Electrolytic capacitors dont have enough ripple rating for load and power level

[treez]

Hello,
We are designing a 60W flyback, (90-265VAC in; 24Vout; 100W peak power).
Unfortunately there is no ambient temperature spec yet, and the SMPS will be mounted in a totally sealed container.

The load is a diaphragm motor (whale gulper 220 pump). The current draw of this is 2.5A average but every 400ms, the current peaks up to 5A for a some milliseconds.

Anyway, the problem is in finding a post diode bridge high voltage smoothing capacitor with enough ripple current rating. The ripple current in this capacitor is 1.7A.

We have identified the Epcos B43505C5107M000 electrolytic capacitor by Epcos….

B43505C5107M000 (£3 UK for 1000 pces)
https://media.digikey.com/pdf/Data Sheets/Epcos PDFs/B41505__B43505.pdf

….however, page 15 of the datasheet does not give information on ESR for the low 2.5Hz frequency of our load. To make matters worse, the ESR graph on page 15 shows the ESR increasing exponentially for ripple currents below 100Hz. We can only predict that ESR will be very significantly high for ripple currents in the 2.5Hz range. –However, we don’t know how high as the datasheet doesn’t say, -As such , it is very difficult for us to give a lifetime expectancy for this capacitor.

It is possible that we may be able to operate the pump on a 5 minutes ON , 5 minutes OFF basis, (or such like basis) but would really like to know the lifetime of this electrolytic when ON all the time continuously.

Cost:
Since there are so many unknowns here (not in the datasheet) and there is the possibility of being able to turn the pump ON/OFF/ON…, we believe that we could just use a much cheaper capacitor here and just do the ON/OFF/ON thing.

We could for example use the EKMQ451VSN101MQ25S capacitor at £1.13 for 100 pces, though of course, we would have to turn the pump ON/OFF/ON….at a suitable duty in order to allow this capacitor to have a decent service life. There s no data in the datasheet which could help us to find the optimum ON/OFF/ON.. regimen.

EKMQ451VSN101MQ25S (100uF, 450V)
**broken link removed**

Mains harmonics:
We find that there are capacitors with much larger capacitance values obviously having higher ripple current ratings, however, as you know, one cannot use too large a primary DC bus capacitance due to the problem of overly large mains input current harmonics (300Hz, 900Hz…etc)

Why so few?
It is rather strange that there is such a shortage of 450V rated , 100uF electrolytic capacitors with high ripple ratings, because you would have thought that it was common for people to get SMPS’s designed at power levels of 60-70W …because it is just underneath the 75W level above which a PFC stage is needed. Therefore, why are there so few high ripple current rated electrolytic capacitors available in the 100uF, 450V rating?

 

[Vbase]

If you parallel two 47uF you will have higher ripple current.
100uF at 1.7A may have too high ripple voltage.

 

[FvM]

I don't see how a SMPS input capacitor could be loaded with relevant ripple current components below 100 Hz. There will be a low frequent modulated 100 Hz (and harmonics) ripple current, not the same as low frequent ripple current.

As suggested in your previous thread https://www.edaboard.com/threads/332014/, it's probably better to determine the capacitor heating under actual load conditions in an empirical than guessing about obviously unspecified product parameters.

 

[treez]

There will be a low frequency modulated 100 Hz (and harmonics) ripple current, not the same as low frequent ripple current.

So you mean that the normal 100Hz ripple current associated with a post mains rectifier bus capacitor, will be in an "envelope" modulated by the 2.5Hz load ripple component........that's the same as saying that there is a 2.5Hz ripple current component..surely?

What I am saying is that the 2.5Hz nature of the load will indeed mean that this electrolytic dissipates more than if the smps were delivering 60W (the same average power), to a continuous, steady load...do you agree?

 

[D.A.(Tony)Stewart]

[hey treez I'm not shouting but See bold answers

QUOTE=treez;1435312]Hello,
We are designing a 60W flyback, (90-265VAC in; 24Vout; 100W peak power).
Unfortunately there is no ambient temperature spec yet, and the SMPS will be mounted in a totally sealed container.

Bad news, no convection creates slow high Tca thermal resistance to outside ambient.

The load is a diaphragm motor (whale gulper 220 pump). The current draw of this is 2.5A average but every 400ms, the current peaks up to 5A for a some milliseconds.

