I've tried this with all the SDK versions (SDK_V11.0.0, SDK_V11.1.0, SDK_V11.2.0) and the same issue persists. What am I doing wrong?

Successful license agreement (received link in email right after seeing this screen)

Checked email right away and clicked on download link, which led to this page:

I am using the BME688 Development Kit with the 8-sensor board and Adafruit HUZZAH32.

My classification model was trained and exported using BME AI-Studio Desktop v2.3.2 for BSEC v2.5.0.2. Therefore, I need the matching BME688 Development Kit firmware v2.1.4 and Mobile App v2.4.16.

However, firmware v2.1.4 is no longer available from the official download page.

When attempting to live-test my exported model, the Mobile App displays:

“Error: There was a problem running BSEC on the board. (2)”

Could you please provide an official archived download link or package for:

• bme688 development kit firmware v2-1-4 / devkitfw_v2-1-4_website_release

• BME AI-Studio Mobile App v2.4.16

Alternatively, if you can send me the Latest and matching package, that should work too?

BME AI-Studio Desktop for Windows

BME AI-Studio Mobile

bme688 development kit firmware / devkitfw_v xxx_website_release

BESC

Etc

Thank you.

I was wondering is there any official list of supported auxiliary devices for BMI270. I wanted to have BMM150 as magnetometer but nowadays it is obsolete part. Is there any other magnetometers that are supported? (As I have understood, e.g. BMM350 is not supported).

what does 'typ.' mean?

I want to know the group delay in various low-pass filter mode

Best regards,

Aarni N.

Our configuration is relatively simple: only the accelerometer is enabled with an ODR of 100 Hz. However, when measuring the actual output data rate, we observe that it is not consistently 100 Hz.

To investigate this further, we compared the BMI323 with another sensor in our system that acquires PPG data. The PPG sensor is driven by LFXO and samples at a stable and accurate 250 Hz. When comparing the accumulated number of samples over time, we observe that the effective output data rate of the BMI323 varies slightly between devices. Some devices accumulate slightly more samples than expected, while others accumulate fewer.

Since this behavior varies from device to device, we believe it is unlikely to be caused by our software implementation. Could you clarify:

• Is the BMI323 output data rate derived solely from its internal oscillator?

• What factors determine the actual output data rate?

• Is device-to-device variation expected, and if so, what level of variation should we expect?

Additionally, in the BMI323 datasheet (Table 1), we noticed that the Output Data Rate Accuracy is specified as 1.7%, with the note that this applies when the gyroscope is enabled in any mode. We would appreciate some clarification on this specification:

• Does the 1.7% accuracy only apply when the gyroscope is enabled?

• What is the expected ODR accuracy when only the accelerometer is enabled?

• Does enabling the gyroscope improve or stabilize the timing reference used by the accelerometer?

We have also observed another phenomenon during long-duration recordings. Our system simultaneously records BMI323 accelerometer data and PPG data. Initially, the two data streams are well synchronized. However, over time we observe a progressively increasing time offset between the PPG and accelerometer data. This drift becomes particularly noticeable during periods of intense motion and continues to increase as the recording duration increases.

Could this behavior be related to the BMI323 ODR accuracy or internal clock tolerance? Are there any recommended methods for maintaining long-term synchronization between the BMI323 output and another sensor sampled using a different clock source?

We would appreciate any clarification or recommendations you can provide.

we are developing a small water-warning device. The pressure sensor should reliably detect temporary immersion at shallow water depths. Exact depth measurement is not required, and the pressure sensor will be combined with a separate water-contact sensor.

We are currently evaluating three concepts:

1. BMP581 behind a hydrophobic, air-permeable waterproof vent.

2. BMP581 inside a sealed air cavity behind a flexible, non-permeable diaphragm that transfers the external water pressure.

3. BMP585 with controlled water access through a protected and drainable wet chamber or short channel.

We would appreciate Bosch’s guidance on the following questions:

• Are the BMP581 and BMP585 generally suitable for reliably detecting shallow-water immersion?

• Which of these integration concepts would Bosch recommend regarding response time, repeatability and robustness?

• Is the BMP585 suitable for repeated direct contact with fresh water, chlorinated pool water and salt water?

• Does Bosch provide application notes, reference designs or engineering support for integrating these sensors into an injection-moulded waterproof housing, particularly regarding the sealing concept, protective grille, wet chamber, drainage and protection of the BMP585 gel surface?

We already tried several ways to find more information about integration suggestions of this sensor family but havent been successful. Before finally switching to another product we would like to give it a last try in this forum. 🙂

Thank you very much for your support in advance!

Best regards,
Kai

I contacted Bosch Sales NA and was directed to Mouser/Digikey. As an independent researcher without institutional budget, I cannot purchase samples at this stage. I am seeking 2-3 BME690 evaluation units to validate heater profile compatibility with my existing PyTorch classifier before clinical data collection.

I commit to publishing detailed integration results, sharing performance notes with the Bosch community, and acknowledging Bosch Sensortec in all publications. Technical documentation: https://leosenges.github.io/Denco/index.html

Is there a developer sample program, health initiative, or LATAM collaboration channel through which I could request these units?