This is the second of what is likely to become a long group of linked articles on the design of PWM controllers. We will take you though practical designs, explaining good points and bad ones.

If you have not already read the page Pulse Width Modulators on our electronics site, public area, you should do so before proceeding as that page is the first in this series.

Clearly, designing any circuit such as this is not trivial and is not really an ideal starting point for absolute beginners. Nevertheless, the interest in robots will cause many beginners to want to start with this project. That will mean that much of the description may be over the heads of some readers. Other articles will be written for 4QD TEC members, explaining how the components and circuits work.

Issue 2 circuit

The first thing that you will note is that the 3524 chip used on the first issue has been dropped. In its place is an LM339. The 3524 and all similar chips are designed for switchmode power regulators rather than for PWM motor control, so they do not allow 0-100% modulation. Using an LM339 allows 100% modulation.

Power supply arrangements

Tr1 is a switch, operated by the presence of the potentiometer, which connects the supply rail to the internal supply rail. There is a 47R in the emitter, to act as supply decoupling. The circuit will work off less than 12v to the limit of the LM339, about 36v.

However a 36v battery can be over 40v when fully charged, and this is above the guaranteed operating range of the LM339. In later designs, we changed the power supply system for this very reason. Note the 12v zener from pin 14 (IC1b) to 0v to clip the MOSFET gate voltage to 12v: 24v is too high for most MOSFETs!

Triangle wave generator

Ic1c is a 'standard' op-amp triangular waveform oscillator: a tried and tested circuit which works well. See Pulse Width Modulators on the public area of 4QD's www site for more information. Ic1b is the modulator which compares the triangle waveform with the speed demand voltage to determine the pwm ratio. Note the 4mA current source (Tr2 and Tr3) pulling up pin 14. This current source provided base drive for the emitter followers Tr4 and Tr5 which drive the MOSFET gates.

A current source here feels 'nice' - the raw switching time out of the comparator is naturally very fast indeed and it is rather necessary to control the rate of rise of the gate turn-on voltage. A current source, charging C9 and C10 in parallel does this nicely. However, it is an unnecessary luxury that we do not use nowadays.

Pole transistor

The output waveform, as well as driving Tr4 and Tr5, is fed to Tr6 which is the pole transistor that controls the hiside switching. C10, as well as controlling the rate of rise of the loside gate turn on, also speeds up the drive to Tr6, helping to turn Tr6 on (therefore the hiside off) before the loside MOSFETS start to conduct.

Current limiting

The way the current sensing works has already been explained in MOSFET current sensing on the public area of 4QD's www site. There is quoted but a small part of this whole circuit. The sensed current develops a voltage across the 47R source resistor of Tr12 which is fed to the base of Tr10, via a resistor divider network.

Hiside charge pump

The 3524 circuit described in the public are of the site used a similar hi-side charge pump. In this case, the output of the triangular oscillator, on IC1c pin 2 is a nice 50% squarewave, just right to operate a hiside charge pump. Tr11 forms the top of a totem pole output stage, to drive C7 which is the pump capacitor.

Hiside Drive

C8 is the reservoir capacitor for the hiside drive. The 12v zener connected across it limits the voltage on it - otherwise the pump and the bootstrap action would add up to give a voltage in excess of what the MOSFET gates can safely handle.

Input buffer and ramp

The pot feeds via a preset, which is to adjust the sensitivity so that full pot rotation only just gives full speed and no more. Remember that the circuit is used of 12 or 24v and the full voltage is present across the pot, so the preset will be adjusted to suit the operating voltage.

Issue 3

This circuit was little different to issue 2 so will not be discussed.

Constructional information

We are not giving constructional information of these early issue boards: the circuit has undergone several updates and it is educational to discuss the reasons for the changes. Constructional information will be given at the end of the series. However, this circuit works fine and there is very little that needs to be altered.

Other relevant pages

Page Information