What's the best way to measure large capacitors?

[Norm Dresner]

I'd like to check out some large capacitors (not supercaps), say, 20,000 -
100,000 uF. What's the best method? Timing the rise (or discharge)
between two voltage levels? Impedence at three frequencies (to try to
separate out ESR & inductance)?

TIA
Norm

 

[Ken Smith]

I'd like to check out some large capacitors (not supercaps), say, 20,000 -
100,000 uF. What's the best method? Timing the rise (or discharge)
between two voltage levels?

I'd say measure both the rise and discharge. Use largish resistor values.
Make the measurement after the charge and discharge has been happening for
several mS. This way you can pull out the leakage and series resistance
and are at low enough frequencies the ESL doesn't matter.

 

[mike]

I'd like to check out some large capacitors (not supercaps), say, 20,000 -
100,000 uF. What's the best method? Timing the rise (or discharge)
between two voltage levels? Impedence at three frequencies (to try to
separate out ESR & inductance)?

TIA
Norm

What do you want to know about them? That's what you measure.
Different applications emphasize different characteristics.
Measuring one parameter may not imply the others are "good".
mike

--
Return address is VALID.
Wanted, 12.1" LCD for Gateway Solo 5300. Samsung LT121SU-121
Bunch of stuff For Sale and Wanted at the link below.
Compaq Aero floppy,ram,battery.
FT-212RH 2-meter 45W transceiver. 2-meter linear
Toshiba & Compaq LiIon Batteries, Test Equipment
30pS pulser, Tektronix Concept Books, spot welding head...
http://www.geocities.com/SiliconValley/Monitor/4710/

 

[Norm Dresner]

What do you want to know about them? That's what you measure.
Different applications emphasize different characteristics.
Measuring one parameter may not imply the others are "good".
mike

--

Okay, since I'm going to use them as filter caps in power supplies, I guess
I need to know
1. Capacitance
2. ESR
3. Inductance
in that order of importance.

Norm

 

[Charles Schuler]

Okay, since I'm going to use them as filter caps in power supplies, I guess
I need to know
1. Capacitance
2. ESR
3. Inductance
in that order of importance.

Norm

Hi Norm: How old are they and were they made by a known and reliable
manufacturer? What happens to some of them is that they get a tad dry
inside and lose much of their capacity. Others, although old, can be
"reformed" by step charging them (start out at 1/4 of their voltage rating
for 6 hours and then 1/2 for 6 hours and so on). After reforming and
checking for leakage, a load test is in order. For example, how long can
they supply the expected current (T = RC and 5RC is close to total discharge
.... easy enough to measure). Last, but not least, can they deliver a peak
current that satisfies a low internal resistance? Some folks use the
screwdriver test for this but that can be daunting and even dangerous. ESR
meters are available. Hope this helps.

I ducked the inductance question since I don't think it is an issue in most
power supplies.

 

[terry]

Hi Norm: How old are they and were they made by a known and reliable
manufacturer? What happens to some of them is that they get a tad dry
inside and lose much of their capacity. Others, although old, can be
"reformed" by step charging them (start out at 1/4 of their voltage rating
for 6 hours and then 1/2 for 6 hours and so on). After reforming and
checking for leakage, a load test is in order. For example, how long can
they supply the expected current (T = RC and 5RC is close to total discharge
... easy enough to measure). Last, but not least, can they deliver a peak
current that satisfies a low internal resistance? Some folks use the
screwdriver test for this but that can be daunting and even dangerous. ESR
meters are available. Hope this helps.

I ducked the inductance question since I don't think it is an issue in most
power supplies.

A good way to measure ESR and ESL is to apply a voltage step with a waveform
generator, and look closely at the rising edge. Assuming that the dominant
time constant (Rpulse*C) is much larger than the timescale at which we
examine the capacitor voltage, we will see the sum of: A voltage step =
Vpulse*ESR/(ESR + Rpulse) and a voltage spike = ESL*dI/dt.

If you terminate a 50R pulse generator into 5R = 10R//10R you get Rpulse =
4.55R, Vpulse = Vunterminated/11 = 0.909V for my waveform generator. A cap
with 36mOhm ESR would have an amplitude step of 0.909V*0.036/(0.036+4.55) =
7.14mV.

