ESR Values for Electrolytic Caps

[ipman]

Resistors seems to be fine, IC is hard to troubleshoot without osciloscope.
The simptoms: normal startup, works for 2-3 seconds blinking some LED's (normal operation) then shutdown.
I think this is not about a resistor, at least from my experience.
If i swap the capacitors back to their original place, it will not go as far as blinking LED's for a short time.

[bitseeker]

There could be multiple issues with the blender's power supply. If the switching IC doesn't get stable-enough power, then it won't start. Moving the primary filter capacitors probably stabilized the power enough for the IC to attempt to start.

After a few seconds of running, if the switching IC doesn't receive feedback from the low-voltage side, then it'll shut down again. Check the components along the path from the output sense resistor back to the IC. The resistor might be damaged, an opto-coupler may be dead, the voltage reference for feedback comparison might be dead.

It's also possible that one or more diodes are shorted or the switching mosfet died. When the supply is on for 2-3 seconds, do you measure the correct voltage at the output of the SMPS? If you get little to nothing, it could be the mosfet.

Those are some of the usual suspects.

[ipman]

@bitseeker Thank you for all of that.
LNK304 is a low-component count chip for SMPS. It integrates the MOSFET and does not have feedback as i can see from reference design. There is no voltage reference also.
Resistors are very unlikely to be burnt in a 1W SMPS (12v, 120mA max), but you are right, i will chech them.
I will try to measure the voltage, but this is extremely ankward because of construction and the fact that my one year old son likes all my colorfull meter probes. It will take some time to do so.
But again, thank you all.

[bitseeker]

Yeah, not having seen the board you're working on, I can only give some general tips of what I've seen in SMPS. With the high level of integration, hopefully there's less to trace through to locate the problem.

[Karlo_Moharic]

.... but when diagnosing caps in a malfunctioning circuit with a lot of caps you will need a ballpark numbers for ESR because desoldering all those caps would be a problem....

You can't properly measure ESR , if your cap is still part of the circuit when you do it.

[nuno]

Resistors are very unlikely to be burnt in a 1W SMPS (12v, 120mA max), but you are right, i will chech them.

One of the latest surprises I had concerning component failure was a 1/4W resistor, clean and 100% immaculate looks from the outside, however, was open (was a 1 or 2.2 Ohm resistor, in series with the rail). Also in a low power (linear) PSU. Didn't take long to find it only because someone on the Internet had already done that job "for me" (many thanks!).

[6PTsocket]

I was considering buying an ESR meter for general trouble shooting but after reading this thread I wonder what "bad" is. Is an in- circuit device only sort of useful? Can it only be trusted to spot a bad cap when readings are very high? Is there some simple guide lines for using one, short of reasearching the factory specs on every cap tested?

Sent from my SM-G900V using Tapatalk

[bitseeker]

Simple guidelines:

1. You can test in-circuit as a quick, but not necessarily reliable, check if you know the circuit topology and how your measurements may be affected by other components on the board. Always follow up with a proper measurement out of circuit.

2. Instead of #1, at least desolder one lead prior to taking a measurement. It's half the work of removing the component, but much better than measuring completely in-circuit.

3. Compare your measurement with the datasheet for the component. If a matching datasheet isn't available, compare it to the datasheet for as similar a capacitor or capacitor series as possible. If the value is significantly different (e.g., very high ESR), then it's probably out of spec.

4. If you have no basis for comparison, replace the component and see if the original problem is corrected.

[ipman]

Some time later: found new caps. Jamicon brand, not LowESR.
They measure (stand-alone, not in circuit) 10ohms, compared to 35-37 originals. Also, they are larger than originals both in diameter and height.
Replaced them, but there is still something else.
Output varies slowly between 3 and 5v and settles near 3v afther a while.
Probably a dead rectifier diode, will continue when i have enough time to do so.

[NikkiC]

Sorry for the members of the holy ESR church but this it the way it should be done.  And yes, in cheap TV's and other consumer crap 99% of the times so much caps are bad that the chance of shunting a good one with a bad one is small. That lonely not completely shot cap that on its own tries co keep thing running will not influence much.

