Quick ESR answer needed

No delayed trigger, and only as much memory as fills the screen. We have a bigger one, 500 MHz, forget the model number, that has more memory and delay.

John

Might I ask to which data sheet you refer? I have looked through Kemet's site, and haven't been able to find ESR figures for their SMT ceramic caps.

steve

The key it the "suggests" word in my posting above. It's from my reading of the graphs of |impedance| versus frequency; the minimum I assume to be where the capacitive and inductive reactances balance and you're left looking at the esr at that frequency. It's not guaranteed to be the same as a DC esr, of course. (Or maybe I should say is guaranteed to not be...) The particular sheet is a pdf that covers C0G, X5R, Y5V and one or two others.

Cheers, Tom

At what frequency? That makes a big difference.

"The Phantom" Jim Thompson

** Not to the minimum impedance the cap can exhibit - which is the series resistance that counts for supply bypass and filtering applications.

No one is too bothered about di-electric losses for that as they only help !

....... Phil

I would definitely say it's guaranteed not to be.

I measured a .27 uF SMT part (I don't know the dielectric). It has an ESR of 40 milliohms up around 1 MHz. The ESR at 100 Hz is 180 ohms, and at

20 Hz it's about 800 ohms. It varies with applied voltage, so it must not be COG; it's one of those dielectrics giving a capacitance that varies with applied voltage.

How would you define (and measure) DC ESR?

Data sheet found at:

$file/F3102_CerPerChar.pdf

Why on earth didn't they provide delayed trigger? It's so easy ....

In single channel the TDS2024 should yield about five times the screen. But that ain't enough for me which is one reason why I bought the Instek (has 25K memory). At least that gets me to 25usec single channel. Still drooling over the 1M or so of the Hameg 2008 but it's just too large in size.

--
Regards, Joerg

http://www.analogconsultants.com

That's the way I have looked at it too.

I don't see how such high values are possible - how are you measuring them? I can see how the ESR might vary with frequency, but the mechanisms I can think of would all produce an ESR that *increased* with frequency (like with inductor losses, e.g. skin effect).

Perhaps you could define it as the limit as the frequency tends to zero. It *ought* to be flat there anyway, so should be identical at

10kHz, 1kHz etc.

$file/F3102_CerPerChar.pdf

--

John Devereux

Ah. Gotcha. I bought a bunch of new, in-the-tape .22 uF Kemet X7R caps for what seemed to be a more-or-less ordinary price, figuring I'd use them for decoupling caps. I hadn't looked at the small print, they turned out to be 100V, 1825-size caps. Mouser has them at $1.44 in quantities of 1,000, I picked mine up for four cents each.

When I saw how large they were, I got to wondering about the ESR of an

1825-sized cap vs. 0805 or 1206. I figure between a greater capacitance than the "standard" 0.1uF, and a lower ESR (presumably), they should work very well. But, because of the size, I have yet to use any of them. Ah, well. :-)

steve

40 milliohms up around 1 MHz. TheESRat 100 Hz is 180 ohms, and at

measuring instrument. Apparently it cannot resolve a real part less than about 1/30th of the imaginary part of the impedance.

Jeroen Belleman/CERN

They may have a lower *ESR*, but for many applications (e.g. decoupling) what matters is the total impedance, which is often dominated by ESL (inductance). This is usually lower with *smaller* devices.

--

John Devereux

I'm measuring them with a Wayne-Kerr component analyzer. The high numbers are an artifact of using a series model rather than a parallel model, which would probably be more appropriate at low frequencies.

Consider the admittance of a .27 uF capacitor (at 20 Hz) in parallel with

1 megohm (due to dielectric loss).

The total admittance would be .000001 + j .0000339292

Take the reciprocal to convert admittance to impedance and get:

867.0 - j 29447.558

that is, 867 ohms in series with a ~ .27 uF capacitor.

Start out with 10 megohms in parallel: .0000001 + j .0000339292, and reciprocate:

86.9 - j 29472.9

that is, 86.9 ohms in series with a ~.27 uF capacitor, etc.

