Best Six-Axis Accel/Gyro?

[Brian DeBusk]

I'm looking at a six-axis Accel/Gyro project and I'd like to work with the beset (within reason) MEMS chip that's commercially available. 

I've seen the MPU6500 and it seems very reasonably priced (including the evaluation boards).  It also touts its own Digital Motion Processor.

1) Is there anything out there better than the MPU6500?  I'm not looking to drop several hundred dollars on a 6-axis chip, but I'd gladly spend $20-$250 if it was truly worth it.

2) How good is the DMP on the MPU6500?  Does it use a sophisticated algorithm or would I want to do my own in software anyway?

[Arthur Benemann]

It may be a little out of your price range, but this is about the best you will be able to get ADIS16488A.

--
You received this message because you are subscribed to the Google Groups "drones-discuss" group.
To unsubscribe from this group and stop receiving emails from it, send an email to drones-discus...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

--
Arthur Benemann

[Brian DeBusk]

That one looks incredible -- but $1,700+ per part?

Would a chip like that have a complete state-of-the-art Kalman filter built into it?  Is that a turn-key 10-degree inertial navigation system or does the user still have to roll their own algorithm?

[Robert Lefebvre]

Hmmm... all that, and it probably still gets confused by turning in circles for a while.  I don't think an IMU can be accurate without GPS data fusion.

[Brian DeBusk]

Do they make a GPS + 10-axis integrated module?

I'm all for doing the Kalman integration myself, but I don't want to reinvent a wheel here.

[Jonathan Challinger]

If you want a closed source turnkey inertial navigation solution, you could look at vectornav.

Otherwise, you should probably go for something similar to the sensors on Pixhawk, but mount them on vibration isolation with a peltier TEC and heatsink providing the weight to make the isolation effective. Control the TEC to keep the board at a constant temperature (25C). This will remove variability of sensor error with temperature so that you can conveniently ignore it.

[Brian DeBusk]

So far I'm at a total loss.  I see the MPU6X00 series that runs < $20 and the integrated systems that run $1,700-$5,000.

Doesn't anyone make a decent $250-$750 inertial navigation system?

[Brian DeBusk]

Here's something in the $130 range. 

http://www.pololu.com/product/2740

What do you guys think?

[Ben Nizette]

Still uses an Invensense sensor; does their EKF algorithm add ~$110
worth of value? Would have to test it and see I guess. What's the
application?

If you want a turn-key AHRS with integrated GPS you could, you know,
use a Pixhawk ;)

Ben.

[Paul Riseborough]

Take a cheap MEMS sensor, provide an interface processor, vibration isolation mount and enclosure that provides EMI shielding, calibrate the heck out of it in a thermal chamber and you have a $1500+ unit. The thermal chamber time required is expensive. Thermal chamber time with a precision motion table for scale factor errors is even more so.

I'm sceptical of manufacturers claiming high performance for cheap units, regardless of the magic algorithms they claim to have on the inside, unless there is breakthrough in the MEMS technology.

[David Pawlak]

Yeah, Ben, Brian, it's not realy something to be integrated into an Ardupilot system.

On another thread here there was talk about how an external MEMS system would improve vibration management... there may be applications...

I'll probably buy a couple to play with, as I am building a "simpler" indoor navegation system, but if you want something to incorporate into Ardupilot, you'd be better off looking for something simpler.

In Arudpilot, a lot of work has gone into the INS, especially with EKF and all. Pololu (or CH Robotics) has obviously done a lot of work preparing this unit.... but you'd have to dig in and get to where you could manage the firmware with a fair bit of authority if you were serious about using it for a product. I don't think you could expect that it would just up and run flawlessly under many varied and difficult situations. And as mentioned, integrating it into Ardupilot or any other Open Source autpilot project would mean undoing/bypassing  a lot of the code that's been done there.

So it depends on the application, and I'm not sure how much you actually gain with their on board EKF etc. You'd have to trust it or get to know it very well. As Ben suggests, might be simpler to use a PixHawk. (Or even an APM2 with just the INS/EKF code on board). At least Ardupilot has a lot of support.

[Robert Lefebvre]

Paul, is my perception right, that *no* AHRS can be accurate without GPS data fusion?  Won't they all get dizzy eventually, no matter how good the gyros are, it's just a question of how long it takes?

[Brandon Basso]

I have no idea what these cost, but cloudcap sells an IMU and an fully integrated standalone INS+GPS:

http://www.cloudcaptech.com/crista.shtm

The specs in the data sheets are interesting; they go into some detail on how they are quoting rate biases, variances, and such.

My guess is that they are <1k since the nano autopilot is around $1.2k.

Brandon

[Doug]

Robert, 

Dizzy?  Not sure if you mean that in the sense Premerlani uses it or something else...  Not all AHRS algorithms will get dizzy like DCM can, but any AHRS algorithm used on a platform which spends non-trivial amounts of time turning or otherwise accelerating need GPS or another velocity measure to correct the accelerometer specific force measurement for translational acceleration.  Otherwise translational acceleration due to turning flight, turbulence, etc. will cause the attitude estimate to be poor.  INS algorithms do not have this issue as accelerometer measurements are used to propagate velocity/position and do not have a direct feedback path to attitude.  However INS algorithms need GPS to correct for velocity/position drift due to integration of accelerometer noise.

So you don't necessarily need GPS to avoid "dizziness" but you do need it if you want a good state estimate.