You wrote this a year ago but I laughed hard reading  your 'holy ESR church' comment.   So true!  I have and still use electrolytic caps that are 40+ years old.   No leakage issues, no high ESR, they are rock solid.  I have concluded that many times they fail due to just being placed into a circuit in a willy-nilly manner, not taking into account heat, ripple current, etc.  I've a lot of old vintage computer power supplies from the 60's 70's and 80's.   Stuff like IBM, Digital, DEC.  I just don't run into very many computer grade electrolytic caps that need replacement. When I do, I replace them with the same exact capacitor that is also 40+ years old. I believe these old caps will likely work flawlessly another 40 years.  To me, blindly replacing them  just because of their age is irresponsible. I have even seen old caps that have been greatly abused, baking in the sun and exposed to harsh weather extremes for years, test good and operate reliably. 

In my opinion, the consumer junk is just copy/paste of previous designs, flaws and all.   It is known, expected and intended for caps to fail in modern consumer electronics.  Since you already know most are going to be bad or marginal, replacing them all makes sense.  I get it.  But this just doesn't  hold true across the whole spectrum of electronic designs.   Not blindly replacing all electrolytic caps due to age is apparently an unforgivable sin resulting in false accusations of being a sorcerer or witch, excommunication and eventual condendamation to  be burnt at the stake and then sent hell for ever and ever amen.  Seriously.   To read the admonishments of  holy ESR church members  to seasoned veterans who know better  and refuse to believe is astonishing. Thanks for the laugh and for sharing your knowledge.

[mvs]

I've a lot of old vintage computer power supplies from the 60's 70's and 80's.   Stuff like IBM, Digital, DEC.  I just don't run into very many computer grade electrolytic caps that need replacement. When I do, I replace them with the same exact capacitor that is also 40+ years old. I believe these old caps will likely work flawlessly another 40 years.  To me, blindly replacing them  just because of their age is irresponsible.

Electrolytic capacitors do have a shelf life. If a piece of vintage equipment was unpowered for decades, it may need a reforming of some caps before it can be used again. Blind replacement of caps is quick solution for this problem, if maintaing of original state is not required.

[wraper]

I've a lot of old vintage computer power supplies from the 60's 70's and 80's.   Stuff like IBM, Digital, DEC.  I just don't run into very many computer grade electrolytic caps that need replacement. When I do, I replace them with the same exact capacitor that is also 40+ years old. I believe these old caps will likely work flawlessly another 40 years.  To me, blindly replacing them  just because of their age is irresponsible.

Electrolytic capacitors do have a shelf life. If a piece of vintage equipment was unpowered for decades, it may need a reforming of some caps before it can be used again. Blind replacement of caps is quick solution for this problem, if maintaing of original state is not required.

Not only that. You don't really know how much life is left in them even if they measure OK at given moment. It might be several decades or just a few months. Replacing all of them is a way to ensure reliability, usually not expensive too.

[WindWalker]

Hi,

I've just received some Rubycon CFX capacitors (3.3uF 400V 10x16mm). I've tested them for ESR using the "method" shown in this video (I don't have an ESR/LCR meter).

With a 2V 100 kHz square wave (zero offset), I've measured a typical 560-580mV peak to peak voltage across the capacitor (1.04V in two of such capacitors). I get a waveform similar to the one at 10:25 in the video. According to the video (12:49), ESR is R=50*.56/(2-.56)=19.4 Ω. According to the datasheet, maximum ESR is 0.20/(2*3.14*2*50*3.3)*1e6=96.5 Ω (mains frequency is 50 Hz here).

Questions:
1) Is any of these calculations wrong?
2) If not, is the ESR supposed to go down in the first hours of use or are these just bad/dry capacitors? The capacitors are supposed to be new (maybe they are from very old stock?)


Thanks

[Jay_Diddy_B]

Hi,

First you have to convert from Tan d (Tan delta), the tangent of the loss angle, to ESR

The datasheet gives tan d = 0.20 at 120Hz

The capacitor value is 3.3E-6

ESR = 0.02 / (2 x Pi x 120 x 3.3E-6) = 8 Ohms

So if we put this in a test circuit:




The result shows



So if you are measuring 550 - 580 mV p-p

The capacitors are good.

