2021-01-17, 05:55 AM
I have a few concerns but otherwise ultrasonic sounds great.
1) wifi latency - from esp32 to esp32 probably ok, but going to raspberry pi I would ensure it is always below 100ms (prefer 10ms) or it will not be suitable for autopilot. Can we determine if this is possible? I think it may require an external esp32 to receive on a dedicated wifi/bt, and this would communicate with the pi on a serial port. The linux wifi stack cannot be relied on for realtime. It may work great for a while then get stuck and not send/receive data for more than a second. Especially if other wifi transfers are going on to another machine or something. Otherwise it will work with degraded performance, and people will just do it anyway and not realize that their boat is not steering as efficiently on wind mode as it really should. So I realize I cannot convince everyone, but this is the reason I do not want wireless wind sensors. You have to run power to them anyway and 4 wires is as easy as 2.
2) accuracy - are these sensors actually accurate to 1 degree? I seriously doubt it, and it's clear from the picture that they are not designed to take air when tilted but are intended to be perfectly vertical. When tilted turbulence will probably mess up the reading. At least a rotating vane feels wind the same as a sail so it compenates for this mechanically. To do this with ultrasonic needs tilt sensor and software compensation which is much more complicated.
3) cost - I'm not sure ultrasonic will ever be as cheap as mechanical. I built mechanical sensors which cost a fraction of the $139 figure and are accurate to 0.1 degrees.
4) power consumption - I don't think these use much power, but how much? For comparison, my mechanical ones are using 13 milliamps at 5 volts, so with linear regulator, 13 milliamps at 12 volts which is 1/6th watt.
5) heavy rain - it's well known that ultrasonic sensors cut out at some point if the rain is heavy enough. I also wonder how the accuracy is affected. Mechanical sensors feel rain the same way the sails do again, so even if the reading is different, it gives what you want.
I could probably make a list for why ultrasonic are probably better too.
1) wifi latency - from esp32 to esp32 probably ok, but going to raspberry pi I would ensure it is always below 100ms (prefer 10ms) or it will not be suitable for autopilot. Can we determine if this is possible? I think it may require an external esp32 to receive on a dedicated wifi/bt, and this would communicate with the pi on a serial port. The linux wifi stack cannot be relied on for realtime. It may work great for a while then get stuck and not send/receive data for more than a second. Especially if other wifi transfers are going on to another machine or something. Otherwise it will work with degraded performance, and people will just do it anyway and not realize that their boat is not steering as efficiently on wind mode as it really should. So I realize I cannot convince everyone, but this is the reason I do not want wireless wind sensors. You have to run power to them anyway and 4 wires is as easy as 2.
2) accuracy - are these sensors actually accurate to 1 degree? I seriously doubt it, and it's clear from the picture that they are not designed to take air when tilted but are intended to be perfectly vertical. When tilted turbulence will probably mess up the reading. At least a rotating vane feels wind the same as a sail so it compenates for this mechanically. To do this with ultrasonic needs tilt sensor and software compensation which is much more complicated.
3) cost - I'm not sure ultrasonic will ever be as cheap as mechanical. I built mechanical sensors which cost a fraction of the $139 figure and are accurate to 0.1 degrees.
4) power consumption - I don't think these use much power, but how much? For comparison, my mechanical ones are using 13 milliamps at 5 volts, so with linear regulator, 13 milliamps at 12 volts which is 1/6th watt.
5) heavy rain - it's well known that ultrasonic sensors cut out at some point if the rain is heavy enough. I also wonder how the accuracy is affected. Mechanical sensors feel rain the same way the sails do again, so even if the reading is different, it gives what you want.
I could probably make a list for why ultrasonic are probably better too.