Instruments¶
RFlect doesn't talk to instruments directly — it reads their saved output. Below is the verified hardware/software the test suite has touched, but the format definitions in File Formats are what really matters.
Chambers¶
| Vendor | Model | Tested |
|---|---|---|
| Howland Company | 3100 Antenna Chamber | yes (primary reference) |
Any chamber producing WTL V5.02 / V5.03 .txt |
— | yes (format-level) |
Open an issue with a sample file if your chamber's export differs.
Vector network analyzers¶
| Vendor | Software / format | Tested |
|---|---|---|
| Copper Mountain | S2VNA .csv |
yes |
| Any 2-port VNA | Touchstone .s2p |
yes |
| Any 2-port VNA | CSV with S2VNA columns | yes |
! Stimulus(Hz), S11(dB), S21(dB), S21(s) are the column names RFlect looks for.
Simulation tools¶
| Vendor | Format | Read | Write |
|---|---|---|---|
| CST Studio | Far-field .txt |
yes | — |
| CST Studio | Farfield Source .ffs |
— | yes (via convert_to_cst MCP tool) |
Gain standards used in active calibration¶
The active-calibration routine has been verified with:
- Howland BLPA (broadband log-periodic)
- Howland HORN
The gain-standard's calibrated bands determine which TRP-cal frequencies route to "Missing Data" — e.g. the BLPA-19 has uncalibrated gaps at 960–1500 MHz, 1610–1710 MHz, and 2170–2300 MHz. See tests/test_active_calibration.py.
What you do not need¶
- A specific OS — RFlect runs on Windows, Linux, and macOS (built from source)
- A specific instrument vendor — formats matter, brands don't
- A specific Python version beyond 3.11+
- Network access — AI features are optional and gracefully disabled when unconfigured