[Robert Lefebvre]

Ok, thanks Doug, that makes sense. I did mean dizzy as in the Premerlani usage.  I hadn't really keyed into the difference between an AHRS and an INS system.  When you use the term INS, does that have an AHRS component as well?  Or do you mean *only* position, with no attempt to judge heading?  I would think that you cannot have an INS system without first having an AHRS, because to be able to do inertial navigation, you have to convert accelerations into earth frame, and to do that, you have to know your vehicle's attitude.

This discussion popped up at a very opportune time, as I am again trying to strike up a conversation with the AlexMos guys that they NEED to work with us to have GPS data fusion into their AHRS for the camera mount, or they will never eliminate the horizon drift problems they have.  The response was typical "Why do you need GPS?  We don't care about position."

So it sounds like you're confirming my question.  All of these systems must have GPS to be accurate, but not for all the same reasons.

Rob

[Doug]

Always a good question as the term INS gets used different ways.  AHRS stands for attitude and heading reference system, as you probably know, so that means a filter which gives you 3D attitude.  INS stands for inertial navigation system.  Some people use that for a filter which just gives position and velocity, but I was using it to mean a filter which gives attitude/velocity/position.

The reason you need GPS (or some other sensor) for an AHRS filter is that you need two reference vector measurements to correct for gyro drift.  Those vectors are (typically) the earth's gravity and magnetic field vectors.  Magnetometers measure the magnetic field vector, but accelerometers do not measure the gravity vector.  Rather they measure the specific force on a proof mass.  The specific force is due to gravity and translational acceleration.  For a quality attitude estimate you have to estimate the translational acceleration (using something other than the accelerometer measurement itself) so that you can remove that component from the accelerometer measurement.  Most methods of doing this use GPS, although people have done it with other things like airspeed.  I have a paper pending for the AIAA Journal of Guidance, Dynamics and Control which goes into the subject of the best way to do this for drones flying in turbulence.

[Robert Lefebvre]

Thanks Doug, that all confirms what I was thinking.

On 8 October 2014 12:16, Doug <dewe...@gmail.com> wrote:

Always a good question as the term INS gets used different ways.  AHRS stands for attitude and heading reference system, as you probably know, so that means a filter which gives you 3D attitude.  INS stands for inertial navigation system.  Some people use that for a filter which just gives position and velocity, but I was using it to mean a filter which gives attitude/velocity/position.

The reason you need GPS (or some other sensor) for an AHRS filter is that you need two reference vector measurements to correct for gyro drift.  Those vectors are (typically) the earth's gravity and magnetic field vectors.  Magnetometers measure the magnetic field vector, but accelerometers do not measure the gravity vector.  Rather they measure the specific force on a proof mass.  The specific force is due to gravity and translational acceleration.  For a quality attitude estimate you have to estimate the translational acceleration (using something other than the accelerometer measurement itself) so that you can remove that component from the accelerometer measurement.  Most methods of doing this use GPS, although people have done it with other things like airspeed.  I have a paper pending for the AIAA Journal of Guidance, Dynamics and Control which goes into the subject of the best way to do this for drones flying in turbulence.

--

[Paul Riseborough]

On planes, EKF does it using airspeed if you lose GPS. It appears to work OK for normal flight, but I suspect it would run into problems if you were doing extreme acro's.

[David Pawlak]

I think "dizzy" was the right term. :) :)

[Brian DeBusk]

Does anyone know what IMU the DJI A2 uses?  Is it thermoelectrically stabilized?

[Charles Taylor]

DJI scratches off the writing on the chips but it's basically 3 single axis accelerometers wrapped around a CNC lead cube for mass and isolated with some very special foam. From what an engineer told me after we opened one up they are pretty old tech compared to mp6k. It didn't look like they are controlling the temp in it. Maybe they cal their sensors at various temps to dial in the bias.

[David Pawlak]

The lead cube could make a great heat mass! Put a small resistor/rtd combo on it!!

[Jason Short]

The HD3D is a heated IMU.

Jason

On Oct 9, 2014, at 5:48 PM, Brian DeBusk <brianc...@gmail.com> wrote:

Does anyone know what IMU the DJI A2 uses?  Is it thermoelectrically stabilized?

[Brian DeBusk]

I may have found a good compromise between the $18 DIY 9 DOF chips and the $1,700-$5,000 stand-alone UAV INUs:

https://www.chrobotics.com/shop/gp9-ahrs

What do you guys think?  It's temperature stabilized, 500 Hz update rates and integrates GPS and baro input. 

For a valuable enough bird, I sort of like the idea of using two of these devices and blending them together with their own Kalman filter.  Then it wouldn't be an issue of failover -- but a failing or errant sensor should inherently produce more noise and contribute less to the overall AHRS calculation.

[Jonathan Challinger]

I'm sure that has the same sensors in it as pixhawk, and who knows what "temperature compensated" means for them.

That thing is just kind of a wildcard. You're probably buying the same sensors that pixhawk has and a closed-source software package that is likely to be orphaned.

If you want a constant temp Pixhawk, you could experiment with adding a TEC on a brushed ESC and controlling it with a PID to keep a constant temperature on the mpu6k

[socrates....@saggezza.com]

David
We are looking for a solution to use drones in a warehouse. This UNMANNED drone has to fly near the stacked boxes and scan a barcode using a wifi enabled barcode scanner onboard to send the data to our system.

I have done some research and found that to gps wont work indoor and we need an alternate system like infrared cameras to get the coordinates and make the drone to fly from one point to other.

We are looking for a partner to work with us.

Have you done any project with indoor unmanned drone?

thanks