What happened to 9-axis IMUs?

[tom66]

This is "weird".

It seems that all three major suppliers of 9-axis capable IMUs (devices containing an accelerometer, gyroscope and magnetometer) have discontinued all offerings in this field.  They only offer discrete sensors (3-axis) or combined twin-sensors (accel+gyro).  Sensors containing a magnetometer in addition appear to be unavailable, but discrete magnetometer devices are still available.

Has there been some kind of patent trolling behind the scenes that has made 9-axis devices non-viable?  It seems weird that manufacturers would discontinue these quite useful devices, that have applications in AR/VR, drones, game controllers, smartphones etc.

Obsolete, no replacement of same performance:

https://www.st.com/en/mems-and-sensors/lsm9ds1.html
https://www.st.com/en/mems-and-sensors/lsm9ds0.html
https://www.bosch-sensortec.com/products/smart-sensor-systems/bno055/
https://invensense.tdk.com/products/motion-tracking/9-axis/

Edit: whoops, can a mod move to "Projects" forum.

[HwAoRrDk]

Just spit-balling here, but three reasons I could imagine:

- Everyone's having trouble reliably making a 9-axis IMU in a single die or package - bad yields, huge costs, etc.
- As you mentioned, patent or other legal troubles. Or egregious licensing costs.
- Some kind of trade restrictions meaning there's no longer a (legal) widespread international market. Wouldn't surprise me if perhaps such devices are classified as 'weapons technology' these days. Although I don't see why they'd stop making them altogether if there was still defence/aerospace demand.

[switchabl]

I don't know for sure, but MEMS is a bit of a minefield when it comes to patents. It wouldn't be the first time either (https://www.eenewseurope.com/en/st-invensense-settle-patent-clash).

[Marco]

Maybe wireless charging and magsafe has made phone manufacturers switch to separate devices (MEMS in the center, magnetic on the top/bottom) and dragged the market with them?

[SiliconWizard]

I have another take: market.

Efficient use of those 9-axis accelerometers, that do little more than giving the 9 values with various filtering options, is inherently difficult (AHRS algorithms) and requires frequent recalibration (especially the magnetometer), which is not ultra-convenient neither for engineers nor for end-users.

IMUs embedding AHRS algorithms directly are starting to appear, and I'm betting that they'll be replacing the more "raw" 9-axis IMUs with no signficant internal processing.

The 6-axis ones are still useful for more basic uses, such as approximate device orientation. On mobile devices, which I'm sure are the biggest market share for those MEMS IMUs these days, the magnetometer is rarely more than a gadget than can be done without for simple orientation purposes, and is mostly used for implementing compasses, which is not all that useful for most people. And since they are rarely used to really implement accurate AHRS, they don't need to be placed inside the same package.

[switchabl]

IMUs embedding AHRS algorithms directly are starting to appear, and I'm betting that they'll be replacing the more "raw" 9-axis IMUs with no signficant internal processing.

Both InvenSense/TDK and Sensortec/Bosch used to offer 9-axis IMUs with built-in sensor fusion. All of them are EOL or NRND with no replacements announced.

[Berni]

The magnetometers are often actually a separate die inside the package. So it is also a cost thing of having to combine multiple dies and wire bond them together. Some pretend like it is the same chip, some don't even bother and it is just multiple chips on the same I2C pins.

So at some point it is easier to just put the magnetometer in a separate package.

[Siwastaja]

I think Berni is close. Cost of producing a combined sensor and integrating it on a customer PCB and software is pretty much the same as using two separate ICs. For the same reason we don't have an IC which combines 7805 and 555 on a same package.

Having the customer to explicitly populate the two parts increases flexibility. Economy of scales is better with more generic, "smaller" units.

Magnetometers specifically are hard to deal with and require careful placement and calibration for the whole PCB assembly, batteries, case, screws, whatnot. Everything that is metal matters. The ideal spot for the accelerometer and gyro is often different than the ideal spot for magnetometer.

So the explanation might be as simple as this: maybe the 9-axis IMU was a bad idea to begin with, and now manufacturers have found it out by working with their customers, who probably just want two separate chips.