A more accurate model would assume a parallel resistance to represent the dielectric losses, and a small series resistance to represent the current in metallization and leads (if any) losses.

The higher the EPR (equivalent parallel resistance), the lower the ESR (equivalent series resistance) at a given frequency of measurement. The ESR is a representation of all the losses as a series resistance.

A constant 1 megohm resistance in parallel with the admittance of a .27 uF capacitor will convert to a higher and higher series resistance as the frequency goes lower and lower. The parallel loss resistance of a capacitor isn't truly constant with decreasing frequency, but you get the idea. This is why the ESR of capacitors increases as you go to low audio frequencies.

$file/F3102_CerPerChar.pdf

40 milliohms up around 1 MHz. TheESRat 100 Hz is 180 ohms, and at

Actually the instrument can easily resolve a real part less than 1/1000th of the imaginary part. See my response to John Devereux.

Le Wed, 15 Aug 2007 08:58:26 +0100, John Devereux a écrit:

Absolutely not. You totally miss the point about *equivalent* series resistance, which is the value needed for a not physically based equivalent circuit (but easy to understand) to behave like the real cap. This esr, as you know, models losses. Dielectric, conduction, and maybe, ahem, radiative losses too.

The minimum impedance at resonance is often dominated by conduction losses, which is the *true* serial resistance. (notice I didn't used the term 'equivalent'). For ex. a 100nF with a 1nH esl will resonate at 100Mrd/s (16MHz) giving Zc=0.1R With a bad 5% dielectric loss, that's only 5mR additionnal resistance. So a cap showing 20-30mR *esr* at its resonance frequency has this resistance to be dominated by conduction losses. Obviously, as you noted, these conduction losses will increase with frequency due to skin and proximity effects (don't we want electrodes to be close to each other in a cap?).

Now, for the *e*sr value at low frequency, you have the dissipation factor DF = Zc/esr = 1/(C.w.esr) or esr=1/(C.w.DF) DF being very roughly constant over frequency, you can see that the *e*sr has to rise at low frequency.

Again *e*sr isn't a true series resistance.

--
Thanks,
Fred.

Le Wed, 15 Aug 2007 12:17:45 +0000, Fred Bartoli a écrit:

This should obviously have been: Now, for the *e*sr value at low frequency, you have the dissipation factor DF = esr/Zc = C.w.esr or esr=DF/(C.w) DF being very roughly constant over frequency, you can see that the *e*sr has to rise at low frequency.

--
Thanks,
Fred.

[...]

Thanks to Fred and the phantom for the explanations - new information for me I have to admit!

--

John Devereux

If frequency makes a difference then isn't it "ESL" rather than ESR?

Application has capacitor charged to +1.8V, it is then connected (4ns full-on connect time) thru a 1.5 Ohm "strap" to -1.8V.

In other words, over 2A peak current.

...Jim Thompson

--
|  James E.Thompson, P.E.                           |    mens     |
|  Analog Innovations, Inc.                         |     et      |
|  Analog/Mixed-Signal ASIC\'s and Discrete Systems  |    manus    |
|  Phoenix, Arizona            Voice:(480)460-2350  |             |
|  E-mail Address at Website     Fax:(480)460-2142  |  Brass Rat  |
|       http://www.analog-innovations.com           |    1962     |
             
         America: Land of the Free, Because of the Brave

Flipping shunt losses mathematically to compute ESR is a very poor model. If one is doing pulse discharge, or hf bypassing, or stabilizing a regulator, such huge ESRs are total nonsense.

You can of course measure the losses of a ceramic chip cap at 20 Hz and decide the "equivalent" series resistance is 800 ohms, but that is a bit deceptive when the actual series resistance is several thousand times lower.

A better model is

---------+ | L1 | | R1 | +--------+----- etc, if you're | | compulsive C1 C2 | | | | | R2 | | | |

---------+--------+------

which has more information than just two numbers.

John

Why not?