4
\$\begingroup\$

I've been trying to make a SMPS for sometime now but I can't get it to work.

In my design it keeps killing the MOSFETs once I short the output with a thin wire that doesn't let more than around 20 A go through and also I calculated the output to be more than 22 V and 5 A but I get around 10 V with I think very low current. Also the voltage rises slowly at around from 0 to 30 % duty cycle (it takes around 15 seconds) and 5 times faster above 30 % duty cycle.

MOSFETs do not heat, they just die.

Can someone please point out what's wrong with my design and how to improve it?

I have little experience in electronics so please easy on me.

Here is the primary side schematic:

Primary side schematic

And here is the secondary side schematic

Secondary side schematic

  • C1: 0.01k 275v~x2 GMF MPX/JFEU

  • T1: interference inductor from a PC power supply - I don't think value is important so I didn't bother to measure it.

  • Bridge: Bridge rectifier which is also from a computer power supply, its rated at 8 A if I remember correctly

  • T2: a transformer also from a computer power supply, it's ETD 39/20/13 I guess

Dimensions:

  • measuring from top left to top right is 39.45 mm
  • width is is 12.50 mm (measuring from the side)
  • middle core is 12.50 mm (the round thing in the middle)

The core is 2 pieces, 1 piece height is 20.26 mm

The core is 2 pieces but I managed to break it and got it to 5 pieces then I glued them back together making sure to keep no gap, I did not put the glue between the pieces just around them to make sure there is no gap.

The other component's names and values are in the schematic.

I'm using an Arduino with 21 kHz PWM signal to drive the MOSFETs with TC4427 driver between the Arduino and MOSFETs.

I tried few times and burned more than 15 MOSFETs so far with single switch and 2 switch forward SMPS.

I didn't bother myself to put protection like thermistor, fuses and so on because this is just a test until I get it to work, but I put a 500 W heater in series with the main and my SMPS to avoid extreme short circuit.

In my earlier design with forward topology if I leave it as it is, it also kills the MOSFETs (I tried with 1 switch too with reset winding) after around 10 seconds the MOSFET dies.

I also tried with flyback but with 40 V input and it worked but not as I calculated, MOSFET heats up quickly and I can only get 5 A at short circuit at the output, maybe because input was 40 V?

Anyway, here is my last calculation which I did using a software package:

Screenshot of calculation

In this software it says I have to have 8 mH inductance at 54 turns at the transformer, but since i broke my core I had to add more turns to get 8 mH, like 100, I forgot.

In this software I selected half bridge, could this be my problem? Because with push pull the software says I need 108 + 108 or 216 total with 31.795 mH.

If I remember correctly I did try with push pull total windings for forward and i got the same results.

BTW, I measured the inductance with this, LCR-T4 ESR Meter Transistor Tester Diode Triode Capacitance SCR Inductance M328 .

This is the best tool I have hopefully it measures correctly.

If I'm not giving enough information please let me know what I have to write and I will when i get back here.

The short circuit thing I do is just like a heavy load it does not do a full short circuit, it's a breadboard connecting wire, i had a pc power supply with short circuit protection and a maximum of 30 A (which it will shut down if i pull more) and connecting this wire from + to - did not trigger the protection it just burns the wire.

After remeasuring the transformer it appears to be 10mH now

As some have reminded me that high side n-channel MOSFET doesn't work in this configuration, I decided to go with only low side n-channel as @TemeV said. I did this and killed 2 more MOSFETs, then I ran out of high voltage MOSFETs so I went down to 28 V AC from a microwave oven transformer and recalculated everything to get an output of 3.7 V at 10 A with 27 kHz then I did some tests and the MOSFET quickly got very hot without an extra load other than those 2x 10 kΩ at the output stage.

Here is the schematic for primary side with one MOSFET (the secondary side is still the same).

Schematic for primary side with one MOSFET

After checking what was wrong for sometime and paying extra attention for everything I noticed the transformer is connected in the wrong polarity in my board.

So I fixed that then I finally got an output with current which goes to around 36 V (35 V is the capacitor's maximum voltage) at around 5 % duty cycle which it supposed to be around 3.7 V but I'm pretty sure it's because it doesn't have regulation (feedback).

Then I decided to test it under load with a 3.7 V 30 W LED.

I went up gradually to around 12 % duty cycle while checking the MOSFET's temperature, I got the LED to around 3.2 V at around 2-4 A.

Then I checked the MOSFET and it was around 40 °C after like 10 seconds, its around 30 °C here, but when I went up to around 16 % duty cycle I noticed smoke coming from the interference inductor, the software stated that I will need around 3 A at the input at 30 V but as we know a microwave transformer is not regulated and the output voltage decreases, depending on the load so my power supply takes more current to compensate with voltage drop which my interference inductor and bridge rectifier isn't designed for such currents so I'm going to stop here until I get more high voltage MOSFETs.