Anyway, the problem is in finding a post diode bridge high voltage smoothing capacitor with enough ripple current rating. The ripple current in this capacitor is 1.7A.

We have identified the Epcos B43505C5107M000 electrolytic capacitor by Epcos….

B43505C5107M000 (£3 UK for 1000 pces)
https://media.digikey.com/pdf/Data Sheets/Epcos PDFs/B41505__B43505.pdf

Vr = 450Vdc higher rating only degrades ESR


….however, page 15 of the datasheet does not give information on ESR for the low 2.5Hz frequency of our load. To make matters worse, the ESR graph on page 15 shows the ESR increasing exponentially for ripple currents below 100Hz. We can only predict that ESR will be very significantly high for ripple currents in the 2.5Hz range. –However, we don’t know how high as the datasheet doesn’t say, -As such , it is very difficult for us to give a lifetime expectancy for this capacitor.

Understand that there is ESR and ESL
If ESR is given at 100 Hz, it applies to 1 Hz
Understand it any cap gives ratings at 100Hz it is intended for rectifier designs and not SMPS i.e. not lowest ESR avail.

It is possible that we may be able to operate the pump on a 5 minutes ON , 5 minutes OFF basis, (or such like basis) but would really like to know the lifetime of this electrolytic when ON all the time continuously.

You must do the math of Thermal resistance and Heat loss, then verify and factor aging increase on ESR x5


Cost:
Since there are so many unknowns here (not in the datasheet) and there is the possibility of being able to turn the pump ON/OFF/ON…, we believe that we could just use a much cheaper capacitor here and just do the ON/OFF/ON thing.

Internal cap temp rise determines failure rate FIT , do the math. x2 FIT every 5 deg C and include aging. and internal ambient rise. then duty factor
FIT x2 or half life every 10 deg for semis


We could for example use the EKMQ451VSN101MQ25S capacitor at £1.13 for 100 pces, though of course, we would have to turn the pump ON/OFF/ON….at a suitable duty in order to allow this capacitor to have a decent service life. There s no data in the datasheet which could help us to find the optimum ON/OFF/ON.. regimen.

EKMQ451VSN101MQ25S (100uF, 450V)
**broken link removed**

0.64A ripple current RMS with f correction. How does that compare to your expectations?
Again if ESR is not spec'd low it is not best choice for high pulse surge applications.

Mains harmonics:
We find that there are capacitors with much larger capacitance values obviously having higher ripple current ratings, however, as you know, one cannot use too large a primary DC bus capacitance due to the problem of overly large mains input current harmonics (300Hz, 900Hz…etc)

This why soft start designs are used.


Why so few?
It is rather strange that there is such a shortage of 450V rated , 100uF electrolytic capacitors with high ripple ratings, because you would have thought that it was common for people to get SMPS’s designed at power levels of 60-70W …because it is just underneath the 75W level above which a PFC stage is needed. Therefore, why are there so few high ripple current rated electrolytic capacitors available in the 100uF, 450V rating?[/QUOTE]



You are using the wrong approach to this design.

High frequency pulses require Low ESR cap and High Voltage using Film X1 rated Caps with a much lower value when using PFC and/or SMPS

but as I stated in your other questions, this motor needs a 120W supply IF it is low rep rate cycles averaging 60W with motor surge currents rated for 2.5A fuse @24V or (28.4V) = 71W
Also I have indicated a 24V small battery will have >100 Farads of low ESR. then you only need a cheap trickle charger perhaps < 25W

 

[Vbase]

Calculating the cap for 2.5Hz is the wrong approach. You have to select a cap that under extreme condition will give you the DC you need. Your extreme conditions are: mains 90VAC and output power 100W. Adding the efficiency makes the input current 1.7A, you also need min 50VDC for your inverter to work properly. The input current and voltage ripple in these conditions require a cap of 470uF. Because of the high temperature you will need 1000uF so in a year time when it drops to about half capacitance it will still work.

If you don't have the right cap then when the input is 90V and the power to your load is 100W even for 10ms the 24v output will dip, I call it bad design.

The reason you find it hard to buy a cap of 100uF with ripple current of 1.7A is not because you aren't lucky, nor because cap manufacturers are detached from reality. Caps are made for 50Hz and if you choose the right value you get the right ripple current.

