High power adjustable switching power supply (SMPS) 3-60V 40A

Hi,

I'm looking for some help on this Adjustable SMPS Circuit found on Danyk.cz,

Schematic

I've built the circuit on a PCB and everything works fine Voltage regulation etc...

But as soon as I attach a load (A 60 Watt 12V Light Bulb) in this case, the IGBTs short out and die, all 3 legs are shorted together.

I tried with 600v Mosfets as well but the same still happens.

It's not on the PCB I designed, but I have soldered RC Snubbers accross the IGBTs

Any Help would be much appreciated.

My PCB
 
Hi,

I'm looking for some help on this Adjustable SMPS Circuit found on Danyk.cz,

Schematic

I've built the circuit on a PCB and everything works fine Voltage regulation etc...

But as soon as I attach a load (A 60 Watt 12V Light Bulb) in this case, the IGBTs short out and die, all 3 legs are shorted together.

I tried with 600v Mosfets as well but the same still happens.

It's not on the PCB I designed, but I have soldered RC Snubbers accross the IGBTs

Any Help would be much appreciated.

My PCB

Hi Richie,

I will try to help you if I can.

First your surprising statement: "I've built the circuit on a PCB and everything works fine Voltage regulation etc...". How do you know that the circuit is operating correctly if it is destroyed when you apply even a very moderate loading? Can it run in idle-mode without the IGBTs shorting?

The power converter is double-switched on the primary and regulation is achieved as in a forward converter. Do you have an oscilloscope in your possession?
Who made the transformers you use?
What is the net-voltage where you are?
 
Hi Faux,


Thanks for the reply, the circuit works in Idle and the voltage adjusts from about 8V to 50V using the 100k Potentiometer, I do have an oscillisope and looked at the gates of the IGBTs and there is a square wave going to them.


The transformer I'm using comes from a 48V 30A Server PSU
 
That's quite a robust well thought out circuit. Normally people present circuits that short a car battery across a bucket with 58 mosfets.

As FF says it's a Two Switch Forward Convertor. Your Max duty cycle is 50% to allow for transformer reset, both drive transformer and main. The UC3845 has that limit built in.

http://www.ti.com/lit/ds/symlink/uc3845.pdf

I could grumble a bit about the layout of your PCB. I assume P$1 is your auxiliary, TNY, supply. However, without going into details, as it stands it's not too horror show. You'll get to see how bad things might be when you probe various points.

Your main transformer might be an issue. Chances are it expects about the input voltage you are giving it and the secondary looks to be correctly scaled. However this type of convertor is generally built with an ungapped core with leakage inductance minimised and Computer power supplies are normally LLC series resonant, Google. The core is gapped and there is a large leakage inductance.

If that is the case then your primary currents are going to be excessive, low primary magnetising inductance. Reset current through the BY diode will be high and power transfer poor... large leakage inductance.

Something to think about later given it still manages to power up and run. I'm impressed.

I would be inclined to check around the current feedback via TR3 into pin 3 of the UC3845. It is a peak current mode control IC and peak limiting occurs when pin 3 rises above about 1V. With R15 as the ballast being 3R3 secondary current in TR3 needs to be 300mA to trip it. D9 does not affect that. However with a 68/1 turns ratio the power switch current has to hit 20 ish amps before limiting occurs. Set P2 to a minimum while working on things and gradually increase it.

First off if your current transformer is not phased correctly then the current limit is not there.

If you scope it, top of ZD3, no load then you should see a low amplitude ramp waveform when the power switches are on and zero volts when they are off. The waveform should be 'clean' and the ramp linear. If it bends/curves up towards the end then your transformer is saturating.

I'd take a guess that for this sort of power the peak primary current will be about 10A and you might run at 10% magnetising current so no load 1A will be the peak magnetising current. With your 3R3 resistor and the 64/1 CT ratio you should expect about 50mV at the top of ZD3.

If it is very much higher than that then your main transformer is not the one you thought you wanted and if it is not there then the CT is phased incorrectly.


OW OW OW. Set P2 wiper towards pin 8, the reference. You will see the offset on pin 3 at about 1V placing the IC at the point of current limit with the ramp on top of it. Someone has put a lot of thought into this circuit.
 
Last edited:
Hi MorbidFractal,


Thanks for your reply I've read through your post and you seem to know your stuff :) If it helps the current transformer I'm using is this one RS PRO Current Transformer, , 15A Input, | RS Components


I can confirm it is wired correctly the right way round becuase if I turn P2 all the way down the Red LED comes on and there is no output voltage.


