You can login if you already have an account or register by clicking the button below.
Registering is free and all you need is a username and password. We never ask you for your e-mail.
[+]rob_white0 points0 points0 points
ago
(edited ago)
[–]rob_white0 points
0 points
0 points
(+0|-0)
ago
(edited ago)
The one I am I having most fun with currently is this it's freaking tiny, has gyroscopes, accelerometers and magnetometers, pressure sensor, F4 processor, supports GPS and does return to home and GPS navigation, position hold is amazing, as good as my DJIs, All for $50 and a very tiny footprint with a case, with the GPS is was around $100. The whole drone cost me $250 as I have a TX and batteries already, besides the advanced features, it is the best drone racing platform I have ever used.
If you want to add a linux board to that you can, it has SPI to do so, this is elegant and flexible, the real time stuff can be done on the STM32F4 and the heavy lifting can be done using an Linux board like the Pi2 or something else but I see no reason for it personally. Qualcomm likely will not just use just their SoC but have either a real time processor or a FPGA on there as well, the SoCs don't even have close to the amount of PWMs required. Sure you can hack it with a I2C to PWM IC, Navio+ does this but then more complexity and no oneshot support (although ardupilot is so backwards it doesn't support oneshot anyway).
Besides that, the Linux stuff has all been done already. It was not that successful, the complexity was the problem, are you aware of the Phenox2? Much nicer design using the ZYNQ-7000 which has an inbuilt FPGA, they used Linux for OpenCV which is where I would use it, there it matters, for normal drone stuff it's overkill. Spiri is another one, long delayed, not even released yet as of course as it is overly complex and they are having issues with it, uses STM32 for real time. How about the Navio+ for the RasPi2, ardupilot based and not very good, patched Linux kernel for realtime support. There are commercial drones using SoCs for years as well, the ARDrone does, it uses real time image processing so needs it, not exactly popular though and can't do much else.
There is zero revolution here, just more drone marketing hype for something that has been done and no one much really wants, extra complexity and extra cost.
FPV requires real time video, any latency kills it so all FPV is currently all analog, it's the digital encoding that adds latency. Yes Fatsharks have always been expensive but they made for a niche market, just the way it goes, they can not be replaced with a tablet if that is what you are saying? Just due to the delay induced by digital encoding. DJI already uses digital video for framing video shots, their app supports it, it's very nice but uses custom modems and gets good range. 3DR tries to do this also but they did it over Wifi so it has little range, the latency is really bad and only 150 meters range certainly isn't good for FPV (range is worse in populated areas) it's not much good for even watching the on board video unless you stay close. You could do this right now with a cheap IP security camera also, there are some with apps.
As for the last gen chips being $20, they are large BGA packages which is going to need an 8+ layer board, complex time sensitive routing for memory timing and that is without factoring in the power chipset they require. It's all complexity for no meaningful return. STM32, ram, flash all on package, SoCs, flash and ram both external. Increasing complexity.
If Linux drones had some amazing utility, I'd be there and using them, they have been around for years already. I have a Navio+, it's big, heavy and clunky and does the same and flies the same as a $40 ardupilot mega.
Sorry for the long reply, I don't want to say you are wrong but I do want to myth bust their marketing, it drives me nuts how much bullshit there is with drones and how many people are getting screwed on places like Kickstarter. People actually believe Amazon will use drones for a lot of deliveries and other such complete bullshit even, it's all nonsense until there is a huge battery breakthrough.
Linux SoC DO have their place, OpenCV usage is a perfect example, collision avoidance would be great but even there a microcontroller and Lidar would work very with less complexity.
view the rest of the comments →
[–] rob_white ago (edited ago)
The one I am I having most fun with currently is this it's freaking tiny, has gyroscopes, accelerometers and magnetometers, pressure sensor, F4 processor, supports GPS and does return to home and GPS navigation, position hold is amazing, as good as my DJIs, All for $50 and a very tiny footprint with a case, with the GPS is was around $100. The whole drone cost me $250 as I have a TX and batteries already, besides the advanced features, it is the best drone racing platform I have ever used.
If you want to add a linux board to that you can, it has SPI to do so, this is elegant and flexible, the real time stuff can be done on the STM32F4 and the heavy lifting can be done using an Linux board like the Pi2 or something else but I see no reason for it personally. Qualcomm likely will not just use just their SoC but have either a real time processor or a FPGA on there as well, the SoCs don't even have close to the amount of PWMs required. Sure you can hack it with a I2C to PWM IC, Navio+ does this but then more complexity and no oneshot support (although ardupilot is so backwards it doesn't support oneshot anyway).
Besides that, the Linux stuff has all been done already. It was not that successful, the complexity was the problem, are you aware of the Phenox2? Much nicer design using the ZYNQ-7000 which has an inbuilt FPGA, they used Linux for OpenCV which is where I would use it, there it matters, for normal drone stuff it's overkill. Spiri is another one, long delayed, not even released yet as of course as it is overly complex and they are having issues with it, uses STM32 for real time. How about the Navio+ for the RasPi2, ardupilot based and not very good, patched Linux kernel for realtime support. There are commercial drones using SoCs for years as well, the ARDrone does, it uses real time image processing so needs it, not exactly popular though and can't do much else.
There is zero revolution here, just more drone marketing hype for something that has been done and no one much really wants, extra complexity and extra cost.
FPV requires real time video, any latency kills it so all FPV is currently all analog, it's the digital encoding that adds latency. Yes Fatsharks have always been expensive but they made for a niche market, just the way it goes, they can not be replaced with a tablet if that is what you are saying? Just due to the delay induced by digital encoding. DJI already uses digital video for framing video shots, their app supports it, it's very nice but uses custom modems and gets good range. 3DR tries to do this also but they did it over Wifi so it has little range, the latency is really bad and only 150 meters range certainly isn't good for FPV (range is worse in populated areas) it's not much good for even watching the on board video unless you stay close. You could do this right now with a cheap IP security camera also, there are some with apps.
As for the last gen chips being $20, they are large BGA packages which is going to need an 8+ layer board, complex time sensitive routing for memory timing and that is without factoring in the power chipset they require. It's all complexity for no meaningful return. STM32, ram, flash all on package, SoCs, flash and ram both external. Increasing complexity.
If Linux drones had some amazing utility, I'd be there and using them, they have been around for years already. I have a Navio+, it's big, heavy and clunky and does the same and flies the same as a $40 ardupilot mega.
Sorry for the long reply, I don't want to say you are wrong but I do want to myth bust their marketing, it drives me nuts how much bullshit there is with drones and how many people are getting screwed on places like Kickstarter. People actually believe Amazon will use drones for a lot of deliveries and other such complete bullshit even, it's all nonsense until there is a huge battery breakthrough.
Linux SoC DO have their place, OpenCV usage is a perfect example, collision avoidance would be great but even there a microcontroller and Lidar would work very with less complexity.