 

[D.A.(Tony)Stewart]

Calculating the cap for 2.5Hz is the wrong approach. You have to select a cap that under extreme condition will give you the DC you need. Your extreme conditions are: mains 90VAC and output power 100W. Adding the efficiency makes the input current 1.7A, you also need min 50VDC for your inverter to work properly. The input current and voltage ripple in these conditions require a cap of 470uF. Because of the high temperature you will need 1000uF so in a year time when it drops to about half capacitance it will still work.

If you don't have the right cap then when the input is 90V and the power to your load is 100W even for 10ms the 24v output will dip, I call it bad design.

The reason you find it hard to buy a cap of 100uF with ripple current of 1.7A is not because you aren't lucky, nor because cap manufacturers are detached from reality. Caps are made for 50Hz and if you choose the right value you get the right ripple current.

i agree, your peak ripple current with pfc on 60W @ 90V or 0.75A rms should be 1.4 x 0.75A, but load cycles thru planetary gear @ 1Hz so it demands almost 2x or pi/2

 

[treez]

The input current and voltage ripple in these conditions require a cap of 470uF. Because of the high temperature you will need 1000uF so in a year time when it drops to about half capacitance it will still work.

..Thanks but wow....you are saying that for a 60W average (100W peak) SMPS from 90-265VAC, we need 1mF of 400V electrolytic capacitance?......that would be extremely costly.

I tooK apart a Fender Mustang IV guitar amplifier once which is 150W at the load , and the input capacitance on that was 110uF. (admittedly that was the 220VAC version)...but its nowhere near 1mF. (by the way, it had no PFC, that 220uF was right after the mains rectifier bridge)

I don't see any manufacturers datasheet anywhere where they give capacitance value dropping to 50% over one year after operation in certain ambient....do you have a manufacturers reference for that?...the Epcos B435XXX range certainly doesn't offer any such info.

 

[Vbase]

..Thanks but wow....you are saying that for a 60W average (100W peak) SMPS from 90-265VAC, we need 1mF of 400V electrolytic capacitance?......that would be extremely costly.
I don't see any manufacturers datasheet anywhere where they give capacitance value dropping to 50% over one year after operation in certain ambient....do you have a manufacturers reference for that?...the Epcos B435XXX range certainly doesn't offer any such info.

To make sure that I don't sound detached from reality I said that 24V dip is a 'bad design' and not that it is bad for you. Of course in your case you can trade of cost for specs. If your motor doesn't stall at say 12V and only stops pumping then you can allow this in your design.
I also thought that the idea of Tony using battery with a charger might be OK, it requires very little design effort and you can get all the parts ready made. If you make large quantity then the design cost is less important.
I will look for data on caps drying out and post again.
If your guitar amp was made to work on 90VAC then the cap would have been 330uF to 470uF. It is also possible that the power supply has large caps on the output so when the voltage dip it will only reduce the power.

 

[D.A.(Tony)Stewart]

You guys are way off base, but that's ok it's a learning process.

THink PFC or SMPS with pulses 2A pk at 30 Khz or near SRF of X1 rated film cap.

how about 0.03 ~ 0.1 Ohm instead of >1 Ohm and <<$1 per cap. in volume?

How many do you need?

 

[treez]

If your guitar amp was made to work on 90VAC then the cap would have been 330uF to 470uF. It is also possible that the power supply has large caps on the output so when the voltage dip it will only reduce the power.

Thanks, my apologies about mistake above, the cap in the fender amp was two pieces of 200V, 220uF in series..presumably they used series because sometimes they connect as voltage doubler.

 

[Vbase]

Tony, don't start pulling your white hairs now, you managed to avoid it all your life so why start now?! This is a learning process for you too.

 

[D.A.(Tony)Stewart]

What me worry? Do you agree with my assessment? or not?

 

[Vbase]

Good Tony, still smiling.
Are you proposing to replace the 100uF with 1uF plastic cap or to add it in parallel? I agree that it can be OK on the 24VDC output.

 

[treez]

Sorry if this sounds a bit smarmy, we are getting to grips with this capacitor situation, the dreaded electrolytic situation....
The attached is an LTspice simulation showing how low the DC bus of an offline SMPS dips down to when supplying a continuous 118W load from 95VAC. (118W is the input power for a load supplied with 0.85 efficiency).
This is with a 100uF smoothing capacitor immediately after the mains rectifier bridge. (which is what we have in our design)

We have designed the SMPS to be able to supply 24V, 100W to the load with an input voltage down to 70V. However, as you can see, when supplying the 100W continuous load, the DC bus dips to 57V.
However, this is with a continuous load, and our load has only temporary intervals at 100W…..I agree that if these “Intervals” happen at the right time, then indeed the DC bus could dip down toward 57V.
However, the 100W interval is only some 5ms in “Interval” (I said 25ms earlier because maybe with a heavy load it could be that).

