why does firmware not appear to address 1 missing motor?

[aartsf]

In other to fly the birth with only 3 motors with a fw/sw fix, it would be necessary for the bird to be able to maintain attitude with the powers of only 2 motors, namely the 2 'outermotors'. The 3th motor would be needed to keep balance and some steering capabilities. All of this just according to the laws of fysics.

 

[cahutch]

Nobody seems to have mentioned the obvious answer.
It's not possible for the controller to know if a motor has stopped. The ESC does not provide feedback on actual motor RPM.
It simply tells the ESC to run the motor at a certain speed and assumes the motor is there spinning at the requested speed.

All other arguments aside, the controller can't compensate for a lost motor if it doesn't know the motor is lost.

The best way to implement this is motor redundancy. Two motors on each arm running off the same control channel.
If one motor dies, the controller automatically compensates by adding power to that channel increasing torque or rpm on the remaining motor.
Example: http://youtu.be/llxtkDJgv2U?t=17m10s

 

[aartsf]

Nobody seems to have mentioned the obvious answer.
It's not possible for the controller to know if a motor has stopped. The ESC does not provide feedback on actual motor RPM.
It simply tells the ESC to run the motor at a certain speed and assumes the motor is there spinning at the requested speed.

All other arguments aside, the controller can't compensate for a lost motor if it doesn't know the motor is lost.

I would be not so sure about that. Wy wouldn't it be possible that the sw / controller recogises that its corrective input to the motors does not have the effect to be expected and 'concludes' a particular motor is failing.

 

[cahutch]

... Wy wouldn't it be possible that the sw / controller recogises that its corrective input to the motors does not have the effect to be expected and 'concludes' a particular motor is failing.

That would take some reasoning. By my estimate, the controller would have about a half second to determine the problem and compensate before it's falling and tumbling and too late to compensate.

If the controller recognizes that the airframe is yawing clockwise, it will increase torque to the opposite two motors to compensate. It probably also decreases torque to the other two to prevent gaining altitude.
If one motor is missing, all the controller can tell is that it's still yawing despite corrective action and it tries to compensate even more. There's no way to tell which of the two possible motors failed.
If it could tell, it would basically have to shut off the opposite motor to balance torque which means we're flying on two motors now and don't have enough thrust to stay in the air.
Edit: The opposite motor spins in the same direction. To balance torque it would have to shut off a motor adjacent to, not opposite to the bad motor which would leave all remaining thrust unbalance on one side. It would quickly flip over and crash hard.

At the same time, it's suddenly tilted and moving at increasing speed in the direction of the missing motor.
How does the controller know it's a missing motor and not just wind or other external force moving it that direction? (maybe because it's yawing too and spiraling downward?)
The controller compensates by increasing power to the missing motor and decreasing power to the opposite motor, to try to correct that lateral movement but it doesn't work.
If it could tell that it lost a motor and it's not just the wind pushing it, the only way to compensate is to shut off power to the opposite motor.
Once again we're unable to stay in the air on only two motors and the two remaining motors are spinning in the same direction causing the airframe to yaw uncontrollably.
It would be uncontrollable and would spin faster and faster until it lost lift and crashed.

Still, it could try to do an automatic emergency descent. Dropping straight down rapidly on two motors is probably better than free falling or spiraling at high speed into the ground.

The future will be interesting. Flight controller computer power and speed will improve. Propeller RPM sensors may give the controller feedback on actual prop speed. We're already seeing ultrasonic altimeters and camera vision based position hold. Sense and avoid systems are just around the corner.
But for now we're flying a cobbled together system based on off the shelf model airplane technology and first generation software.

 

[yorlik]

In other to fly the birth with only 3 motors with a fw/sw fix, it would be necessary for the bird to be able to maintain attitude with the powers of only 2 motors, namely the 2 'outermotors'. The 3th motor would be needed to keep balance and some steering capabilities. All of this just according to the laws of fysics.

