Stabilising an ON/OFF controlled SMPS that has low output capacitance.

[ocset]

Hello,
We designed and layed out a 10W, offline, isolated  Flyback converter  using the TNY287 controller as in the attached schematic.  We were not allowed to use any electrolytic capacitors. We only have room for 30uF  of ceramic capacitance on the output.  The  output has a 4V pkpk oscillation on it at some 4kHz…but this is   solved  when we use the RC network R4 and C3  as in the schematic.
TNY287 is an ON/OFF controller, and the manufacturers recommend high levels of electrolytic capacitance on the output…in our case, the  manufacturers  software program  (PI Expert) recommends 250uF  of electrolytic capacitance….but we are just not allowed to use that…and we  can only fit 30uF on the board.
Anyway, I was wondering if we  could alternatively avoid R4 and C3, and instead achieve “stability” by having a capacitor  connected from cathode to anode of the TL431  (U17)?
(I  use the term stability with some reserve, because I realise that true  “stability” isn’t   on the menu for an ON/OFF  converter like this.)


Tinyswitch 4 datasheet (TNY287)
https://ac-dc.power.com/sites/default/files/product-docs/tinyswitch-4_family_datasheet.pdf

[xavier60]

Have you tried removing all compensation?, R25 and C12.

[David Hess]

Adjusting the compensation networks may be enough to make it stable with 30 microfarads of ceramic output capacitance but if not, a low value of series resistance will need to be added to the ceramic output capacitance.  Regulators which use ceramic output capacitors take special steps to deal with their low ESR.

[ocset]

Thanks, i see, kind  of  bringing the ESR zero down in frequency by "adding ESR".
I think another way is to add in a pole to counteract the ESR zero.......i suppose that is what i have kind of done with the RC network across R28 (the upper divider resistor).
Though with this being an on/off  converter, i suspect i have trampled over some elements of feedback loop theory there.

However, all we want to  do is reduce the "motor boating" of the output (the output looks a bit like a motor boat riding through wave sea).

I think we can kind of immitate  an big electrolytic capacitor bank by having a capacitance across anode to cathod of the TL431.

But i will also try and remove the other RC as kindly described.

[mzzj]

Didn't you have friend living in Northern UK who is specialized in SMPS design and looking for challenges?
This and your other topics lately sounds like a perfect task for him! 

[David Hess]

Thanks, i see, kind  of  bringing the ESR zero down in frequency by "adding ESR".  I think another way is to add in a pole to counteract the ESR zero.......i suppose that is what i have kind of done with the RC network across R28 (the upper divider resistor).

In practice that is not actually how it is done.  With such a low ESR, compensation just becomes impractical; the output resistance and shunt capacitance already have 90 degrees of phase lag starting at effectively DC.

If you do not want to increase the output capacitor's ESR with respect to the load, then linear voltage regulators designed to use low capacitance "zero" ESR output capacitors use a trick which I think can be applied here.  The series resistance is added between the pass element output (switch in this case) and the output capacitor and AC coupled feedback is taken *before* the series resistance while DC coupled feedback is taken across the capacitor.

Now as far as the feedback network is concerned, the capacitor has a high controlled ESR making frequency compensation feasible while as far as the load is concerned, the capacitor has "zero" ESR.

I think we can kind of immitate  an big electrolytic capacitor bank by having a capacitance across anode to cathod of the TL431.

That is called dominant pole compensation and it will work but it severely limits line and load transient response.  The same result is produced if the output capacitance is increased to a very large value; AC gain falls below unity before the phase reaches unstable values but this also limits control loop bandwidth.

[ocset]

If you do not want to increase the output capacitor's ESR with respect to the load, then linear voltage regulators designed to use low capacitance "zero" ESR output capacitors use a trick which I think can be applied here.  The series resistance is added between the pass element output (switch in this case) and the output capacitor and AC coupled feedback is taken *before* the series resistance while DC coupled feedback is taken across the capacitor.

Thanks, yes i think the power.com recomended design circuit does that to an extent with inductor L2, as attached.

[David Hess]

Thanks, yes i think the power.com recomended design circuit does that to an extent with inductor L2, as attached.

If so, it looks like it was an accident due to limited power supply rejection of the TL431.  I am not sure why they would power the TL431 from the input side of the filter though unless that was the case.  I would not do it like that.

