Thru Calibration in Power Integrity Analysis: What, Where, When, & Why

In power integrity analysis, we primarily concern ourselves with the impedance response when evaluating the frequency domain. Specifically, we focus on the magnitude of the impedance response and tend to neglect the phase of the impedance response. After all, the phase can be extracted from the magnitude:

• Flat impedance magnitude across frequency is purely resistive with 0° phase

• -20 dB/decade slope is purely capacitive with a -90° phase

• +20 dB/decade slope is purely inductive with a +90° phase

Figure 1 - Impedance Magnitude and Phase of Ideal R, L, and C Elements

This is true of the ideal DUT (Device Under Test) measurement. But the DUT is never measured in pure isolation. The low impedance power integrity measurement method of choice is the 2-port shunt-thru with a low frequency VNA like the Omicron Bode 500, Keysight E5061B, or Rohde & Schwarz ZNL. The VNA measures what is in between the two 50Ω ports, including any cables, connectors, mounting fixtures, ground loop isolators, injectors, amplifiers, and so on. All of which introduce electrical delays between the ports and the DUT.

What is the thru calibration?

To remove these electrical delays, a thru calibration is performed. The thru calibration measurement is made with everything necessary to measure the DUT, except for the DUT itself. That is, all the cables, connectors, mounting fixtures, etc. are included in the thru calibration measurement.

Note that multiple measurements will be performed for calibration, which means connecting and disconnecting several fixtures like mount boards. This means repeatability is imperative and obtaining very high consistency with components like the Picotest UC10 calibration kit and mount boards and the CTF fixture and mount boards are required.

Figure 2 shows a simple representation of a VNA measurement and the S21 phase response as simulated using Keysight ADS. In the leftmost circuit are the two 50Ω VNA ports with 0.5 ns delay transmission line elements to the DUT. The center circuit represents a thru calibration measurement with the DUT removed. The right circuit represents an ideal measurement of the DUT with no delay elements. The DUT itself is a resistor, inductor, and capacitor in series (SRLC), representing a typical capacitor with ESR and ESL parasitics.

Figure 2 - Delay Removal with the Thru Calibration Measurement

With a thru calibration measurement, the electrical delays can be removed by the VNA’s error correction, which effectively manifests as subtracting the thru phase from the measured phase. The rightmost S21 phase plot in Figure 2 shows the DUT phase obtained from the ideal circuit and the calculated DUT phase by a simple subtraction.

The thru calibration on a VNA also adjusts any scaling so that we can measure much closer to S21 = 1. This enables the measurement of higher impedances. To keep things simple, our simulations here focus only on the phase correction.

Also of note, is that our simple simulation is a 50Ω system. When the measurement system has impedance mismatches, the calibration becomes more complicated and different techniques are needed.

Where does the phase error occur in impedance measurement?

It may not be intuitive what causes the phase error by looking at the frequency domain alone. Especially with the phase wrapped. Consider how the S21 parameters are acquired for impedance measurement with the simple circuit in Figure 3.

Figure 3 - Simple VNA S21 Measurement Circuit

A sine wave source is applied from VS and the frequency is swept. The magnitude of S21 is the magnitude of V2/V1. If we only look at the magnitude of the V1 and V2 voltages, the transmission line elements (if matched to 50Ω) are transparent and the magnitude is not impacted by any electrical delays.

Figure 4 shows a snapshot of this circuit in the time domain, where VS is at 10 kHz. This image is to emphasize that what is actually measured at V1 and V2 are sine waves and that their magnitude and phase difference can be determined independent of one another.

Figure 4 - Oscilloscope View of V1 and V2 at 10 kHz

Let's direct our focus to the higher frequencies where the phase error occurs. Figure 5 shows some time domain snapshots at a few of the higher frequencies. Since the magnitudes of V2 at these frequencies are so small (< -50 dB), the magnitude of the V2 sine waves shown are normalized to 1 Vpp. The ideal V2 sine wave (with no phase error) is also plotted in the time domain for reference.

Figure 5 - Oscilloscope Views of V1 and V2 (with Normalized Magnitude) at Higher Frequencies Show 1 ns Phase Error

At each of the frequencies observed, there is a constant 1 ns shift in V2 from where it should be if there were no phase error. This is attributed to the two 0.5 ns delay transmission line elements in our simulation.

At 10 MHz, where the phase error is small, the period of the sine wave is 100 ns, which is very large compared to the 1 ns error. At 50 MHz, the 1 ns error begins to be significant compared to the 20 ns period. At 500 MHz, the period is 2 ns and the 1 ns time shift puts V2 180° out phase from where it should be.

So, the phase error is negligible at lower frequencies where 1/f is much larger than the electrical delay. The phase error becomes substantial at higher frequencies where the electrical delay is significant compared to 1/f.

When does the phase error matter?

There’s obviously a huge error in the S21 phase at higher frequencies, however the S21 magnitude shown in Figure 6 has no error.

Figure 6 - Ideal DUT Measurement vs with Delay - S21 Magnitude and Phase

For power integrity analysis, we often observe the impedance response of the DUT, which can be calculated from the S21 parameter as

When the impedance is plotted and compared against the ideal case, we see the same phase error. The magnitude again has no error, except for small periodic ripple at the higher frequencies, which becomes apparent when zoomed in, as shown in Figure 7.

Figure 7 - DUT Impedance Magnitude and Phase Calculated from S21

These ripples begin around 1 GHz and repeat at 1 GHz intervals, which exactly corresponds to the round-trip delay of our 1 ns transmission lines. This is the signature of standing waves in the measurement system.

