How would *you* design a capacitor ripple current tester?

[TimNJ]

Hi everyone,

First off, I'm not 100% serious about starting this project, but I've been kicking around the idea for a while. And, for the sake of brainstorming, thought I'd ask here. Chroma (and other manufacturers) already make electrolytic cap ripple current testers, if you have $10,000 to spare, but in the event that you don't have that kind of money...

I'm wondering what approach you might take to DIY a reasonable ripple current tester for electrolytic caps. The goal is to verify datasheet endurance specifications, i.e. 10,000 hours @ 105C,  full rated ripple current, rated voltage, etc. Forget the temperature aspect for now...

My first guess:

Power op-amp (TO-220 type) with some current control loop based on 100KHz sine wave oscillator
Add DC voltage bias...how?

Generally, the RMS current range requirement would be 0 - 6A or 7A, if possible. I have yet to see a polymer cap with a ripple current rating much above 6A.

Thanks,
Tim

[Gyro]

My first guess:

Power op-amp (TO-220 type) with some current control loop based on 100KHz sine wave oscillator
Add DC voltage bias...how?

You could always AC couple the output of the oscillator. 


... actually, joking aside,  maybe you could if you used one of those beefy high ripple polypropylene foil capacitors (Epcos/TDK?)

EDIT: https://www.tdk-electronics.tdk.com/en/1641622/products/product-catalog/film-capacitors/power-capacitors

[floobydust]

I would use a decent audio power amplifier, DC-coupled and derated for 100kHz use. There are designs that are tough and fast enough.

[Ian.M]

I'm probably being thick but what's the problem with applying DC bias?  Cant you simply apply it to the other side of the capacitor, fed through a resistor from a variable bias supply, with a massive parallel array^* of decoupling caps to provide the return path to ground for the current injected by the power OPAMP. 

* The reason for a massive decoupling array is to split up the ripple current between enough capacitors to avoid fatigue testing the test jig rather than the DUT.

[TimNJ]

I'm probably being thick but what's the problem with applying DC bias?  Cant you simply apply it to the other side of the capacitor, fed through a resistor from a variable bias supply, with a massive parallel array^* of decoupling caps to provide the return path to ground for the current injected by the power OPAMP. 

* The reason for a massive decoupling array is to split up the ripple current between enough capacitors to avoid fatigue testing the test jig rather than the DUT.

Don't worry, 100% chance that it's me who's being thick.    I barely even sketched this down on paper. My original thought was that some how you'd have control loops in conflict with each other. But uhh..yeah, probably just a resistor from the DC bias source small enough to counteract the leakage current of the cap.

[TimNJ]

My first guess:

Power op-amp (TO-220 type) with some current control loop based on 100KHz sine wave oscillator
Add DC voltage bias...how?

You could always AC couple the output of the oscillator. 


... actually, joking aside,  maybe you could if you used one of those beefy high ripple polypropylene foil capacitors (Epcos/TDK?)

EDIT: https://www.tdk-electronics.tdk.com/en/1641622/products/product-catalog/film-capacitors/power-capacitors

Haha, thanks! I feel like you got it right the first time? A foil capacitor can also be a polypropylene capacitor, as far as I know. Foil should have the lowest ESR/ESL, best ripple performance. Metalized film, where they spray one or both sides of the dielectric (poly) film is cheaper and more widely used, I think. But, thinner "plates".

[TimNJ]

I would use a decent audio power amplifier, DC-coupled and derated for 100kHz use. There are designs that are tough and fast enough.

Any recommendations? I checked the usual suspects like TDA1875, LM3886, and so on, but no characterization above 20KHz. I'm not sure what to expect.

[Jay_Diddy_B]

Hi,

I would take a buck regulator controller that operates in the forced continuous mode. The LTC3891 is a good example. I would set the output voltage to be 1/2 of the input voltage. The value of the inductor determines the ripple current.

The ripple current will be triangular, not sine, but this really doesn't matter. Because of the symmetry, there is no second harmonic. The third harmonic is 1/9 th. the amplitude of the fundamental.

