Firstly pumping the system down is often a variable. The pumping is balanced against the various system leaks, real and virtual, as well as the gas load from the outgassing of the substrate and system. This outgassing is not a constant. The rate of water outgassing from the surfaces depends on the surface roughness and the tie and temperature that the system was kept at atmospheric pressure before the pump down started. The more water vapour that was condensed onto the surfaces the longer it will take to pump the system down. The longer the time between cleaning the surfaces the higher will be the surface porosity and the greater capacity for collecting water vapour. It is common practice to pump the system down to a given base pressure and then start the process. There are two changes that starting the process will cause, one is to provide a source of heat and the other is to start unwinding the roll of substrate, which will release more water vapour as fresh surfaces become exposed to the vacuum. The heat will come from either the radiation and condensation heat load from any evaporation source or from the energetic particles and condensation heat load from sputtering sources. The heating will cause an initial spike as the water vapour at the surface is released more quickly as a result of the heat energy. Once this surface water vapour has been released the heat continues to cause an increased gas load as the more deeply adsorbed water takes time to diffuse to the surface and be released.
The heating from any radiation from the source and as a result of the condensation heat load will cause many surfaces to increase in temperature throughout the deposition cycle. The temperature of the surfaces, such as deposition shields, will continue to rise in temperature and will progressively increase the heat load to the substrate as the amount of radiated heating from the shields increases with the increasing shield temperature. In some systems a number of these shields are water cooled and so the heat load to the substrate is reduced or limited but there are usually other surfaces where the temperature rise is unconstrained.
If there is any part of the process where gas is introduced and is controlled by the pressure it is possible that the pressure may remain stable but the composition of the gas present may vary. The background gas will be a proportion of the total and if the quantity of background gas varies the proportions will vary in the process area. In some cases this may not matter. In fact some systems where argon is added for plasma treatment the background gas can actually be a benefit with the oxygen from the water vapour helping in the plasma cleaning process. Where argon and oxygen are added as a controlled proportion the addition of water vapour becomes an uncontrolled variable. Again this may not matter for many processes. Where it might have an adverse effect is in some critical reactive deposition processes where the final coating stoichiometry needs to be controlled to a few parts per million or finer.
The system temperature as a whole will rarely reach a stable temperature. Even individual items that are cooled may take a long time to come to equilibrium. Some metalizing processes may only take an hour to complete whereas some sputtering processes may take days to complete the process. The short metallization process makes it very difficult to try to stabilise the temperature of the system fast enough. The very long process times make the process similar to some of the flat glass coating systems that run for weeks. These glass coating systems often take a couple of days or so to stabilise enough for the product to become reproducible.
The cooling of shields and sources needs to be done with care and ideally the cooling should also be symmetric about the substrate centreline. If there is any imbalance in cooling this can change the heat load across the substrate width and in any reactive deposition process this can affect the reactivity and potentially the stoichiometry of the coating. This includes the cooling of the deposition drum as well as any magnetron sputtering sources where I have seen some sources where the cooling enters at one end and leaves at the other end and so there is a temperature gradient down the length of the source. This will be reflected in the front surface temperature of the target and this in turn can affect the secondary electron emission. Depending on the magnet type used this type of temperature gradient can also affect the magnet field strength, which in turn affects the sputtering rate.
So even before we get to looking at the stability of pressure measurement, coating properties and any feedback control loop from the a measurement of the coating to the source power or wire feed or both, we can see that the process takes place under anything but constant conditions. What it is very hard to establish is how much coating variation is caused by any of these various variables. Until more of these variables are measured it will be impossible to know which could be controlled better to improve the reproducibility of the coatings.