This has been puzzling me for a while too. L298 and SN75440 designs are required to have external diodes to deal with back-EMF, and people seem to use physically large ones ( dfrobot.com/image/data/DFR0004/Layout.pdf ). The TB6612FNG like the L293D has internal diodes but the size disparity makes me wonder about on-die diodes. The sparkfun discussion on the SN75440 has the dry remark:
The “TI tech support answer” Limor Fried got was “I never trust internal kickback diodes”, which doesn’t say anything one way for the other.
I’m abusing/wasting motor controllers a different way right now (uh, PC fans; doing a single high-side drive per channel through an RC filter, so I have my own problems and suspect I need to clamp at least once externally.) But for people using the H-bridge the right way, yet under somewhat abusive testing conditions, would external diodes be worthwhile, or as useful as racing stripes, or somehow harmful?
This is yet another reason that small linear regulators are almost never suitable for driving motors. I highly recommend using a motor supply rated for the stall current of your motor and a separately regulated logic supply.
Hm. Should I stop using the 9v transistor battery as a test supply for small loads? (EDIT: yeah, in the FAQ that they probably won’t work for real motors, but it doesn’t say “don’t”.)
BTW, don’t feel compelled to say anything about my app; “I bet you’ll be ordering another controller shortly” is as fine response as any. If I expected a design service I’d pay for it.