[ocset]

Thanks,
We can stabilise the TNY287 output with a 47uF capacitor....
Anyway, We need to implement the 47uF , 35V capacitor with a small-as-possible, cheap-as-possible capacitor which has preferably no lifetime issues, or has a long lifetime.
Film capacitors for 47uF , 35V are about 31mm by 18mm by 28mm high…too big and too pricey

So the only capacitor we can go for is as follows……the ZA series Aluminium polymer capacitor…..

ZA series polymer electrolytic capacitor
https://industrial.panasonic.com/cdbs/www-data/pdf/RDD0000/ABA0000C1221.pdf

…these are not “wet” electrolytic, so they will not have anything like the low lifetime issues of “wet” electrolytic capacitors, whose electrolyte just gradually vaporises away.
Are there any new arrivals to the capacitor scene that can do the job better?

[David Hess]

I would also consider polymer solid tantalum and hermetically sealed solid and wet tantalum capacitors.  Traditionally wet tantalum capacitors would be used in an application like this if the best reliability was required and high price was acceptable and they are still used in military, aerospace, and extreme environment applications.

[T3sl4co1l]

Polymer capacitors degrade just the same as electrolytics, due to moisture ingress rather than electrolyte loss.

Dry tant would be fine if you don't mind the conflict minerals, and manage the ignition risk.

Hermetic caps are fine but they aren't cheap.

You get what you pay for.

Tim

[ocset]

Thanks, all the Tantalum caps i saw on digikey have limited lifetimes of say 2000hrs at 125degC.
eg this one..
https://content.kemet.com/datasheets/KEM_T2009_T495.pdf

[David Hess]

That just means that after 2000 hours of operation, the capacitance will have changed by at most +/-10% and the leakage will have increased by up to 25%.  Unlike polymer electrolytic capacitors and aluminum electrolytic capacitors, there is no wearout mechanism other than failure.

[ocset]

Thanks @ David Hess, thats great info, i wish i had more on that issue of failure in tantalums, and when that failure may occur....we are not allowed to put components in to our lamp PCB that have "lifetime" issues...so Electrolytics are  disalloweed for us....if we can proove that the tantalums will not fail in 20 years say, ( (2500 hrs of night time useage per year) then thats great.

[David Hess]

Thanks @ David Hess, thats great info, i wish i had more on that issue of failure in tantalums, and when that failure may occur....we are not allowed to put components in to our lamp PCB that have "lifetime" issues...so Electrolytics are  disalloweed for us....if we can proove that the tantalums will not fail in 20 years say, ( (2500 hrs of night time useage per year) then thats great.

Tantalum capacitors are electrolytic capacitors also.  They just do not have the lifetime operating limits that aluminum electrolytic and polymer electrolytic capacitors do.

Tantalum capacitor reliability is a complex question.  Surge related failures *only* occur when charging so in a current limited application like yours with the capacitor on the output of a switching power supply, this is not an issue.  High voltage tantalum capacitors suffer from "field crystallization" around dielectric defects according to NASA but this is a long term problem and I think it only becomes an issue when combined with surge related failures.

Ripple current rating is likely to be the largest problem with a switching power supply especially if a low output capacitance is used so pay particular attention to this.  You may need to use a larger capacitor to get a high enough ripple current rating.

Other than that, voltage derate the tantalum capacitor by at least 33% for higher reliability.  50% would be even better.  The thicker dielectric layer make self healing more reliable.  Consider a premium high temperature and high reliability part.

One trick for high reliability is to burn in the solid tantalum capacitor at its full working voltage *after* soldering.  Temperature cycling during soldering causes stress fractures which "reset" the infant mortality part of the reliability curve.  Burning in the capacitor at its full working voltage forces self healing and weeds out defective parts.

[ocset]

Thanks David Hess that's great.
BTW, the fix i had in the top post... i re-wired it nice and small when i was putting the prototype together......but with the smaller , neater wiring, it no longer worked!...(didnt stop the ripple) ..so i just went for the 47uF capacitor solution.
I suppose we coudl do a frequency compensated feedback design, but it would be a bigger solution size....and this thing is only 3W continuous, and 10W in short  bursts.