I will try scoping it when I get some new IGBTs as I've blown all the ones I have lol


Also the Gate drive transformer I'm using is this 1013C Murata Power Solutions | Mouser United Kingdom


And the Transformer comes from a Tyco 300-1632-06.
 

rayma

Member
2011-04-29 8:37 pm
the circuit works in Idle and the voltage adjusts from about 8V to 50V using the 100k Potentiometer,
I do have an oscillisope and looked at the gates of the IGBTs and there is a square wave going to them.

You can't use regular probes for the gate-source nodes, which are floating at AC line potential,
and not referenced to ground (unlike your scope). This is extremely dangerous, please do not attempt.

Only a HV differential voltage probe can be used in a circuit like this.
 
Last edited:
Looks like good choices for the current and pulse transformers but you might want to look harder at the notes about isolation particularly on the pulse transformer in respect of regulatory approval.

Taking pin 3 to 1V using P2 will put the IC permanently into current limit without reference to the actual output of the current transformer. Under those circumstances you will get 200nS drive pulses, delay time of the internal comparator so you have not really confirmed that things are right in that area. Having said as much I get the impression that you would not have made many if any mistakes.

Try to find a different load for testing. Light bulbs are low impedance until they are lit. If the main transformer is questionable you may already have large primary currents due to a low magnetising inductance without much having to appear at the output. The other side of the coin is a leakage inductance primary to secondary may be the thing that saves the day at start up.

Current limit in these things is King, ground is God. You really should be keeping the layout of the associated loop, and that of others, nice and tight. Short tracks, ground planes and so on. Ideally you would rotate/move the power devices up against the transformer and route the control to the left of them, burble burble. Inches matter.
 

rayma

Member
2011-04-29 8:37 pm
I Powered the board from an 230V Isolation transformer when probing with the scope
so it was floating and not connected to mains.

Still not safe enough to recommend, but regardless you can only probe one gate at a time
using unbalanced probes, since the gates are floating at different potentials.
Is the current transformer (and others) connected in the right polarity?
 
Last edited:
You already have two very competent members commenting on the control circuit design and on how to measure on the primary live side with high, potentially lethal, voltages.

My focus is on why you burn the IGBTs and why the power section does not seem to function. You clearly have some experience with SMPS and I may mention things you already know. I write this also with the aim to give advice to less experienced readers among some there is the belief that an SMPS can be designed and made even by “noobs”.

When you design an SMPS, you tend to work in three working phases:
Conceptual design with decisions on the conversion principle and the modulation/regulation principle -> design and test of the power circuit using manually variable control signals (no closed loop) -> design and implementation of the control circuit such that you end up with a closed loop regulated system.

The first part, you implicitly jumped by taking an existing design you have not made yourself, the two following parts you turned into one by connecting everything together and putting power on. Even a very well designed control circuit cannot correct a poorly designed power circuit and you loose oversight by the control circuit changing control signals without your control.

The design you have chosen is probably OK as a concept such that you can progress to the second phase of work - the power circuit. The reason why your IGBTs burn is in the power circuit. Before you can operate the double switched driver circuit with a duty-cycle from close to zero and until 50% (without load on the secondary), there is no point in thinking about the control circuit. When you are successful with this, you start loading the power circuit on the secondary until you can do 50% duty-cycle and full loading of the output. Only then it is worth concentrating on the control circuit.

You seem to have been VERY optimistic in assuming that you can just take a transformer (T1) from a power converter and use it in a different SMPS, without thoroughly checking the properties of the transformer and how these properties fit with the design of the new SMPS. The transformer is one of the crucial (particularly adapted) designs in an SMPS. In general, it has to be ensured that the transformer core can handle the SMPS operating frequency and that the winding structure does not have an unsuited capacitance. If any of these two parameters are not right, it will not work. Next, it has to be checked, or the transformer designed such, that the transformer core will not saturate under any operational condition. This second criterion is the most difficult to ensure and means either thorough calculations and detailed knowledge of the transformer core and windings, or through analysis of the transformer using more extensive measurement tests.