Regards,
Jay_Diddy_B

[WindWalker]

The datasheet gives tan d = 0.20 at 120Hz

The capacitor value is 3.3E-6

ESR = 0.02 / (2 x Pi x 120 x 3.3E-6) = 8 Ohms

Thanks a lot for the images/simulation, but I couldn't figure out why you used 0.02 in the ESR calculation while the datasheet says 0.20. Isn't it 'ESR = tan d/Xc'?

Also, could you tell me what software you used for the simulation?

[Jay_Diddy_B]

Hi,
Sorry I made a typo

The ESR should be

ESR = 0.20 /(2 x Pi x 120 x 3.3E6) = 80.3

This is the maximum value allowed by the datasheet.

I used LTspice for the simulation.

I would expect new capacitors to be better than this.

Regards,

Jay_Diddy_B

[Shock]

To answer the other question yes leakage and esr is less when capacitors have been recently used or are warm. Likely to bounce back somewhat to their original state when they have been sitting unused for while. Just ensure you discharge them before testing.

[The Electrician]

Hi,
Sorry I made a typo

The ESR should be

ESR = 0.20 /(2 x Pi x 120 x 3.3E6) = 80.3

This is the maximum value allowed by the datasheet.

I used LTspice for the simulation.

I would expect new capacitors to be better than this.

Regards,

Jay_Diddy_B

Since the max DF is specified at 120 Hz, 80.3 ohms is the max ESR also at 120 Hz.  At the lower left corner of the first page of the data sheet is a chart labeled "Multiplier for ripple current".  The different allowable ripple currents at different frequencies tells us that ESR is varying with frequency.   The allowable ripple current at 100 kHz is 5 times what it is at 120 Hz.  Since the dissipation in the ESR is proportional to the square of the ripple current, the max ESR at 100 kHz is apparently 1/25 of the value at 120 Hz, namely 3.2 ohms.

WindWalker made his measurement at 100 kHz, so the max ESR at 100 kHz of 3.2 ohms is what he should be comparing his measurement to.

[WindWalker]

Since the max DF is specified at 120 Hz, 80.3 ohms is the max ESR also at 120 Hz.  At the lower left corner of the first page of the data sheet is a chart labeled "Multiplier for ripple current".  The different allowable ripple currents at different frequencies tells us that ESR is varying with frequency.   The allowable ripple current at 100 kHz is 5 times what it is at 120 Hz.  Since the dissipation in the ESR is proportional to the square of the ripple current, the max ESR at 100 kHz is apparently 1/25 of the value at 120 Hz, namely 3.2 ohms.

WindWalker made his measurement at 100 kHz, so the max ESR at 100 kHz of 3.2 ohms is what he should be comparing his measurement to.

Thanks a lot for this valuable information. I didn't know what these multipliers meant. So, I suppose these multipliers relate to the rated rippled current (e.g., if the multiplier is 1.0 and rate ripple current 180mA @ 100kHz, then if the multiplier is 0.2 for 120Hz ripple current should be derrated by a factor of 1/0.2=5, so 180*0.2=180/5=36mA at 120Hz).

I did some testing today. I put the capacitors as close as possible to the function generator output terminal (see attachment). Then I applied a 1V 100kHz square wave. Then I proceeded to heat the capacitor with a heat gun (400ºC at a distance of 20-30cm). I have a video here. As you can see, the peak-to-peak voltage at the capacitor dropped from ~120mV to ~40mV. This means that ESR went down from 50*.12/(1-.12)=6.8 to 50*.04/(1-.04)=2.1 Ω.

I was able to achieve the same decrease for all other capacitors (~40mV). So, I guess ESR will most likely decrease to a value within spec during use

[The Electrician]

Note that the "Dissipation Factor (MAX)" chart on the first page of the datasheet specifies that the measurement is to be made at 20C.  Even though the allowable ripple current is specified at 105C, the ESR at 100 kHz we calculated would be the ESR measured at 20C, so this is the max allowable ESR at 100 kHz when measured at 20C.

[WindWalker]

I see... then they should all be out of spec, as after a few minutes resting the voltage went back to ~120mV/6.8Ω, two of them even to 200 and 300mV (12.5 and 21.4Ω). So my guess is that they are old stock or even rejected stock (re)sold cheaper.

[The Electrician]

Try reforming them.  Apply a variable voltage (through a protective resistor, maybe 100k) increasing gradually to rated 400 VDC.  Leave the cap on the voltage supply for an hour or so.  Then discharge the cap with a 10k resistor connected to the cap for several minutes.  Disconnect the cap from the 10k resistor, wait several minutes and measure the cap voltage, which will recover somewhat.  Make sure the recovered voltage won't hurt your ESR meter before you measure it.  If the recovered voltage is more than a few tenths of a volt, reconnect the 10k bleeder resistor and wait several more minutes, and try again.

If you can't apply 400 VDC, use the highest DC voltage you can come up with.

I've occasionally had NOS capacitors end up with much inproved ESR after this treatment.

[WindWalker]

Try reforming them.  Apply a variable voltage (through a protective resistor, maybe 100k) increasing gradually to rated 400 VDC.  Leave the cap on the voltage supply for an hour or so.  Then discharge the cap with a 10k resistor connected to the cap for several minutes.  Disconnect the cap from the 10k resistor, wait several minutes and measure the cap voltage, which will recover somewhat.  Make sure the recovered voltage won't hurt your ESR meter before you measure it.  If the recovered voltage is more than a few tenths of a volt, reconnect the 10k bleeder resistor and wait several more minutes, and try again.

If you can't apply 400 VDC, use the highest DC voltage you can come up with.

I've occasionally had NOS capacitors end up with much inproved ESR after this treatment.

Thanks again for your input. I'm assuming that by protective resistor you meant series resistor (to limit the charging current on the capacitor) and not parallel resistor (to induce some ripple current).

I put 5 of these capacitors in parallel () and charged them through a ~500kΩ resistor up to 300-320V (output of a bridge rectifier, mains voltage here is 230V single-phase so 230*sqrt(2)~320V). They stayed like this for almost one hour and a half.  After discharging (the simple voltage measurement with a multimeter was enough to almost fully discharge them) I did some 2-3 "bleedings" with a 10kΩ resistor like you suggested. Voltage before bleeding went up to ~1,1V and ~0,5V.

I don't have an ESR meter, and don't want to buy one to have it sitting unused for months after these experiments, that's why I use the oscilloscope plus function generator method (at a laboratory). It may take some time to have access to them again because of quarantine issues

[WindWalker]

Try reforming them.

My experiments with this reforming have been very interesting.

One the one hand, these 3.3uF 400V Rubycon capacitors still give me 120mV after "reforming" (they were charged at ~320V DC from several hours to about a full day), so clearly out of spec (6.8 vs 3.2Ω at 100kHz).

On the other hand, I've also received some nichicon capacitors (4.7uF 400V LD series and 6.8uF 400V CS series). Interestingly enough, some of these capacitors (not sure if the 4.7uF or the 6.8uF ones) measured about 11uF (yes!). So I put some them in parallel to do "reforming"/burn-in (with the 500kΩ resistor), and it took over 24 hours to bring the voltage up from 60V to the full 320V (across several hours, I measured 60,160,200,250,290 and 300V). After this, I got 70-80mV on them (3.8-4.3Ω), and if I did the math correctly I should expect (the squared multiplier ratios come from the datasheets):

1) 4.7uF LD: 0.24/(2*pi*120*4.7*1e-6) * (1.00/2.00)^2 = 16.9Ω
2) 6.8uF CS: 0.24/(2*pi*120*6.8*1e-6) * (0.50/1.00)^2 = 11.7Ω

Considering these reference values, my guess is that modern 400V capacitors will actually have an ESR in the range of 1-10Ω. It is a bit strange, since usually I read that proper ESR values should be less than 1Ω, eventually down to 0.1Ω, so I guess this applies only to lower rated voltage capacitors.

BTW, the protective resistor is a very good idea to 'burn-in' capacitors. It is able to limit the current to less than 1mA, allowing for capacitor self-healing without "big" currents.

[wraper]

I read that proper ESR values should be less than 1Ω, eventually down to 0.1Ω, so I guess this applies only to lower rated voltage capacitors.

This is nonsense when you apply it to capacitors in general. ESR heavily depends on capacitance and to lower extent on rated voltage. Not to say some ultra small capacitors may have significantly higher ESR than capacitors with same ratings in general. For many capacitors 0.1 Ohm is completely unacceptable, for others 10 Ohm is completely normal.