The standard servo signal that's used by 90% of surface servos is based on an on-off pulse. The width of that pulse determines the position of the servo, with 1.5ms pulse width being the standard for centered position. The maximum and minimum pulse widths usually seen are 2.4ms and 0.6ms, but we usually use less than this range because we need less than 90 degrees of motion in either direction. This concept is called Pulse Position Modulation, or PPM.

In the days of analog servos, the rate at which these pulses were sent to a servo was at around 60 times per second. This rate limits the response of the servo, because there is a period of nearly 16.67ms between pulses (1/60 = 16.67ms), so the latency of the system could never be less than 16.67ms.

Digital signal processing used in digital servos allowed the pulse rate to be increased, reducing latency. It should be clear then, that the faster the rate the pulses are sent, the quicker the servo (or ESC) can react to a change. The fastest radio systems based on the 1.5ms neutral pulse width have a pulse frequency of around 384Hz, which means that pulses are being sent at 2.6ms. (1/384 = 2.6ms).

The limiting factor now becomes the 1.5ms neutral pulse width, because this standard allows the pulse to be up to 2.5ms wide, and if the pulse rate were to be faster than every 2.6ms, the width of the pulse could cause an overlap!

For this reason, a new 'Narrow Pulse' standard is available, where the pulse width is 0.3ms wide at neutral. If we assume a similar range as the 1.5ms pulse width, the largest pulse width for a narrow pulse system would be somewhere just under 0.5ms. Theoretically then, it should be possible to build a radio system with a less than 1ms response time. I've not seen that yet, but I have a big hole in my data because I've not yet measured the M12 and M12S... Hopefully tomorrow night I'll get a chance.

Before I get to my results, I should state that I'm measuring the lowest theoretical response time of the radio system. The actual radio transmission adds no measurable latency, but there could be some processing in the transmitter and receiver. I'm not able to measure that, and not making any claims about that either. I would think however, looking at the trends in the data, that at least of the Sanwa / Airtronics kit, the processing is what limits the PPM rate, so this is probably a very accurate way of measuring latency.