There can be a number of issues as to why instability in a reading. Below is a link that as you scroll down you will find a lot of information.
A pH meter is a very sensitive volt meter. A voltage is generated by a pH electrode when it is placed in a solution.
There are two wires that are in the measuring circuit. They are separated by the glass pH bulb (indicating membrane) and the reference junction. The link below brings you to a pH electrode page and if you scroll down there is a diagram comparing a single junction vs. double junction pH electrode. Understanding the parts will help in determining the problem
For a reading to drift there has to be a change to the measuring circuit. These changes can be from:
1) Probe is not properly hydrated. A hydrated layer forms on the glass pH membrane over time. It takes 3-4 hours to fully form. It is one of the reasons that we store the pH electrode in a storage solution. A dry electrode will generate a completely different voltage than a hydrated one. So if the probe is dry you will read one pH value and then it will drift over time until the hydration layer is formed.
2) Related to storage solution is also the junction potential. The junction (barrier from reference cell to sample) has its own potential that is part of the measured potential. It is important that a liquid junction potential be maintained. Also if the junction is allowed to dry then some of the reference electrolyte will precipitate clogging the junction. Any clogging of the junction will result in erratic/unstable readings.
3) The probe had a slight build up on the glass at the beginning that came off as it was in the sample. Again any changes to the glass will result in a different measured potential
4) Most likely, a build up is forming on the glass as it is left in the sample. If the probe was hydrated and clean in the beginning a a slight layer forms on the bulb as it sits in solution then the reading will drift because three is a change to the bulb that has not been accounted for.
5) The sample is not conductive enough to carry the flow of current. Most likely not the problem but it is still important to know that a standard pH electrode needs 100 uS/cm of EC to work. Without salts present then the electrical current cannot flow. RO water equals 20-30 uS/cm. Lake Michigan is around 300 uS/cm. Groundwater varies by location. In Chicago suburbs upwards 1500 uS/cm. If using RO water with Bru'n additive then this is not a problems since the salts will increase EC.
If you take a paperclip and short the BNC connector on the meter then it is possible to do a default calibration that can be helpful in determining if there is a build up on the probe.
1) Place one end of the paperclip into the center hole and then wrap the other end around the outside of the connector. This will cause a short that = 0 mV
2) Calibrate the meter to pH 7.01 with the connector shorted. Theoretically pH 7.01 = 0 mV (known as offset). It is the deviation from this point that we compensate for by calibration.
3) After calibrating to pH 7.0 then reattach the pH electrode and place in fresh pH7.01 buffer. Record the pH value.
4) Determine the mV from the pH value displayed. The probe should read between pH 6.5 and 7.5 (+/-30 mV). Outside this value indicates the probe might need to be cleaned.
A pH electrode theoretically generates 59.16 mV/pH unit at 25 oC. This is known as slope. A pH of 6 = +59.16 mV and a pH of 8 = -59.16 mv. So each 0.1 pH = 5.916 or basically 6 mV. A pH reading of 6.5 is 0.5 pH units away from pH 7 which means that 0.5 pH x 6 mV = 30 mV. If the meter displayed pH 6 then that would be a 60 mV offset error.
It is important that the pH 7 buffer is fresh. If the solution is not fresh then it is possible that it is actually not pH 7 but a different value. It would make the test above irrelevant since there is not a known standard to compare to.
Lastly, manufactures will allow an offset of +/-60 mV or greater to be accepted for a pH 7 calibration. Accuracy will suffer if the value is outside +/- 30 mV. The same for slope. when a slope is less than 90% or 53 mV/pH (59.16 x 0.9) then accuracy will suffer.
If you have questions then feel free to message me.