The inductance can be calculated from the spike that sits atop the ESR step,
by measuring the area of the spike - but I lent someone my copy of "High
Speed Digital Design" which details this method, and I cant recall how the
trick works, but its pretty easy - you want a pulse generator with a pretty
fast risetime though - faster than Rpulse*ESL - which for 10nH is about
45ns. If you have a couple of smt inductors with a known low inductance
(10nH, 47nH, 100nH) you can manually "calibrate" your test fixture.

Cheers
Terry

 

[Norm Dresner]

A good way to measure ESR and ESL is to apply a voltage step with a waveform
generator, and look closely at the rising edge. Assuming that the dominant
time constant (Rpulse*C) is much larger than the timescale at which we
examine the capacitor voltage, we will see the sum of: A voltage step =
Vpulse*ESR/(ESR + Rpulse) and a voltage spike = ESL*dI/dt.

If you terminate a 50R pulse generator into 5R = 10R//10R you get Rpulse =
4.55R, Vpulse = Vunterminated/11 = 0.909V for my waveform generator. A cap
with 36mOhm ESR would have an amplitude step of 0.909V*0.036/(0.036+4.55) =
7.14mV.

The inductance can be calculated from the spike that sits atop the ESR step,
by measuring the area of the spike - but I lent someone my copy of "High
Speed Digital Design" which details this method, and I cant recall how the
trick works, but its pretty easy - you want a pulse generator with a pretty
fast risetime though - faster than Rpulse*ESL - which for 10nH is about
45ns. If you have a couple of smt inductors with a known low inductance
(10nH, 47nH, 100nH) you can manually "calibrate" your test fixture.

Cheers
Terry

Thanks for the info. I'll check our library tomorrow for a copy of that
book.

Norm

 

[Harry Dellamano]

A good way to measure ESR and ESL is to apply a voltage step with a waveform
generator, and look closely at the rising edge. Assuming that the dominant
time constant (Rpulse*C) is much larger than the timescale at which we
examine the capacitor voltage, we will see the sum of: A voltage step =
Vpulse*ESR/(ESR + Rpulse) and a voltage spike = ESL*dI/dt.

If you terminate a 50R pulse generator into 5R = 10R//10R you get Rpulse =
4.55R, Vpulse = Vunterminated/11 = 0.909V for my waveform generator. A cap
with 36mOhm ESR would have an amplitude step of 0.909V*0.036/(0.036+4.55) =
7.14mV.

The inductance can be calculated from the spike that sits atop the ESR step,
by measuring the area of the spike - but I lent someone my copy of "High
Speed Digital Design" which details this method, and I cant recall how the
trick works, but its pretty easy - you want a pulse generator with a pretty
fast risetime though - faster than Rpulse*ESL - which for 10nH is about
45ns. If you have a couple of smt inductors with a known low inductance
(10nH, 47nH, 100nH) you can manually "calibrate" your test fixture.

Cheers
Terry

Hey Terry,
Sounds like we need a simple Cap_Driver box. BNC input, driven by pulse
generator (5V, 1.0uS PW, 100uS RR), our favorite Zeltex PNP/NPN transistors,
10uF MLC, 9V battery and on/off switch. BNC output for scope and pads for
CUT (cap under test). Output R of driver trimmed to 0R50. Those transistors
should pump +/- 10Amps, <10nS rise and fall. We could also check cables with
this puppy. Call it a Z/100 box!
Cheers
Harry
..

 

[Winfield Hill]

Harry Dellamano wrote...

terry wrote ...


Hey Terry,
Sounds like we need a simple Cap_Driver box. BNC input, driven by pulse
generator (5V, 1.0uS PW, 100uS RR), our favorite Zeltex PNP/NPN transistors,
10uF MLC, 9V battery and on/off switch. BNC output for scope and pads for
CUT (cap under test). Output R of driver trimmed to 0R50. Those transistors
should pump +/- 10Amps, <10nS rise and fall. We could also check cables with
this puppy. Call it a Z/100 box!

A post I made seven years ago yesterday, reposting an earlier post...

From: Winfield Hill
Subject: Re: ESR of large caps?
Date: 1997/08/15
Message-ID: <[email protected]>#1/1
References: <[email protected]>
<[email protected]>
Organization: Rowland Institute for Science
Newsgroups: sci.electronics.design

Kendall Castor-Perry says...

