Why Your RF Numbers Change in the Shielded Room
A switch that measures 0.35 dB insertion loss on the bench measures 0.42 dB in the chamber. Nothing about the part changed. What changed is the environment, and the difference is large enough to break a link budget if you do not know where it comes from.
This is a short list of the effects that move an RF measurement when you move it into a shielded room, and how to tell them apart from a real problem in the hardware.
1. Ambient coupling disappears, and your baseline moves
On an open bench, a coax run picks up whatever is in the air: cell towers, Wi-Fi, a colleague's soldering iron, the switching supply on the next bench. Most of it is small, but it is not zero, and it does not add coherently. When you move into a shielded room, that contribution goes away, and the part suddenly looks slightly better or slightly worse depending on whether the leakage was adding to or subtracting from the signal at the detector.
Practical consequence: your reference measurement has to be redone in the room. A calibration done on the bench and applied in the chamber carries an error that no amount of averaging removes.
2. Cable routing changes, and phase goes with it
The chamber forces you to route cables differently: through a bulkhead, around a turntable, along a ground plane. A phase-stable cable is only phase-stable along the path it was characterised on. Bend radius, coiling, and how close the cable runs to a metal surface all change the electrical length slightly.
At 18 GHz a 10 mm change in electrical length is on the order of 2 degrees. If you are measuring phase balance on a DPDT switch, that is not noise, it is your result.
3. The room changes the load, and the load changes the reading
A shielded room is a cavity. At frequencies where the room dimensions are a few wavelengths, it supports resonances, and the field at the antenna is not uniform. This matters most when the measurement depends on what is connected to the unused ports: isolation, return loss, and anything involving a termination.
Absorber on the walls and floor reduces it, but does not eliminate it below a few hundred MHz. If your isolation figure is measured in a room and quoted without the room's absorber treatment, you are quoting the room as much as the switch.
4. Grounding schemes get decided for you
In a chamber, everything bonds to a single reference, usually at one end only. On a bench, instruments are grounded through their mains leads and the rack. The difference shows up as common-mode current on the outside of the coax shield, which is a well-known cause of a measurement that changes when you touch the cable.
The test for this is simple: put a clamp-on ferrite around the cable near the instrument and see if the reading moves. If it does, you were measuring a ground loop, not the part.
5. Temperature stabilises, and that is usually good news
A chamber is thermally stable in a way a bench is not. Insertion loss and isolation both drift with temperature, and a bench with a door opening every ten minutes has a few degrees of cycling that show up as scatter. In a chamber the scatter drops, which makes real problems easier to see, and also makes marginal parts look worse than the bench suggested.
What to do about it
- Re-establish the reference in the room, with the same cables and the same torque, before measuring anything.
- Log the cable routing, and re-characterise when it changes. A photo of the route is not overkill.
- Terminate unused ports with loads that are matched at the frequency of interest, in both environments, so the comparison is like for like.
- Check for ground loops before trusting any number that moves when you move.
- Quote the conditions next to the number, in the datasheet and in the test report. A number without a room, a termination and a frequency is a rumour.
Where I sit
I work on RF switches, attenuators and terminations at DB Design / MeiXun in Wuxi, so most of my measurement time is spent on the manufacturing side of this problem. We build coaxial switches from SPDT to SP12T, fixed attenuators for level setting, and coaxial loads that are matched across the band, which is what most unused-port measurements actually depend on.
If a part measures differently in two rooms, the part is the last thing I would change.
Top comments (0)