Variac on SMPSU

[PETERg0rsq]

From another thread

Don't use a variac on switch mode power supplies.

Why is this?

I was planning on using my variac to test a variable 20A smpsu that I couldnt get to start!

The last SMPSU i had, I ended up fitting a small linear psu to replace the starting circuit (it contained two smpsu's, one small constant voltage to power the chopper in the main SMPSU for the 5-20V output) as every time I plugged it into the mains (even after changing every component in the starting circuit) it blew the main chopper (£5 ) instantly as it hadnt started running! It was impossible to work out why?

Since then I have avoided smpsu's like the plaugue, but thought if I came across another that needed repairing I could use a variac to limit the potential for destruction. (I would also use an isolation transformer).

Peter

[kalee20]

Variacs are an essential item if you are developing an SMPS, because you can wind up the input slowly. But you need all the 'housekeeping' circuitry active, usually by external DC supplies, to apply proper drive to the power devices.

If you apply reduced mains voltage and let normal start-up circuits get you going, then at the reduced mains you may get insufficient drive to the power switches (FET or bipolar transistor), so they overheat quickly; with reduced mains the duty cyle will ramp up to maximum; currents will be at their highest; stresses will thus also be highest.

A good SMPS will cope with all this. Unfortunately not all SMPS's 'out there' are good!

[Skywave]

Originally Posted by kalee20

If you apply reduced mains voltage and let normal start-up circuits get you going, then at the reduced mains you may get insufficient drive to the power switches (FET or bipolar transistor), so they overheat quickly.

With reduced drive to the power switches, I would have thought that they would then draw less current and would thus not overheat. So, what have I failed to understand, please?

Al.

I would have thought that they would then draw less current and would thus not overheat

Not enough gate (grid) drive so they have a lot more resistance.

[Skywave]

Originally Posted by merlinmaxwell

Not enough gate (grid) drive so they have a lot more resistance.

If 'they' - power MOSFETs - have a lot more resistance (presumably that is resistance drain to source), then surely they will then draw less current?
Now I'm even more confused!

Al.

When MOSFETs are fully on there is very little power dissipation in them because they have a very low resistance, in the order of milliohms, (Power = Current squared times Resitance) up that resistance a small amount (to an ohm or so) and then you have lots of power dissipated in the MOSFET. Less current will be drawn but most of the voltage will be across said MOSFET.

[kalee20]

SMPS's are generally regulated, ie they give a constant output voltage with changes of load and input voltage.

To give the same output at a lower input voltage, the input power must be the same, so the current drawn from the mains must be higher. That current flows though the switching devices, thus at lower mains voltage they pass a higher average current.

Now consider the effects of reduced drive. Sure, the resistance of the switching devices increases. Unfortunately, the current through them is virtually 100% defined by their transformer load. The increase in resistance makes hardly any difference to the overall current. But the voltage drop across them can increase from virtually zero to several volts. It's a bit like trying to reduce the current flow to your electric fire by changing the 13A plug to one with really dirty pins - the current flow might drop infinitessimally, but the plug now catches fire.

[maninashed]

This is partly the reason why many tv sets fail during power cuts, and voltage reductions on the mains. If the smps is not up to scratch, the poor regulation and heavy load combine to detroy the mosfet, and then, as often happens a catastrophic chain reaction takes place and takes out many other smaller components. If you want to power a smps from a variac, you should power the control circuitry independantly.
Bill

[PETERg0rsq]

Thanks for the info. That should save me from damaging a switching transistor/FET unnecessarily in the future.

If I understand correctly the problem is that if the input voltage is too low, then there is not enough voltage to fully turn on the FET (or transistor), which introduces a small (but insignificant) resistance, which causes power to be dissipitated within the FET!

I therefore guess that the input voltage could be reduced to an arbatory figure, so long as the switching of the FET was fully saturated (is that the correct phrase!), however knowing what this level was is the uncertanty.

If I ensured that the switching voltage was always high enough (by using an external PSU (Isolated??) presumably fed through a diode) would it then be OK to fully vary the input with the variac, or even not have any supply connected at all?

I dont have a smpsu to look at at the moment, so no circuits to describe, but I am sure one will turn up in a piece of test equipment one day that needs looking at.

Thanks

peter

[Wendymott]

In my experience of SMP PSU's they are all designed to operate from approx. 90V AC up to 250V, obviously the current required is higher, say double that of 240V. In my workshop all suspect SMP's are tested at approx. 200V, and then raised if operating normally. I accept that older SMP's were VERY tetchy with odd loads or input voltages, the NEI CE25 was one typical animal that destroyed the SMP easily. One test when evaluating new products is to ensure the SMP operates cleanly down to 110V and up to 260V AC. Regards
Wendy G8BZY

[Radio Wrangler]

Well-designed SMPS should have an undervoltage lockout which disables their starting if the line voltage is so low that excessive currents would be taken, or if there isn't sufficient voltage to fully turn on MOSFET gates.

I've had the job of developing SMPS, and I fully support what Kalee said.

The very best thing for development is a small power supply to power thecontrol circuits and a high voltage (which will also go down low) regulated supply to stand in for the mains. Development is relatively tame this way, emitting the minimum of smoke and screams. Without a nice reg supply you go straight to stage two:

The next best way is to have a small psu to run the control circuits, then a variac and isolating transformer to bring up the power circuit gently.