Actually I believe the 3rd motor would be running the same high prop speed and producing almost the same lift as the other 2, so in reality, 1 missing motor means 3 remain to lift, not 2. I think all 3 would modulate to rotate the unit around center to equalize the lift so all 3 would share the same load still.

Not to beat a dead horse, but assuming dji watches these forums as I suspect they do, they should see the writing on the wall for potential future upgrade paths. I say potential paths because they may elect to ignore the work already done in this area by others who have proved it works and have youtube videos showing it working. A lot of objections have been brought up here why it cannot be done, no need to reiterate them. But a classic one has been 2 or 3 motors are not powerful enough to keep the pv2 from falling out of the sky. All the rest had decent possible work arounds already explained, except this one.

So I decided to test that theory today. I precisely weighed my unit: 50 oz with the battery installed.

I attached 52 oz of dead weight (wood boards) to the bottom of the unit and took picture of it trying to lift: https://www.dropbox.com/s/qfd5eptoomojh ... 1.mp4?dl=0 Not quite possible. Since 4 motors almost lifted 2x the weight, I suspect 2 motors would allow the pv2 to SLOWLY descend. Of course we are talking about 3 motors, not 2 for this exerciser....

I removed 1 board to reduce dead weight to 44 oz extra: it was able to not only hover but lift the unit: https://www.dropbox.com/s/t9nxcs7l7rdxo ... 2.mp4?dl=0

Moral of the story in my book is that a pv2 can ALMOST lift its own weight, and CAN definitely lift 88% of it. BTW, this means its vertical acceleration is almost 2G. Since in normal flight 3 motors lift 75% of it, this says we have a good 13% EXTRA lift that is not even required. This tells me that 3 working motors will produce more than enough lift to allow not only a safe landing with one errant motor, but also adjustment of the flight path during the emergency landing process.

 

[cahutch]

Weight constraints aren't the problem. Assuming that two motors could in fact lift the whole 800 grams ...

If you have an odd number of motors, the extra torque in one direction will cause the platform to yaw in the opposite direction.
For it to be stable, torque needs to be balanced.
The flight controller constantly re-balances torque by re-distributing power between M1-M3 and M2-M4
If motor M1 stopped, the flight controller will reduce power to M2-M4 and/or increase power to M3 to balance torque and stabilize yaw.

If you have three motors running, that third motor will end up lifting one side causing the system to flip over.
The platform is balanced. If you have unbalanced thrust, you flip over.
For it to be stable, thrust needs to be balanced.
The flight controller constantly re balances thrust on all four motors to keep the platform level.
If Motor M1 stopped, the flight controller would reduce power to M3 to keep the platform level which would mean reducing M3 to the same as M1 which is zero.

Tricopters are only possible because the platform is balanced on three arms and the tail rotor tilts side-to-side in order to steer and offset the torque imbalance.

IF you try to run it on two motors, they would have to be the two across from each other over the center of gravity. M1 and M3 or M2 and M4.
The problem is, those two motors spin in the SAME DIRECTION. See the diagram above.
The torque would be unbalanced and the platform would spin uncontrollably.

In order to do what you suggest, all four motors would need to be tiltable to re-direct thrust and compensate for lift and torque imbalance.
This is impractical and obviously can't happen with a software update.

Although running on two motors at full power WILL cause the whole system to spin like a top and will be completely uncontrolleable, it may be able to stay level and may provide a softer landing than a dead drop from altitude.

Try this ...
Look at the motor diagram in my previous post above.
Now imagine one motor is dead and try to keep the platform balanced using three motors.
Next, try to keep the platform and torque balanced using only two motors.

 

[skyhighdiver]

In other to fly the birth with only 3 motors with a fw/sw fix, it would be necessary for the bird to be able to maintain attitude with the powers of only 2 motors, namely the 2 'outermotors'. The 3th motor would be needed to keep balance and some steering capabilities. All of this just according to the laws of fysics.

Actually I believe the 3rd motor would be running the same high prop speed and producing almost the same lift as the other 2, so in reality, 1 missing motor means 3 remain to lift, not 2. I think all 3 would modulate to rotate the unit around center to equalize the lift so all 3 would share the same load still.

