Budgeting Pulse Width Error Across a Digital Isolator

A digital isolator can delay both edges of a pulse while also changing its width. These are separate effects. If a pulse represents a measurement, PWM command, or timing window, the difference between edge delays can matter more than the overall latency through the device.

Read the correct timing parameters

The TI ISO7721 data sheet defines pulse width distortion from the difference between low-to-high and high-to-low propagation delays. It separately specifies propagation delay and other skew quantities. Use the parameter that matches the information carried by the signal.

Record supply voltages on both sides, output load, input edge conditions, and temperature range. A typical delay at one supply does not establish a worst case pulse width error over every supported combination.

Relate edge delays to pulse width

For a noninverting path, output high time equals input high time plus the falling-edge delay minus the rising-edge delay. If both edges are delayed equally, the pulse moves in time without changing width.

A hypothetical 100 ns input pulse with a 4 ns difference between those delays changes width by four percent. The same absolute error is a much smaller fraction of a 10 microsecond pulse. Evaluate the shortest informative pulse, not only the nominal repetition frequency.

Include the receiving threshold

Measure timing at thresholds consistent with the specifications and the actual receiver. Unequal output rise and fall slopes can make apparent width depend on the selected voltage crossing. Include additional buffers or pullups in the end-to-end path where they affect the receiving decision.

Keep pulse width distortion separate from channel-to-channel skew. A pair of signals transmitted through different channels may have acceptable individual widths but an unacceptable relative arrival time.

Verify the encoded quantity

Transmit representative minimum, middle, and maximum pulse widths, including the intended spacing between pulses. Measure both input and output with characterized channel timing so the instrument’s own mismatch does not dominate the result.

Repeat at the required supplies and loads. Check that a very short pulse remains within the isolator’s supported input timing rather than assuming the distortion specification applies below its valid range.

Save edge thresholds, timing conditions, width error, and any calibration method used by the receiver. A useful budget states both latency and information error. It does not accept a pulse measurement interface merely because its isolated output toggles at the correct average rate or because the propagation delay is small compared with a full PWM period.