Hijacking an ESC

A straight test of RPM vs PWM showed the dreaded stair stepping.


A test of minutes/mile using the most precise, slowest RPM feedback still showed stair stepping above RPM's where it oscillated. So either the Tamiya ESC wasn't precise enough or the mechanics had some voodoo. It's probably a limitation of all ESCs whether brushed or brushless. The next solution was to install the H-bridge from the G-buggy.





With some widening of vias, a cheap toy becomes a hobby grade ESC. Unfortunately, that overheated instantly & shut down.




The next step was to break open the Tamiya further & try to access the H-bridge. 4 MOSFETs control the brushed motor wires. They're all 4806NG N channels with 2.5V threshold voltage, 20V GS, 30V DS maximums. Low side gates are connected through an unknown LC circuit.


Q20's gate goes to Q14's collector.
Q22's gate goes to Q16's collector.
Q23's gate goes to Q17's collector.
Q21's gate is connected through a tented via to an unmarked Q.




A pad marked 18V pulls all MOSFET collectors high through 50k resistors.
All BJT bases are connected to microcontroller outputs through 1k resistors. The low side BJT bases are also connected to Vdd through 40k resistors, so they're on by default, keeping the low side MOSFETs off.
The 18V pad is not on by default, but switched. There is another 18V which is always on & fed by a charge pump.

Q5 turns on the 18V pad by pulling down the base of a P transistor on command from the microcontroller. With Q5 removed & its collector grounded, 18V stayed on. It was time to disconnect the 1k's.

With the 4 resistors removed, it did as expected, with the high MOSFETs staying on & the low MOSFETs staying off. The final step was connecting the BJT bases to the new microcontroller through 1k resistors. Turning on the H-bridge required logic low. A current limiting power supply was key to making this work without frying it.





With all that soldering, there was finally a smooth RPM curve.


In the 2Hz feedback mode, speed control was not so lucky, wit h a stair step still in the key 9.5-8.5 minute/mile range. It may have been the feedback falling behind the target speed.


In the 10Hz feedback mode, it was bang on the target speed.



Loaded it with 10Ah of batteries to try to get more range. It liked flipping over.


The 1st drive with the hijacked ESC was bang on 10min/mile to 10m20s/mile, but it took setting it at 11min/mile. The 1st battery went 7.75 miles without headlights. The power calculation seems wrong, since it showed 12W on the downhill & 14W on the uphill. At that rate, it should have gone 4 hours, which it didn't.

The erroneous power calculation was still lower than the stock Tamiya ESC. The hacked H-bridge probably is more efficient.

Among the bugs, it tended to go into a mode where the low speed steering was turning real fast. Resetting the gyro got it out of this mode. Sometimes the heading spun wildly. Resetting the gyro got it to stabilize. The gyro hasn't been as stable as hoped. A magnetic compass guided heading is a good idea.