Will Rose writes[snip]
input and across-cap signals on X and Y, scale them so they fit across
the screen properly, then do a Lissajous. Wind up the sensitivity to
'open out' the small error gap; for any particular scale you're using...

What's with all these 50 ohm or even 1k-ohm resistors, fellas? When
I want to measure 0.01 ohms, etc. I want AMPS! not mA or uA of test
current! For modestly interesting results I used a HIFI amplifier with
a 2.5-ohm resistor. Several amp drive is a piece of cake! But for
more useful capacitor data, I created the apparatus described below,
which can achieve 0.0001-ohm resolution:

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

Subject: Re: 0.01ohm impedance measurments ?
From: Winfield Hill
Date: 1997/06/22
Message-Id: <[email protected]>
Newsgroups: sci.electronics.equipment

Adam Craig Seychell said...

How can I measure impedances below 10 milli ohms at frequencies
of about 500kHz or less?

I measure impedances to 0.01 milliohm by using a fast high-current
pulse. Although it takes a little construction and setting up - as do
all high-current measurements - it's actually quite easy to do. Note,
this technique may damage devices which can't carry a high current for
a short time, but such a device likely won't have milliohm impedances
anyway!

The concept is simply to force a 10 to 100 amp current step through
the device and measure the voltage drop as a function of time. For a
10A pulse with a known dI/dt (say 10A/us) the voltage step (lasting
1us) is mostly due to the device inductance, L = v/(di/dt), then a
lower voltage plateau reveals the series resistance, r = v/i, and
finally an increasing voltage measures series capacitance,
C = i/(dV/dt).

The high current is supplied from a few large computer electrolytics
(e.g. I used two 33,000uF 15V units in series - note that the C * esr
figure-of-merit simply scales with physical size - select _large_
capacitors!). First, a modest power supply charges the capacitors,
then the cap's are used to charge an inductor with the help of a few
high-current MOSFETs, turned on with a single-shot pulse generator.
The pulse generator must supply a negative-going OFF-state discharge
current so as to set a sub 1us current rise time (adjusted with Roff)
in an avalanche diode, which steers the current to the device under
test (D.U.T.).

3 to 50 10 to 35V
--- 100 --+--- uH ----+---- avalanche ----+
+ | | diode |
computer D +------ scope probe
capacitor --G D.U.T.
| | S +------ gnd clip
- | | | |
----------+---------+-+-------------------+---- SCOPE chassis gnd
| |
200us etc -- Roff -- |
pulse gen -------------+

All the measurements are made with a scope probe with the groung clip,
acting as a 4-terminal sensor. For example, 0.1m-ohm resolution
implies a 10mV measurement capability (1/5 div with a 5mV/div range
and a 10:1 scope probe) and a 10A current.

This method easily measures the inductance and resistance of say,
shunt resistors and high-current inductors, as well as the ESR,
inductance and capacitance of big computer capacitors. Although these
are time-domain measurements, for most purposes, the results are
easily translated to the frequency domain.

For my purposes, larger 100 to 500A current pulses from the same setup
provided very useful data for avalanche breakdown investigations, but
that's another story.
---------------------------- end quote -----------------------------


inductance spike
_ /
| |
| |
flyback | |
voltage | | due to dV/dt = i/C
| | \ ______
| |___,,,....-----'''''
| \
_____________ / v = iR due to ESR

Here's the relevant waveform for the first 10us, before the current
in the coil has fallen too much.

 

[Winfield Hill]

Winfield Hill wrote...

inductance spike
_ _/
| |
| |
flyback | |
voltage | | due to dV/dt = i/C
| | \ ______
| |___,,,....-----'''''
| \
_____________ / _ _ \ v = iR due to ESR

Here's the relevant waveform for the first 10us, before the current
in the coil has fallen too much.

The height of the inductance spike about the esr voltage step is
V = L di/dt, which means to measure L you need to know the di/dt
of the high-current MOSFET switch. It's convenient to make the
FET switch fast (but not too fast), and to give it a well-defined
dI/dt by controlling the gate-voltage waveform. The di/di can be
measured with a current probe.