Tek-Tips is the largest IT community on the Internet today!
Members share and learn making Tek-Tips Forums the best source of peer-reviewed technical information on the Internet!
-
Congratulations cowski on being selected by the Eng-Tips community for having the most helpful posts in the forums last week. Way to Go!
How to measure capacitor ripple current
- Thread starter outlawed
- Start date
Capacitor ESR varies widely with frequency, and there is also ESL, so measuring the ripple voltage across the capacitor and dividing by the datasheet value for ESR will not give you a very accurate ripple current measurement.
You could lift one leg of the capacitor and put a current meter in series. I remember seeing another thread on EngTips where someone had a small current sense loop that fit around the legs of SMD components, but I cannot find that thread now.
You could lift one leg of the capacitor and put a current meter in series. I remember seeing another thread on EngTips where someone had a small current sense loop that fit around the legs of SMD components, but I cannot find that thread now.
No, you cannot measure the voltage and divide by the ESR.
Anything in series with the cap will greatly reduce the actual ripple current messing up the measurement.
You would need something like a clamp-on current probe.
Ripple heats capacitors. It limits the system's ultimate running temperature. If you can run your system steady-state until the temperature is stable and determine the cap temp, you can assess whether or not your ripple level is going to be an issue. If the temp rise is too great, you will need more capacitors to share the ripple.
Keith Cress
kcress -
Anything in series with the cap will greatly reduce the actual ripple current messing up the measurement.
You would need something like a clamp-on current probe.
Ripple heats capacitors. It limits the system's ultimate running temperature. If you can run your system steady-state until the temperature is stable and determine the cap temp, you can assess whether or not your ripple level is going to be an issue. If the temp rise is too great, you will need more capacitors to share the ripple.
Keith Cress
kcress -
well
we are used to hearing how academic calculations are useless.
but not in this case.
U dont say topology so i assume boost.
U know your duty cycle by scoping fet gate etc.
u work out the boost diode trapezoid current shape.
then either use integration
..or go to google and search for "rms of smps waveforms" or similar.
"smps waveform calculator" etc etc........
...put in peak and pedestal current levels of the trapezoid, and duty cycle and it will give you the rms, the dc and the AC.
the AC is what your caps take.
of course if you want to se what each of the paralleled caps take then it will be in proportion to their faradic values.
so you can try this. it definetly give u the tops value.
we are used to hearing how academic calculations are useless.
but not in this case.
U dont say topology so i assume boost.
U know your duty cycle by scoping fet gate etc.
u work out the boost diode trapezoid current shape.
then either use integration
..or go to google and search for "rms of smps waveforms" or similar.
"smps waveform calculator" etc etc........
...put in peak and pedestal current levels of the trapezoid, and duty cycle and it will give you the rms, the dc and the AC.
the AC is what your caps take.
of course if you want to se what each of the paralleled caps take then it will be in proportion to their faradic values.
so you can try this. it definetly give u the tops value.
electricpete
Electrical
You could capture the voltage accross the cap digitally and then calculate current as a function of time from that (itot = c * dv/dt where dv/dt computed by numerical differentiation). Then subtract out the average current to get the ripple current. Then compute whatever statistics you like such as rms of the ripple current and true peak of the ripple current.
There may be easier ways if you have advanced knowledge of the circuit/waveform, but this should work in general.
=====================================
Eng-tips forums: The best place on the web for engineering discussions.
There may be easier ways if you have advanced knowledge of the circuit/waveform, but this should work in general.
=====================================
Eng-tips forums: The best place on the web for engineering discussions.
electricpete
Electrical
I guess I should add there are obviously limits to the linearity of the capacitor which should be considered if using that approach in very high frequency ranges.
=====================================
Eng-tips forums: The best place on the web for engineering discussions.
=====================================
Eng-tips forums: The best place on the web for engineering discussions.
electricpete
Electrical
Yes, clamp-on probe. Something like a hall effect (?).
Makes a lot more sense than what I suggested.
=====================================
Eng-tips forums: The best place on the web for engineering discussions.
Makes a lot more sense than what I suggested.
=====================================
Eng-tips forums: The best place on the web for engineering discussions.
Skogsgurra
Electrical
Even less influence with a PEM rogowski coil. None, as a matter of fact.
Gunnar Englund
--------------------------------------
100 % recycled posting: Electrons, ideas, finger-tips have been used over and over again...
Gunnar Englund
--------------------------------------
100 % recycled posting: Electrons, ideas, finger-tips have been used over and over again...
Because it's a ripple current it might have some odd waveform which will effect a normal ammeter.
I think the best method would be to place a low value series resistor or ammeter shunt in series with the capacitor then measure the voltage with a scope. Of course your series resistor will have some small effect on the current but then so would almost any other form of ammeter.
BTW the poster is not very specific, does he want average, peak or something else?
Roy
I think the best method would be to place a low value series resistor or ammeter shunt in series with the capacitor then measure the voltage with a scope. Of course your series resistor will have some small effect on the current but then so would almost any other form of ammeter.
BTW the poster is not very specific, does he want average, peak or something else?
Roy
Roy,
That's why the Rogowski coil is such a useful device: no insertion loss to note, tremendous bandwidth, and immune to magnetic saturation. The negatives are that they can't detect a DC level, and they need signal conditioning which requires a power supply. The problem with adding a resistor is that a shunt of a size large enough to be useful is likely to be equal or larger than the ESR of the cap under test, and will shift the circuit operating point so far that the readings are potentially meaningless.