Update:

i got some high voltage mosfets and after burning 3 of them heres what i did

  • Added some more filters at the input with few more protections like thermistor and fuse
  • Some recalculations, now transformer is 12mH and operates at 50 khz frequency
  • Removed diodes, 10k resistors and capacitors at the output
  • added new components at the output(Secondary side):
    • bridge rectifier(KBPC3510)
    • 2x 1.5k ohm resistors
    • 2x 470uf 100v capacitor
    • 1x 470uf 80v capacitor

So when i turn it on and set the duty cycle to 0.5% my multimeter goes out of scale even at 600v DC when i measure after the bridge and before the inductor, i have around 90 turns primary and 14 secondary.

when i measure after the inductor the voltage starts at few hundreds then settles at around 90v for few seconds before i shut it down

Could this be because the inductor at the output can't filter 50 khz frequency?

Update:

After doing some tests i can successfully turn on 16 watt led at 12v at 13% Duty cycle without killing the mosfet and everything stays cool except for the bridge rectifier at the output, it gets too hot to touch.

\$\endgroup\$
8
  • 3
    \$\begingroup\$ why are you shorting the output? Are you sure this design can handle that? You say you calculated that it can handle 5 amps, but you're just shorting it? \$\endgroup\$ – Hearth May 29 '19 at 0:23
  • 6
    \$\begingroup\$ Would I be correct in guessing that Q2 dies an awful lot ... ? A high-side switch needs a high-side driver - and you don't have one. It's just a source-follower. \$\endgroup\$ – brhans May 29 '19 at 0:55
  • \$\begingroup\$ What voltage/current are you attempting to get out of this? The most obvious problem is that you need a level-shifter for the high-side MOSFET. See: eas.uccs.edu/~cwang/ECE5955_F2015/PowerElectronics_f2015/… If you don't have a specific goal and are just playing around, consider playing with a buck or boost regulator, these are more suited to a beginner. \$\endgroup\$ – Mattman944 May 29 '19 at 1:19
  • \$\begingroup\$ @Hearth its like somesort of a test i do xD, those circuits can't handle this kind of shorts? its actually like a big load rather than a short \$\endgroup\$ – Adam May 29 '19 at 10:14
  • \$\begingroup\$ @brhans yea n mosfets arent good at high side unless you have around 10v at gate above source as i know, But forward smps schematic shows 2 n mosfets on both side so i said there could be something i dont know \$\endgroup\$ – Adam May 29 '19 at 10:17
1
\$\begingroup\$

Your high side fet is not driven with high enough voltage. You can't connect the gates of the FETs in parallel like you have done. TC4427 is intended to have a P-channel FET on the high side. And actually the high side FET is not even needed here, you can use only the low side.

So change to P-channel FET on high side or remove it completely

\$\endgroup\$
10
  • \$\begingroup\$ So the calculation i did was right?and using p mosfet or n mosfet with bootstrap or n channel high side driver will work? \$\endgroup\$ – Adam May 29 '19 at 10:25
  • \$\begingroup\$ And using only the low side i will need a reset winding right?or i can get away by keeping those 2 diodes d2 and d3? \$\endgroup\$ – Adam May 29 '19 at 10:29
  • \$\begingroup\$ okay i tried with only low side with the same setup but by connecting vcc directly to t2(transformer) and now i burned 2 mosfets, i picked 30% duty cycle with first mosfet and killed it instantly, and with the second one i tried to go up gradually from 4% duty cycle but the mosfet died right after i set the duty cycle to 4% \$\endgroup\$ – Adam May 29 '19 at 12:43
  • \$\begingroup\$ oh i forgot to say, last test was with 45khz signal \$\endgroup\$ – Adam May 29 '19 at 12:47
  • \$\begingroup\$ This is forward conveter, sorry i didnt know adding a "flyback" tag would show it as flyback, i removed it now. \$\endgroup\$ – Adam May 30 '19 at 20:14
0
\$\begingroup\$

Apart from the high side MOSFET not getting a high enough gate voltage, it also is redundant.

Forward converts usually use a half bridge and a dual rail sypply or an h-bridge and a single rail supply or a center-tapped transformer. I.e. a push-pull configuration, this is important because otherwise you will get a DC current in the primary which may cause the core to saturate.

When using a push pull confoguration, it is essential to have some deadtime between conducting in one direction and then the other. Without dead time both transistors will conduct at the same time and short the supply. (This will also happen with a center-tapped transformer since the two currents will cancel out and all reactance will be lost.)

And flyback diodes or a carefully calculated snubber network is always required. When the transistors turn off, there must exist a path for the current to flow through the inductor without inducing a too high voltage.

I would suggest you look at some ATX PSU schematics and figure out a way to alternate the pulses in the PWM signal to two different signals to drive a push pull stage.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.