Even if the load is 100W continuous, the dip down below 70V at the input lasts only for 1ms….and to be honest, that isn’t going to drastically affect the flow rate of the pump.

So we believe that using huge input capacitances like 1mF, 400V is vastly over-expensive. Not to mention giving overly high mains input RMS current, causing high inrush current, and causing an unnecessary noise issue due to the high peak current ‘spikes’ at the mains input.

I am pretty certain that in the entire world market of offline , 90-265VAC SMPS's, there is no 60W average SMPS’s in the world with anywhere near 1mF of input smoothing capacitance.

I believe that what we need to do is get ambient temperature data and then do some suitable ON/OFF/ON’ing of the pump to extend the electrolytic capacitor lifetime.

It would be nice if we could use a PFC stage and thus avoid this post mains rectifier DC bus capacitor, however, the PFC stage would be more expense.

In many cases, the pump will not be working at 60W average and 100W peak, and in fact, its more like 30W average and 70W peak. In fact, when the pump has pumped all of its water out and it goes dry, it then draws very little current at all. (Diaphragm pumps can run dry no problem).

The unfortunate thing is that the lifetime of an electrolytic capacitor can only be determined by equations supplied by the manufacturer, there are no laws of physics to use. The only “general” rule is that every ten degree decrease in ambient temperature results in doubling of lifetime. Do you agree that our 100uF input capacitance is reasonable?

 

[D.A.(Tony)Stewart]

Good Tony, still smiling.
Are you proposing to replace the 100uF with 1uF plastic cap or to add it in parallel? I agree that it can be OK on the 24VDC output.

Replace it and drive with 30~50kHz PFC Buck on primary side X1 rated , fuse protected, adequate ripple current for power required.

100uF is totally wrong idea.

 

[treez]

PFC buck I like.

I also like the idea of a "dummy boost PFC" stage......this is like a PFC boost, except its much cheaper since you don't bother with a proper PFC chip, but just use a cheap constant off time current mode controller, and boost the incoming mains up to 200V (considering a 90-265vac mains).

When the mains peak goes above 200V, then the mains current just bypasses the boost inductor and goes through a diode parallel with it into the output, which obviously then goes above 200V......

So one ends up with not needing the post rectifier smoothing capacitor, and one has a nice high voltage DC bus to use for input to your flyback SMPS........so one much reduces the issue of ripple current rating of the input electrolytic capacitors.

- - - Updated - - -

100uF is totally wrong idea.

So you are saying we must use a PFC stage?, even though the power is <75W

I mean, this app note shows a 65W, 100-265vac input flyback with a 120uF input smoothing capacitor.....
https://www.st.com/web/en/resource/technical/document/application_note/CD00252755.pdf

..are they wrong to do this?

The rubycon KXW 120uF capacitor that they use doesn't have anywhere near the required ripple rating for 65W from 100VAC....
https://www.rubycon.co.jp/en/catalog/e_pdfs/aluminum/e_kxw.pdf
...even from 115VAC there KXW series 120uF capacitor suffers a ripple current if 1.6A ac RMS, whereas its only rated to 0.74A.
-so this kind of puts in perspective the request for a 1mF input electrolytic capacitor.

 

[D.A.(Tony)Stewart]

Just a simple PFC buck with 1-2 uF cap front end , $15 @ 60W in volume or make your own and get it certified.

 

[Vbase]

treez, you convinced me with your argument about the 100uF, you are likely to be OK with that.
Tony, I don't know enough about PFC to comment about it.

 

[D.A.(Tony)Stewart]

Vbase, sorry but you also dont seem to acknowledge that the Motor cycles to at least 50% more than the average rate. every 1/ cycle per second.
The existing concept is doomed to fail under worst case load.

- - - Updated - - -

Just a simple PFC buck with 1-2 uF cap front end , $15 @ 60W in volume or make your own and get it certified.

If I were to offer you a 24V 5A solution now CE rated for above price, would you buy? Not that I'm selling , just asking.