----------------------------------
If we learn from our mistakes I'm getting a great education!
That's why the Rogowski coil is such a useful device: no insertion loss to note, tremendous bandwidth, and immune to magnetic saturation. The negatives are that they can't detect a DC level, and they need signal conditioning which requires a power supply. The problem with adding a resistor is that a shunt of a size large enough to be useful is likely to be equal or larger than the ESR of the cap under test, and will shift the circuit operating point so far that the readings are potentially meaningless.
----------------------------------
If we learn from our mistakes I'm getting a great education!
Yeah, but ESR and ESL are both poorly defined parasitic effects, and you will end up trying to solve a series of equations with too many unknowns: unknown ESR, unknown ESL, unknown current, and a known voltage. Not the easiest piece of maths. ![[wink]](/data/assets/smilies/wink.gif "[wink] [wink]")
Draw out the equivalent circuit of the capacitor including the parasitics and insert it in your equation i = C dv/dt.
----------------------------------
If we learn from our mistakes I'm getting a great education!
Draw out the equivalent circuit of the capacitor including the parasitics and insert it in your equation i = C dv/dt.
----------------------------------
If we learn from our mistakes I'm getting a great education!
Scotty,
I didn't think a shunt resistor would add that much, you could probably get a decent signal with < 10mV and it meets KISS rules eh!
A clip-on ammeter ammeter would likely add as much impedance and have the effect of damping out any spikes due to the iron. Perhaps some sort of Hall Effect transducer (not ammeter) reading the waveform with a scope.
Is a Rojowski coil something readily available?
Cheers
Roy
I didn't think a shunt resistor would add that much, you could probably get a decent signal with < 10mV and it meets KISS rules eh!
A clip-on ammeter ammeter would likely add as much impedance and have the effect of damping out any spikes due to the iron. Perhaps some sort of Hall Effect transducer (not ammeter) reading the waveform with a scope.
Is a Rojowski coil something readily available?
Cheers
Roy
It doesn't have to add much, although if you can get away with a shunt of a few hundred [μ][Ω] then it 'might' be achieveable. That's a small signal if it's a noisy environment though! The following is the spec for a good quality electrolytic:
250V DC, 6800uF, 43.5A @ 40[°]C, ESR = 20m[Ω] @100Hz, ESR = 12m[Ω] @100kHz, ESL = 16nH
I suppose you might get away with using a part of the interconnecting wiring or PCB trace as a shunt. A PCB trace should be of a known thickness so it should be very easy to determine a convenient length to act as shunt.
Rogowski coils are readily available, but it's simple to wind your own too. Any bit of flexible tubing will do as a former providing it can be wound with fine magnet wire. The inner bore provides a return path so both ends of the coil are conveniently placed together for connection to the outside world. The governing equations are readily available - I ain't fighting TGML to post them here, ha-ha - but the trick (if there is one) is usually the integrator design, making sure that it is behaving correctly at the frequencies of interest. For precision work I'm told it is possible to add compensating networks to the coil to get a response accurate down into the handful of PPM level - the guy who owns Ro-Coil has done some pretty specialised stuff and is very knowledgable about the subject, and he's been kind enough to improve my understanding of these transducers.
----------------------------------
If we learn from our mistakes I'm getting a great education!
250V DC, 6800uF, 43.5A @ 40[°]C, ESR = 20m[Ω] @100Hz, ESR = 12m[Ω] @100kHz, ESL = 16nH
I suppose you might get away with using a part of the interconnecting wiring or PCB trace as a shunt. A PCB trace should be of a known thickness so it should be very easy to determine a convenient length to act as shunt.
Rogowski coils are readily available, but it's simple to wind your own too. Any bit of flexible tubing will do as a former providing it can be wound with fine magnet wire. The inner bore provides a return path so both ends of the coil are conveniently placed together for connection to the outside world. The governing equations are readily available - I ain't fighting TGML to post them here, ha-ha - but the trick (if there is one) is usually the integrator design, making sure that it is behaving correctly at the frequencies of interest. For precision work I'm told it is possible to add compensating networks to the coil to get a response accurate down into the handful of PPM level - the guy who owns Ro-Coil has done some pretty specialised stuff and is very knowledgable about the subject, and he's been kind enough to improve my understanding of these transducers.
----------------------------------
If we learn from our mistakes I'm getting a great education!
Roy it does add too much...
It all hearkens back to the capacitor not actually caring about waveform or magnitude. No particular ripple current will hurt the capacitor. It is the temperature that hurts the capacitor. Run your circuit and measure the cap temperature as that is the only thing you should ultimately care about. Why go to painstaking, expensive, effort if the result is a result that then gets inferred to yet something else, (temperature). Measure the actual end result directly.
Keith Cress
kcress -
It all hearkens back to the capacitor not actually caring about waveform or magnitude. No particular ripple current will hurt the capacitor. It is the temperature that hurts the capacitor. Run your circuit and measure the cap temperature as that is the only thing you should ultimately care about. Why go to painstaking, expensive, effort if the result is a result that then gets inferred to yet something else, (temperature). Measure the actual end result directly.
Keith Cress
kcress -
- Status
Similar threads
- Locked
- Question
