ZMOD4410 - Post-Processing Algorithms

I just have a few questions with respect to the data output and post-processing algorithms.

Is the classification of the different types of odors, VOC’s, etc. done on the sensor using the embedded AI, or will this need to be done on the user’s MCU using the raw data output from the sensor?

And with the neural network training feature - if I’m using this sensor in an environment that is starting to observe a new recurring odor that isn’t specifically identified in the firmware already, can the sensor be trained to identify this as a new odor (and would this require a firmware update?) or would I need to do this as part of the post processing on my own MCU?

An I’m not sure if you can elaborate on this - but what MCU is used on the actual sensor for running the embedded AI, and is it using DRP-AI or a third party IP?

Thanks in advance.

  • Hi There, all data processing sits externally - nothing is done on the ZMOD4410.  Everything is contained with the host SW .. <runs a wide range of target MCU> - and of course Renesas MCU such as RL78 and RA family.

    The algorithms are set-up for specific odor within the range of capabilities of the sensor, so anything outside of what we have characterised - thats' difficult to judge and would need deeper analysis.

    Not sure if you have access to the programming manual mentioned here, it gives a fairly wide explanation of what is supported and MCU requirements. 

  • Thanks JE.

    I’m working on a small project which is focused on low power consumption for a battery powered device, so I need to be able to minimize the processing at my end.

    The tech data for ZMOD4410 mentioned “embedded AI” which I took to mean that there is “pre-processing” done on the sensor before my own MCU receives the data.  And by that I mean that the flow of data goes from the gas sense element to the CMOS signal conditioning IC which is on the gas sensor module, before then being output to my MCU.

    It would be great if the IC on the gas sensor module could use AI to complete classification of the VOC’s and only output this data to my MCU when there is a change.  That way I’m not running the post-processing algorithms on my MCU constantly, especially if the air quality is remaining constant.  This will help reduce power consumption.

    Can you confirm if this is the way it works, or is all of the raw data output to my MCU constantly (at the 3min intervals) with no pre-processing.  Because if this is the case then I don’t quite understand what role the embedded AI plays on the gas sensor module?