Choosing pulse capacitor for parasitic AC ripple suppresion

[artis]

So long story short I have a power supply (DIY) for some time and on the secondary side filter/reservoir capacitors there is a small ripple (about 1v PP) of 50khz (the switching frequency of the PSU).
Aside from others methods that I will implement to reduce this I am thinking of adding a shunt from the positive rail to negative by using a pulse capacitor as some years ago @Baluncore suggested.

While looking for such caps I realized that it's not only the ESR that I'm interested and the capacitor ripple amperage rating but also the reactance of the capacitor that i need to take into account right?

Because if I take my ripple at the frequency of 50Khz then a capacitor with 100nF/0.1uF rating will have a reactance of around 32 ohms , so very little current will be able to flow through it due to the high resistance correct?
So I think I need a capacitor of larger capacitance?

Is there much point in going for the lowest ESR or should I rather get a higher capacitance rating? Since higher capacitance decreases the AC resistance far more than getting a few miliohms lower ESR.

 

[berkeman]

there is a small ripple (about 1v PP)

 

[DaveE]

So long story short I have a power supply (DIY) for some time and on the secondary side filter/reservoir capacitors there is a small ripple (about 1v PP) of 50khz (the switching frequency of the PSU).
Aside from others methods that I will implement to reduce this I am thinking of adding a shunt from the positive rail to negative by using a pulse capacitor as some years ago @Baluncore suggested.

While looking for such caps I realized that it's not only the ESR that I'm interested and the capacitor ripple amperage rating but also the reactance of the capacitor that i need to take into account right?

Because if I take my ripple at the frequency of 50Khz then a capacitor with 100nF/0.1uF rating will have a reactance of around 32 ohms , so very little current will be able to flow through it due to the high resistance correct?
So I think I need a capacitor of larger capacitance?

Is there much point in going for the lowest ESR or should I rather get a higher capacitance rating? Since higher capacitance decreases the AC resistance far more than getting a few miliohms lower ESR.

With such a large amount of ripple, I think you're right about needing lots of capacitance. It's not just filtering, I think, it sounds like you need energy storage.

However, any advice you get isn't worth much since we don't know anything about this power supply. They come in many different flavors.

 

[DaveE]

Here's a very rough approximation. The capacitor follows this law $C⋅ \frac{dV}{dt} = I$ (where I is the load current). So:

$dV = 0.1V$ the desired amount of voltage drop during discharge.
$dt = \frac{1}{2} \frac{1}{50KHz} = 10 \mu sec$ the duration of cap discharge, half the switching period.
$I = 1A$ the load current that discharges the capacitor.

Then you want $C = I \frac{dt}{dV} = 1A⋅\frac{10 \mu sec}{0.1V} = 100 \mu F$

Of course you'll want to put in your own numbers. Also, it may be way off depending on the PS type. But it should give you a crude idea about the magnitude.

 

[artis]

Small is a relative term i guess...

Here's a very rough approximation. The capacitor follows this law $C⋅ \frac{dV}{dt} = I$ (where I is the load current). So:

$dV = 0.1V$ the desired amount of voltage drop during discharge.
$dt = \frac{1}{2} \frac{1}{50KHz} = 10 \mu sec$ the duration of cap discharge, half the switching period.
$I = 1A$ the load current that discharges the capacitor.

Then you want $C = I \frac{dt}{dV} = 1A⋅\frac{10 \mu sec}{0.1V} = 100 \mu F$

Of course you'll want to put in your own numbers. Also, it may be way off depending on the PS type. But it should give you a crude idea about the magnitude.

Well I have 3000uF per side (rail to ground), given it's a bipolar supply it would mean 1500uF across the +-.
I have a single transformer but the output is divided into two full bridges , each bridge has 1500uF.
Voltage is +-80 or 160v DC.Now years ago I made one mistake though, I attached my DC power wires on the PCB board between the filter caps and rectifier bridge.
Instead I will now have my rectifier bridge , then the thick traces passing along the filter caps and only then will I attach the output wires after the caps. This might reduce the ripple because the ripple on the traces before the caps might be higher than after.

Now I understand you assumed the 0.1v of drop as one you would aim for, but where did you got the 1A of load current? Is that also assumed?
As for my case the maximum load current would be around 10A.

But from what I understand the formula you gave deals with the total capacitance to support a sudden peak current without excessive voltage drop? In terms of voltage drops I already have done the tests, the storage capacitance is enough to support it, even using the formula.
The problem is that I get the parasitic ripple at no load, it just rides atop the DC, and I think that is because with just electrolytics is not enough as they have rather high ESR and impedance and cannot effectively "shunt" aka short circuit the ripple that gets through from the high speed diodes in the bridge rectifier.

This is why I am redoing the traces and connection points and considering shunting the main electrolytics with low ESR high current polypropylene pulse caps.
My main concern was how to choose them, I though about reactance and how that could be more important than hunting for the most expensive lowest ESR one. At 50Khz I reckon i would need somewhere between 20 to 30uF to get around 0.1 ohm of reactance.

 

[DaveE]

Sigh... Why I probably shouldn't respond to EE questions with schematics...

First, you need to understand that answers to questions without data are always vague, usually wrong. I was just trying to show you about estimating energy storage in capacitors. I made up the numbers, those were the only numbers I could get, they mean nothing.

So, I'm kind of busy today, you won't get a lot from me. But, it appears that you know a lot about this PS, please tell us some of that. Can we assume it's a buck topology operating in continuous conduction mode? In that case we need to know something about the inductor. Shunt capacitors work as filters in conjunction with a series impedance of some sort, as I'm sure you know.

1V ripple for a 160V output isn't too bad. I think we were all assuming you had something like a 5V, 1A PS. The difference being whether we are improving a pretty good output filter or a really crappy one. Now I know you are more in the pretty good, but not good enough, arena.

Better questions get better answers, especially when you are getting down to component value details. There are several others at PF that know about this stuff too. Maybe they can get back to you before I do.

PS: What sort of output ripple do you need?

 

[Tom.G]

The problem is that I get the parasitic ripple at no load, it just rides atop the DC

Please measure the ripple with the measurement probes right on the capacitor leads.

If there is ripple on the capacitors themselves with no load, then the capacitors have failed and must be replaced.

If no ripple there, move one probe at a time towards the rectifier. Where the ripple increases, you have passed a high-resistance connection. Fix it and repeat the tests.

If you have an oscilloscope, then repeat all of the above with maximum load.

If you still have 'too much' ripple, please post images of:

• the schematic
• the physical construction of the power supply
• the oscilloscope display,
  • each labelled with:
    • the point on the schematic of both Common and Probe placement
    • the point on the circuit construction of both Common and Probe placement
    • the voltage you read
    • the range setting and whether AC or DC coupling
    • the Time-per-Division or sweep rate

Without the above information, all we can do is guess. The above information is needed to make logical conclusions (or at least 'educated guesses' ).

Cheers,
Tom

 

[artis]

Thanks @Tom.G

Sigh... Why I probably shouldn't respond to EE questions with schematics...

No need to be upset, I appreciate your input, that formula you gave I had long forgotten about, will come in handy will write it down.

As @Tom.G said , well there is no need to guess because this thread's main question was how to better choose a pulse capacitor, but before I do so I will do the test Tom suggested as I actually had it in plan before

 

[alan123hk]

This link has another formula for calculating the filter capacitor to remove the ripple. But since I haven't used it in my actual design, I don't know if it's accurate.
https://www.homemade-circuits.com/calculating-filter-capacitor-for/