After having introduced an automated latency measurement method in the last blog post, we used it to test several combinations of monitors and FPV systems, both digital and analog.
Transmission systems tested:
- Analog: standard 5.8 GHz system with a CCTV camera (XK-3089).
- Analog: standard 5.8 GHz system with a GoPro camera, running at 720p and 100fps.
- Digital: Amimon Prosight, in both HP (High Performance) and HQ (High Quality) modes.
- Digital: fpv.blue, using the HDMI input (GoPro, 720p@50fps) and custom camera (at 720p and 480p).
- Digital: Wifibroadcast with default settings, running on a Raspberry Pi 1 on the transmitter and a Raspberry Pi 2 on the receiver.
Display devices tested:
- Analog: a simple FPV monitor (RMRC-LCD-12).
- Analog and Digital: Headplay HD googles.
- Digital: a simple HD monitor (part numbers LQ121K1LG52, M.NT68676.2A).
The previous blog post goes into the methodology with more detail, but we are using a photodiode, a Texas Instruments OPT101P, to measure video latency. The measurement starts from a random point in time by toggling a LED via an Arduino and checking how long the photodiode takes to detect a change in light (code here). This has several advantages over the standard stopwatch method, as already explained.
The raw results can be downloaded from here. The following is a graph visualising them that we invite you to take as a comparison point next time you are thinking about latency.
- The display device matters. A lot. If you flown with an analog system and Headplay glasses you already experienced more latency than many digital FPV configurations.
- Analog systems are still the only ones that can claim close to zero latency. Digital systems are, at best, two times slower.
- Increasing GoPro framerate over 50 fps hoping to decrease analog output latency does not help. In fact, latency increased by a few milliseconds in our tests when going from 50 to 100 fps.
- The analog video output of a GoPro at 50 or 100 fps is also slower than many digital FPV solutions.
- HDMI display devices are not immune from display latency. The Headplay glasses introduced an average of 24 milliseconds latency when feed with a 720p stream and only 16 ms when feed with a 1280×800 stream. The additional 6 milliseconds are explained by the time the Headplay takes to resize the image from 720p to 1280×800, something that does not need to happen if the native resolution is correctly provided (well done, Wifibroadcast!).
A word of advice if you decide to replicate the test. Do it with battery powered devices only. Displays, oscilloscopes, etc., were all creating too much noise in the photodiode voltage sensor, making measurements tricky. If you need some help replicating those numbers don’t hesitate to send an email.