Double difference carrier phase observation

Hi everyone,

I am new in the world of GPS and I am using two ublox GPS LEA-6T precision timing receivers for my project for the relative positioning. currently I am doing the offline processing by using RINEX files and both of the receivers are stationary. I am investigating how much the baseline will change with time using a basic geometric model for relative positioning.

I took the double difference carrier phase measurement as following.
1. appointing a common satellite (hi)
2. DDCP = (CProver[prn] - CPref[prn]) - (CProver[hi] - CPref[hi])

The application requires that the initial position of both the receivers are known. In my case both of the receivers are static so I did the following 2 different cases to get rid of the ambiguities as I have not gone to the step of ambiguity resolution yet.

1. I took the known GPS receiver position and found the arbitrary initial ambiquity value by finding the distance between known receiver and satellite positions. although these ambiquities are not accurate but I just want to see the baseline variation with time so will this variation be effected due to this?

2. If above case is not correct then in other case I placed both of the antennas closed together within one wave length distance so that both receivers experience the same ambiquities so they should be canceled in differencing.
I hereby require help on following two issues.

1. I want to ask that if both of the GPS receivers takes measurements at full GPS second then is there still a need of extrapolation for receiver data synchronization for differencing?

2. Secondly should not it be that Double Difference observation remains constant with time or they vary a bit due to receiver noises or multipaths but they should not have increased or decreased linearly with time? I am having the case that although I took observation in open area with clear sky, double differences linearly increase with time. what can be the possible cause for this linear drift? Is it something ok and can be handled in more advance solutions like kalman filtering or should I resolve this issue in the basic model?

Bundle of Thanks.
Mohsin