Voltage Doubler Capacitor Values

Thread Starter

SamR

Joined Mar 19, 2019
3,490
Playing around with voltage multipliers and modeling a doubler and ran into something that is counterintuitive for me. I normally think of using large capacitors (470-1000uF) to smooth the ripple for converting AC to DC. Instead, modeling it I come up with 0.1uF as giving me a very low inrush current and millivolt output ripple! What am I missing here? Using 470uF caps I am getting ~10A of inrush. Will the low-value caps affect the output current in any way?
1617615961852.png
1617616000835.png
 

Thread Starter

SamR

Joined Mar 19, 2019
3,490
Yep, that was the answer I was looking for. Under load, it needs a lot of capacitance! Thx!

EDIT: Will have to design for the ~10A inrush for the first few cycles...
 
Last edited:

Yaakov

Joined Jan 27, 2019
3,492
Yep, that was the answer I was looking for. Under load, it needs a lot of capacitance! Thx!

EDIT: Will have to design for the ~10A inrush for the first few cycles...
Current is pesky stuff. If we only needed voltage things would be so much easier!
 

MrAl

Joined Jun 17, 2014
8,475
Playing around with voltage multipliers and modeling a doubler and ran into something that is counterintuitive for me. I normally think of using large capacitors (470-1000uF) to smooth the ripple for converting AC to DC. Instead, modeling it I come up with 0.1uF as giving me a very low inrush current and millivolt output ripple! What am I missing here? Using 470uF caps I am getting ~10A of inrush. Will the low-value caps affect the output current in any way?
View attachment 234540
View attachment 234541
Hi there,

Well what exactly did you expect the inrush current to be when you app;y an ideal voltage source across a capacitor? To make matters worse the AC line voltage does not always start out at zero volts when a switch is turned on such as when turning on a light bulb. The AC line can start out at any phase angle and as you know the peak occurs at 90 degrees and for a normal household line that would be 170 volts. Put that across a cap with just a diode in series and you have got a very low impedance to the cap which means very high surge current. That is why they started using surge current limiting thermistors.

It is very hard to calculate the actual surge too because the diode dynamic resistance is hard to predict in real life when the current goes very high for a short time. This even happens during normal operation of a full or half wave bridge rectifier circuit with capacitor filter.

In theory the surge should be the same for any cap value it is just the duration of the surge that will change. In a real life circuit, the cap ESR plays a large role in the surge too though.

As you know, a capacitor follows this:
i=C*dv/dt

where i is the current through the cap and dv/dt is the rate of change of voltage across the cap.

Now in sinusoidal steady state, the current at the peak of the sine wave is zero because the rate of change in voltage is zero. But during transient operation the voltage is not sinusoidal it jumps up from zero to possibly the maximum peak in near zero time, which means the dv/dt can be very very high. In theory, the dv/dt is infinite so the current is infinite. It is only the series resistance that limits the current at that point.

Because there is some series resistance in a real circuit, that becomes the significant feature. If the cap is considered a short for very short time periods, all we have left is that series resistance for that short time period. Thus the current could be as high as:
i=Vpeak/Rs
and because diodes have relatively lower resistance we dont get much help from that, so the cap ESR becomes a big factor in the surge current and also the efficiency of the conversion. That means the quality of the cap comes into question. High quality caps would have higher surge but higher efficiency as long as they dont move the significant part of the efficiency loss mostly to the diode which would then place the burden of the efficiency spec on the diode.

So there is a little bit more to it then we might think at first glance because we are dealing with things that are hard to predict in real life and component strains that we usually dont allow in circuits. It takes some serious study and testing time to be sure everything works well.

Very often in these situations a little math tells us more than a simulation because we can pinpoint what is signficant in the equations and that can be hard to notice in the simulations. I guess maybe the best bet would be to look for the places of power loss and go from there.
 

Thread Starter

SamR

Joined Mar 19, 2019
3,490
Continuing on a bit... As to output voltage ripple taming. I know this is a bit of an extreme example going from 120VACpp to ~120VDC for the model. I played a bit with cap values in a 2 capacitor model and came out with this. The C1 does a good job limiting the inrush and C2 limiting the ripple. Further increase in C2 value has little effect.
1617654796298.png
1617655062066.png
Which has ~0.55VDC ripple, which isn't bad @ 0.45%. If I increase the load from 2k to 200k it becomes ~0.1VDC. Other than increasing the load, are there any other tricks to reducing the output ripple. Further, will it always be dependant on the load or is there a way to avoid that?
 

