Signal integrity is critical for reliable operation of high-speed interfaces, and the MachXO5-NX family defines overshoot/undershoot limits on input pins to protect long-term device reliability. Those limits are expressed against the Unit Interval (UI). This article explains how UI is determined for static and aperiodic control signals — such as EN, PGD, and RST — that assert only once and have no fixed operating frequency, and it corrects a common misreading of the 20 µs condition.
For periodic signals, where UI is derived from the signal frequency, see the companion article Overshoot and Undershoot Specs Explained (UI Calculation + Examples), which walks through worked periodic examples.
The Unit Interval is defined as UI = 1/f. For a periodic signal it comes straight from the frequency — for example, a 500 kHz square wave gives UI = 1/500 kHz = 2 µs. The overshoot/undershoot table then expresses each allowable voltage level as a percentage of one UI, and that percentage is the duration the signal is permitted to remain at that level.
"100% UI" therefore means the full UI duration — a time budget, not a fixed pulse width. Importantly, the %UI rules only apply when UI is less than 20 µs (equivalently, f greater than 50 kHz, since 1/20 µs = 50 kHz).
A frequent misreading is that the "UI < 20 µs" condition means the overshoot or undershoot pulse itself is 20 µs wide. It does not.
The 20 µs is a threshold for applicability of the table, not a pulse width or bandwidth. When UI < 20 µs (f > 50 kHz), the percentage-based timing rules apply. When UI ≥ 20 µs (f ≤ 50 kHz), the table no longer applies in the same way, and designers should instead confirm the peak stays within the absolute maximum ratings. Real overshoot transients are typically only a few nanoseconds wide regardless.
The correct UI depends on whether the signal repeats:
Periodic signals derive UI from frequency (UI = 1/f). Faster signals have a shorter UI and therefore a tighter absolute time budget for the same %UI allowance; slower signals allow more time for the overshoot to decay, as long as UI stays under 20 µs. The companion article shows this with two 500 kHz examples — one that exceeds its allowed duration and fails, and one that stays within it and passes.
Static and aperiodic signals — EN, PGD, RST and similar control pins — assert only once and have no fixed frequency, so there is no 1/f to compute. For these signals the specification fixes UI = 20 µs. In effect, the applicability boundary itself becomes the reference UI, giving the %UI allowance a consistent basis even when no clock is present. This is the key difference from the periodic case.
To check compliance, compare both the peak level and its duration against the table using the fixed 20 µs UI:
The voltage and timing figures above are illustrative to demonstrate the method — they are not measured results.
Both "UI < 20 µs" and "100% UI" refer to the allowance window, not the pulse width. For periodic signals, compute UI = 1/f (see the companion article). For static and aperiodic signals such as EN, PGD, and RST, use the fixed UI = 20 µs. In either case, convert the applicable %UI into a time budget (%UI × UI), then confirm your measured peak and its duration fit within it.