Not to beat a dead horse, but assuming dji watches these forums as I suspect they do, they should see the writing on the wall for potential future upgrade paths. I say potential paths because they may elect to ignore the work already done in this area by others who have proved it works and have youtube videos showing it working. A lot of objections have been brought up here why it cannot be done, no need to reiterate them. But a classic one has been 2 or 3 motors are not powerful enough to keep the pv2 from falling out of the sky. All the rest had decent possible work arounds already explained, except this one.

So I decided to test that theory today. I precisely weighed my unit: 50 oz with the battery installed.

I attached 52 oz of dead weight (wood boards) to the bottom of the unit and took picture of it trying to lift: https://www.dropbox.com/s/qfd5eptoomojh ... 1.mp4?dl=0 Not quite possible. Since 4 motors almost lifted 2x the weight, I suspect 2 motors would allow the pv2 to SLOWLY descend. Of course we are talking about 3 motors, not 2 for this exerciser....

I removed 1 board to reduce dead weight to 44 oz extra: it was able to not only hover but lift the unit: https://www.dropbox.com/s/t9nxcs7l7rdxo ... 2.mp4?dl=0

Moral of the story in my book is that a pv2 can ALMOST lift its own weight, and CAN definitely lift 88% of it. BTW, this means its vertical acceleration is almost 2G. Since in normal flight 3 motors lift 75% of it, this says we have a good 13% EXTRA lift that is not even required. This tells me that 3 working motors will produce more than enough lift to allow not only a safe landing with one errant motor, but also adjustment of the flight path during the emergency landing process.

nicely tested just remove the prop guards and you have even more saving lift and less chance of vortex ring

 

[yorlik]

cahutch I thought we put the software fix being impossible behind us and just had issue of amount of total lift left. Sorry.

What you are missing is that the 3 of 4 motor quadcopter does not sit still; it yaws to even the lift all the way around - via software routine. You can go back in this thread and see more than 1 example of folks who thought outside the box and made it work and have youtube videos to prove it. Think of it as a rotating bird - sure it would want to fall where the missing prop is - but instead, it simply turns and now the prop is over there holding it up. But wait, now it needs to be over there and it turns again, and again, and again. Watch the videos and you will see how this routine works. This software only 'fix' simply works. Hopefully there are still dji engineers who can think outside the box - they sure did to come up with the phantom!

 

[cahutch]

I see. I've seen that done with a quad copter since it can re-balance lift with 5 motors.

No doubt it will yaw uncontrollably with only three motors since torque will be unbalanced.
The faster it spins, the less lift it will have since it is spinning in the opposite direction of the propellers.
But the extra lift on one side will always be on that one side no matter how fast it spins. So it will still flip over.
I suppose if it's spinning fast enough, centripetal force will keep it balanced like a gyroscope.

In any case, the flight controller doesn't know that one motor is dead. It has no feedback from the motors and can't know if one is stopped.
So the flight controller will assume the dead motor is still running and try to balance lift and torque normally causing the system to flip over and crash.

In order to stay in the air with three motors, the flight controller would have to switch to a new flight model, which would require some way of knowing when it has lost a motor.
I don't see a practical way for it to detect a lost motor by flight behavior alone. It would have to know the difference between a missing motor and a gust of wind or bump against a tree branch.
I don't think there is a way to accurately differentiate that in time before the whole thing is upside down and unrecoverable. Unless you are at a very high altitude.
But maybe it's possible. I'm no expert.

 

[yorlik]

Guess this is why Mark Mueller is having such a hard time selling his subroutine to drone builders. Even when he shows them how in white papers and proves it with video, there are too many who still think what they see is not possible. His routine is creative; he can tell immediately when one motor fails by it falling in that corner; there is absolutely no need for any hardware motor feedback change - it is all software driven when it goes beyond normal wind/pitch, he kicks in his rotate failsafe subroutine. Eventually there will be a software engineer @ dji who can see how this routine works in his mind and will add it. Until then, the soothsayers will keep him from even considering it for a longer time. Shame.