Why did your IGBTs burn? Either due to excessive current or voltage (often in the form of “spikes”). You cleverly used snubbers but those values on the schematic are of no use as they relate to a different transformer (designed for that specific purpose), with different leakage inductance, different winding capacitance etc. When you plug-in a rather unknown transformer into an SMPS circuit, you can only expect a high level of voltage and current spikes due to (very likely) non-optimal operation. Are you sure that you do not have nasty voltage spikes across your IGBTs?
Very likely is your transformer saturates at a certain moment when the control circuit regulates in order to compensate for a load. If so, saturation can cause very high currents and the IGBTs may burn in a split second. When one IGBT shorts, the other is exposed to even worse conditions and burn as well.

As I see it, you now have 3 options:

1) Have a transformer designed and made that fits the SMPS circuit you have chosen for, including its operational frequency.

2) Find an SMPS circuit that uses the transformer, you already have, in a way very similar to how it was used in the circuit it came from.

3) Eventually tedious analysis of the transformer you have in order to find out if there is a chance that it may be used successfully in your present SMPS circuit. The burned IGBTs hint it is not likely.
 
Hi Faux,


I understand what you are saying, and yes I have jumped the gun a bit in assuming the Transformer will work.


This is my first SMPS I've made even though I am competent in Electronics.


So what I can gather from your post is I should of tested the Power side by applying an adjustabe external quare wave to the IGBT gates from an Arduino or something similar to see if the transformer outputs in the way it should, correct me if I'm wrong.



The transformer that was used by the guy who posted the schematic came from an old 56V PSU High power adjustable switching power supply (SMPS) 3-60V 40A thats why I assumed it would work.
 
You can initially test the power side by just having the auxiliary supply running without the HV bus connected to the power switches themselves. This will let you look at the gate drive waveforms without the devices carrying any power and confirm that the levels and phasing, both devices are turned on at the same time, are clean and correct. Things may/will get dirtier when you apply the main power. In respect of that a variac is a handy lump of iron to add to your test bench.

The tell from the original author beyond dimensions and winding is...

"Transformation ratio of primary to secondary is about 3:2 to 4:3 and the ferrite core (EE shape) has no air gap."

No air gap. If you can see it, you might have to remove a couple of layers of tape from the outside of your transformer and possibly a screen foil, then if it is gapped there will be a visible shim where the two halves of the core mate with each other. That's not guaranteed because sometimes the gap is placed in the centre leg.

If it is there then you have to think harder about things. Assuming the core has sufficient area and number of primary turns with the gap being shimmed then you might be able to pull it apart, remove the shim and put it back together again with no gap. That will raise the magnetising inductance but leakage will probably still be high.

The hope would be, probably not, that things are much of a muchness. It's certainly the case that two switch forward, ungapped, vs flyback, gapped, designs work out to have more or less an equivalent number of primary turns and core size for a given power throughput from a given input voltage. LLC may not be so forgiving.

Either way it gets you closer to a working result.
 
Hi Richie,

With our traditional use of net-transformers, we are used mainly to care about the primary voltage and the winding relation to the secondary winding(s). That core saturation is avoided the producer takes care of in his marking of the transformer and the operating frequency of the net (50Hz/60Hz) is substantially constant all over the world. Secondary parameters like leakage inductance are normally negligible and the input signal is necessarily symmetric. Your assumption about the recovered transformer is logical from such experience.

For SMPS transformers, you clean the blackboard and start with all parameters free to be decided though the basic laws of physics evidently still apply. The core has to be chosen for the intended operational frequency, the core size and core type for the operational max. power and EMI circumstances. Then you have to consider how the flux is building up in the core, how to use this energy and in particular how to reset the magnetic flux in the core such that you have no slow flux build-up that in the end will make the core saturate on top of a dynamic flux.
Second order effects like ringing of the leakage inductance and winding capacitance suddenly play an important role. You hinted good electronic knowledge in general but you violated a basic concern in SMPS design because it is your first SMPS.

In case of malfunctioning of an SMPS that has never worked before, the transformer is a frequent suspect.
I know of many good SMPS designers but compared to the number of electrical engineers, it is in general small I believe. Most engineers I know of are not comfortable with how magnetic components behave when switched at high frequency.
Evidently, the most simple Buck and boost converters most can make work.

At the time where I designed SMPS, we used manually variable generators to control the power switches during test of the power stage. No Arduino in those days. If you had to build a constant-speed regulation system for your car, you would check manually that the engine would run stably and predictably before including it in an automatic regulation loop.

My guess is that constructing a correct power transformer (T1) is the only way ahead for you now.
NB: Follow the experienced advice of MorbidFractal first. He has an impressive knowledge also on this transformer.
 
Last edited: