A third sensor is necessary to know which IMU is failing. At the moment APM code cannot differentiate which IMU is bad ( with 2 IMUs). It can only compare 2 reading but it will not know which one is necessarily the bad one. 3 IMUs is ideal but APM not supporting 3 as of now.
I totally agree with you sir.
This is actually not true. The most common failure would be a single chip reporting zeros on all channels due to internal failure or cold soldering joint. This is very easy to detect.
Also EKF can determine which IMU reports reasonable data by cross checking if data matches what other sensors report - GPS, baro and based on history of measurements.
2 Likes
I am talking about ,measurement differences not a total failure⌠obv thats easy to detect.
EKF adds into account more than just gyros⌠so its not a fair comparison.
Adding one IMU theoretically improves the drift by integration by 30 % which is not bad, a thrid one just by 13 % (if I am correct). For flight behavior a third one will not pay off. The mentioned point that the accelerometer is still needed is not fair, because it will be needed always, even if the gyrometer would be perfect, and because the noise reduction by a second IMU takes place before the filter (DCM or whatever) is executed.
Deviation
- 2 IMUs : 71 %
- 3 IMUs: 58 %
Regarding the sudden fail of one gyro during flight. We are not in space where heavy radiation will cause a bit-flip or whatever. I am not sure how abundant a sensor reporting random stuff is, but I never saw one.
The point that the accelerometer still would be needed hasnât been made. The accelerometer will always be needed, as the gyro measures angular velocity and not orientation. The need for this accelerometer has nothing to do with the fact that noise reduction happens before the filter. A reduction of noise must happen before it is integrated into random walk.
That being said, I believe even a reduction in noise of say 50% would fail to yield any noticeable difference in attitude estimation. The usual mems sensors (i.e. 9250, 6050) already have low enough levels of noise for attitude estimation of sufficient accuracy, especially if linear acceleration from i.e. large banking turns with fixed wings is corrected by (limited derivation of) GPS velocity measurements.
Having spent little time inside the pretty nasty navigation code of APM, I do not know if the correction from the accelerometer is lowered if one uses dual IMUs. If this is correct, the less noisy gyro measurements from dual IMUs leading to lower drift after integration is being overcorrected with noisy (from the failed assumption of zero linear acceleration) accelerometer data. This can only yield sub-optimal performance, and the theoretical 30% improvement will stay theoretical.
Improved attitude estimation, which at this point really means increased robustness to non-zero linear acceleration, especially non-zero mean over longer periods of time, would need measurements with noise (and bias) at least one (preferably several) order(s) of magnitude lower than current levels. These gyros exist, but they are a little more expensive.
you know what? its OK, just the fact youd even consider that is so wonderful, since i posted that ive been getting to grips with my navio+ and i have no idea what i was on about, its perfectly fine for my needs and im even happier i got it now knowing youd look after your customers like that and even happier to join the community. im just learning to code really, and as usual getting ahead of myself, but when the time is right ill upgrade properly. i plan on video streaming so ill properly upgrade in the near future once ive got the hang of it since you mention improved camera performance, and who knows? raspberry pi 3 may be out by then for even better performance. so excited
thanks so much again.
Hello
Do you have any update on the Navio 2 Shipment date? Are you on time or do we have to expect a delay? 
Canât wait to receive it !
Very straight forward question - can RTK (accurate) position info be sent to Arduplane for precision flight?
The aim is to have very accurately geolocated photos. My setup would be Rpi2 + Navio2, LTE modem, correction data from local NTRIP server. Is that possible?
@gelazanskas You have a full Linux box there, so sending the data around is not an issue, however for RTK you will need an external RTK receiver.
Why donât you put M8T on NAVIO(x) and have all in one RTK solution?
The price is significantly higher and Navio will not be affordable as a result. We have covered this question many times on the forum.
If the question is covered many times in the forum, I believe there might be a demand 
Consider having NAVIO-RTK version.
1 Like
Hey there!
I just wanted to ask If the Navio2âs are allready beeing shipped, or when that will take place
Will the current APM ArduCopter release support the new features in the new Navio2 (i.e. the dual IMU and the high resolution barometer) or do we need to wait until APM releases new code that fully supports the Navio2?
One extra comment. 3DR and PIXHAWK are starting to move away from the DF13 connectors and they will start using JST GH connectors on their new boards (e.g. Pixracer); which meets the safety standards. Hopefully, future Navio board release use them as well.
I myself am curious as to when the Navio2 will be shipping. I place a pre-order 5 days ago under the impression that they would be shipping mid January(18th?). Now it appears that they wonât be shipping until late February? What exactly is going on?
@Al_B we have the code that supports Navio2, it will be merged into APM master soon. Yeah, weâll probably switch to JST GH in the future, but for now itâs better to stick with DF13 for some time since most of the periphery uses it too.
@jlrogers2 all Navio2 orders will be shipped today.
Just to clear up the situation about shipping. All orders placed for mid january has been shipped today, the boards were in stock only for 2 hours until we ran out. After that we have placed a note that all new orders will be shipped late February.
If you need to get Navio2 sooner I suggest to monitor our dealers, most of them are going to have stock within a week.
Making our products constantly available on stock is our priority and we have made a big effort to ensure uninterrupted availability of Navio2 after February.
Thank you for choosing Emlid! 
1 Like