Standing waves occur when forward and reflected signals interfere on a transmission line. A key property of transmission lines is that impedance varies periodically along their length when standing waves are present.

Even small reflections from the DUT create waves that travel back and forth through the 1 ns delay path. At certain frequencies, these reflections arrive back at the measurement point either in-phase (making impedance appear higher) or out-of-phase (making it appear lower). This creates a periodic ripple pattern.

Near 9 GHz for example, the ideal DUT impedance is 6Ω, but the measurement with the delays varies by about +/-1Ω (a +/-17% variation) as standing waves oscillate between constructive and destructive interference.

We can verify this standing wave mechanism by varying the delay length. Figure 8, compares the impedance responses calculated from S21 for measurements where the round-trip delay between ports is 0.5, 1, and 2 ns.

Figure 8 - Calculated S21 Impedance Responses with Varied Delay Lengths

As predicted, the ripple frequency is inversely proportional to the delay:

• 0.5 ns delay: ripples at 2 GHz intervals (1/0.5 ns = 2 GHz)

• 1.0 ns delay: ripples at 1 GHz intervals (1/1.0 ns = 1 GHz)

• 2.0 ns delay: ripples at 0.5 GHz intervals (1/2.0 ns = 0.5 GHz)

This confirms the impedance errors are directly caused by standing waves in the electrical delay path, and the thru calibration eliminates these errors by removing the delay itself.

The S21 magnitude remains smooth because it only measures transmission. The standing waves however, affect the calculated impedance which is dependent on phase since S21 is a complex value. So, phase-related effects can introduce errors in extracted impedances even when the S21 magnitude appears perfect.

For measurements out to frequencies well below the first standing wave resonance (in the 1 ns case: below ~500 MHz where λ/2 = 1 ns), the impedance errors from standing waves are negligible. In these cases, if you only need impedance magnitude and not phase, you can obtain accurate results without thru calibration - provided your fixtures are consistent and you understand your measurement uncertainty.

However, as frequency increases and standing wave effects become significant, thru calibration becomes essential for accurate impedance extraction. The thru calibration eliminates these standing waves by mathematically removing the electrical delay.

Further, if you are measuring capacitance, inductance, or the imaginary part of the impedance by using cursors on a VNA, you are likely to get incorrect values when the impedance phase is incorrect. Figure 9 shows an example inductor measurement from Steve Sandler [1] using the Bode 500.

Figure 9 - Inductance Found by Cursors is Incorrect When the Phase is Incorrect

The impedance at 100 MHz is 791 mΩ which calculates to an inductance of 1.26 nH. However, by applying the VNA's inductance math function at the cursor, it reports 601 pH. This is because the phase at 100 MHz is 151°, where it should be near 90°. The VNA is likely calculating the inductance strictly from the imaginary term given that L = Z/jω.

Why thru calibration matters for accurate measurements

It’s always important to understand what you’re measuring and how it’s measured to know what to expect and how to interpret the results. As we continue to push power integrity analysis measurements into the μΩ and pH, we need high confidence that what we’ve captured reflects reality.

At Signal Edge Solutions, we’re continually pushing measurement boundaries and therefore tend to approach our measurements from first principles. The thru calibration makes for a good example, because it’s a standard procedure that can easily be performed without fully understanding the underlying mechanism.

By visualizing the problem in the time domain and simulating it in Keysight ADS, we can see exactly how electrical delays create phase errors that scale with frequency and standing waves that create periodic impedance ripples. This fundamental understanding helps identify when calibration is critical.

For measurements at low frequencies (well below the first standing wave resonance), you may be able to work without thru calibration if you only need impedance magnitude. In fact, when you know what you're doing, it's possible to get an accurate measurement without any calibration.

A critical point here is that calibration corrects small imperfections in the setup, not fundamental problems. It is not meant to correct large errors introduced by a bad cable, poor contact resistance, or over-tightened connectors. This is why we emphasize high-quality, repeatable fixtures like the Picotest UC10 and CTF mount boards. They ensure your VNA calibration works.

The real value in understanding these mechanisms is developing the intuition to troubleshoot when measurements look wrong, to identify whether you're seeing fixture problems versus calibration issues, and to become a more proficient VNA user and engineer.

Learn More at DesignCon 2026

The information presented here was learned from conversations with Steve Sandler and his 2025 Omicron Symposium presentation. We’ll also be giving a tutorial with Steve Sandler and Heidi Barnes at DesignCon 2026 covering this topic and much more. Stop by and see us at booth 601!

Have a Measurement Challenge?

Whether you’re working on PDN impedance characterization, signal integrity verification, or any other high-speed measurement challenge, we have the expertise to help:

• Oscilloscope ground loop isolation: J2115A

• VNA ground loop isolation: J2113A and J2102B

• UC10 cal kit and mount boards

  • Custom-made 3.5 mm connectors for high accuracy, high bandwidth, and repeatability

  • Calibrator and mount board made from highly-controlled PCB materials for excellent matching and repeatability

• CTF cal kit and mount boards

  • High-quality design and highly-controlled materials for measurement repeatability

  • Connector-less mating of mount boards to the fixture

  • Built-in coaxial isolators

  • DC bias available

• Complete solutions: Ground Bounce Measurement Bundle

Contact us today for a consultation on your critical measurement challenges.

References

1. Calibration, Embedding and De-Embedding - Achieving Highly Accurate Impedance Results by Steven M. Sandler - Picotest at the 14th Power Analysis & Design Symposium 2025, April 9, 2025
   

2. Next-Level Measurements: Easier, Faster, and More Accurate with the New Picotest Component Test Fixture
   

3. Impedance Transformation, All About Circuits