The bias voltage is equal to the output voltage.

Only the losses need to be supplied. Most of the energy circulates.

Here is a quick model and the FFT.








Regards,
Jay_Diddy_B

 3891.asc (3.28 kB - downloaded 70 times.)

[TimNJ]

Hi,

I would take a buck regulator controller that operates in the forced continuous mode. The LTC3891 is a good example. I would set the output voltage to be 1/2 of the input voltage. The value of the inductor determines the ripple current.

The ripple current will be triangular, not sine, but this really doesn't matter. Because of the symmetry, there is no second harmonic. The third harmonic is 1/9 th. the amplitude of the fundamental.

The bias voltage is equal to the output voltage.

Only the losses need to be supplied. Most of the energy circulates.

Here is a quick model and the FFT.

Regards,
Jay_Diddy_B

Thanks for the recommendation. This is probably the most "logical" option in the sense that the capacitor under test is...literally just being used in a switch-mode power supply.

The question is: Can you adjust the pk-to-pk ripple current simply by adjusting the current sense feedback signal, perhaps by programmable gain amplifier? I'm pretty confident you'd need to physically adjust the inductance value to change the ripple current, which is obviously not so convenient.

[Jay_Diddy_B]

TimNJ,

you can trim the ripple current in the capacitor by adjusting the 24V input.

New model




I haven't stabilized the control loop properly 

Jay_Diddy_B

[Jay_Diddy_B]

Hi,

You can add a blocking capacitor and a bias supply if you want to:




I have stabilized the loop a little better.

Jay_Diddy_B

[NiHaoMike]

If you're OK testing capacitors in pairs, connect one side directly and the other side to a transformer. Then add a H bridge to drive the transformer and a resistor to connect a bias supply. Easy to scale up to very high currents and voltages.

[Gyro]

My first guess:

Power op-amp (TO-220 type) with some current control loop based on 100KHz sine wave oscillator
Add DC voltage bias...how?

You could always AC couple the output of the oscillator. 


... actually, joking aside,  maybe you could if you used one of those beefy high ripple polypropylene foil capacitors (Epcos/TDK?)

EDIT: https://www.tdk-electronics.tdk.com/en/1641622/products/product-catalog/film-capacitors/power-capacitors

Haha, thanks! I feel like you got it right the first time? A foil capacitor can also be a polypropylene capacitor, as far as I know. Foil should have the lowest ESR/ESL, best ripple performance. Metalized film, where they spray one or both sides of the dielectric (poly) film is cheaper and more widely used, I think. But, thinner "plates".

Ha, yes. The reason I struck through 'foil' is that Polypropylene foil only tend to available at low capacitance values (or are presumably massive in large values). I think most of the ones that I linked are film types, but at a random pick they were good for around 70A 'continuous' at 320-350uF. Clearly all films aren't as as fragile as X caps! The linked ones are the upper end of the range (M6 terminals), but I'm sure I've seen smaller heavy duty Epcos snubber ones with mounting tabs.

[Berni]

I would also just make a switchmode buck converter with a really under speced inductor that would cause a large ripple current on the capacitor trying to filter it. The amount of ripple current only limited by how beefy of a switchmode chip you get, since the inductor can easily be hugely overspecced by making it a air core inductor made from thick wire.

If the aim is for testing non polarized capacitors with no bias on a sinewave then i would build a ZVS oscillator to make a LC circuit making this capacitor oscillate with a nice continuous sine wave, again choosing a huge overspecced inductor with the correct inductance to create the desired ripple current. Since most of the current is recirculating between the inductor and capacitor means that very large currents in many 10s of amps is easily achievable without any powerful switching components. These are the sort of ripple currents that make even large beefy film capacitors get rather toasty.

[TimNJ]

Hi,

You can add a blocking capacitor and a bias supply if you want to:

I have stabilized the loop a little better.

Jay_Diddy_B

Thanks. This is probably conceptually straight forward to implement for my mind.

