L
Admin
A spectrum analyser + tracking generator (or a classic scalar network analyser) is still the fastest way to tune a filter, check a cable, or walk a PA chain — and a reliable way to invent a 1–2 dB “insertion loss” that is really the jumper you flexed. This is about **scalar |S21|-style habits**, not full two-port VNA error models.
**1. What the instrument is actually doing**
- Tracking generator (TG) follows the SA LO so the detector sees a swept CW through the DUT. You get a **magnitude** path (plus whatever detector / RBW / attenuator chain you set) — not phase, not a full 12-term correction.
- Habit: label every plot **scalar |S21| (TG+SA)** or “scalar analyser,” and write RBW, attenuator, and whether normalisation was thru-only or thru+short/load if the box supports it. Do not paste it into a report as “VNA S21.”
**2. Normalisation is not calibration theatre — freeze the path**
- A thru normalisation (TG → cable → SA) removes the **static** path at the moment you pressed Normalise. Flex the cable, swap an adapter, change SA attenuator, or warm up 20 minutes and the “flat” line is a different measurement.
- Symptoms the normalisation is stale: ripples that move when you touch the jumper; a filter IL that improved after lunch with no tune; a “0 dB” reference that is 0.5 dB high at band edge only.
- Habit: re-normalise after any mechanical change, and again after the SA/TG has been on long enough that the first 15 minutes of drift are over.
**3. Dynamic range and the filter that looks too good**
- SA input attenuator + TG level + RBW set how deep a notch you can trust. A deep cavity notch measured with too little attenuator can be mixer compression or LO breakthrough looking like “60 dB rejection.”
- Habit: raise SA attenuator one step and demand the notch depth stay put; if rejection collapses, you were measuring the analyser, not the filter. Same spirit as mixer-level discipline on spur hunts.
- Wide RBW smears narrow notches; narrow RBW slows the sweep and can make a tuning hand look noisier than the DUT.
**4. When scalar is enough — and when to stop and get a VNA**
- **Scalar wins:** quick filter tune, relative IL compare of two cables, “is this jumper broken,” production go/no-go against a golden scalar template.
- **Stop and use a VNA:** match / return loss that drives the tune, phase / group delay, multiport isolation with real directivity, anything you will publish as absolute S-parameters, and any argument where 0.2–0.5 dB is the product margin.
- Habit: if the scalar IL and a later VNA |S21| disagree by more than your scalar uncertainty, believe the VNA **after** you have ruled out connector torque and a bad thru — do not average the two numbers.
**5. A short scalar / TG sanity checklist**
1. Thru normalise with the **production** adapters and cable dress you will keep.
2. Flex test: gentle cable move should not move the trace outside your gate.
3. Attenuator step test on a deep notch / high-IL DUT.
4. Spot-check one known pad or golden filter against its dated scalar (or VNA) trace.
5. Log TG level, SA attenuator, RBW, and normalisation timestamp next to the plot.
**Discussion**
On your last filter tune or cable screen with a TG/scalar box, what lied first — a stale thru normalisation after a cable flex, SA attenuator inventing rejection, or trusting scalar IL as if it were VNA |S21| — and where do you draw the line to switch instruments?
— rfforum 编辑部 / Editorial Team
**1. What the instrument is actually doing**
- Tracking generator (TG) follows the SA LO so the detector sees a swept CW through the DUT. You get a **magnitude** path (plus whatever detector / RBW / attenuator chain you set) — not phase, not a full 12-term correction.
- Habit: label every plot **scalar |S21| (TG+SA)** or “scalar analyser,” and write RBW, attenuator, and whether normalisation was thru-only or thru+short/load if the box supports it. Do not paste it into a report as “VNA S21.”
**2. Normalisation is not calibration theatre — freeze the path**
- A thru normalisation (TG → cable → SA) removes the **static** path at the moment you pressed Normalise. Flex the cable, swap an adapter, change SA attenuator, or warm up 20 minutes and the “flat” line is a different measurement.
- Symptoms the normalisation is stale: ripples that move when you touch the jumper; a filter IL that improved after lunch with no tune; a “0 dB” reference that is 0.5 dB high at band edge only.
- Habit: re-normalise after any mechanical change, and again after the SA/TG has been on long enough that the first 15 minutes of drift are over.
**3. Dynamic range and the filter that looks too good**
- SA input attenuator + TG level + RBW set how deep a notch you can trust. A deep cavity notch measured with too little attenuator can be mixer compression or LO breakthrough looking like “60 dB rejection.”
- Habit: raise SA attenuator one step and demand the notch depth stay put; if rejection collapses, you were measuring the analyser, not the filter. Same spirit as mixer-level discipline on spur hunts.
- Wide RBW smears narrow notches; narrow RBW slows the sweep and can make a tuning hand look noisier than the DUT.
**4. When scalar is enough — and when to stop and get a VNA**
- **Scalar wins:** quick filter tune, relative IL compare of two cables, “is this jumper broken,” production go/no-go against a golden scalar template.
- **Stop and use a VNA:** match / return loss that drives the tune, phase / group delay, multiport isolation with real directivity, anything you will publish as absolute S-parameters, and any argument where 0.2–0.5 dB is the product margin.
- Habit: if the scalar IL and a later VNA |S21| disagree by more than your scalar uncertainty, believe the VNA **after** you have ruled out connector torque and a bad thru — do not average the two numbers.
**5. A short scalar / TG sanity checklist**
1. Thru normalise with the **production** adapters and cable dress you will keep.
2. Flex test: gentle cable move should not move the trace outside your gate.
3. Attenuator step test on a deep notch / high-IL DUT.
4. Spot-check one known pad or golden filter against its dated scalar (or VNA) trace.
5. Log TG level, SA attenuator, RBW, and normalisation timestamp next to the plot.
**Discussion**
On your last filter tune or cable screen with a TG/scalar box, what lied first — a stale thru normalisation after a cable flex, SA attenuator inventing rejection, or trusting scalar IL as if it were VNA |S21| — and where do you draw the line to switch instruments?
— rfforum 编辑部 / Editorial Team
