Transformless Transformers

Google on "flying capacitor". This idea has been around for ages. In integrated form, the ICL7660 (late 1980s?) uses it, but it's been used with mercury reed relays and such like since the dawn of electronics.

It's of limited use for power conversion, IMHO, partly because the isolation (using typical semiconductor switches) is not good enough (should be thousands of volts, not just hundreds) to meet safety standards for mains-connected supplies).

You could use an inductor as well, just make sure you short it during transfer. I'm sure this has been done as well.

More resistance and even less efficiency if there is power involved.

You can look at their "Claims" and see if you can avoid them, or demonstrate prior art.

Best regards, Spehro Pefhany

Do you mean that the breakdown voltage is too low? Here the sensing is what is the major saftey effect. It only allows the left or right side to be on at any one time by making sure current through but cannot occur.

e.g., suppose you have the live input side and the live output side connected. Because only one high side switch is on at any given time it will still be isolated(assuming the low side isn't connected).

But here, suppose one of the switches was shorted(failure). In this case current would always flow on that side which would prevent the other switch from being openned. In the case that no current was flowing(capacitor drained) then both switches would be momentarily opened but immediately closed.

So in some sense the new thing is to add the current sensing into the control. Maybe this is more trouble than it's worth though?

yes, It can be anything thing can act as a temporary power source. I was thinking one could "insert" a SMPS circuit in between the switches. Any configuration should work. After all, a SMPS can be disconnected from the input when it's switch is off. So just by adding a low side switch should fix this. This doesn't give isolation though unless one adds switches on the "secondary" side.

Yes but one can just oversize the mosfets.

It's a very basic idea and I'm curious how much I have to "change" in the circuit so that I wouldn't get sued. IF, say, I move the SMPS part of the circuit inside the switches then would it be ok? or if I used a different filter network?

The problem is that with most MOSFETs "off" is only off to a few hundred volts, then it turns on again. You need kV for safety. If you try to do that with semiconductors a transformer starts to look very, very attractive in almost every way. A transformer can get kV with a bit of tape and insulated wire. And it will pass safety agency mandated hipot tests with ease. I can't think of a single place I've seen a semiconductor allowed between hot and ground, but maybe someone else can.

Imagine the hot is connected to S1, and your toddler is sitting in a grounded metal bathtub chewing on S3/S4. Then the fat mechanical contactor in your beer fridge switches off (the microcomputer determines Tb e.g., suppose you have the live input side and the live output side

Transformers are usually cheaper and better. For MOSFETs, as voltage rating goes up, Rds(on) goes up, for a given die area. Cost goes up with die area.

You have to read and understand each of their twelve "claims" and evaluate the validity of each of them, and if any of them apply to your work-around. There's not really a shortcut.

It's been 14 years since that patent-- what commercial applications use it?

Best regards, Spehro Pefhany

I recall mentioning our "re-invention" of the flying-capacitor power isolator in a recent post here.

If you use IC-size mosfet-type SSRs as the switches, you can get good isolation and pretty low noise to about 400 volts isolation. Most of them are break-before-make, so just drive the banks from complementary-phase square waves, in the hundreds of Hz maybe.

A capacitor as "T" should be fine. Caps don't wear out from reasonable charge/discharge cycles.

There's no great advantage over using magnetics here, and transformers are safer as regards isolation. The flying-cap isolator is potentially quieter, with relatively soft switching at low frequency.

I suppose you could combine the isolation with voltage stepup/stepdown: say, charge some caps in parallel, reconnect them in series to the isolated load. Lots of switches.

John

yes, as I mentioned the circuit senses the current through each branch so that no current can be between both. As I mentioned in my original post, if S1 is shorted for some reason then current will always be flowing through that branch which will preclude S3 and S4 from being open.

So even if S1 and S3 were both shorted, as long as the "saftey circuit" was functioning then everything would be ok. If that circuit is well designed then it should allow the circuit to be completely isolated. i.e., assume the saftey circuit is ideal... then in no case will any significant amount of current flow.

Now I do not know if one can practically implement the saftey circuit to approximate the ideal case enough for your example but it is a fairely unreasonable one too. You could easily come up with off the wall cases for transformers too.

Of course what you are really asking is, if S1 or S3 fails then will it still be isolated.. This is much more reasonable because S1 and/or S3 will eventually fail hence the "saftey circuit".

e.g., the saftey circuit might control a circuit breaker before the isolation that is tripped when current flows through both branches which will shut off the circuit. One could add multiple switches in series to significantly decrease the chance of failure. (this doesn't avoid catastrophic failure though). e.g., if p is the probability a switch will fail then p*p is the probability of 2 switches failing. (of course p is really a distribution)

I do agree that transformers are not useless(I never said this would replace them for all applications) but transformers are not cheap and do not out perform this topology in many regards. One has to look at the complete picture. We already see non-transformer based power supplies all the time now. My "idea" only adds to that by adding more isolation.

Computer SMPS use transformers for isolation after the switching circuit because it can be made smaller. By adding my isolation method the isolation is increased and potentially the transformer can be removed. Of course this might not be a great idea but might be applicable in some cases.

On only needs to design the safety circuit to have a failure rate that of transformers and ability to sense accurately enough. This may or may not be impossible.

Well, for my specific applications which involve audio I'd probably what something a bit higher(smaller caps) or low(larger caps).

From what I read, some caps can loose around 3% of there capacitance on a complete charge/discharge cycle. That is pretty significant if true and makes them worthless for such an application if the switching frequency is high and they are not chosen large enough that such a condition might occur.

Do you mean SMPS with a transformer or transformer based supply? I am thinking of transformer only which requires large transformers and large caps.

