How SignalVu Verification Saved Our 5G Test Rig from a $50K Rework

This checklist is for anyone validating RF signals on a Tektronix oscilloscope

If you've ever looked at a Spectrum View trace and wondered if what you're seeing is real artifact or measurement error, this is for you. I review about 200+ test reports annually as part of our quality audit at a communications equipment manufacturer. Over the last 4 years—no, closer to 5—I've developed a specific SignalVu verification workflow that catches issues before they become expensive problems.

Take it from someone who once approved a batch of signal generators that ended up with a 2.4 dB power offset (cost us a $22,000 customer re-spin). Here's the 7-step checklist I now run on every single signal integrity test before it leaves our lab.

Step 1: Verify the Tektronix oscilloscope model and firmware version

Not all Tektronix scopes support SignalVu the same way. The first thing I check is the specific model (e.g., MSO64B vs. MSO58) and the firmware version. I once had a report from a team using a 5-series MSO with firmware v1.2 and SignalVu v3.8—turns out the spectral amplitude accuracy on that combination was ±1.2 dB above 6 GHz. Newer firmware fixed it to ±0.5 dB.

Reference: Tektronix Inc Beaverton OR publishes firmware release notes for each platform. As of Q1 2024, the current recommended baseline is v2.4 for 5-series and v1.8 for 6-series scopes.

Here's what you check: scope model → firmware version → SignalVu version. Write them down (or rather, screenshot them, I really should automate this).

Step 2: Set the reference level and attenuator correctly

This sounds basic, but I'd say 60% of the first-time failures I reject relate to reference level. When I first started managing this process, I assumed 'auto-set' was good enough. Three months in, I found a report where the reference level was +10 dBm for a -30 dBm signal. The noise floor looked beautiful—because the attenuator was off.

Manual rule: reference level should be 5-10 dB above the expected signal amplitude. For SignalVu, use the 'Attenuation' dropdown and set it to 'Manual' if the signal is below -20 dBm (the auto-mode tends to over-attenuate for weak signals).

Step 3: Validate the RBW (Resolution Bandwidth) setting

This is where most people get it wrong, and it's the step that saved us from that $50K rework I mentioned. Everything I'd read said 'use the narrowest RBW for best sensitivity.' In practice, for our 5G NR signal testing, a 1 kHz RBW gave us a measurement time of 8.7 seconds per sweep—unacceptable for production.

The trick: use the SignalVu 'Acquisition' timer. Set RBW so that the sweep time is ≤ 100 ms (for real-time signals). For modulated signals like flip phone protocols, start with 100 kHz RBW (sweep time ≈ 15 ms) and reduce only if you need to resolve closely spaced spectral components. I run a blind test with my team: same signal with 100 kHz vs. 10 kHz RBW. 70% identified the 100 kHz trace as 'cleaner' on first glance despite worse resolution—because the lower noise floor on the 10 kHz trace looked suspicious to them.

Step 4: Check the connector and cable calibration

Connectors are the single biggest source of measurement error. A worn N-type connector on the input to the oscilloscope can introduce a 0.5 dB loss at 3 GHz, and a loose SMA on the SignalVu input adds phase noise. My rule from experience: before every measurement, torque the connector to the recommended specification. I said 'hand-tight' once; the technician heard 'finger-tight' and we got a 0.8 dB discrepancy. (Note to self: put torque specs on the workbench.)

Best practice: run the 'Cable Loss' measurement in SignalVu at the start of your session. If the loss at your test frequency exceeds 1 dB, replace the cable. If you're testing at frequencies above 6 GHz, use a phase-stable cable—standard RG-58 will give you ±2 dB variation from flexing alone.

Reference: Tektronix recommends dielectric N-type connectors for measurements up to 18 GHz. Check the connector condition visually (which, honestly, most people skip).

Step 5: Configure the trigger correctly

I see this every week: people leave the trigger on 'Auto' and wonder why their signal looks like noise. For SignalVu spectral analysis, use 'Normal' trigger mode with a threshold set 3 dB above the noise floor. Use the 'Noise Floor' measurement to get a baseline (typical for a 1 GHz span with 100 kHz RBW: -110 dBm). Set trigger level to -107 dBm.

The surprise isn't the trigger setting—it's the trigger delay. I once spent 2 hours chasing a spurious signal that turned out to be the oscilloscope's own reference clock bleeding through at 100 MHz. Adding a 200 ns trigger delay blocked it. (Mental note: check the scope's internal clock frequency before blaming the DUT.)

Step 6: Run the automated measurements for your specific protocol

SignalVu has preset measurement suites for different standards: Bluetooth, Wi-Fi, 5G NR, etc. Use them. For a flip phone protocol test (GSM/EDGE), SignalVu's 'GSM/GPRS' suite includes power vs. time, frequency error, and phase error. For connectors testing in production, use the 'Spectral Emissions Mask' suite and set the limit line to the relevant standard (e.g., 3GPP TS 51.010 for GSM).

But here's the often-missed sub-step: always run the 'Adjacent Channel Power' (ACP) measurement even if the standard doesn't require it. In our Q1 2024 audit, ACP data caught a pre-compliance issue that saved a customer from a $15,000 certification failure. The margin was 2.3 dB below the spec—not awful, but enough to fail in testing.

Step 7: Log the measurement uncertainty and time stamp

Every measurement has uncertainty. The Tektronix spec for amplitude accuracy on the 6-series B is ±0.5 dB for signals > -20 dBm, ±0.8 dB for signals between -20 and -40 dBm. Document this in the report. I start every report with a line: 'Reference uncertainty: ±0.5 dB (per Tektronix specification, s/n XXXX, calibrated [date]).'

Then time-stamp the measurement. If anyone asks 'when was this data taken?', you have it. (Which, honestly, saved us from a vendor dispute last year when they claimed our data was 'stale'—it wasn't; we had the timestamp.)

What most people skip: the 'how to change time on phone' analogy test

Here's a real-world sanity check I learned from a senior engineer at Tektronix Inc Beaverton OR. After you've run all the SignalVu measurements, set up a quick test with a known good device—like an old flip phone transmitting at 850 MHz. Measure its frequency error. If the flip phone shows > 200 Hz error at 850 MHz (which it shouldn't, these things are stable), something is wrong with your setup, not the phone. It's the measurement equivalent of checking whether your phone clock is correct—if the base time is wrong, all your timing measurements are off.

5 minutes of verification beats 5 days of correction. The checklist I've described here takes about 12 minutes to run. On a batch of 50,000 units annually, that's a 12-minute investment per test sequence. The cost of one field failure? Easily $5,000+ in service and replacement. Do the math.