All projects

Project · 2025

Analog design · High-speed PCB · VNA

High Speed Active Oscilloscope Probe

A 1 GHz probe developed from a differential OPA858 simulation into single-ended hardware, followed by stability and input-impedance measurements.

Overview

This project explored a compact 1 GHz active oscilloscope probe, beginning with a differential architecture and continuing through two single-ended hardware iterations. The work combined circuit simulation, spectrum analysis, and VNA measurements to understand both amplifier stability and the probe’s input impedance.

Circuit Design

The original design was supposed to be differential, with two OPA858 input stages, and a third stage to convert the difference to a single-ended output for the oscilloscope. The OPA858 is a decompensated amplifier with a gain-bandwidth product above 5 GHz (which would later turn out to be very important). LTspice schematic of the original differential active-probe circuit using three OPA858 amplifiers.

The original differential OPA858 design in LTspice.

The simulated response was essentially flat through the hundreds of megahertz, reaching -3 dB at approximately 1.10 GHz.

LTspice frequency response of the differential probe, showing a minus-three-decibel bandwidth of approximately 1.10 GHz.

Simulated differential-probe response, with the -3 dB point at approximately 1.10 GHz.

To simplify the hardware, I reduced the design to a single-ended signal path. The 971 kΩ/14 kΩ input divider and its parallel capacitors provide frequency-compensated attenuation, the 453 Ω/75 Ω feedback network gives the OPA858 a closed-loop gain of about 7 V/V, and the 50 Ω source-and-load network produces a nominal 20:1 probe response.

LTspice schematic of the simplified single-ended OPA858 active probe.

The simplified single-ended circuit used for the hardware prototype.

Unfortunately, the very large GBW also made layout critical. The undamped trace between the compensation capacitor and the input of the amplier caused the amplifier to oscillate at 1.184 GHz; the spectrum analyzer measured the narrow tone at approximately -4 dBm.

Spectrum-analyzer measurement of the active probe oscillating at 1.184 GHz and approximately minus 4 dBm.

The layout-induced 1.184 GHz oscillation measured on the first PCB.

Hardware iterations

To fix this, I made a second version with an open loop BUF602 buffer I had on hand. Unfortunately, this would ultimately mean that the input impedance wouldn’t be as good, since the input divider plays a big role in reducing the input capacitance.

A small hand-etched copper PCB for the active probe.

The etched probe PCB before assembly.

The first active-probe prototype on a green PCB, with a coaxial output and twin probe pins.

The first prototype PCB.

The second handmade active-probe prototype inside a soldered metal shield.

The second, handmade revision.

Input-impedance investigation

I used a VNA to measure the input impedance. I measured the impedance with nothing connected to the VNA to find the parasitic capacitance before measuring with the probe, and then corrected to get only the probe impedance in Excel.

The shielded active probe connected to a LiteVNA on a workbench.

Bench setup for the VNA measurement.

Zmeasured=ZVNA∥Zprobe,Z=Z01+S111−S11Z_{\text{measured}} = Z_{\text{VNA}} \parallel Z_{\text{probe}}, \qquad Z = Z_0 \frac{1 + S_{11}}{1 - S_{11}}

VNA impedance measurement of the input fixture without the probe attached.

Input impedance without the probe, showing the VNA-input parasitic.

VNA impedance measurement with the active probe connected.

Input impedance with the probe connected.

The second revision’s input impedance was also compared with an equivalent capacitance before damping was added. The measured curve did not behave like a simple capacitor at high frequency.

Impedance magnitude sweep from 100 MHz to 1 GHz.

An impedance-magnitude sweep used during the input-model investigation.

Measured probe impedance compared with the impedance of an equivalent capacitance.

The second revision before damping, compared with an equivalent capacitance.

Modeling the resonance

I reproduced the impedance feature in LTspice with an equivalent RLC network. That model identified input inductance as the cause of the resonance and provided a direction for damping the next revision.

LTspice plot of an equivalent RLC network reproducing the probe's impedance resonance.

Equivalent-network simulation used to isolate the effect of input inductance.