This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

DRV5053: Sensitivity Range Min/Max Thresholds

Part Number: DRV5053
Other Parts Discussed in Thread: DRV5032

We've been working with the DRV5053 in our product, we've successfully qualified it on the bench, but have been going back to the data we've been accumulating on our manufacturing line to ensure that at volume our firmware algorithms will remain robust with all the product that we are producing.

In short, we are inducing a magnetic field incident on the DRV5053 and taking a set of measurements with our main processor, and then averaging those as a one-point calibration value which we can utilize in the field to make up for the vast sensitivity range specified. However, from the distribution of the values we're seeing over the past ~4k samples, there is ~1% of parts that would statistically fall outside of the expected sensitivity range per the datasheet. Below is a histogram of the distribution we are seeing, including the secondary distribution well above our expected range of values.

These values are repeatable, and even swapping subassemblies with different mechanical parts (and new magnets) has not been able to sway these units within what should be expected based on the peak of our distribution and the tolerance of sensitivity on the datasheet.

Is there something to attribute this wider than expected tolerance to? Are these defective parts?

  • Tyler,

    Can you provide details what the horizontal axis in your plot represents?  Also, can you describe how your tests were carried out to evaluate the sensitivity of the sensors?

    Often in our own test setups it is necessary to characterize the part with uniform magnetic fields while removing nearby ferromagnetic materials to help prevent any unintentional fields from interacting with the device. 

    Thanks,

    Scott

  • Hi Scott,

    The horizontal axis is in mV, and it is the offset from 1V that we are reading on our microcontroller. We read the voltage input from the sensor, then take the absolute value of 1V minus that value (to be able to utilize either polarity sensor or magnet if we so choose).

    Our tests are run on completed products, where a magnet is in proximity to the sensor and we force a set of readings to be processed. Particular product which reads very high has been binned out as failures for further testing. I've tested these units with a number of other parts from known "passing" units, replacing the magnets as well as the plastics which locate the magnet proximal to the sensor. These changes have no considerable effect on the values read from the sensor when the test is reran.

    We are assuming that if there were other materials or devices which could affect this magnetically, that we would see that more widely, or that it would follow a specific part of the unit/test. As far as we can tell, this has just followed the PCB which has the sensor installed on it.

    Let me know if this clears things up or if you need more explanation

    Regards,
    Tyler

  • Tyler,

    Thanks for the details, this helps me.  Looking at your nominal voltage shift, for any particular sensitivity I see the max min range would capture a normal range of about 0.1 - 0.4 V for device variant.  In addition to this variation due to the sensitivity, it would also be good to look the the Quiescent output voltage of one of these devices.  The voltage range for the output with no magnetic input is shown here:

    An additional 0.13 V to the output range we see due to sensitivity error might account for some of these units. Can you power a few of these units and see what the output voltage is without a magnetic input?

    Also, I expect it to be less likely, but was there any chance of a significant temperature variation for the outlier devices?

    Thanks,

    Scott 

  • Scott,

    Sorry for the delay - I was out sick at the end of the week last week.

    I was able to test three samples that are in the higher distribution, and below are the results:

    • SN-288: Quiescent = 4mV; Active = 584mV
    • SN-889: Quiescent = 20mV; Active = 637mV
    • SN-890: Quiescent = 5mV; Active = 650mV

    I was able to reliably stay within the same range for multiple attempts, and these values are the averages. I also took one of the samples (SN-288) and placed multiple other subassemblies with multiple different magnets (all of which were in units which were NOT in the second, higher distribution, those results are below:

    • Sub #1 (original subassembly) = 572mV
    • Sub #2 (from failed low unit) = 585mV
    • Sub #3 (from normal passing unit) = 498mV
    • Sub #4 (from beta product) = 583mV

    Across all those different subassemblies, the same high values occurred, though there is some tolerance for variation in placement and magnet strength. I'm going to also take the entire unit apart and reassemble to see if anything changes there (which I'm not expecting). I will follow up on results from that when it is completed.

    To answer your final question, we are manufacturing in a temperature controlled environment, and we are also taking temperature measurements on the fixture, which will fail if our temperature is outside a 10C tolerance from 25C, so I don't think there is any element of temperature skewing this data.

    Regards,
    Tyler

  • After having fully disassembled and re-assembled the unit, I'm still seeing the same results, averaging at/above 575mV when the magnet is present.

  • Tyler,

    Which package sensitivity variant are you using?

    Thanks,

    Scott

  • We are using the -EA sensitivity variant 

  • Would it be possible for you to capture the package markings on one of your outlier devices?

    Also, can you share a schematic or describe any components you have connected on Vcc and Out?

    Thanks,

    Scott

  • Scott,

    Below is the marking I saw on the failing unit, which looks to me like "+AKVA." Out of curiosity after reading that, I broke down a known good unit and found its marking was "+AJEA," I've also added that part's markings as well below.

    I assume the last two characters of the marking represents the sensitivity variant, but both of these units were from the same order from our CM, so they should have parts from the same reel, or at the very least reels from the same order of parts from a distributor.

    Could it be that we somehow got the wrong sensitivity variant mixed into our reel? 

    As for the schematic, I'd prefer not to share on a public forum, but I will describe the connections for you. VCC is driven off of a GPIO from our micro, which can source 6mA. VCC also has a 0.1uF capacitor as a bypass element. OUT is directly routed to another GPIO on our micro, but it has a 1kOhm current limiting resistor to prevent any over voltage effects on our micro.

       

  • Tyler,

    It looks like you did somehow get a mixed shipment from your CM. The sensitivity range for the VA device is about 2x that of the EA, which correlates to the data you provided.  Do you know how the units were shipped?  Were these cut tape or did you receive them in a reel?  Do you still have any of the TI original packing related to these devices?

    Thanks,

    Scott

  • Scott - I will relay these questions to our CM and verify how and when they bought these parts. I'm not sure if they still have any packaging from the purchase, but I will also check into that.

  • Tyler,

    In the event that the units were repackaged by the CM, the most likely explanation is that they gathered units from the wrong reel when filling your order.  

    Thanks,

    Scott

  • Scott,

    We have received response from our CM about the how and when we ordered these parts. We have only ever ordered the DRV5053EAEDBZTQ1 part number.

    Our first shipment was ordered in February 2022 from Rochester Electronics. These were purchased as a stop-gap in our supply chain for our usual digital sensor, DRV5032, which was unavailable at the time. We purchased a quantity of 37299, in as complete of reels as we could get.

    We build our product serially with respect to our serial numbers, and below you can see the sporadic nature of the serial numbers we'd associate with the incorrect variants based on our data. 

    This looks to me that the reel given to us by Rochester, or the reel given to them by TI was at fault and had parts of the wrong variant placed into them.

    Regardless, this answers the question of why we were getting these values, and we will be able to work from there.