Measuring with .meas

Turning a curve into a number

meas runs after an analysis has finished. It reads the saved vectors and prints one line per result, such as tdelay = 6.931367e-07. Inside a .control block the result also becomes a vector, so a later let or meas can use it.

The crossing forms take a count. RISE=2 means the second time the signal rises through the value, FALL=1 the first time it falls, and CROSS=3 the third crossing in either direction. LAST picks the final one. TD=20u ignores everything before 20 µs.

Most forms also print where they found their answer. vglitch = 7.843741e-01 at= 2.507000e-06 says the glitch peaked at 0.7844 V at 2.507 µs, just as the current pulse ended. A TRIG ... TARG line prints both crossing times, trig= and targ=, which is the quickest way to see which edges a delay was measured between.

A meas line cannot compute an expression of vectors inside .control. Build the vector first with let, then measure it. The lab does that for the power the source delivers, let p = -v(in)*i(V1). The minus sign undoes the SPICE convention that a delivering source has a negative current.

Name the crossing

The hook's line asked for the first rise of v(out) past 0.5 V, and the glitch supplied it. Two small changes ask the question that was meant.

meas tran tbad TRIG v(in) VAL=0.5 RISE=1 TARG v(out) VAL=0.5 RISE=1

meas tran tdelay TRIG v(in) VAL=0.5 RISE=1 TARG v(out) VAL=0.5 RISE=1 TD=20u

The rise time in the lab uses the other fix. Its trigger, TRIG v(out) VAL=0.1 RISE=2, skips the glitch's rise by counting it. The result, trise = 2.197201e-06, is the 10 % to 90 % time of E1.3, ln 9 · τ.

Prefer TD= when you know when the event starts, and a count when you know how many edges come first. Never trust "the first one" on a waveform you have not looked at.

Check that it worked, and check it on a known answer

A meas that finds no crossing prints a line ending in failed! and creates no vector. A later let that uses it stops with an error, as below, and every line after it that needs the value fails too. Read the log for failed, and never let a script reuse an old number.

meas tran t15 WHEN v(out)=1.5
let x = t15*2
* the log:
* meas tran t15 when v(out)=1.5 failed!
* Error: RHS "t15*2" invalid

A script usually turns a number into a verdict with if. Inside .control, comparisons are written gt, lt, ge and le. A condition on a vector that a failed meas never created counts as false, so the else branch runs. Write the test so that only a good number can reach PASS.

if tdelay lt 0.65u or tdelay gt 0.75u
  echo FAIL tdelay
else
  echo PASS tdelay
end

if tdelay gt 0.65u and tdelay lt 0.75u
  echo PASS tdelay
else
  echo FAIL tdelay
end

Then test the measurement on a circuit whose answer you know. Charging a capacitor C to V through any resistor takes energy C·V² from the source. Half is stored in the capacitor, and the other half is burned in the resistor, whatever its value.

The lab measures esrc = 9.99837e-10 from the source and eheat = 4.99841e-10 in R₁: C·V² = 1 nJ and half of it, to 0.02 %. Without the glitch source I2, change R₁ to 2 kΩ: the charge takes twice as long, and esrc still reads 0.9999 nJ. The small shortfall is the 1 ns input edge: a ramp charges a little more gently than a step.

Try it

Run it and read the seven lines. tbad is negative, tdelay and trise match the RC milestones of E1.3, vglitch shows how high the glitch went, and e5 shows what a short window misses.

Now delete the I2 line and rerun. tbad turns into the same number as tdelay, and trise fails: with no glitch there is no second rise. Change its count to RISE=1. Then change R1 to 2k and predict tdelay, esrc and eheat before you run.

Check yourself

* count the edges
V1 in 0 PULSE(0 1 10u 2n 2n 20u 50u)
R1 in 0 1k
.control
tran 10n 100u
meas tran t2 WHEN v(in)=0.5 RISE=2
.endc
.end