step_kit: a channel's absolute bandwidth from a fast-edge pulser

CH1What it needs

• Python 3.8 or newer with numpy, on a PC that reaches the scope's SCPI port (TCP 5555). Tested against the DHO800, DHO900 and MHO900 command set.
• A fast-edge pulser whose edges are well under the channel's rise time; we use a Leo Bodnar LBE-1322 (30 ps edges). Set it once in the maker's own utility to about 0.5 Vpp at a frequency that is not a round number, its sync input left open.
• Its output straight into the channel at 50 Ohm through a short SMA-to-BNC cable. A 1 MOhm input needs a 50 Ohm feed-through.

python3 step_kit.py selftest
python3 step_kit.py measure --host SCOPE --ch 1
python3 step_kit.py analyse records.npz

The README has the wiring, every option, what measure sends in order, and the known limits. METHOD has the arithmetic step by step and the self-test results.

CH2Files

MIT licence (LICENSE).

The same sums are in SHA256SUMS; check them with sha256sum -c SHA256SUMS. The kit's own tests run against a fake SCPI scope: python3 -I test_step_kit.py (15 tests).

CH3What we measured

The bandwidth of an MHO934 with the band limit OFF, under Rigol's firmware and under NovaOS, measured on 2026-10-17 with the same pulser and cable: the results, the method and the raw runs are on NovaOS vs stock. Those runs were taken with our bench tool's step method, whose analysis this kit carries unchanged: on synthetic records the two give exactly the same numbers.

Choose the pulser's frequency so its edges sweep the sample interval several times a record. Ours, 1,003,017 Hz, swept it only about 1.7 times in each 250,000-point record Rigol's firmware returns, and on synthetic records with so few sweeps the -3 dB reads up to about 1 % low at 4 GSa/s. The kit counts the sweeps and warns below five; if it warns, move the frequency by a few hundred hertz.