An IRMS with those stripped off is not a number, and the two TDK editions above differ on exactly that basis rather than on the part.
Fitting bigger cans to be safe. It is the one instinct this page exists to break. At a fixed capacitance the larger part carries fewer amperes per nanofarad, and a bigger case is not even a reliable guide within one series.
Laying the strings out however they came out. The current divides on ESR and path resistance, so the string nearest the bus takes more of it and gets hotter, and matched capacitors do not help.
What goes wrong
One string runs hot and the others do not. Layout, not parts. Measure the path lengths before you measure the capacitors.
The bank is within its rating and dies anyway. Check which edition of the sheet the rating came from, and check whether the number was VRMS or VR, DC.
A part fails and the rest look fine. A series stack divides the voltage between its members. Losing one hands its share to the survivors, so the stack that killed one part is now over its rating for the rest.
Where next
The bus bank behind the bridge, which is a different capacitor with a different job and is genuinely sized by easy arithmetic, is the DRSSTC bus capacitor bank, the page this one was cut out of.
What the same part does on a ramped machine, where the current is deliberately lower and held fifty times longer, is the QCW tank capacitor.
Whether the tank resonates with the primary at all, which is not a given on every topology, is the capacitor that decides what you built.