But, still, to get a range of even 1A - 5A RMS ripple current, the adjustable input voltage range needs to have a similar range, something like 12 - 54V range, let's say. So I need some flyback/forward based power supply with a good adjustable output, or a fixed high-ish voltage supply with an intermediate buck after it. Now it's getting more complicated.

The blocking capacitor, I assume,  should be high ripple current film/foil type.

[TimNJ]

If you're OK testing capacitors in pairs, connect one side directly and the other side to a transformer. Then add a H bridge to drive the transformer and a resistor to connect a bias supply. Easy to scale up to very high currents and voltages.

I'm failing to visualize the setup. Any similar circuit on google images you can point me to?

[Jay_Diddy_B]

Hi,

You can add a blocking capacitor and a bias supply if you want to:

I have stabilized the loop a little better.

Jay_Diddy_B

Thanks. This is probably conceptually straight forward to implement for my mind.

But, still, to get a range of even 1A - 5A RMS ripple current, the adjustable input voltage range needs to have a similar range, something like 12 - 54V range, let's say. So I need some flyback/forward based power supply with a good adjustable output, or a fixed high-ish voltage supply with an intermediate buck after it. Now it's getting more complicated.

The blocking capacitor, I assume,  should be high ripple current film/foil type.

Remember you only need to supply the losses, so the power required isn't very high.

You only need the bias supply and the blocking capacitor if the voltage you need is not equal to the output voltage.

You can do course steps by inductor selection and fine steps with the supply voltage.

IMHO, it is heat that ages the electrolytic capacitors.

Heating effect is (ripple current)² x ESR

The bias voltage will have very little effect.


Jay_Diddy_B

[TimNJ]

Hi,

You can add a blocking capacitor and a bias supply if you want to:

I have stabilized the loop a little better.

Jay_Diddy_B

Thanks. This is probably conceptually straight forward to implement for my mind.

But, still, to get a range of even 1A - 5A RMS ripple current, the adjustable input voltage range needs to have a similar range, something like 12 - 54V range, let's say. So I need some flyback/forward based power supply with a good adjustable output, or a fixed high-ish voltage supply with an intermediate buck after it. Now it's getting more complicated.

The blocking capacitor, I assume,  should be high ripple current film/foil type.

Remember you only need to supply the losses, so the power required isn't very high.

You only need the bias supply and the blocking capacitor if the voltage you need is not equal to the output voltage.

You can do course steps by inductor selection and fine steps with the supply voltage.

IMHO, it is heat that ages the electrolytic capacitors.

Heating effect is (ripple current)² x ESR

The bias voltage will have very little effect.


Jay_Diddy_B

True true. Probably a 10 or 20W rated power supply would be plenty of overhead. The only other issue is maintaining good stability over a wide range of input voltage and a wide range of DUT capacitance and associated ESRs. Usually the control loops are tuned around a given capacitance and ESR. Now we are asking to deal with virtually any capacitance and virtually any ESR.

[ogden]

I would take programmable DC supply, programmabe DC load, three DUT capacitors, two powerful & fast enough MOSFET switches and make programmable charge transfer switcher. Middle capacitor would be subject to positive/negative [nominal] pulses and actual subject of the test. Supply and load side capacitors - more or less sacrificial, to not stress supply or load. Current would be measured on middle cap. Such setup seems to me more flexible than buck converter which I find as great idea indeed.

[NiHaoMike]

I'm failing to visualize the setup. Any similar circuit on google images you can point me to?

Take two capacitors and connect the negatives together. Now take the positives and connect them to the secondary of the transformer. The transformer is used to lower the voltage but increase the current so that the power electronics will be a lot easier to design. That takes care of pushing AC through the capacitors, now for the DC bias, you connect the negative side of the bias supply to the negatives of the capacitors and the positive side to one of the capacitor positives through a resistor, the resistor sized such that there would be negligible voltage drop across it once the capacitors charge yet be very large compared to the ESR. Something like 1k will work great for most setups.