 

[cahutch]

I'd need (like) to know more about how he can detect a motor failure.
For example, say one corner drops. How do you know that's a motor failure and not an updraft raising one side? And how do you know that quickly enough?
If he's figured that out, he should patent it quickly.

IIRC, DJI has already added such a feature to it's high end hex and octo-copter platforms.
Since the consumer models are selling like hotcakes without that feature, it will probably require another manufacturer to take some market share from DJI before it would make financial sense for them to add it to their low end models.
If in fact, it is even really possible with a quad.

To me, it would make more sense to improve the overall reliability of the motors, ESCs and the whole platform in general so that the likelihood that this would be needed becomes so low that it doesn't matter.
The motor design is already practically foolproof. The bearings are sealed and the only thing that can really go wrong on this kind of brushless motor is if a magnet comes loose from the housing.
That's more likely to happen when the motors are still than when they are running.
The ESCs are normally the culprit in a motor failure so if they can beef those up to the point where they don't break a sweat at full power, the likelihood of a motor failure in a reasonable time frame would be extremely low.

 

[gaspimp]

While I'm sure a sophisticated algorithm could figure it out, it seems far more reliable to rely on hardware. Ideally, there should be redundancy in the sensor to ensure that the motor isn't falsely detected as offline. A rotary encoder in addition to power consumption / current monitoring would work. I'm a bit surprised that at least the current monitoring isn't already offered by ESCs. In fact I would go so far as saying that the feature probably exists in many ESCs out there, but just isn't interfaced/connected to the flight controller.

I think an equally worthwhile investment would be redundant ESCs and power subsystems. 2 half-capacity batteries vs. one big battery would provide some redundancy. It could even be abstracted by a more complex version of DJI's smart battery technology. Redundant ESCs wouldn't be too costly either. We need to analyze the probability of a motor actually failing compared to all sorts of supporting circuitry. I highly doubt that the motor itself has anywhere near as high a failure rate as anything else.

 

[cahutch]

Agreed gaspimp. I would just add that the rotary sensor should look at the propeller, not just the motor. A lost prop is more likely than a motor failure and the results would be the same.
I envision an IR sensor that detects reflections from the prop and comparing the RPMs of the prop to the motor RPMs. So that if the motor RPMs stay the same but the prop RPMs just dropped in half, you know just lost half a prop.

 

[yorlik]

Be suprised then as there is no current feedback in our esc's.

I know from a study I recently did showing a potential reason for so many props flying off. About to send to dji sdk group for comment....

 

[gaspimp]

Be suprised then as there is no current feedback in our esc's.

I know from a study I recently did showing a potential reason for so many props flying off. About to send to dji sdk group for comment....

I guess I'm more disappointed that surprised. It seems so obvious.

 

[cahutch]

Be suprised then as there is no current feedback in our esc's.

I know from a study I recently did showing a potential reason for so many props flying off. About to send to dji sdk group for comment....

Question re: Props flying off...
I always torque my props down with the wrench lately to avoid my props flying off. However if a motor stops mid flight that's not going to help much.
Is this a temporary slowdown of the motor ?

 

[yorlik]

Max decel rate (32amp) decel kicks that may be happening almost every time full throttle is applied For fast ascent....

 

[cahutch]

Are you saying that if you are full or high throttle, when you back off the throttle quickly, the motors slow down too quickly and the prop momentum causes the props to spin off?
That would make sense since brushless motors don't "coast".
This seemed to start happening to more people recently. Perhaps after a certain firmware update.

I'll be careful about backing off the throttle to quickly and I'll keep torquing down my props.
Thanks

 

[yorlik]

No, it may be happening IN THE MIDDLE OF (during) max throttle. Ya, seems 2.1 esc fixed easy-bake I mean easy pop FETS, but might have given us this equally bad velocity loop instability.... why when folks try their bird after prop off crash and all motors work fine..... this is very typical tuning issue I see in my servo world by inexperienced engineers every day......