I used a DATAQ data acquisition board to capture the current pulses so I would have a record of the pulse width and amplitude. The sample rate and resolution are low but sufficient to the task. I would include the plots here but they are disgustingly boring because Estes igniters do not burn through at these currents.
One problem I had was that after 9 trials I realized I had failed to invoke my code with the flag indicating that the probability distribution function should be log normal rather than normal. So I replaced the expended igniters and started over. The initial conditions were lower and upper bounds of 3.0A and 5.0A with a standard deviation of 0.5A. The result was this test series:
My initial guess at the average current of 4 Amps wasn't too bad but my guess of the standard deviation was much too high. This results in testing far from the mean early on but the algorithm zeros in on the correct value reasonably fast. Note that it takes until trial 15 to get overlap. (a success at a lower level than a failure)
The estimated mean at the end of the test was 3.85A with a standard deviation of 0.092A. All fire current (99.9%) was 4.15A.
This is of course limited to a current pulse of 100ms but you can use it to get a rough idea of what would be required at different pulse widths. The one complication is that longer durations allow for the igniter to dissipate some of the energy it is getting so the actual current required will be greater than a simple scaling would indicate.
For example, a one second pulse would need 1/10th of the power or about a third the current based on scaling alone.
An addenda about the new (2014) starters. I am not sure why Estes is removing the pyrogen from their tried and true and very reliable igniters but it didn't improve them. Digging up a MSDS for old igniters lists potassium nitrate, carbon, corn starch, and hide glue as the ingrediants. Those igniters were classified as UN 0454 (igniters) explosives. The new starters have no UN designation on the packaging so they are not considered to be explosive by any standard.
After seeing a video of a side by side test and noting that the new coating doesn't burn (maybe a little bit where it contacts the bridgewire directly but it doesn't propagate) I suspected that the new coating might be plain hide glue. I tested this by attempting to remove it using a couple of solvents. Acetone did nothing but soaking in distilled vinegar (a Google search suggested that this would soften hide glue) and light brushing took it right off.
Because the coating is not a pyrogen and doesn't burn, repeating these tests with the new coating is pointless. The operation is totally dependent on the hot bridgewire contacting the BP propellant in the rocket motor. While the coating helps to keep the wires in the right place, it separates the bridgewire from the propellant making it harder to ignite the motor.
If you need a high reliablity motor igniter for clusters, I think that removing the coating and dipping in a real pyrogen would be advisable. But it seems likely that the ATF would consider that to be a regulated igniter with all of the hassles that involves. Note that back in 1997 the ATF told the NAR that they would exempt all igniters with 35mg to 50mg of pyrogen. That exemption was never formalized in the regulations and I suspect that the ATF never wrote an official ruling on it.