Only way to defeat that flow is a blocking diode on the Vcc input, only allowing the normal power into the chip, but pay attention to the minimum Vcc.
Careful use of a mosfet may also help to reduce this voltage drop, but pay attention to the substrate's parasitic diode influence.
Thanks a lot.
[1] Today I found a weird phenomena in my old DSP system, which uses a Texas Instruments TMS320C31 DSP Starter Kit (DSK).
[2] The DSK is connected to PC via parallel port for programming and debugging.
[3] A 12-bit digital output board (DOB) is interfaced to the DSP Starter Kit.
[4] Each digital output is buffered with a 74HCT14 and a LED is used as indicator.
[5] When a 12-bit data (let's 1100 0000 0000 0000) is sent to the DOB, the LEDs connected to bit-11 and bit-10 are ON.
[6] Then, I turn OFF the DSP system (i.e. the DSK and the DOB are OFF and the computer is still ON). When I turn ON the DSK and the DOB again, the LEDs connected to bit-11 and bit-10 are still ON.
[7] It seems that the system has a "non-volatile" memory which can remember the last digital output state.
[8] Step [3] to [5] are repeated. Similar to step [6], but this time I turn OFF the DSK, DOB, and the computer. Then, all are powered up again. The LEDs connected to bit-11 and bit-10 are OFF.
[9] I really don't know where does the "memory" come from?
[10] Step [3] to [5] are repeated. Similar to step [6], but this time, while keeping the computer alive, I turn OFF the DSK and the DOB. Then, I disconnect the parallel port cable from the DSK and reconnect it. Then, both DSK and DOB are powered up again. The LEDs connected to bit-11 and bit-10 are OFF.
[11] It looks to me that the "memory" is from the "parallel port".
[12] Why can this happen?
Thanks.