L
Admin
[Technical article / FAQ, rfforum Editorial Team]
Phase and group-delay claims often get blamed on the DUT while the test cable is still settling from the last bend. Above a few GHz, flex, temperature, and connector torque can move phase more than the device under test you are trying to freeze.
1. What the cable actually owns
- Semi-rigid and formable lines still show phase walk after a bend; the "settled" number may take tens of seconds, and some cables never fully recover to the previous path
- Phase-stable marked cables reduce wander but do not erase torque or adapter-stack residuals — they change the budget, they do not delete it
- A cable that looks fine for |S21| can still dominate arg(S21) and group delay, especially on long fixtures or through adapters
2. Freeze list before you publish a phase number
- Cable family / length / bend history on the same traveler line as the phase or delay result
- Torque (or engagement) at every mating face, and whether the cable was re-routed between cal and DUT
- Temperature at the cable and at the DUT — a few °C on a long run can look like a DUT drift
- Whether you re-check thru residual after the final cable dress, not only after ECal / kit cal on the bench
3. Habit
Treat cable dress as part of the method. If moving the cable by a hand-width changes phase by more than your pass/fail margin, you do not yet have a DUT number — you have a fixture number. Archive a short "cable moved, phase delta" note next to every published group-delay or phase-match claim.
Discussion: on your last phase disagreement, did the fix land in cable family, torque/dress, or a true DUT change — and do you require a post-dress thru check before freezing S21 phase?
— rfforum 编辑部 / Editorial Team
Phase and group-delay claims often get blamed on the DUT while the test cable is still settling from the last bend. Above a few GHz, flex, temperature, and connector torque can move phase more than the device under test you are trying to freeze.
1. What the cable actually owns
- Semi-rigid and formable lines still show phase walk after a bend; the "settled" number may take tens of seconds, and some cables never fully recover to the previous path
- Phase-stable marked cables reduce wander but do not erase torque or adapter-stack residuals — they change the budget, they do not delete it
- A cable that looks fine for |S21| can still dominate arg(S21) and group delay, especially on long fixtures or through adapters
2. Freeze list before you publish a phase number
- Cable family / length / bend history on the same traveler line as the phase or delay result
- Torque (or engagement) at every mating face, and whether the cable was re-routed between cal and DUT
- Temperature at the cable and at the DUT — a few °C on a long run can look like a DUT drift
- Whether you re-check thru residual after the final cable dress, not only after ECal / kit cal on the bench
3. Habit
Treat cable dress as part of the method. If moving the cable by a hand-width changes phase by more than your pass/fail margin, you do not yet have a DUT number — you have a fixture number. Archive a short "cable moved, phase delta" note next to every published group-delay or phase-match claim.
Discussion: on your last phase disagreement, did the fix land in cable family, torque/dress, or a true DUT change — and do you require a post-dress thru check before freezing S21 phase?
— rfforum 编辑部 / Editorial Team
