Mechanical Installation

Determine the cross-sectional plane in your microwave installation where the Homer should be inserted (Load Plane). For best performance, the Homer should be installed as close to load (applicator) as possible. For more details, see the topic How Long Is Too Long? below.
Separate the microwave system at Load Plane such that the Homer can fit in the space.
Insert the Homer such that the DUT label on its nameplate faces the load (applicator, workspace) and the GEN label faces the generator (magnetron).
Secure the Homer using appropriate screws. Use washers on both sides. Use all screws.
If your Homer is a water-cooled model, install water-cooling hosepipes.

How Long Is Too Long

Suppose for a moment that an applicator has a constant input reflection coefficient G_A. Let the applicator be connected to an Autotuner via a section of essentially loss-free waveguide of length L. The question arises: how long can the waveguide can be without adversely affecting the tuning performance?

The reflection coefficient at the connecting waveguide input is

where f is the generator frequency, f_c is the waveguide cutoff frequency, l_g is the guide wavelength at f, and c₀ is the speed of light. This is the load reflection coefficient G_L that the Autotuner percieves at its output and attemps to match.

While the connecting waveguide does not affect the magnitude of the reflection coefficient, i.e., |G_L| = |G_A|, the phase at the waveguide input varies as

If the frequency of the generator is unstable and fluctuates between f_min and f_max, as is often the case with magnetrons, then the phase of G_L varies in an interval of width

The variation is proportional to the waveguide length L. A numerical example for WR975 and WR340 waveguides and the generator frequency instability of ±0.1 % is presented in the table below.

Waveguide	WR975	WR340
Nominal frequency	915 MHz	2.45 GHz
Frequency instability	±0.1 %	±0.1 %
Frequency variation	±915 kHz	±2.45 MHz
Waveguide length L	10 m	3 m
Phase variation width Dj_L	59°	50°

Hence, even if the applicator reflection coefficient G_A is constant, the Autotuner percieves a varying reflection coefficient, and hence must keep adjusting the tuning stubs for matching. This may lead to incessant stubs movements or even stub swapping. An additional inconvenience is that if a match has been achieved at a certain instant, this match may be perturbed sooner than the next matching attempt, thus resulting in a reduced overall matching efficiency.

Returning to the title question: the maximal acceptable length L depends on the amount of Dj_L you are willing to tolerate. As a rough estimate, Dj should not exceed about 30°, but perhaps even 60° may be acceptable.

It is worth noting that this problem is virtually non-existent when using the highly frequency-stable solid state power generators.

Simulation

To simulate this effect, use the HoSim simulation program. On the DUT Page of the Homer Model window, check Add Line Length and set a desired intermediary waveguide length. On the More page, set a frequency uncertainty. Then you can start measurement without and with autotuning and observe the effect.