Hi folks,

Looking for some feedback on the Pixhawk 4.  

There seems to be a vacuum out on the newest iteration from the Pixhawk team. I know many of you are veterans of this and I can tell from following the forums that a lot of basic info is requested that was blatantly available with a little effort.  I'm trying not to be that guy.   I've been reading and watching for about a month and haven't found much on it except release notices being re-churned.  2 youtube videos of setup is what I've found that wasn't complete clickbait.  Nothing much about how well it's handling noise and feedback from the field.  I'm a little leary of IMUs and such that haven't been soak tested on a new platform.  

Maybe I'm looking in the wrong places or it's just not making it to market.  

This will be my first full DIY build for UAV.  Planning to run with a hex cross platform for payload and would like to invest in recent production hardware.  I will run an 8ch RX/TX since I'm already invested in it.  I don't have a flight controller yet, so I'll also be learning hands-on with the ground control as well.   

The flight controller is my first dedicated acquisition before I buy a single piece for the rest of the project.

Any info or insight would be greatly appreciated.  

Thanks.

Views: 1601

Reply to This

Replies to This Discussion

Ask youself why you want it ?

Pixhawk and Pixhawk 2,1 are open hardware, well designed, quality components used.  Very well tested.

Then you have lots of knockoffs , with significantly fewer sensors, some have saved away active power redundancy selector, and IO buffers to safe a few dollars.

Sure Pixhawk "4" , saves good money on not having the same amount or quality of sensors, no internal damping and heating like 2.1 , much simpler design.   - but you'll never know for sure, as schematics or BOM is not open.

So I guess you want it because it's a "4" , and 4>2.1
I suggest you wait for "Pixhawk 999"  it should spawn really soon, that is, if increasing a version number alone attracts customers :)

Thanks for taking the time to provide feedback.

I have been leaning towards Pixhawk 4 because it allegedly offers true 2M memory.  I narrowed it to Pixhawk because I'm impressed with what I've learned about the Pixhawk project and it's goals align with what I desire.   I have been concerned about how long I can continue to utilize previous iterations of the hardware before being forced to re-invest my efforts and start over.  

The IMU's and dampening  you touched on are exactly my concern.  Working with industrial IMUs, gryos, DGPSs, other sensors and PME's for dynamically positioned vessels, I've witnessed first-hand how a poorly performing IMU (we still call them "VRUs" most of the time in this field; although that term is incorrect and a pet peeve) can affect a position solution.  (ETA:  Sorry, PME is "position measuring equipment" , I reread this and realized I acronym-ed it to death).

Garbage in, garbage out.  I'd prefer not to have to tune my PID's until my FCM is oblivious because of garbage IMU's.  My intent is work with (3) each motion and references so that the system can better weigh the inputs.  If Pixhawk 4 doesn't offer this, then I'm done looking at it.  As I read the marketing, it implied that it did but perhaps I assumed too much.

I'm a newb to this particular discipline and and still trying to determine if this system is capable of merely averaging the the sensor input or if it can truly vote.  That's where this is going to go if it isn't already there.  If it is the case that these cannot vote out a bad reference, then the increased memory is going to become mandatory for the flight stack.  If not, then I'm off on the wrong tangent.  That's fine too.  I'm just getting into the weeds and trying to sort this out but seems to me that the redundancy is being utilized as an average.

My suspicion of averaging and the lack of voting is simply due to the fact that I'm not seeing redundant GPS/DGPS being employed as well.  No surveyor in my field would even be allowed to legally operate without the bare minimum of 3 and most likely 5 (tertiary backups for at least 2 sources) PME's and at least 3 of each motion references as well.  3 sources would be the bare minimum to keep us legal in accordance with several regulatory agencies.  From there it only gets more boring as I try to pick apart the flight stack and sort this out.

I do have to admit; very shallowly, I'm also concerned that Holybro appears to be more toy racing orientated.  Maybe that's foolish and I'm certainly not planning to be a start-up, but when pitted against companies that are robotics industry related, Holybro seems to have more in common with Nickelodeon than they do with autonomous mission planning.  That said, it's very commonplace for companies that are tooled up for a different market to successfully provide solutions in another.  IBM; for example, didn't turn it's nose up to the US government around WWII when it required small arms and cost wasn't an issue.

Andre K. said:

Ask youself why you want it ?

Pixhawk and Pixhawk 2,1 are open hardware, well designed, quality components used.  Very well tested.

Then you have lots of knockoffs , with significantly fewer sensors, some have saved away active power redundancy selector, and IO buffers to safe a few dollars.

Sure Pixhawk "4" , saves good money on not having the same amount or quality of sensors, no internal damping and heating like 2.1 , much simpler design.   - but you'll never know for sure, as schematics or BOM is not open.

So I guess you want it because it's a "4" , and 4>2.1
I suggest you wait for "Pixhawk 999"  it should spawn really soon, that is, if increasing a version number alone attracts customers :)

Since I'm still not seeing a lot of feedback on this, I'm probably just going to go with the 2.1 Cube Standard ed.  Seems like it would do the job.   Shame to not integrate 3-sensor selective voting with alarm feedback for the PID's.  I'll probably just build on a DJI flamewheel 550 Hex frame (not a clone) since it's ubiquitous and robust.  Staying away from Carbon fiber for now due to the fact that graphite happens to be an RF mask.  There should be plenty of real estate to get them to play together.

Down the road,I'll probably look at something a bit more custom.  The shop I've built next to our home is Disneyland for tinkerers.  Machine shop + welding shop + wood shop + auto shop in a 2100sq ft facility to play.  Between this and a few other projects, I've got plenty to do and in no rush.  

The sky's no limit. 

-Fixxer

I just got a pixhawk 4...working great, dont regret it....and who said they are lower quality sensors?

I have a pixhawk 4 as well. Definitely like it more than my cube. Wires is great (forget the power brick, that’s a mess, use something else) but the initself is slim reliable and I think you’re right. Holybro was probably just tooled up to do it. That’s the only hint that makes sense. There is likely a connection or relationship somewhere but since these boards overall are not exactly the most complex thing out there I don’t see why they couldn’t build. Shame about no design files. I actually don’t care but I don’t like something being sort of quasi masked on the open sourced side and then not. My suspicion is they’ll release the files after they have the lead as he sole supplier or recoup their r&d costs.

Also, for the cube. That’s old old at this point. Go look st the Cuav pixhack v5. That’s a newly built (or designed, I don’t know) cube shaped unit but with new sensors and a floating imu. 

Thanks Edward. I'll keep looking i to tbe Cuav stuff. I'm looking for tertiary redundancy IMU schemes for possible voting. The specs are kind of difficult to sort out. Some seem to hint at 2 IMUs for averaging. That seems to be the norm. Adding an IMU would be problematic at best due to latency issues with non-native hardware. It could be done but it will have to be an all or nothing option where I would have to add 3 and tweak the flight stack to ignore the the native units or else it would lagging the input from the serial bus units vs the instant feedback of the onboard units and sluggish at best.

Reply to Discussion

RSS

© 2019   Created by Chris Anderson.   Powered by

Badges  |  Report an Issue  |  Terms of Service