Simple switching power supply filter cap value

Thread Starter

RichardO

Joined May 4, 2013
2,270
I found this schematic at
http://electronics.stackexchange.com/questions/106862/smps-with-two-outputs-12v-3amax-24v-2amax
It is similar to the wall wart I have.

What I don't understand is how the filter cap can be such a low value. It seems to me that the ripple would exceed the power line peak voltage at 120 volts in. Any thoughts would be appreciated.

Also, does anyone have an LTspice simulation for this or a similar circuit so I can see better what is happening?

Simple switching PS.png

edit: Oops. I meant this to be in the chat forum. Can a Mod move it?
 

tindel

Joined Sep 16, 2012
936
Rich - my back of the envelope calculations also have 10uF being too small. 100uF is a bit more realistic - but I'd say 470u is probably best.

I think of it two ways.
First think of the Power out - 60W. Say you get 70% efficiency (pretty standard for this type of flyback) - that means your input has to supply about 85W - or a resistor of 345 ohms hanging right off of the primary DC rail. The attached plot shows that the vin rail tanks with this large of a load (also note that the average power is nowhere near 85W - further indicating that the input voltage will tank).

As a double check - think of it in terms of energy -
with a 60W load, you expect 60J/s output energy - over about 8ms - that's about 0.5J per 120Hz cycle. Again assume 70% efficiency - or about 0.7J. A 10uF capacitor only stores 145mJ at 170Vdc (0.5*C*V^2) - that's not enough energy storage in the cap to supply the load for a full half cycle - again confirming that 10uF is probably not up to the task at hand. 100uF stores 1.445J at 170Vdc - so you'll expect 50% voltage ripple with a 100uF cap. A 470uF cap is probably the best option for reasonably low ripple... as it will store about 6.8J of energy and you'll see about 10% ripple.

Of course - this doesn't account for the energy that the line provides directly - which will further reduce the ripple, but this is a good 'first-order' calculation.
 

Attachments

tindel

Joined Sep 16, 2012
936
Another thing that makes the situation worse is that the supply input is negative impedance (i.e. input current actually goes UP as voltage goes down to maintain a constant output power).

Finally, (I hope, and you probably do too) - Flybacks aren't great at these kind of power levels. Somewhat because Capacitors become so large and expensive that they quickly become uneconomical. A quick digikey search had the cheapest, in-stock, 400V 470uF cap at $2.62 in 1000 pc quantities. Yikes!

http://www.digikey.com/product-detail/en/nichicon/LLG2G471MELB35/493-7237-ND/3768575
 
Last edited:

Thread Starter

RichardO

Joined May 4, 2013
2,270
Rich - my back of the envelope calculations also have 10uF being too small. 100uF is a bit more realistic - but I'd say 470u is probably best.
I think that the original circuit(s) are trying to minimize the capacitor size as a trade off with ripple voltage. Because of this, I suspect they would not even be happy with your 100uF cap. :eek:

Since I posted the question, I did a crude simulation.
AAC_Simple_switcher.png

Many of the values are guesses. Especially the transformer.

The circuit is reverse engineered from a 12 volt, 1 amp wall wart I bought at MicroCenter. I think they claim that it will operate anywhere from 100 volts to 240 volts at the input. It does not seem to work as spec'ed and I was curious why...
 

Attachments

tindel

Joined Sep 16, 2012
936
Rich - I made an egregious error last night when I posted this. I though deltaE=0.5*C*deltaV^2. I realized today that I was wrong, wrong, wrong! Sorry to steer you astray. DeltaE=0.5*C*(Vmax^2-Vmin^2). I'm going to blame it on being tired.

So, solving for Vmin - you get
Vmin=sqrt(Vmax^2-2*deltaE/C)

Note that the 2*deltaE/C factor ends up being larger than the Vmax^2 resulting in a negative sqrt (and a imaginary number) when the capacitances are very low - and not adequate for the tasks.

So Vmin and ripple the attached image. Note that this is not the whole truth - but is closer. The line actually provides up to 50% of the energy theoretically. The line adding energy adds a sine factor in the deltaE - and my algebra/trig is not good enough to solve for Vin for that equation.

Screenshot from 2016-10-13 17-24-56.png

I'd say a of 47uF is the absolute minimum that reasonably work. A 100uF cap should work fine - with less than 20% ripple. However, you can also run into problems with wide voltage changes due to changes in frequency or duty cycle (based on if it's constant frequency, or not).
 
Last edited:

tindel

Joined Sep 16, 2012
936
And another quick screenshot on the 12W converter that you simulated. 15uF is pretty low - but you show that it might work with as low as 4.7uF. I know I wouldn't put my name on it.

Screenshot from 2016-10-13 18-14-58.png
 

ronv

Joined Nov 12, 2008
3,770
I think that the original circuit(s) are trying to minimize the capacitor size as a trade off with ripple voltage. Because of this, I suspect they would not even be happy with your 100uF cap. :eek:

Since I posted the question, I did a crude simulation.
View attachment 113579

Many of the values are guesses. Especially the transformer.

The circuit is reverse engineered from a 12 volt, 1 amp wall wart I bought at MicroCenter. I think they claim that it will operate anywhere from 100 volts to 240 volts at the input. It does not seem to work as spec'ed and I was curious why...
If you just want to check the cap size you can model it as a plain buck like this. You should be able to get the same results from the flyback, but this is easier to see. This is just a kludge of another circuit so I didn't have to draw and design, so don't think the parts are correct.:rolleyes:
 

Attachments

Thread Starter

RichardO

Joined May 4, 2013
2,270
And another quick screenshot on the 12W converter that you simulated. 15uF is pretty low - but you show that it might work with as low as 4.7uF. I know I wouldn't put my name on it.
Interesting results. I would not put my name on it either.

I am trying to get the simulation more correct. When I put an actual model for the MJE13005 in the sim it does not oscillate. :(


I did a search for:
ieca-inc.com switching

I found this application note and other interesting papers:
www.ieca-inc.com/images/smps_design_guide.pdf


www.ieca-inc.com/images/Conducted_EMI_filter_design.pdf
www.ieca-inc.com/images/powersupplywhitepaper_10042010.pdf

Moderators note: Please do not use bold in links as they will not work.
 
Last edited by a moderator:

Thread Starter

RichardO

Joined May 4, 2013
2,270
http://www.mouser.com/pdfdocs/2-8.pdf

Here is a design guide that I found helpful. I think it helped me identify a problem that was causing me to not be able to use the entire input voltage range on my flyback that I'm designing right now.
Thanks. There appears to be a lot of down to earth information there.

On a related note. I finally got around to putting a "real" model for the ST13005 (MJE13005) into the LTspice simulation. The circuit would not oscillate.
There was a bogus value in the Central Semiconductor model!

I found a Motorola model but it does not work as expected in the circuit. It does not give enough peak current through the transformer primary. :(

I guess I will look for yet another model...
 

Thread Starter

RichardO

Joined May 4, 2013
2,270
Here is a design guide that I found helpful.
I found a (probably soon to be end-of-lifed) part that looks interesting. What really caught my eye was all of the links to ap-notes in the data sheet.

A lot of the ap-notes apply to the Fairchild chip but I suspect that they still contain good general information.

edit: I looked at the ap-notes and they seem quite good. Lots of detail on general flyback switcher design. I have attached the ap-notes since the links in the data sheet did not work.
 

Attachments

Last edited:
Top