How do modern open source autopilots compare to aerospace-grade IMUs?

I noticed that Digikey is now selling Honeywell's newest aerospace-grade IMUs, which cost $1,328 each (note that's just for the IMU; it's not a full autopilot). How do the specs of these aerospace IMUs compare to those we use here? Are they worth the extra money? 

In terms of overall approach, the Honeywell IMU seem very similar to modern autopilots such as Pixhawk 2.x and 3.3: they both have MEMS sensors with internal environmental isolation and temperature compensation.

As for the sensors themselves, I'm no expert on specs, so I'll just post the basics here, comparing the Honeywell sensor to the Pixhawk 3

On the face of it, the Invensense and ST sensors in the Pixhawk 3 appear at least as good, if not better. But I imagine that there are some other factors that may be more important, such as gyro drift and vibration filtering. The Honeywell specs in drift are shown here: 

Meanwhile the Invensense ICM-20602 sensor in the Pixhawk 3 gives its drift in different units: ±4mdps/√Hz. I really don't know how to compare those.

Finally, I'm sure that a lot of the performance depends on the software running on the Pixhawk boards, be it PX4 or APM, both of which use GPS to augment the raw IMU data to compensate for drift, along with a lot of other smart filtering. 

So for those IMU experts out there: how do you think these two approaches compare? Are aerospace-grade IMUs worth the extra money?

Views: 5465

Comment by Rafael Mendy on February 15, 2019 at 10:26pm

@ Tom Jakel, Anyway, the price for such room temperature performance ( second table of datasheet ) is very high. I think the datasheet must show the advantages. Maybe if other manufacturers indicate their performance in full temperature range, HG1120 shows its advantages.

Comment by DavidJames on February 16, 2019 at 3:18am

@ Tom B.

""one of the reasons that the Pixhawk autopilot has been so successful has been the remarkably low noise level of the MPU6000s." ...  Well probably actually because the software/hardware is adequate to do the job and is CHEAP at < $200."

That the APM2.X worked fairly well using the MPU6000 is pretty clear evidence that for a small drone autopilot like the Ardupilot, the low angle random walk of the MPU6000 was an important part of it's success.     You certainly can't say that the APM2.X autopilot's success was due to superior computing power.

Comment by Tom Jakel on February 16, 2019 at 11:31am
@Rafael Mendy there are quoted values on the HG1120 for full temperature parameters. What we have found is that users confuse them for room temp specs and then discount the IMU based on that. I will say that most of our customers are users who tried the chip scale parts and failed to meet their system requirements for one reason or another.
Comment by Curt Olson on February 16, 2019 at 11:48am

Assuming apm/px4 is running their ekf with a 10hz gps, those systems only have to coast on IMU for about 1/10 of a second between corrections.  I know vibration can cause weird things to happen, but outside of that, the big driver of attitude error (in my experience) is the gps (errors and latency) and mag (calibration/interference.)  This is because the EKF is trying to force all the sensors and bias estimates and state predictions to agree with whatever the gps spits out.

Comment by Thomas Butler on February 16, 2019 at 6:55pm

@Tom HW specifies at room temperature too as do most vendors, and most vendors (Invensense,et al) have a "percent sensitivity change vs temperature" error correction; same difference as HW's stated value at operating range. 

@Rafael, yes I misspoke I know the HW products are IMUs. I meant the Pixhawk+discrete sensors constitutes an AHRS. One would have to take the discrete sensors and feed that into IMU firmware and output that in order to compare the HW devices to the Pixhawk. However, the HW devices also output integrated sensor data at 600hz (the "control data output") which would be normalized data (rate, acceleration, etc.) from the HW's discrete sensors. This would be the simplest to compare code in the Pixhawk (aka ArduPilot). Then we would have comparable data. Come to think of it, the AduPilot code probably can take the HW control data as input if there were an abstraction for the device...maybe HW will create one! Should be fairly easy to hack the MPU6000 driver...the HW should win since it contains integrated hardware low-pass filtering within the housing. The hardware low-pass filter (silicone pad mounting, etc.) on the Pixhawk is usually an installer devised hack and has probably NOT been engineered like the HW devices. Sidenote; the Pixhawk 2.1 and the few other PX4 devices with integrated hardware low-pass filtering would be another bunch of apples to compare. 

The real question is what is the specification of the MEMS sensor(s) in the HW device(s). Something HW does not release (for all we know, there is an MPU6000 die IN the HW devices). Probably using the 5983 mag die too.These WOULD directly be comparable to the MPU6000, Pixhawk sensors, etc. and there would be no software greasing of the specs!

If anyone has a spare $13,000, I'll buy an HW device off of Digikey and crack it open!

@Chris, It's not common knowledge that 3DR put any money into software or hardware development (I'm not talking about developing production capabilities or supply chains or "bringing to market" which are costs the clone makers have too) in the ArduPilot code or Pixhawk design. Perhaps an article about how that came about is in order and how it in effect, gave the clone makers free IP.

Enough said on this thread! How fun!

Happy flying!

Comment by Thomas Butler on February 16, 2019 at 7:10pm

@Chris, while I have your attention, how about upgrading this arcane forum software used by DIYDrones. At least, add the ability to print ALL of the posts for a thread (very common on the majority of forums today) so we don;t have to wade through pages and pages...just a thought!

Comment by Tom Jakel on February 17, 2019 at 5:25am
@Thomas Butler it is no secret that the Honeywell HG4930 uses Honeywell's own bulk micromachined MEMS Inertial Sensors. The HG4930 has noise performance that is 5x better than the ICM20602. The HG4930 also has an RMS gyro bias over temperature of 20 degrees/hour.

If you want to see evidence that Honeywell's IMU line is the best in the world check out the NovAtel SPAN lineup. In the brochure the SPAN systems with Honeywell IMUs perform the best
Comment by Thomas Butler on February 17, 2019 at 8:24am

@Tom, I think I can speak for everyone listening in and commenting on this thread that your participation is excellent. It is rare that a person working at a component manufacturer ever gets involved in this sort of dialog. Thanks for being in the thread Tom.

To continue, Tom, you are comparing apples and oranges. Comparing the unprocessed output of the discrete component to the signal processed output or the HW devices is not a valid comparison. The specifications of the HW IMU is AFTER signal processing. Let's see the specs of the sensors IN the HW devices. It is peculiar that HW's devices (i.e., their MEMS sensors are) are "the best in the world". If that were the case, why isn't HW marketing the sensor?

Again, comparing to another manufacturer's ( Novatel) post-processed output is not a valid comparison of the sensor's capabilities.

Comment by Tom Jakel on February 17, 2019 at 9:32am
@Thomas no problem I love these types of discussions. It just so happens that I do what would be a hobby for me as a job. The value in the HG4930 is the sensors AND the calibration. On the HG1120 the value is in the calibration.

As far as the value of the IMU the most critical specs are the navigation parameters like the ones (both real time and post processed) in the NovAtel specification. The intrinsic properties of the sensors are things that an IMU designer considers when selecting sensors for the IMU. I have tested the ICM and a lot of other sensors as well as IMUs from high end navigation grade to low end Bluetooth IMUs. The one thing I've learned from all of this testing, use and simulation of IMUs is that it is never as easy as looking at a single parameter.

As far as the question of making Honeywell MEMS sensors available for sale. The answer is that the investment to make the MEMS sense elements and ASICs commercial products hasn't been worth the investment.


You need to be a member of DIY Drones to add comments!

Join DIY Drones

© 2019   Created by Chris Anderson.   Powered by

Badges  |  Report an Issue  |  Terms of Service