Here one would just need to add the switches to the standard SMPS topology.

If you take one of the non-isolating converters shown and add 3 switches to it, 2 after the output and 1 on the lowside of the input(below Q1) then one would get isolation with it.

i.e., when Q1 is off then one doesn't need the circuit connected to ground... hence one can "isolate" the input side. One only needs to add the switches to the output side to add isolation to that side.

One then must add the safety circuitry so that both sides can never be connected simultaneously. Some smoothing caps afterwards should do the trick. This is very similar to the one in the patent except he puts the converter after the isolation and I'm putting it inbetween.

The main thing here is the safety circuitry that will prevent both sides from being connected at the same time and if current through both do occur(say one or both high side switches fail) then it would trip a circiut breaker before the isolation that would be equivialent to blowing a fuse.

Or it could check the voltage on the switches and make sure that when it's closed it should read ~0 and when it's open it should read >> 0. If not then it knows the switch is bad and would through a breaker and could report which switch failed.

One could have a simple differential current comparator on the two branches(S1 and S3) which would trigger a circuit breaker/relay. This would add an additional measure if the main control circuit failed.

Of course what really is important is the probability that the safety circuitry will fail and or fail catastrophically(not some nice failure state). If this can be made low and the proper components can be chosen for a wide variety of applications then it should be a decent design.

BTW, they do have some high voltage fets and bjt's that can handle a significant amount of power. Maybe sooner than later we might get some even larger ones.

I suggest that you take a look as transformer less version of the telephone subscriber line interface circuit (SLIC). Some of them are already over 20 years old now. It provides:

Battery to the line Overvoltage protection (for equipment, not so much people) Ringing voltage insertion / transfer Signaling Hybrid (2 wire to 4 wire conversion) and Test

(borsht)

The conversion efficiency of a switched capacitor circuit is about

RL

That patent falls into my "not worth the powder to blow it to hell" category... easily challenged.

Are you sure about that?

...Jim Thompson

That's only true if the two capacitors are equal, so the equilibrium voltage is half of the initial voltage.

In general, the efficiency is given by the ratio of the voltages; E=QV/2, so E1/E0 = (Q.V1/2)/(Q.V0/2) = V1/V0.

The total charge before equalisation is Q = C1.V0, at equilibrium Q = (C1+C2).V1, so the ratio of the voltages (and thus efficiency) is given by C1.V0 = (C1+C2).V1 => V1/V0 = C1/(C1+C2).

If C1 = C2, E1/E0 = C1/(2.C1) = 1/2.

IOW, the efficiency is exactly the same as if you just discard the excess energy with a series voltage drop, e.g. a resistor, diode, or linear regulator.

I've seen a couple other ideas used for very-high-voltage isolated power transfer. We're talking megavolts.

One is to have a motor at one end of a long insulated shaft and a generator at the other end. Works well for a few watts to a few hundred watts. Flexible shafts allow some but not perfect mechanical decoupling between platforms.

Another, that is replacing the motor-generator setup, uses high efficiency lamps at one end of a light guide and high efficiency solar cells at the other. This is true genius, because there's no mechanical linkage, it can be done with flexible fiber optics. While the applications I've seen for it are limited to a few watts, that is often enough for modern instrumentation.

Tim.

Potential energy that is not transferred does not enter the efficiency equation.

I like this 'eg a resistor' bit. The energy lost doesn't just go 'poof'and vanish, it heats up all non-ideal conductors in series with the current flow, including the materials in the capacitor construction and the switch.If you can confine most of the losses to a 'real' resistor, placed intentionally, you might get by with lower-powered parts operating at lower stress margins in the expensive positions.

Note that in the flying capacitor, the energy transfer has to occur twice.

RL

MOV ?

between hot and ground is not the issue.

the issue is between hot and isolated.

what makes you think that?

It will approach 100% efficiency if the input and output voltages are close. For a given load, you can do that by making the flying cap big enough and/or the frequency high enough, even if both caps are the same.

If both caps are relatively big, neither changes voltage much during operation, so Vout is close to Vin and the losses in the switch are small. It's just like carrying a battery back and forth.

That doesn't look right to me. With equal value caps and a very light load, Vout will approach Vin after a while.

The efficiency is given by the voltage ratio; you can get that from conservation of energy. As a black box, you indeed can't tell it from a linear regulator.

John

The voltage difference is created by power consumption in the load and variations in input/output voltage ratios - the voltages that need to be transformed in the first place.

It might be educational to consider capacitor volume requirements for an acceptable ripple component at suitable conversion frequencies. Voltage differences between input and output may not even be supportable in such an analysis.

Regardless of technology improvements in components, the issue is the same for switched capacitor conversion.

RL

While not actually true in every case, the basic prejudice is formed by basic switching conversion rules that warn against switching from one voltage sources to another - effectively shorting the two voltages through the switch. Capacitors are, effectively, voltage sources, at t=0, which can create practical problems.

If you do a few rough calculations of ideal charge transfer between capacitors, who's individual energy storage in stasis is C x V^2 / 2, you will find that at the end of charge transfer, the total system energy is less than before the charge was transfered, without the presence of any identifiable load.

Other posters are correct in pointing out that if no voltage difference is present on the capacitor(s) involved, then no losses will occur. Obtaining practical intermediate conditions that are acceptably low-loss for the components involved is the sticking point in using switched capacitors for actual energy conversion. It is also the reason why the topology is currently restricted in practise to low-power circuit applications that do not require a voltage or an impedance transformation.

RL

There are some switched-cap voltage converters that achieve non-unity voltage ratios at good efficiency. For example, you can charge two caps from the input source, disconnect them, and reconnect to the load in series, for a 2:1 step-up.

John