Re: [Paparazzi-devel] New IMU in YAPA

[Hector Garcia de Marina]

On Sat, Feb 16, 2013 at 5:15 PM, Mauro Garcia Acero <address@hidden> wrote:

Dear Hector,

 

yes, kalman filtering improves very much the performance on attitude estimation, but using better sensors, the accuracy of the measurement increases also. The thing is using fuel engines, vibrations may completely change the scenario of attitude determination because of the noise in the measurements.

 

AD claims to have produced, with this unit, a very reliable unit outperforming other military grade IMUs on vibration rectification.  

http://www.analog.com/en/press-release/11_08_11_10-DoF_MEMS_IMU_Makes_Tactical_Grade_Perf/press.html

I totally agree. Indeed the accelerometers will welcome this feature. In my experience I use to fly in a regime where some pwm commands are "forbidden" for the motor. Some frequencies propagate better throughout the fuselage ruining

the accelerometers measurements. Nevertheless, they can be always low pass filtered (but usually you need a high sample rate for avoiding aliasing), and also but with this you are omitting some information from the gravity vector as well.

Moreover, for the attitude probably low pass filtering is enough in a small plane, but for 3D position this is a little bit more compromising.

 

 

Looking at MPU6000 specs, they claim for a consumer grade cost effective IMU, which means to me that the stability of their gyros is far away from something reliable considering missions of several hours. ADIS16488 claims for a ARW of 0.3 °/√hr. But I don't find this specs of MPU6000. Same for accel specs.

Mmm, I have never flown for more than 45min in a row. So I can not say anything about that employing low cost MEMS sensors for a really long missions. However, if you employ some kind of fusion (such as Kalman), you should not suffer of drifting, and the uncertainty in the gyros and attitude should be bounded, even for really long missions. This is what I maths says and I have observed in practice.

 

 

In the webpage here after, there is an interesting description on the different grades of IMUs and their uses:

http://www.vectornav.com/support/library?id=76

 

I'm not saying that MPU6000 is a bad IMU, it is a good IMU for determined kind of missions and uses, as well as ADIS16488 is more adapted to other ones (still to be confirmed).

Well, it is within the range of low cost sensors. So the bound for the gyros measurements will be always higher than a ADIS164xx, this is what I can confirm from my side. 

However as I said, the differences between low cost and the ADIS module were not so significant. This is due to (as you have pointed out) vibrations from the motor, not correct compensation with

respect to the CoG of the plane, nonlinearities, temperature compensation, etc... for a small UAV. By the way, Analog Devices claim that they compensate in temperature their ADIS. A simple test 

in a freezer with three ADIS16405 I have found that this is not true in a really significant way. I have contacted to them, and the response was basically that without the compensation, the result would be worst.

I did the same experiment with low cost sensors, and I have to say that the result was a little bit worst. Nevertheless, with a simple linear interpolation for the calibration, ADIS16405 and MPU6000 almost had the

same performance with respect temperature. With same I mean, for attitude estimation in a small fixed wing, the temperature is not an issue anymore, there are other important problems.

 

 

Anyway, talking about integrating this SPI IMU into paparazzi, what do you propose? To begin from scratch with a new module or to reuse other existing module, for instance a SPI IMU?

mmm, I would say that a separated ADIS module would be better. ADIS family is big and I guess that they share all of them the same protocol. So you can always reuse it for other ADIS modules.

 

 

Best regards,
Mauro.

 

Best regards,

Héctor

 

 

---

From: paparazzi-devel-bounces+m.garcia=address@hidden [mailto:paparazzi-devel-bounces+m.garcia=address@hidden] On Behalf Of Hector Garcia de Marina
Sent: sábado, 16 de febrero de 2013 16:11
To: address@hidden
Subject: Re: [Paparazzi-devel] New IMU in YAPA

 

Hi Mauro,

I have a long experience with the ADIS16405.

In my opinion for the control loops without taking into account a physical model of the plane. The performance is not so big compared with other imus such as MPU6000.

The estimators based on complementary filters or Kalman are far enough for a nice estimation of biases for the gyros. In fact the performance of the accelerometers is quite nice for small airplanes.

I can not speak for other applications. But for small fixed wings. I really have to say that I have not found any differences.

On 16 Feb 2013 15:56, "Mauro Garcia Acero" <address@hidden> wrote:

Hello everyone,

 

After some discussion with TU Delft people, I will try to integrate the new ADIS16488 from Analog Devices into the YAPA in order to verify the behavior of paparazzi with a high end MEMS IMU:

 

<From the AD site>

6°/hr in-run bias stability

0.3°/√hr angular random walk

0.01% nonlinearity

 

Because this IMU communicates through SPI port, do you have any suggestion about how could I go quicker in the development of the new module? I was thinking to reuse one existing IMU module already communicating through SPI. But I have just begun to analyze the doc of the web site, so, I'm open to any kind of possibility.

 

Thanks in advance,

 

---

 

_______________________________________________
Paparazzi-devel mailing list
address@hidden
https://lists.nongnu.org/mailman/listinfo/paparazzi-devel

_______________________________________________
Paparazzi-devel mailing list
address@hidden
https://lists.nongnu.org/mailman/listinfo/paparazzi-devel

--
Héctor