The very worst PSU I've come across was in a highly reputable Japanese brand of 21 inch monitor, which was OEM'd in and fitted with the badges of a highly reputable American brand of computers. It had a rectifier that switched between voltage doubling and bridge modes for 110 and 230v supplies. Unfortunately it checked once on rise of power, selected a mode and then stuck with it come exploding zeners and everything. Unfortuntely, a mains brown out dipped so low that the monitors thought power was off, then it came up to somewhere in the 110v range, the monitors chose to enable the voltage doubler, thenthe mains came up to 240v. The monitors had made their minds up and were sticking with it! Fortunately someone had added protective zeners across the main reservoirs. The zeners pulled the fuses, but the zeners exploded with the energy in the reservoirs. Made a lot of folk jump.
Quite an oversight in design. Running one of these up slowly on a variac had the predictable result, I might add. Officially the monitors (hundreds of them at a couple of thousand pounds each) were not repairable and not covered by any warranty (being in a branch of the firm whose badge they bore) but folk were stuck without them. A look inside, a fuse and a pair of zeners and soon monitors were up and firing on all three nitrons.

Never run a whole SMPS on a variac.

David

[woodchips]

See other SMPS thread?

Umm, thing about a light bulb is that it has a considerable non-linearity of resistance, hence will both provide near full line voltage and be a safety thing all at the same time.

Obviously the initial start up circuitry will need effectively a full voltage, but once the thing has started oscillating the control power comes from a winding on the o/p transformer. Can't see why you can't run that through a variac, light bulb, whatever?

David, how was the selected supply voltage latched after startup? Can't be a relay, SCR? Do these 90-260V supplies all use this supply latching? I had always assumed it just had a wide input range.

[kalee20]

Originally Posted by PETERg0rsq

I therefore guess that the input voltage could be reduced to an arbatory figure, so long as the switching of the FET was fully saturated (is that the correct phrase!)...

No it isn't - 'saturated' in connection with FET's is associated with operation such that the current flow is device-limited (lots of power dissipation), rather than external-circuit limited. Same as a valve, with insufficient heater voltage.

Originally Posted by Wendymott

In my experience of SMP PSU's they are all designed to operate from approx. 90V AC up to 250V

Lots are, but not all.

Before the wide-range input became commonplace, many were designed tooperate from 110V mominal (87 - 132V) or 230V (187 - 264V), selectable with a link. This changed the input circuit from a voltage-doubler to a simple bridge.

David Radio Wrangler effectively described this system.

One manufacturer - I think it was SGS-Thomson - introduced an IC intended to drive a thyristor or triac, to do the changeover automatically according to a sensed input voltage. The result of this was that there is a no-mans-land area betwen the ranges.

Originally Posted by woodchips

Obviously the initial start up circuitry will need effectively a full voltage, but once the thing has started oscillating the control power comes from a winding on the o/p transformer. Can't see why you can't run that through a variac, light bulb, whatever?

Sometimes you can. It depends.

Often, the control circuitry is fairly low power - the winding to supply this, for economy, peak-rectifies the waveform. The main output winding, however, may have an averaging choke, so that regulation by pulse-width modulation is possible.

If the control power is simple peak rectification, then these circuits will get powered from a variable voltage according to the mains. It's quite possible for the control supply to be dangerously low, even when the output is still within regulation.

Good power supplies do indeed have a low-voltage shutdown, but I have seen many that don't.

[Radio Wrangler]

Most SMPS have a single wide range and switch nothing. They are usually designed to work from 88v (Japan on a heavy day) to 260v (UK) without a gap. Any voltage between will do.

The big trinitron monitors had a thick-film hybrid circuit which switched a rectifier bridge into a voltage doubling mode. This is quite an unusual arrangement. Quite a bit of HP gear had a manual switch for 110/230V ranges, and smaller stuff tended to be single wide range. The switch was more economical on reservoir capacitor ratings, though the ripple current in doubling mode made me doubt the benefit was as large as some people claimed.

Light bulbs are good if there is a short in the input mesh, but aren't really dainty enough to protect things if they can handle the initial surge. You can get a self-oscillation of start-stop-start. Some lower resistance thermistors have been used as inrush protection for power meshes though.

If debugging an SMPS you need an isolating transformer, a power supply to run the control circuits and a lot of care as a minimum.

David

[kalee20]

Originally Posted by Radio Wrangler

If debugging an SMPS you need an isolating transformer, a power supply to run the control circuits and a lot of care as a minimum.

Yes. And a 'scope. And a variac. And a variable load.

[McMurdo]

Lots of modern SMPSU's also have PFC correction and this can self destruct if you wind them up slowly.

[Radio Wrangler]

True. PFC without an undervoltage cutout is a disaster waiting to happen. I have a high power 'function generator' with several kW output and a PFC single phase input. I was very careful to pick one with a properly designed input section. A fault could have been spectacular.

David

[broadgage]

To add to the points above, I forsee no harm in operating a SMPS from a variac provided that the voltage supplied from the variac into the SMPS is within its stated operating range.

With a minimum design input of say 85 or 90 volts, it would be unwise to gradually increase the voltage from zero, but no harm should result in turing the SMPS on with the variac already set to 90 volts.
The SMPS does not "know" the difference between Japanese mains at the lower end of the range, and a UK supply reduced to 90 volts by a variac.

One might do this either for test purposes, to verify that a SMPS does in fact work correctly over the claimed input range, or because one had a SMPS with an input range from 85 volts to 130 volts only, and a UK supply.

[Radio Wrangler]

I have run a number of new SMPS designs from variacs and from monster amplifiers (to produce missed cycles, interruptions and surges) for the purposes of validation testing. All were fine. BUT, I knew the designs had been thoroughly done and the low voltage case was covered.

When fixing an SMPS, my normal technique is to run the power section from low voltage (like a current limited 30v lab PSU) rather than from AC, and to put a second PSU in as the source of the control system power. Under these circumstances there are no hazardous voltages, I can use a normal grounded scope, and the possibility of smoke and flame is controlled. Once I've got it running at scale-model voltages and checked a few waveforms, then a variac and isolating transformer can replace the first lab PSU.

Two different circumstances, two different approaches.

David