Wednesday, May 6, 2009

Regulated Power Supplies, Part 2 by Lee Bond, N7KC

May 6, 2009 Educational Radio Net, PSRG 50th Session

This is the second of a two part series on the subject of regulated power supplies. The first part dealt with the, so called, linear power supply and the feedback scheme used for regulation. This final part will introduce the idea of using pulse width modulation for voltage regulation as well as describe the various schemes for handling DC to DC conversion using switching mode circuitry. Finally, we will be in a good position to contrast the linear approach and the switching mode approach to producing regulated DC voltages. Neither of these two series parts is intended to delve into the actual circuit particulars in depth but rather describe the underlying principle of the process.

Previously we talked about unregulated power supplies and various voltage sources which were lightly loaded and then heavily loaded. A voltage source was defined as a ‘perfect’ voltage generator in series with an impedance. A AA battery, for example, can be modeled as a perfect voltage generator in series with some internal impedance. How well the AA battery performs in the real world is related to how heavily it is loaded. If the load on a battery, or other voltage source, is constant and never changing then the regulation will be excellent provided that battery chemistry holds up. On the other hand, constantly changing loads on a power supply with a large internal impedance will cause the output voltage to change wildly and the load, possibly a radio transceiver, will be very unhappy.

Lets set up an example using a human as the ‘control’ element in a power supply scheme. Imagine yourself in front of a large panel with a big switch handle in front of you which controls a perfect switch with zero on resistance and infinite off resistance. This switch has a built in hold off function such that, when turned on then off it cannot be turned on again until a 10 second interval has elapsed. To your left is a large battery and you notice that it is labeled as a 12 volt unit. To your right is a load that requires 6 volts to operate properly. Above your switch is a voltmeter which is attached to the load. One last item is required before we turn this contraption on. Lets add a giant capacitor across the load. You are sitting in a chair in front of the switch and your task is to simply turn the switch on and off while watching the meter and, hopefully, manage to hold 6 volts across the load by momentarily connecting the large battery to your left. Notice that you have more voltage available than required at the load.

Ok, the moment of turn on has arrived, time zero if you please. You close the switch and the meter across the load and energy storage capacitor starts from zero volts and just as it passes 6 volts you turn off the switch. Now only the load and energy storage capacitor are connected and the capacitor supplies voltage to the load. Given that you cannot reactivate the switch until 10 seconds have elapsed from first turn on the output voltage starts to droop below the required 6 volts. After the hold off period you again close the switch and the output voltage rises above the required 6 volts a bit faster than the first time since the energy storage capacitor did not have to start from zero volts. Once again you open the switch when the output voltages passes the nominal 6 volt reading. After 10,000 operation cycles you are plenty tired of this process but you have learned something interesting. The average time that the switch is closed is 5 seconds.

Hummmm… you start to wonder if there is some relationship between the 5 second on time and the 10 second interval and the 12 volts input and the 6 volts output. So, for fun, you deliberately reduce the ‘on’ time to 2.5 seconds and watch the output voltage meter. Sure enough, the voltage falls to 3 volts average. Wondering if this could work the other way too you increase the switch ‘on’ time to 7.5 seconds and magically the output voltage increases to 9 volts average. In technical terms we are ‘modulating’ a pulse ‘on’ time or ‘width’ in relation to a fixed interval or width so complete control is possible by using PWM as in Pulse Width Modulation techniques.

You have deduced a very important principle. In a, so called, bang-bang circuit where the peak is constant and the switch is totally on or off the average output is always peak value times duty cycle. Clearly the peak voltage in our scenario was the 12 volt battery. The switch ‘on’ time compared to the interval is the duty cycle so 5 seconds divided by the 10 second interval is just ½ or 50% and the average output voltage is ½ times 12 volts or 6 volts. We have not talked about the ‘ripple’ in the output voltage. The average output is 6 volts but the actual voltage may be 7 maximum and 5 minimum depending on the size of the energy storage capacitor in relation to the load. So, even though we can produce an average of 6 volts the use may be limited by the ripple voltage present.

Another good example of this is the microwave oven. When the magnetron is turned on it supplies a constant energy to the oven cavity. To reduce the heating function, or average oven power, the magnetron is duty cycled just as in our example above. If we measured temperature of the food very carefully then we would see a temperature ripple just as we see a voltage ripple in the above example.

Using a long 10 second interval as in the example without a near infinite energy storage capacitor across the load yields an intolerable voltage ripple on the output. Shortening the duty cycle interval by using a very high switching frequency will maintain the average but reduce the maximum and minimum variations from the average output voltage.

Thus far we have not talked about actually locking the output voltage to some reference as is normally done in a ‘regulated’ power supply. The scheme for both linears and switchers is much the same. Sample the output voltage and compare it to some internal reference voltage. If there is an error voltage in the switcher then control the ‘on’ pulse width with negative feedback to null the error. If there is an error voltage in the linear then use negative feedback to control the output voltage of a follower circuit to null the error.

Now we can contrast the difference between the linear power supply and the switching mode power supply. The linear uses a ‘lossy’ element in series with the load to control output voltage vs the switching power supply which uses the average equals peak times duty cycle to control the output voltage.

The beauty of switching mode operation is that modern MOSFET semiconductor switches have ‘on’ resistance in the milliohm range so they dissipate very little energy when ‘on’ and their ‘off’ resistance is near infinite and, clearly, do not dissipate anything when open. This is in direct contrast to the linear pass element which by design must dissipate energy to operate properly. As a result, the relative efficiencies are 40 to 60% for the linear vs 85 to 94% in a well designed switcher.

The 10 second interval I used in the example was only to make it obvious and easy to compute the duty cycle. In reality the switching interval is normally very short since the switching frequency is well above 20 Khz so that acoustic output is inaudible to humans. It is not unusual to see switching frequencies in the 100 Khz range or higher with corresponding intervals of 10 microseconds or less. There are some huge benefits to high frequency operation such as very small filter elements in both value and physical size plus ferrite core inductors and transformers which weigh a mere fraction of their iron core cousins that are used in the typical 60 Hz linear supply. High frequency ripple effects can be tamed with very modest components.

Both the linear supply and the switching supply must have a reservoir of energy preceding the control element which exceeds the requirements of the load. For the linear device this energy supply will be at a voltage greater than the desired output voltage. For the switcher device this is not necessarily so given that some circuit configurations actually boost the output voltage above the input voltage.

There is great flexibility in design when using switching mode devices since they can ‘buck’ or reduce the input voltage, ‘boost’ or increase the output voltage above the input, both ‘buck-boost’ with the same circuit, invert the input voltage, or forward transfer much the same as the linear circuit. When the boost or buck type of circuitry is used the output is not isolated from the input however the forward transfer type of circuit uses a ferrite coupling transformer so offers complete isolation between input and output.

As mentioned in part 1 the major downside of the switching mode device is the possible high frequency noise generated by the very fast current transitions in the circuitry. This is in contrast to the linear which normally operates at 120 Hz from the AC line. Another downside to consider is the possibility of the output voltage going to maximum input voltage in the event of a switching fault. The major upside of the switching mode device is the very high efficiency obtained with light weight components.

This concludes the set up discussion for regulated power supplies part two. Are there any questions or comments with regard to tonight's discussion topic?

This is N7KC for the Wednesday night Educational Radio Net

No comments: