This page discusses important PWM controller design tips on layout, timing ramp, output dead time, bootstrap circuits, output driver, and controller generations. If you don't take care of these details, power circuits can and will fail.

Layout of the components around the controller is critical. Handbooks tell you this but it cannot be emphasized enough.

Priority 1 is the location of the timing capacitor - it must be right next to the pins on the chip. Not 1/4"; away, but as close as you can possibly get it. If you don't do this, you can see strange events, including very high-frequency clock rates (e.g. 1 MHz when you designed for 100kHz).

Priority 2 is the location of the bypass capacitors for the Vcc and output stage (if applicable) - they must be right next to the pins on the chip.

These events may be sporadic, occurring only briefly during start-up or under transient conditions, for example. Just because it doesn't occur on your one breadboard doesn't mean it won't occur in production-- keep the layout tight and you won't have any problems.

The PWM chip Timing Ramp does not like to be loaded. Be very careful in trying to take a sample of the timing ramp for use as the compensation ramp in current-mode control. Even a very high impedance (100 k in some cases) can interfere with the clock operation.

The safest approach is not to use this timing ramp at all; generate the current-mode compensation ramp with an alternate approach.

 

The 3842 and 3843 control chips allow you to adjust the output dead time from 5% to 30%. The 3844 and 3845 chips are intended for less than 50% duty cycle, and let you adjust the dead time all the way to about 70%. Choose the right family for you particular application.

What if you need more dead time than this?
Maybe you planned on a certain topology which would use more than 50% duty cycle and have plenty of 3483s in stock. Suddenly marketing changes the requirements (yes, it's been known to happen), and you need to change topologies and now need to limit the duty cycle to 45%. Usually you'd have to wait until some 3844 or 3845 chips are ordered - it can sometimes cost a few weeks, especially for surface-mount packages.

You can increase the normal available dead time - just adjust the timing R and C to get the most you can out of the chip (don't go less than 1 k on the resistor, it doesn't help any). Then connect another resistor from the RtCt pin to Vcc (nominally 12 V). This will shorten the on-time, and lengthen the off time, extending the operation range where you want it.

This is probably not recommended by the chip manufacturers, but it can save you precious time while waiting for parts to arrive. The setting of the maximum duty cycle will now be dependent upon the Vcc supply, but that's OK for working in the lab.

 

Bootstrap Circuits

Power supplies face the big problem - what's going to provide the power for the power supply when you first turn on? The answer is the bootstrap circuit, method of direct powering from the input voltage for a short time. Bootstrap circuits can be passive and cheap, using just resistors and capacitors, or active with turn-off circuits to save power when the power supply is up and running, and self-powering.

If you want to use a passive bootstrap circuit, use a controller which has a wide hysteresis band on its undervoltage lockout setting. This will help you minimize the size of the bootstrap components, and give the converter time to start properly before the bootstrapped Vcc runs out of steam. The 3842 and 3844 controllers, for example, are designed for this.

With an active bootstrap circuit, a narrow hysteresis band is usually preferred since the bootstrap voltage is regulated, and you do not want to dissipate too much power in the controller. The 3843 and 3845 controllers fit into this category.

 

Output Driver

PWM chip Output Drivers do not like to see inductive loads. If you are driving a gate drive transformer you may see a problem in the drive signal, manifesting itself as a double output pulse with a brief dropout period.

If you don't see this problem in the lab, don't assume it won't be there in production. The magnetizing inductance of the gate drive transformer can vary substantially over time, core lot, temperature, etc. Make sure you consider the worst case scenario before deciding you don't have this problem.

Unitrode recommends the use of catch diodes to eliminate the problem.
Another alternative is to parallel the transformer with a resistor.

 

Controller Generations

Watch out if you are trying to use Early Generation Controllers. You might be tempted to do this to use up some old lab stock or to get a special deal from your favorite component distributor.

Apart from subtle changes in tolerances on references, some of the early controllers can exhibit wide changes in clock frequency with temperature, making them unusable for many power supply applications. Make sure to check the suffix number on the parts to know whether you are getting the latest and greatest.

©copyright Ridley Engineering, Inc. 2007