[T3sl4co1l]

Hah, I breadboarded one just the other day, to test a resonant circuit.  Half bridge into series resonant tank, returned to ground through a cap divider.  The supply bypass and C divider are all 1000uF 25V caps (supply is 12V) and the half bridge is a IR2101 into a pair of IRFZ34N, plus an inverter to complement the high side drive, and a pair of RCD networks to skew the rising edges (setting dead time).  Signal source from the function generator -- about as simple as could be.

Driving an inductor in bursts, I've got peaks up to 25A, not too shabby for breadboard work.  (Measured with a current transformer, of course.)  The transistors aren't too happy up there, and neither are the capacitors; I started with the divider being a pair of middling 220uF 50V types, which got rather hot, rather quickly, their ESR also limiting my output current.  So, I moved to the larger, low ESR types.  Down at a more modest 5-10A RMS, everything is cool enough for continuous operation.

Anyway, the resonant tank is a good way to get nearly sinusoidal current.  The Q isn't very high for the inductor I'm testing, so the peak isn't very sharp, easy enough to track by hand.  To design something general-ish in purpose, you'll need to start with a range of ESRs, with the worst case minimum setting how much base ESR your circuit needs (to limit maximum Q and current) and the maximum setting what resonant impedance (Zr = sqrt(L/C)) you need, and whether you'll need multiple L and C to select from, to cover that range, given available supply voltage and whatnot.

In the other tests above, triangular or square current isn't a bad thing, by itself.  ESR should be relatively insensitive to harmonics, as far as I know, and the RMS value of all harmonics in a square wave only amounts to, what was it, 18% of the total?  And of a triangle, even less.

Tim

[Berni]

The issue with a series resonant circuit is that your mosfets need to carry all the current.

The ZVS oscillator makes use of a parallel LC cirucit (Here formed by C1 and TR1 ) that keeps the high current nicely contained:
http://www.kiblerelectronics.com/bob/app_notes/note11/note11.html


This means the mosfets only have to carry the current needed to keep the circuit going by covering the losses. Its a pretty popular circuit for cheep DIY induction heaters due to the ability of small mosfets being able to handle a lot of power (they switch at zero crossing so they are very efficient). The massive recirculating currents in the LC circuit is what makes it nice for a induction heater since the strong field the coil ends up making is sure to create hefty losses in the workpiece you want to heat. But the C side of the LC cirucit also has heavy work to do, typically its made from many foil capacitors in parallel and even those get toasty.

You could repurpose the same circuit by placing your device under test capacitor in series with the LC tank capacitor, forcing all the circulating current to go trough it. Its also AC coupled so you could inject a DC bias onto the tested capacitor via a resistor.

[T3sl4co1l]

Indeed; what we're really doing is matching a resistance to a source.  ESR tends to be low relative to typical amplifier ratings, therefore some matching is desirable.  Can be simply a transformer, can be parallel resonant, LLC, whatever.

And as you note, series resonant, and the series-L above, doesn't do us any favors as the total (reactive and real) current goes through the transistors.  The only thing you get, is some control over output current by adjusting frequency, rather than varying supply voltage.  So, it may be better than direct drive from a linear power amp, assuming the waveforms are an acceptable compromise; but it's not a whole lot.

Tim

[Jay_Diddy_B]

Hi group,

Here is an LTspice model of a ZVS design optimized for ripple current testing.




I will let you run the model and measure the waveforms.

Regards,
Jay_Diddy_B
 ZVS.asc (3.25 kB - downloaded 163 times.)

[Berni]

Hi group,

Here is an LTspice model of a ZVS design optimized for ripple current testing.

I will let you run the model and measure the waveforms.

Regards,
Jay_Diddy_B

Nice touch on the Q1 gate supply undervoltage protection. Most people wire that directly to the supply and this makes the power MOSFETs prone to exploding if the circuits power supply is not brought up fast enugh to get it into oscillation. This causes both mosets to turn on, huge currents to flow and BANG.

Lovely cirucit.