QC testing power film capacitors 1

[MagicSmoker]

Greetings all - what a site!

Assuming the tomatoes aren't launched my way in force I will probably be bouncing a bunch of ideas off you all as I have a few projects starting up in the next few weeks.

But for now I'm looking for some ideas on ways to do incoming QC testing of large dc link film capacitors, e.g. - DigiKey part number 338-1923-ND. Specifically, I want to run the rated ripple current through it at, say, 10kHz. The best idea I have come up with is to make, essentially, a big Royer oscillator and use the capacitor under test as the resonating element along with the inductance of the secondary of a step down transformer.

I like the Royer idea because it makes the capacitor part of a resonant tank. This will let me get the circulating current up high enough to be useful - in this case, somewhere between 75 and 100A - without requiring a lot of power.

Why are we doing this testing? Well, let's just say we had some problems with one film capacitor vendor. Their end spray metallization wasn't quite up to carrying the 500A of ripple they promised and a few of our motor controllers (basically a buck converter on steroids) popped off in the field. We were not happy.

Anyway, one potential problem with the Royer idea is that the current will be sinusoidal, while in a buck the current is square (well, ramp on a step).

So, I'm open to suggestions. Mainly just want a way to run 75-100A through the caps at 8-10khz for 1 hour. That should cook them to a nice golden brown, no?

 

[IRstuff]

Obviously, they need to be tested, but I'm unclear why you didn't require the supplier to do the testing.

TTFN

FAQ731-376

 

[MagicSmoker]

IRstuff - what was that phrase Ronald Reagan used w/r/t whether the Soviets were really reducing their nuclear weapons? Trust but verify.

Basically, if this capacitor fails it causes immense collateral damage. Given our recent experience, we would much rather have the caps fail in testing, rather than in the field.

 

[IRstuff]

I understand the issue, but like Schroedinger's cat, the more you look at it, the deader it gets.

You've given no information about how much margin you've designed for, but it's common practice to not run such devices at their operational limits. Unless Digikey is sending me to the wrnog datasheet:

http://www.cde.com/catalogs/944U.pdf

they don't look like they should be run at the currents you've indicated.

I'm not sure I've got all the math right, but it appears that running these capacitors at the currents you're suggesting will not only cook them, but seriously degrade their expected lifetime, so that you won't have a few infant mortalities, you've have hugh numbers of latent failures, particularly if there's no airflow.

TTFN

FAQ731-376

 

[ScottyUK]

You might want to look at a single phase current-source inverter circuit, particularly the type with a resonant load such as used in induction heating applications. That gives you the resonant tank circuit with the ability to bring the Q up pretty high so your external power requirements are low. One of the CSI's nice features is that the rate of rise of fault current is fairly low, being limited by the DC link inductor, so there is time for a protection circuit to activate before things go in to meltdown.


----------------------------------

If we learn from our mistakes I'm getting a great education!

 

[MagicSmoker]

ScottyUK - thanks for the CSI suggestion. That might be even easier to cobble together with parts from the junk bin than the Royer. I definitely wanted to go with a circuit with intrinsic fault tolerance - which pretty much means some sort of resonant oscillator - but with good circulating current capability without a crazy transformer design - which sort of argues against the Royer.

IRstuff - I don't think you got the math right, but me providing some more detail certainly wouldn't hurt.

The application is, essentially, a buck converter. Worst case ripple current for the single switch (non-interleaved) buck is at 50% duty cycle and maximum allowed output current (Iripple = 0.5 * Iout), but Iripple is still at 40% of Iout with duty down to 0.20 and up to 0.8, so, a hard life for this capacitor, indeed.

However, the application, more specifically, is a dc traction motor controller, so the process variable is torque (current) and, of course, it neither spends all of its time at maximum output nor can it run 24/7 (batteries have to be recharged, after all).

In my experience, you derate capacitors for temperature or current, but not both unless you have no budgetary or space restrictions and need maximum reliability over all else. Not really a luxury that can be afforded in every instance, I'm sure you realize.

So, the datasheet for this capacitor says it can handle 73A[rms] at 55C ambient. It also says that the max ESR is 8m? at 10kHz and the total R[?] is 7.5C/W.

Plug those numbers into the venerable I²R equation and you get 4.3W of loss which leads to a hot-spot temp of 87C at a 55C ambient. At 25C I could run 100A[rms] through the capacitor and keep the hot-spot to the same temp.

Maximum bus voltage is ~340VDC, but 180-240VDC is much more common, so there is plenty of margin there.

 

[zappedagain]

I can't get the datasheet to download at the moment; so what is the current capability at 87C?

John D

 

[MagicSmoker]

Hey zapped... a graph says it can handle 20Arms of 10khz ripple at 85C and survive 5000 hours. That would be approximately 10000 operating cycles in this application which is eminently acceptable.

 

[zappedagain]

"it can handle 20Arms of 10khz ripple at 85C and survive 5000 hours"

"Mainly just want a way to run 75-100A through the caps at 8-10khz for 1 hour."

Sounds like your testing may significantly shorten the operating life of the part. I hope you are only sampling from a lot.

Also,

20A^ * 0.008 ohms = 3.2W, or 24C rise at 85C (100C max temp)

73A^2 * 0.008 ohms = 43W, not 4.3W, for a 320C temperature rise at a 55C ambient (375C max). I hope that's only a typo in there, otherwise the data looks suspicious. These two sets of data should give the same maximum temperature; if the data is correct then I suspect I^2R is an inadequate model.

John D

 

[MagicSmoker]

Sorry zapped - yeah, a typo.. ESR is 0.8m?. not 8m?.

Were you ever able to get the datasheet? It says it can 73A[rms] of 10khz ripple at 55C ambient; running 75-100A of 10khz ripple through it at 25C ambient should be fine. I don't think I'm missing anything here, but if so, please point it out!

Otherwise, no, this testing will be done to EVERY capacitor incoming.

 

[zappedagain]

I finally got the datasheet to download. It is a little suspect because the chart on the last page shows an ESR of about 1.9 mOhm at 55C; the table on the first page says 0.8 mOhm.

That makes me a little suspicious of their data plots - what temperature are they showing, ambient or device (core) temperature? The chart on p. 3 leads me to believe they are talking core temperature for all graphs, but that is only an assumption. That can push your current capabilities down significantly, so I believe a call to the manufacturer is in order.

John D

 

[zappedagain]

This isn't an exact fit to your situation, but it is timely!

http://www.edn.com/article/509092-What_a_cap_astrophe_.php

 

[MagicSmoker]

Yep, I will be having a little chat with each manufacturer of this style capacitor, but setting aside the issue of whether it is appropriate, etc., I'm still interested in hearing suggestions on how to test this. I can't make up my mind to go with a series or parallel resonant circuit. Royer or driven (but not hard switched). Adapt an induction heater to the job, or totally roll my own.

I posted here to hopefully get some ideas

Ideally the circuit is tunable, intrinsically safe (capacitor shorting or opening results in non-operation, not destruction), and easy to adjust the AC current.

The Royer still seems to fit this bill, as does a CSI induction heater circuit.

And, yeah, tantalums in a low-impedance circuit is always asking for trouble, no matter what the mfg says.