Lab Power Supply - Pre-regulator

The next part of the lab power supply that I wanted to tackle is the pre-regulator circuit. The problem with making a power supply without a pre-regulator is that the main pass transistor may have dissipate a lot of energy as heat in order for the circuit to operate.

You may recall my design has two modes - 0-15V and 16-30V. In 16-30V the worst case is if you set the output to 16V @ 3A in which case the MOSFET must dissipate (44-16) * 3 = 84W. Or in 15V mode you set the output to say 1V @ 5A in which case it is (22 - 1) * 5 = 105W.

These are pretty insane numbers - even a huge 0.35 degrees per Watt fan cooled heatsink will get 30 degrees above the ambient temperature.

If we instead use a pre-regulator with a switching element that dissipates little heat when it is on and nothing when it is off then we can control the voltage on the bulk capacitor and reduce the energy waste.

One approach to this problem is to use a switching pre-regulator which is essentially a switch mode power supply placed between the rectifier and the bulk capacitor. This will chop up the voltage using a MOSFET and maintain the capacitor voltage through PWM.

The problem with using a switching pre-regulator is that they create a lot of switching noise at frequencies that are hard to filter out of the final power supply output.

Linear Technology AN32

AN-32 describes an interesting circuit that uses SCRs to implement a pre-regulator. An SCR is essentially a diode with an extra leg. The SCR won't conduct unless a current is applied to the gate but once it is conducting, it doesn't matter what happens to the gate and it will keep conducting until the current flowing through the diode stops.

The way these are used is to switch the current to the bulk capacitor and thus reduce the voltage on the capacitor. The SCR is basically a diode when it is on and so dissipates little heat. When it is off it dissipates no power at all.

The voltage coming out of the bridge rectified is basically a sine wave where the negative half has been flipped up. This creates a big ripple with a cycle at double the mains frequency (100Hz where I live). The trick is to delay passing the current to the capacitor until later in the cycle. This reduces the voltage in the capacitor and the later you wait the lower the voltage.

An SCR is ideal for this as you fire the gate at the time point where you want to pass voltage to the capacitor. Then when the voltage dips back down to zero again, the SCR resets ready for the next cycle. Here is a photo from the Agilent PSU Design Handbook that illustrates the effect of varying the firing time.

The way the circuit in AN32 does this is by first generating a sawtooth voltage that is synchronised with the line frequency (a line synchronous ramp). This is generated using a comparator to detect when the voltage on the bridge rectifier is near zero and then charging a capacitor/resistor network. If you then use a comparator to compare the ramp voltage with another set voltage then the output of the comparator will turn on when the ramp goes higher than the set voltage.

Here is the line synchronous ramp I designed using a LT1716 comparator. This comparator is handy as it is safe up to 44V which is the voltage level of the transformer in series connected mode.

The inverting input is set to one diode drop above ground by the resistor and diode network. The AC input is diode-ored and then clamped to one diode drop above ground by the diode/resistor network on the right. When the output of the comparator pulls low (i.e. when both the AC signals are less than 0.6V) it pulls the capacitor low and then the capacitor slow charges back up until the cycle repeats again.

The next step is you have an op amp that compares the bulk capacitor voltage with the target voltage and this generates a voltage for a second comparator. It is this second comparator that fires at to turn on the SCR. Again I used a LT1639 for the op amp as it can handle 40V and I used another LT1716 for the other comparator.

The op amp has a capacitor/resistor network to slow down its operation enough that it acts like a servo and controls the firing point to keep the output voltage around the target. The size of the capacitor is a bit tricky since I found that if I make it too small then the circuit doesn't maintain the voltage as well (it gets low). If I make it too big then the circuit doesn't react very quickly to changes to the desired output.

This is what the voltage on the bulk capacitor (the pre-regulator output) looks like when the pre-regulator is set to generate 10V and the output is drawing about 3A. At turn on the op amp hasn't reacted yet so the capacitor gets over charged. This then bleeds off and the circuit falls into a regular firing pattern. Note the output does fall below the target 10V so I probably need to set the regulator to a few volts above the target.

Here is a close up of the AC signals, the output of the comparator driving the SCR and the output voltage. The output was set of 25V in this trace. You can see the firing point half way through the cycle.

In this trace I set the target voltage to 40V (which it can't achieve). This is as much as it can generate and unfortunately it is quite a bit below the voltage the transformer is capable of. The output of the SCR comparator is constantly at the maximum here but it still doesn't achieve even close to what I would have expected the full voltage to be.

I'm not sure why this is as in this state the SCR gate is fully on the whole time.

I also experimented with using a MOSFET to switch the voltage and an SCR just to generate the MOSFET gate signal but I couldn't get it to increase the voltage at all. This still needs some more work.

Until next time...