capacitive isolation

[Alex555]

I plan on making a power supply that has isolated outputs by charging capacitors in a sequence. The problem is, I want to have the lowest switching rate possible for 5-10% deviation from 15v. For a 2200uF capacitor, what frequency should I switch it at for a constant 5A draw? And how can I caclulate this>

[matseng]

A lot of theory on capacitor charge/discharge: http://www.allaboutcircuits.com/vol_1/chpt_16/4.html

A simple calculator: http://hyperphysics.phy-astr.gsu.edu/hb ... apdis.html

Plugging in your figures
R =3 Ω,
V0 = 15 V
C = 2200 uF
in the calculator gives the following:

After 1 millisecond the current is 4.3 amp and a voltage of 12.9 volts
After 2 milliseconds the current is down to 3.7 Amp and the voltage will have dropped to 11.1 volts.
After 5 milliseconds the current is 2.3 amp and a voltage of 7.0 volts

So in reality you could say that you need to swap in a newly charged capacitor at least 1000 time every second. This is with an ideal capacitor with 0 ESR, in the real world the figures will be a bit worse due to the ESR in the cap and the resistance in your switches.

[Bertho]

You are using ~75W of power? That would also create a large ripple current on the capacitor(s), which means they need to be of serious dimension to dissipate the powerloss. Your switches are under great stress too, in which you need both high- and low-side switches.

It may be a lot more easy and efficient to use a DC-DC converter with a transformer as isolation.

[Alex555]

I plan on using large, rather old capacitors. If I have 50% duty cycle, it would charge at 10 amps, right? Is this too much for a standard cap? They are rated at 40v and were pulled from very old switching power supplies.

[Bertho]

I plan on using large, rather old capacitors. If I have 50% duty cycle, it would charge at 10 amps, right? Is this too much for a standard cap? They are rated at 40v and were pulled from very old switching power supplies.

It depends on what specs the caps have and whether they can cope with the load.

If they are from a switching PSU, then they are probably rather low ESR, which will increase the surge currents dramatically. You actually may need to limit the current in that case with a series resistor or an added LC dampening circuit. At 0.25 Ohm you'll get a start-up surge current of 60A (if your ESL is not in the way) and an LC filter can limit that to better levels..

[Alex555]

I think I have found a way to calculate esr. I will use a 555 oscillator and a resistor that will make the time high of the oscillator equivilant to the capacitance. I will plug that into an online calculator to find the reactance of the cap at 60hz. i will put the cap in series with a .1ohm shunt to measure current. I will then measure the voltage across the cap, and the current going through it, and I can calculate additional resistance that way. About what resistance should I expect?

[Alex555]

i got .5 ohms when I should have seen a little over 1 ohm for the capacitor's resistance. That was a miserable failure due to not having any way to calibrate my 555 circuit, my meter not being accurate enough, and not having a proper ac supply/ resistor.

[Bertho]

One of the problems of ESR is that it is an impedance. That means frequency is a significant factor. The second problem is with (electrolitic) capacitors where you have the U*C=constant dependency; the capacitance is not the same at different voltages.

So, how you measure and where in the curve you measure is a significant factor. Also, some capacitors are specifically made for low frequency (50/60Hz) and others more for high frequency (for switching PSUs).

It would help to estimate the different aspects that pop up if you could attach a (coarse) schematic diagram of your design. It may help to point to the factors to take into account.

[Alex555]

Very coarse indeed. I plan on using an IRF3205 for the transistors. Two of these circuits will be attached to the same supply and the transistors will be switched alternately. The two capacitors will be connected in series for a positive and negative supply voltage.

[Bertho]

Ok, so you need a helper PSU (or trampoline) to switch the high-side switch. BJTs will do a very poor job indeed, so mosfets is the way to go. You will need a PWM or FM modulation scheme to stabilize the output to a fixed level. And, with switches at each side, you need to think about dead-times. Efficiency will be highly dependent on switch and capacitor losses. Output stability may not be satisfactory at high currents without an LC output filter.

However, you are not creating an isolation of any significant value. So why are you using this setup?

[Alex555]

I will have two of them switched alternately. The idea is to make two outputs isolated from each other, bot necessarily from the input.I am putting the two isolated outputs in series for a positive and negative supply. I can build an oscillator straight off of the input with enough delay to account for the turn off times being longer than the turn on times. What value inductor and capacitor should I use? And what do I need pwm for?

[Bertho]

I will have two of them switched alternately. The idea is to make two outputs isolated from each other, bot necessarily from the input.I am putting the two isolated outputs in series for a positive and negative supply. I can build an oscillator straight off of the input with enough delay to account for the turn off times being longer than the turn on times.

Ah, you are building a dual PSU with one inverting PSU. I still think you should consider using a "standard" DC-DC converter, which is much easier to control and stabilize. It will probably have a better efficiency too.

Switched capacitor setups are mostly used in low-power applications. The major drawback is that the output voltage is dependent on the load and is hard to keep at a constant level. Mostly you will find them where the load is known to be relatively constant.

What value inductor and capacitor should I use?

That depends entirely on the ripple-current and ripple-voltage you want to achieve. I would suggest you pick up (LT)Spice or equivalent simulation package and simulate your schematic to see what will happen (with or without inductor). There are too many unknowns and there are many differential equations to be solved to give you sufficient values. The inductor should be small though. If it is too large, you will get a very high voltage spike when the switch turns off.

And what do I need pwm for?

None at all... I was thinking switching PSU.

[Alex555]

I found a capacitor that would be better, 50v, 33,000uF, rated at 17 amps continuous current with 12.5 millohms. I am also not too concerned with ripple, I just care that the minimum voltage is kept to a maximum, if that makes any sense. I plan on feeding the output through an adjustable linear regulator.

[Bertho]

I found a capacitor that would be better, 50v, 33,000uF, rated at 17 amps continuous current with 12.5 millohms

Turning that beast on will give you 15V/12.5mOhm ~1200A surge current...

I do not think you want that.

[Alex555]

I would put a resistor in series with it for charging...