MrAl

Joined Jun 17, 2014
8,475
Continuing on a bit... As to output voltage ripple taming. I know this is a bit of an extreme example going from 120VACpp to ~120VDC for the model. I played a bit with cap values in a 2 capacitor model and came out with this. The C1 does a good job limiting the inrush and C2 limiting the ripple. Further increase in C2 value has little effect.
View attachment 234621
View attachment 234622
Which has ~0.55VDC ripple, which isn't bad @ 0.45%. If I increase the load from 2k to 200k it becomes ~0.1VDC. Other than increasing the load, are there any other tricks to reducing the output ripple. Further, will it always be dependant on the load or is there a way to avoid that?
Hello again,

Try using a cosine wave. Look at the inrush close to t=0 seconds. That is probably worst case because the voltage rises very quickly and that is why either a low value resistor is used or a inrush surge thermistor is used in series, unless of course there is some series inductance.
Capacitors have a given impedance for steady state AC:
ZC=1/s/C which can be looked at as 1/(w*C) which can be significant, however for a step wave the "pseudo equivalent impedance" is:
PEZ=0
Yeah, that is zero, except you can add the ESR to that in real life:
PEZ=0+ESR
 

Thread Starter

SamR

Joined Mar 19, 2019
3,490
low value resistor is used or a inrush surge thermistor is used in series
I was thinking of modeling with a low value resistor to see what it would do. If the input is high enough to give a large input surge I would think that a low value resistor would not take much off of the "head" at the output. 10-12A surge for a few cycles is probably well within the capacity of the 1N4007 or one of their ilk. Need to check the PDFs. Thermistors are something I need to look into and find out more about. Kinda know what they are but no real study of them. Something else to learn about. Thx Al!
 

MrAl

Joined Jun 17, 2014
8,475
I was thinking of modeling with a low value resistor to see what it would do. If the input is high enough to give a large input surge I would think that a low value resistor would not take much off of the "head" at the output. 10-12A surge for a few cycles is probably well within the capacity of the 1N4007 or one of their ilk. Need to check the PDFs. Thermistors are something I need to look into and find out more about. Kinda know what they are but no real study of them. Something else to learn about. Thx Al!
Hi,

Yes the diode may take it and for low voltage old style unregulated wall warts they use bridge rectifiers off the output of a small power transformer and directly into somewhat large caps like 3300uf, but the transformer always has some leakage inductance.
For higher voltage circuits the diodes do blow however. They last for a while then all of a sudden one day one blows out in the bridge and the device stops working. I had this happen in an old CRT type TV a while back. All i had to do was replace one diode though i didnt bother to mod the circuit.

But yeah it depends on other things too like line impedance, and also how many surges appear on the line over time while the device is running. So it is a bit of a risk but sometimes it works.
For me, i prefer to know for sure something is going to work before i ship it. When i worked in the power converter industry we always used an inductor in series with the caps because that reduces stress on the diodes even while running steady state and also smooths the output a little better too.
With an inductor, the conduction phase is much wider than without so the diodes get much less of a peak current to worry about.
 

Thread Starter

SamR

Joined Mar 19, 2019
3,490
This was just an illustration in the book I'm working on of creating a DC source without a transformer and voltage multiplying. Something I would not normally want to do. I put a 1N4007 in becuase that is overkill for what I normally work on. It is way out of spec as it is limited to 1A and even steady state this model is running very close to that so I would shoot for at least 2A rated diodes. Also way out of spec V wise as it is a 1000V diode. When I looked in my online PDFs I found a 2A diode and in the specs it showed a series ~7Ω resistor to be used to tame the inrush. When I modeled with the series resistor, I lost a couple of volts on the output but it really tamed the inrush. Point being this is new territory for me. The book illustration has no values on any of the components so I put it into LTS and gave some to model with. Which in a way is better because I'm having to think about it instead of just doing the calculations for what has been given in the book. I also need to look at typical chokes. I know what they are but never really studied or used them. Anyway, this was an exercise about voltage multipliers using diodes, and once again it started me asking and looking at it in more depth. I get inquisitive like this...
 

MrAl

Joined Jun 17, 2014
8,475
This was just an illustration in the book I'm working on of creating a DC source without a transformer and voltage multiplying. Something I would not normally want to do. I put a 1N4007 in becuase that is overkill for what I normally work on. It is way out of spec as it is limited to 1A and even steady state this model is running very close to that so I would shoot for at least 2A rated diodes. Also way out of spec V wise as it is a 1000V diode. When I looked in my online PDFs I found a 2A diode and in the specs it showed a series ~7Ω resistor to be used to tame the inrush. When I modeled with the series resistor, I lost a couple of volts on the output but it really tamed the inrush. Point being this is new territory for me. The book illustration has no values on any of the components so I put it into LTS and gave some to model with. Which in a way is better because I'm having to think about it instead of just doing the calculations for what has been given in the book. I also need to look at typical chokes. I know what they are but never really studied or used them. Anyway, this was an exercise about voltage multipliers using diodes, and once again it started me asking and looking at it in more depth. I get inquisitive like this...
Hi,

Oh yes that's good the more curious you get the more you find out about all things.
Voltage doublers used to interest me a lot too long time ago and then i went to voltage doublers that are driven with square waves for times when there is no AC source present to work from. Very interesting the different ways we can do this.

Also of interest is the dual current path voltage doubler and the voltage multipliers.
The dual path type can get you voltage translation.
The voltage multipliers can take the voltage up several levels if the current output demand is not too great.
I think you would really like looking at these two which are really just modifications to the basic voltage doubler if you havent already.
 
Top