Sherif:

The 'HC165 will present a QH to the MCU that is a copy of the H input to the shift register very shortly after the SH/LD pin has gone low. If there is a CLK L->H edge before the MCU reads the initial QH pin then that data is lost and replaced by the internal QG data.

As someone else pointed out your code sample also has the problem in that you do a CLR of the port pin to which the 'HC165 will be sending in the serial data. This CLR places that MCU pin as a low output. You should be doing a SETB of this port pin to place it into the High Input state so that the 'HC165 can validly drive the serial data signal to the MCU.

There is a possibility that this circuit gave the appearance of working due to the problem with the MCU pin being CLRed as I mention above. In this case the 'HC165 shift register QH output would be fighting with the MCU pin and there could be a resultant long rise/fall time on the signal line as the QH tries to change to the QG value right after the first clock time. Since your code reads the serial data pin within one instruction time of the CLK edge it is possible that you read (by total luck) the previous level of the QH pin before it transitions to the QG value.

If you still have your test board for this circuit it may be a good idea to re-look at it and adjust the code a bit. If you do that then you will likely gain some good insight.

Michael Karas