Switched Mode Power Supply SMPS by SM5UIU   -   In 220VAC / Out +/- 40 VDC

(Best viewed with 1025x768 pixel's)

(Note: The component value/type's in the schematic's below may have been changed to other values or type's.)

After I had build't an SMPS for +13V in +/40V out i decided to take the next step and design another SMPS for home use. The experience was sure nice to have in hand from the lo voltage design but sure, new problem's came up wich had to be solved. Challenges are always welcome so I jumped right into the datasheet's again here's the result.

The input line filter consist's of 2 separate filter's. First the filter that's included in the line connector (does it work?) and then C8 / C11 along with T2 wich purpose is to filter out all possible voltage spikes / overtone's  from the SMPS. Then we have a rectifier along with 2 filter cap's in serial (since they only have a 200VDC rating) giving a total capacitance of 235uF. (The cap's were removed from a computer power supply wich operated in half bridge mode) R16 / R17 help to keep the voltage over each cap as equal as possible and also work as bleeder resistors discharging the cap's when power is turned off. There is also a small pcb transformer wich is used to power the switch regulator ic. The transformer's output is rectified using standard 1Amp diodes, filtered and then regulated down to +12VDC.

First we apply the input voltage to Vin and Colector A / B (the IC has two transistors intregrated which can be configured in many different way's, here we tie their collectors directly to +12V) Each transistor output has a voltage divider R6/R10, R5/R9 wich center tap is fed to the switching fet transistor's. VREF is at +5.1 V and is divided by R1 & R4 which in turn supplies the NONINVERTING input of the error amplifier with VREF/2 Volt. (The regulator's error amp stage has an operating range of 1.8-3.6V, so 2.5V will be perfect).

R3 and C1 set's the operating frequency (~140kHz). C3 & R2 are connected to the compensation input and is to get the whole loop stable. Q1, R13 and C14 is a slow start circuit. On power on C14 will charge slowly (3-4 sec) up to +5V though R13 wich is connected to the Vref output and Q1 will hold down any error signal output from the error amplifier.

The error amplifier output is the same as Comp pin 9 and when the error voltage is below 1V we won't have any switching action at all. When the voltage is at 3.5V we have full pulse width (if needed). R8 is connected via the ON/OFF switch so when unit is turned off the slow start cap will quickly discharge. This is for protection, if one has full output load and would quickly turn the unit off and then back on again, then all cap's would proboably be pretty much discharged and C14 would not have had enough time to discharge wich could result in transistor burnout due to the the high start current needed to charge all cap's. (Full error signal from the PWM controller error amp.) 

U3 is an optocoupler wich led diode is connected to the +31V and -31V rail's. If voltage is below 62V the optocoupler transistor will not conduct and hense the inverting input will be held lo which the controller will correct by increasing pulse width to the switching transistor's. When the voltage reaches 64V (this due to the 2 series connected zener diode's) the led diode current will increase and the optocoupler transistor will start to conduct and the voltage on the inverting input will rise up to a point were it's at the same potential as the non inverting input (reference).

When looking at datasheet's and comersial SMPS they have often a special reference ic on the secondary output so that the output voltage will be precise the expected and nothing else. When using zener diodes we will have a slight variation depending on the temperature (on the diodes) and the current needed on the output transistor. Let's say that we need a current of 1mA on the output we would most likely have to supply at least 1 mA to the led diode and hense 1mA x 560ohm (R7) = 0.56 Volt's plus the additional voltage drop over the led in the optocoupler. So depending on current needed and temperature we will get and output voltage from about +30VDC to +32VDC, wich is fair enough.

(The above schematic has no snubber circuit) Drive pulse's from the PWM controller enter the gate's of Q3 & Q6 wich in turn pull their respective source to ground. The transistor's will have (due to the "transformer effect") double the dc input voltage over the source pin's so 2 x 325V = 640V. By knowing this we choose the fet's to be min 800V type's.

We will also have some additional voltage spikes which we could use a snubber circuit to minimize the spikes. If we would use one the snubber cap's would charge upp to >600V and then discharge again though the fet's. This causes a pretty large power loss in the serie's resistor's. This loss is mostly dependent on cap size and with a 1.5 nF cap (wich is the smallest cap you can use to get the desired result's) the loss will be minimum 9 watt's with a switching frequency of 2 x 58 kHz on each resistor. Some say that you dont need snubber's at all if you have fast enough fet's so that's why I decided to not use one..

What about power loss in the SMPS?

Switching losses in the transformer core? ETD44 (3F3) core has a loss of >2.2 W at 100kHz and 100 milliTesla load, this configuration get's us up to >130 mT. Hey, we must not forget the losses in the transformer winding's which can be significant. The calculated dc wire resistance is much lower than the actual resistance for ac current's. This is why it's a must to use several winding's in parallell. I have experimentet with using only one winding which resulted in a temperature of >140°C. With 2 windings 0.7 mm copper wire in parallell the temperature is still very high and the recomended number of windings is 3 or more. Rectifier losses, 2-4 W (depending on type),  the losses in the output inductor and filter cap's. The transistors that I use have a on resistance of about 1.5 ohm and keep in mind that the current that the fet's can handle get's lower as the temperature increases and that the ON resistance will also increase wich leed's to even higher voltage drop over the transistor's and increasing power loss!

How to measure actuall efficiency? Huh, by measuring the input current on the 220VAC side a "standard" Fluke77 / 0.85 A, a Fluke112 (TRMS) / 1.36 A and this just tell's us that the TRMS one we can't trust at all when it comes to pulse type current so I hooked up a 1 ohm resistor in series  (1 V =  1 A) a scope, and tried to calculate the RMS current, result  0.7 A / 220 V. Input power 154 W / 133 W out = 86 %. These numbers indicate that the configuration work's pretty ok, even if 86% perhaps could be degraded to 80%?

The transformer's primary consist's of 2 x 60 turn's and the secondary is 4 x 10 turn's, core used is Philips ETD 44 / 3F3. L1 / L2, 50uH choke wound with 8 turn's / 1 mm wire. Core blue Philips 3F3 TN13/7/5.5 AL 900.  Cap's used are lo ESR type's (ELNA, type RJH, 2200uF/50V. Impedance 100kHz/24milliohm's.Size, 18x40 mm)

As the frequency increase there is not only a problem with capacitors (lo ERS), transistors and so on. We will also need special switching type diodes capable of handeling the current. "Normal" diodes are to slow for this kind of high frequency rectifying. Diodes used here is BYW29 with a TO220 case rated for 150V / 8 Amp average current - 16 Amp repetive peak current. Diodes D1-4 are ofcourse mounted on a heat sink.

The test's that I have done is to load the SMPS with 70W continuous output and it work's prefect. The core temperature will rise up to about 70°C which indicate that my 2 windings in parallell should be 3 or more... So that's pretty much it..

(I will not give out any pcb layout since this is only an experimental SMPS and I can not guarantee the reliablity of it.)

This page has had visitors and was last updated 2002-06-22.