I went ahead and yanked that wimpy regulator, connected the input to output and fed the board with 5V. Everything worked great. Some measurements showed that with both motors running, along with the wifi and the USB hard drive, the system was pulling 1.6 amps, well above the capability of the Raspirobot’s regulator even with a heatsink. By turning off the LEDs and the hard drive, current consumption dropped to 1.2 amps with both motors freely running. The lowest I saw was 900 mA with the motors off, the wifi running and the disk inactive. In short, this board is marginal or inadequate for powering a robot and the RPi. I’d recommend using an LM323T regulator instead. It’s pin compatible and will handle up to 3 amps. At 1.6 amps it will be cruising, even without a heatsink.
As far as why I have a USB hard drive – the SD card is a miserable choice for a hard drive substitute. The write times are slow and non-deterministic and it is prone to corruption. SD cards were designed to hold pictures in camera, not to be operated as general purpose read-write drives. In my setup, I’ve modified the kernel parameters in cmdline.txt and some startup files so that the RPi boots from the SD (in read-only mode), then mounts its root file system from the hard drive. That gives me fast, reliable storage with no chance of SD card corruption. But it does add 200 ma or so to the power budget, as well as moving parts.
For the final robot, I’ll be using a RAM disk-based system where the SD card will be the boot device and the root file system will be mounted on a RAM disk loaded once from the SD card. But this means that I won’t be able to use Raspbian because it’s too bloated. This will require a custom-built distro. I’ve done this on x86 and gotten the kernel, root file system and app into 6 MB. With this approach, there will be no moving parts, no chance for SD card corruption and minimal power consumption.
Cheers all.