Thermal Testing of Motor Drivers

(continued from previous page)

Temperature tests are performed on Roundabout PCB version 1.5 and version 2.0 (to be released with the second edition of Intermediate Robot Building). As previously discussed, PCB v2.0 has additional copper which we'll test to see if it helps the motor drivers run cooler.

Measuring motor driver chip temperature with thermistors embedded into DIP sockets on a Roundabout printed circuit board.

The motor drivers are IXYS IXDN404PI dual non-inverting power MOSFET driver ICs in a DIP package. They're rated at 4 amps of peak current (to start a motor rotating), with a continuous output rating of 1 amp.

The motor drivers operate from 4.5 V to 35 V, with a typical voltage of 18 V. They're driven by a 3 amp laboratory desktop power supply at 8.8 V (unloaded) to simulate a 9 V alkaline battery under load. A 33% reduction in resistance (and thus better performance) could have been achieved if the chips were run at their typical voltage (18 V), but that would not match most hobby robots.

The motor drivers are tested with a single chip, two chips in parallel (Dual), and two chips in parallel with a heat sink made of copper foil (Dual HS).

Test Procedure

Three thermistors are monitored in parallel by the temperature testing board:

• Motor driver chip being tested under power. (Inner socket.)
• Motor driver chip powered but not driven to act as a control and indicator of board temperature. (Outer socket)
• Ambient room temperature attached to an aluminum block.

In all tests, the room temperature measured 73.2 to 74.2 degrees Fahrenheit, with 73.5 being the most common. For this reason, I believe that the slight changes in the room's temperature did not affect the test results.

The testing process is:

1. Power on the Roundabout board for 10 seconds without commanding the tested motor driver chip to power the motor. That is, both motor drivers are idle.
2. Push the microcontroller button to command the inner motor driver chip to apply continuous power to the load (various resistances of power resistors, one at a time).
3. Allow the motor driver chip to supply power for 3 minutes and 10 seconds (the 200 second mark of testing). However, this step is stopped and the next test step is executed immediately if the measured temperature reaches 185° Fahrenheit. In this case, the chip is considered "overheated". (By the way, newer data sheets indicate the chip can go as high as 257°, but this was not the limit in my reference datasheet at the time of testing.)
4. Power off the circuit board and wait one minute.
5. Turn on a box fan aimed at the front of the circuit board (allowing some air to flow below the circuit board) until the chip temperature reaches 75° Fahrenheit.

Temperature Profile of Single Motor Driver under Various Loads

This first graph shows five complete tests with a single motor driver chip on a v1.5 (no extra copper) PCB with loads:

• 600    Ω or  15 mA @ 8.8 V (Roundabout's motors lifted off of the ground)
• 150    Ω or  59 mA @ 8.8 V
•  39    Ω or 226 mA @ 8.8 V
•  19.5  Ω or 446 mA @ 8.7 V (Desktop power supply drops 0.1 V)
•   9.75 Ω or 882 mA @ 8.6 V (Desktop power supply drops 0.2 V)

Note: The actual current flowing through the circuit ends up lower than the maximum current stated above, due to the resistance of the motor driver chip.

Chart motor driver chip temperature by load.

Interesting findings:

• Except when overheating, the chip cools off faster than it heats up, even without a fan.
• The fan cools the chip faster initially, but the temperature improvements begin to decline as the chip approaches room temperature.
• Based on the curve leveling off, the unassisted single IXYS404PI motor driver chip can probably drive a 19.5 ohm load continuously. The driver needs to dissipate 0.675 watts, which is just under the published limits for the DIP package of 0.730 watts at room temperature. (I measured the voltage delivered to the motor at the 200 second mark and calculated that the current was 346 mA.)

Temperature Profile of Variations of a Motor Driver

Now let's test variations of the circuit boards (v2.0 has more copper) and stacked motor drivers (dual, dual with heat sink) under the maximum load of 9.75 ohms.

Chart motor driver chip temperature based on single, dual, heatsink, and copper layer.

Interesting findings:

• The single chip still overheats on the v2.0 circuit board, but takes longer to heat up and is faster to cool down. The extra copper on the PCB has an effect.
• The dual chip (stacked) has much better performance. Calculations indicate that the resistance of the double driver is half the single driver, confirming the theory of parallel resistances. However, the chips are approaching their combined thermal limit with 1.37 watts. That can only be sustainable if the heat is equally divided between the chips. I'm going to say "no."
• The little strip of copper foil ("Dual HS") further increases the performance of the motor driver. This motor driver delivers 0.683 amps at 77% efficiency with less heat than the dual chips without foil. This level of output could likely be sustained continuously at room temperature.
• Wow! Combining a dual chip with copper foil on a circuit board with copper filled areas doubles the power output compared to a single chip with the same load (6.88 V * 0.705 amps = 4.8504 watts of electrical motor power vs. 4.8 V * 0.492 amps = 2.3616 watts). Yet the heat profile equals the single chip with half the load.

Temperature Affect of Copper Fills on Other Components

As you've seen, the filled copper areas do a good job of allowing heat to flow away from the chip under load. But, does this heated copper have a measureable negative influence on other components on the board?

During the previous tests, the second motor driver chip was powered on but not commanded to supply power to a motor. The following graph shows the temperature of this idle chip.

To be fair, note that the vertical scale of this chart is much smaller than the previous chart. There is up to a one degree of difference in starting temperatures that may fool your eyes into thinking some peaks are higher than others.

Chart temperature of an idle single motor driver chip (B) when various motor driver arrangements are tested under load.

I didn't expect to see any change in temperature for the idle chip. Yet, there is a multiple degree increase in temperature where the heat had to travel through the board and into the idle chip itself to be measured by the thermistor. Or, the surrounding board had to be radiating that much heat into the thermistor that is buried inside the DIP socket.

• The peak temperature occurs a minute after the circuit board is turned off. It takes that long for the heat to travel across the board.
• After adjusting for the starting temperature, the extra copper on the v2.0 board does not appear to place a significant thermal burden on the other components. Although more testing would be helpful in reaching a solid conclusion, it appears the copper fills are radiating most of the heat rather than simply redistributing it to other components.

Conclusion and Useful Tips

So what have we learned?

• Filling in empty spaces on your printed circuit board with copper areas is significantly beneficial for heat dissipation. This is true even if the copper fills are not electrically connected to ground.
• In open air within specified wattage, the chips cool off faster than they heat up. Giving a robot a brief rest might be an effective heat reduction technique. In fact, if the robot monitored the motor driver chips with a thermistor in their DIP sockets, the robot could rest the motor drivers somewhat intelligently.
• Stacking multiple CMOS/MOSFET chips on top of each other actually works. If you need a moderate performance boost, this is a relatively simple adjustment that can be made without redesigning a robot or circuit board.
• Not only do heat sinks prevent a chip from becoming damaged due to overheating, but also they improve performance of CMOS/MOSFET chip by avoiding or delaying heat that increases resistance. Use heat sinks on components that get hot.

All of that being said, these results are largely academic for MOST hobby robots. The voltage drop and power usage of most electronics has decreased to the point that voltage regulators run fairly cool. And, if your motor driver is getting hot, you should switch to a highly-efficient power MOSFET H-bridge (see Chapter 9 of Intermediate Robot Building) or even a power bi-polar H-bridge instead of hacking with dual chips or heat sinks.

However, if your competitive robot needs a slight edge, or your existing design almost meets your needs, you now know how to get a little more out of your chips without too much extra effort.