About RRAM – 5 Pulse measurement and impedance matching

There is a mystery that many RRAM researchers face. "When measured with a DC Sweep, the On/Off ratio was good and the operating voltage was nice at 1V, but it won't budge with a 100ns pulse. And when I raised the voltage, it suddenly died."

To put it coldly, DC measurement data is only half the truth.This is because memory in actual computers operates with short pulses in the nanosecond (ns) range.

The critical difference between DC and Pulse environments lies in 'Heat' and 'Signal Reflection'. In this article, we will reveal the voltage-time dilemma that occurs during high-speed operation and practical know-how to solve impedance mismatch problems that distort signals.

1. The Critical Difference Between DC and Pulse: Heat Accumulation

Filament formation and rupture in RRAM are a joint effort of the Electric Field (E-field) and Joule Heating.

  • DC Sweep: Voltage is increased slowly (in ms). Since the current flows for a long time, heat accumulates sufficiently.Thanks to this, filaments grow well (Set) or melt well (Reset) even at relatively low voltages.
  • Pulse Mode: Voltage is applied only for a very short time (in ns). There is insufficient time for heat to generate.
  • Result: A device that turned on at 1V in DC requires 2V or 3V to barely turn on in Pulse mode.

Voltage-Time Dilemma

The switching speed(𝛕) of RRAM and the applied voltage (V) have an exponential relationship.

τ=τ0exp(V/V0)\tau = \tau_0 \cdot \exp(-V/V_0)

This means "to operate faster (with shorter pulses), you must apply exponentially higher voltages."This is why the device does not respond if you apply a pulse with a low voltage relying only on DC data.

2. The Identity of Ghost Signals: Ringing and Overshoot

The biggest enemy of pulse measurement is not the device itself, but the Cable. When the frequency exceeds a few MHz, the wire acts not just as a simple conductor but as a Transmission Line.

At this time, if Impedance Matching is not achieved, the signal reflects and returns.

Situation: Mismatched Case (Most Labs)

  • Source: Pulse Generator (Output impedance 50Ω)
  • Cable: BNC/SMA cable (characteristic impedance 50Ω)
  • Load: RRAM device (impedance Rdevice: KΩ ~ MΩ) -> Incorrect!!

The signal travels along a 50Ω cable and suddenly hits a high-resistance (~MΩ) wall (RRAM). The energy, having nowhere to go, is 100% reflected,combined with the incoming signal, creating Overshootwhere the voltage jumps to double, and Ringing.

  • Risk: I sent 2V, but momentarily 4V is applied to the device. -> Hard Breakdown
  • Solution: Reflection must be eliminated.

3. The Solution: Impedance Matching

To prevent signal reflection, the RRAM side must also accept it as 50Ω. However, since we cannot change the RRAM resistance, we perform matching circuit-wise.

Method 1. Utilizing the Oscilloscope's 50Ω Mode (Parallel Connection)

The easiest way is to connect the oscilloscope in parallel with the RRAM and set the input impedance of the oscilloscope to 50Ω instead of High-Z (1MΩ).

  • Path: PGU -> Splitter -> (1) RRAM / (2) Oscilloscope (50Ω Termination)
  • Principle: Since the end looks like 50Ω from the signal's perspective, reflection is significantly reduced.

Method 2. Using an Active Probe

This is the most accurate but expensive method. Active Probes have very low input capacitance, measuring high-frequency signals without distortion. The key is to place the probe tip as close to the device pad as possible.

Active probe

4. Practical Measurement Setup Guide (Setup Diagram)

RRAM 측정 회로
RRAM measurement circuit

This is the equipment connection sequence for successful pulse measurement.

  1. Pulse Generator Unit(PGU): Equipment that fires pulses. Set the Rise/Fall Time as fast as possible (must be faster than the device speed).
Capacitive current during pulse operation
  1. Bias Tee (Optional): Used when AC pulses need to be superimposed on a DC bias.
  2. RRAM Device: Keep the cable length as short as possible. (Parasitic capacitance increases as the cable gets longer.)
  3. Oscilloscope: You must visually check the 'Real Voltage' applied across the RRAM. Do not trust the voltage displayed on the PGU screen.
    • Channel 1: Current monitoring through the device (Measure voltage drop by attaching a resistor in series)
    • Channel 2: Input pulse monitoring

5. Conclusion: Clean Waveforms Yield Good Results

DC measurement shows the 'result', but Pulse measurement shows the 'process'.

If the RRAM doesn't work or keeps dying, check the Oscilloscope Waveform first before blaming the device.

  • Is the waveform dancing? -> Do Impedance Matching.을 하세요.
  • Did you apply the pulse with the same DC voltage? -> Increase the voltage or extend the pulse width.

Just remembering these two things will elevate your RRAM data quality to Publication Quality.

References: Tektronix

Similar Posts