L
Admin
A step or programmable attenuator (relay turret, mechanical rotary, or PIN/solid-state) is how most benches set SA input level, sensor range, and ATE path loss — and one of the easiest ways to invent a 0.2–0.5 dB “DUT drift” that is really a dirty contact or a stale flatness table. A few habits that keep the steps honest.
**1. Say which physics you bought**
- **Mechanical / relay steps:** low IL and decent power handling; failure mode is **contact** — repeatability walks with cycle count, humidity, and how hard the last switch slammed.
- **PIN / solid-state:** fast, no relay wear; failure mode is **bias, temperature, and compression** — steps look fine at −20 dBm and lie at the top of the rated range.
- Habit: label the path **relay step** vs **PIN step** in the traveller. Do not paste a PIN flatness table onto a relay box after a repair.
**2. Flatness is a table, not a sticker**
- “0–70 dB in 10 dB steps, ±0.5 dB” is usually a **band + temperature + VSWR** claim, not a promise that every 10 dB click is 10.00 dB at your frequency into your mismatch.
- Habit: keep a dated **per-step, per-band** check against a VNA or a characterised pad stack at the frequencies you actually use. Re-run after any connector change on the attenuator ports.
**3. The 0.1–0.3 dB that is wear or bias**
- Relay: same step commanded twice should overlay inside your gate (often ~0.05–0.2 dB setup-dependent at microwave). If it walks, cycle the contacts a few times and re-check; if still open, schedule contact service or retire that decade — do not “average” it into the DUT cal.
- PIN: change case temperature or supply bias 10% and demand the step stay put; if IL moves, you were measuring the driver, not the RF film.
- Habit: after a suspicious DUT IL shift, **command the same attenuator code again** (or toggle away and back) before you redesign the product.
**4. Power, mismatch, and the SA story**
- Steps ahead of an SA or diode sensor set both level and **mismatch**. A relay attenuator with poor output VSWR at a particular decade can move a power reading even when the step IL is correct.
- Habit: for absolute power, prefer a short characterised pad + sensor, or correct with a mismatch factor; do not assume every attenuator setting is a perfect 50 Ω thunk.
- For SA ranging: verify that raising input attenuator (instrument or external) moves the noise floor and the spur the way the manual predicts — same spirit as mixer-level discipline.
**5. A short acceptance gate**
1. Zero / thru reference with attenuator at 0 dB (or minimum) and production adapters.
2. Spot-check each decade you use at low, mid, and high band against a dated table.
3. Repeatability: same step three times; scatter inside the agreed gate.
4. One hot / one cold spot-check if the box lives in an unconditioned rack.
5. Log relay vs PIN, max input power, and the flatness file revision next to the plot.
**Discussion**
On your last ATE or SA setup, what moved first — a relay decade that would not repeat 0.1 dB, a PIN step that walked with temperature, or trusting a brochure flatness as if it were a cal factor — and do you re-verify steps on a calendar or only after a connector incident?
— rfforum 编辑部 / Editorial Team
**1. Say which physics you bought**
- **Mechanical / relay steps:** low IL and decent power handling; failure mode is **contact** — repeatability walks with cycle count, humidity, and how hard the last switch slammed.
- **PIN / solid-state:** fast, no relay wear; failure mode is **bias, temperature, and compression** — steps look fine at −20 dBm and lie at the top of the rated range.
- Habit: label the path **relay step** vs **PIN step** in the traveller. Do not paste a PIN flatness table onto a relay box after a repair.
**2. Flatness is a table, not a sticker**
- “0–70 dB in 10 dB steps, ±0.5 dB” is usually a **band + temperature + VSWR** claim, not a promise that every 10 dB click is 10.00 dB at your frequency into your mismatch.
- Habit: keep a dated **per-step, per-band** check against a VNA or a characterised pad stack at the frequencies you actually use. Re-run after any connector change on the attenuator ports.
**3. The 0.1–0.3 dB that is wear or bias**
- Relay: same step commanded twice should overlay inside your gate (often ~0.05–0.2 dB setup-dependent at microwave). If it walks, cycle the contacts a few times and re-check; if still open, schedule contact service or retire that decade — do not “average” it into the DUT cal.
- PIN: change case temperature or supply bias 10% and demand the step stay put; if IL moves, you were measuring the driver, not the RF film.
- Habit: after a suspicious DUT IL shift, **command the same attenuator code again** (or toggle away and back) before you redesign the product.
**4. Power, mismatch, and the SA story**
- Steps ahead of an SA or diode sensor set both level and **mismatch**. A relay attenuator with poor output VSWR at a particular decade can move a power reading even when the step IL is correct.
- Habit: for absolute power, prefer a short characterised pad + sensor, or correct with a mismatch factor; do not assume every attenuator setting is a perfect 50 Ω thunk.
- For SA ranging: verify that raising input attenuator (instrument or external) moves the noise floor and the spur the way the manual predicts — same spirit as mixer-level discipline.
**5. A short acceptance gate**
1. Zero / thru reference with attenuator at 0 dB (or minimum) and production adapters.
2. Spot-check each decade you use at low, mid, and high band against a dated table.
3. Repeatability: same step three times; scatter inside the agreed gate.
4. One hot / one cold spot-check if the box lives in an unconditioned rack.
5. Log relay vs PIN, max input power, and the flatness file revision next to the plot.
**Discussion**
On your last ATE or SA setup, what moved first — a relay decade that would not repeat 0.1 dB, a PIN step that walked with temperature, or trusting a brochure flatness as if it were a cal factor — and do you re-verify steps on a calendar or only after a connector incident?
— rfforum 编辑部 / Editorial Team
