Hi All, not sure if this should be in the test or the sensor section, apologies if I have mistakenly placed it here.

I am currently trying to calibrate this Pressure Sensor ( https://www.te.com/commerce/Documen...pdfEnglishENG_DS_SM9000_A1.pdfCAT-BLPS0062) The SM9336 250 Pa differential sensor (range -250/+250 pa). We are using it in a venturi meter, and calibrating it with a known volume syringe (3L). There is a picture of our test set up below. Essentially when we started our test, the count of the sensor (the reading which we receive from the sensor) was 5 at 0 pressure, which converted to pressure was 0.05 pa, so we thought this was our zero error and added this into our formula. The test setup was moved during our test, and each time the zero error would change, ending up at about 20 counts, 4 x our initial zero error. I understand that it shouldn't have been moved, but this sensor is supposed to be going into a handheld device, so it will be moved daily in any case.

I would like to know if anyone has any insight into why this could be happening, is it because the sensor is moving relevant to the MCU ? Whereas in a closed product they wouldn't be moving relative to each other. Would digital signal processing be able to help this ?

Thanks for any help

 

hi tee,
Its possible due to that very loose wiring assembly on the test jig, that it is subject to ambient electrical field pick up, that is caused by the changes in position of the project..
I would retest with a 'tidier' test assembly.
E

 

Welcome to AAC.
If I'm reading the datasheet correctly, the rated error for your sensor is in the range -1.5%FS to +1.5%FS and the FS count is ~ ±26000. You can therefore expect an error of up to ~ ±390 in the count. Your actual zero error seems well within this range. Any drift in the count may be due to temperature changes, as although the sensor has on-board temperature compensation it is not clear from the datasheet just how effective that compensation actually is. The position sensitivity of the sensor is rated as 0.1%FS/g, so that may also account for part of your zero error drift.

 

hi tee,
Its possible due to that very loose wiring assembly on the test jig, that it is subject to ambient electrical field pick up, that is caused by the changes in position of the project..
I would retest with a 'tidier' test assembly.
E

Hi Eric, Thanks for the response, I will try that out. Thanks for your help, could signal processing help with this also ?

 

could signal processing help with this also ?

You could set the value to zero at the start of each test.

 

Welcome to AAC.
If I'm reading the datasheet correctly, the rated error for your sensor is in the range -1.5%FS to +1.5%FS and the FS count is ~ ±26000. You can therefore expect an error of up to ~ ±390 in the count. Your actual zero error seems well within this range. Any drift in the count may be due to temperature changes, as although the sensor has on-board temperature compensation it is not clear from the datasheet just how effective that compensation actually is. The position sensitivity of the sensor is rated as 0.1%FS/g, so that may also account for part of your zero error drift.

Hi Alec, thanks for response. If we were testing in a room at 25C I doubt it was changing that rapidly that it would be noticeable, if it was is this something that could be filtered out? The problem is that I am now calculating the zero error before taking a reading but it is changing so frequently that it is changed by the the end of the reading (<5 seconds). Any idea on how I can account for this?

 

You could set the value to zero at the start of each test.

Hi Crut, Thanks, that is what I am now doing but it is changing so frequently that by the end of the test ( < 5 seconds) the zero error has changed and it is throwing off readings. Any idea on how I can account for this, is there a way I can clean up the data either via hardware (OpAmp/Filter) or by Digital filtering?

 

t is changing so frequently that by the end of the test ( < 5 seconds) the zero error has changed and it is throwing off readings. Any idea on how I can account for this,

Sure. Measure the zero error right after the end of the test and subtract that from the final answer.
Then the only drift that can affect the answer is that which might occur between the last two readings, which should be a lot less than occurs during the test.

is there a way I can clean up the data either via hardware (OpAmp/Filter) or by Digital filtering?

None of that will correct for offset error.

 

If the offset error is reasonably random then averaging a number of measurements would be beneficial. How rapidly/frequently do you need to take measurements?

 

Sure. Measure the zero error right after the end of the test and subtract that from the final answer.
Then the only drift that can affect the answer is that which might occur between the last two readings, which should be a lot less than occurs during the test.
None of that will correct for offset error.

Hi Crut, as we are using the pressure sensor as a flowmeter, we are reading a few hundred values from the sensor and the zero offset is changing while we are taking the values, meaning that we are getting lower than expected values (for the overall flow rate). and then the last few values where 0 pressure is applied we are seeing the readings at (-10) meaning the offset error has lowered during the test, and we think this is why we are seeing lower overall values.

 

If the offset error is reasonably random then averaging a number of measurements would be beneficial. How rapidly/frequently do you need to take measurements?

Hi Alec, Currently I read in a 100 samples at 0 pressure/flow and then average them to get the zero error before we do a test. Then we record 500 sensor readings in 5 seconds and send them via bluetooth/load into Matlab. After shortening the wires and using better quality wires and reflowing the solder on the pressure sensor to board mount the offset Error has increase to an average of 130 (ish) but varies every few seconds between 125 - 135 (ish). The problem we have ( we think) is that in the middle of our readings ( the zero error could have difference of ten (or more) on what was calculated before it went into the loop. Leading us to not get the readings we are looking for.

Is this common with sensor and has to be designed out with code and clever product design or should a high quality sensor not be acting like this?

 

Could it be possible that this isnt an off set error but rather a (threshold error) in that some values below the (proposed) off set value cannot be seen? As the offset has increased and we are subtracting this from all of our readings, our overall value (volume of air that has passed through the Venturi meter has gone down. Could this be possible? Thanks for all the help, I am struggling to understand how the readings are going so wrong, when yesterday they were close to the expected value?

 

we are seeing the readings at (-10) meaning the offset error has lowered during the test, and we think this is why we are seeing lower overall values.

So why wouldn't measuring the offset error immediately after the reading help that problem?

 

So why wouldn't measuring the offset error immediately after the reading help that problem?

So why wouldn't measuring the offset error immediately after the reading help that problem?

Hi Crut, I will give this a try, was resistant to do this as I was previously filtering the values as they were being read, but this way I will have to create an array of all readings and then apply average of the error before and after. Thanks