I have just received this sensor and have done some preliminary testing of the device.
I have a reference static water pressure of 200 kPa confirmed by two independent pressure gauges as well as a land survey that has determined water levels. When using the DFRobot sensor at the same tap point, I measure a signal level of 1.17 volts which equates to a pressure of 287 kPa (after adjusting for a zero pressure voltage of 0.443 volts). This equates to an error of about 5% FS.
I have also tested a pump boosted pressure which is measured by the gauges as 475 kPa. The DFRobot sensor indicates a signal level of 2.47 volts which equates to a pressure of 799 kPa and an error of about 20% FS.
I have used a well regulated 5.00 volt supply to drive the sensor. The claimed accuracy for this sensor is 0.5-1.0 percent FS. I think I have covered all bases in testing this sensor but I am interested to see if anyone has a suggestion on what else to do.
It might help to see my bench for context.
That is certainly a great test setup. It’s probably a fair assumption that your DMM is accurate.
This is could be within the FS error. We don’t have charts for this device, but Full-Scale error for sensors of these types is often non-linear throughout the range of performance. The value itself is measured at the Full Scale of the device, so 1% would be 16kPa at FS. Here’s an image I found that shows FS Error:
I’m not sure what the non-linear error is leading up to the FS error. You could map (and share!) the results so we can add your findings to the product page. Though correlation with a second sensor would be helpful as well.
What concerns me a little is that you may be dealing with a bad batch scenario. Do you have a customer that has recently verified an accuracy of 1%FS or better?
No known issues and the most recent review on the product page reports a similar non-linearity that you might be seeing:
Works according to the data sheet, but don’t buy for “liquid level detection” as stated in the “features”. The full scale is 1600kPa, which is 163m of water depth. My tank overflow is 2.2m above the base, which is about 1.3% of full scale. the error is quoted as up to 1% of full scale - so the error is 77% of what I am measuring. Still a 5 start product.
(Posted on 12 Mar 2021)
Whatever the outcome; the differences could be mapped and corrections applied. Especially if you have the convenience of other gauges to work with. There could be factors with the transducer that I’m not aware of that may introduce errors even with offsets applied, though I would suspect it would be a lot more accurate than using the formula alone.
I have recorded 28 data points between 100 kPa and 620 kPa for each of the two analog gauges that I have. I would like to have gone higher but I cannot produce higher pressures at present. The response is linear between 0 and 620 kPa. Extrapolating the response up to the 4.5 volt grid line would intersect at a pressure of around 1.0 mPa. This is suggesting to me that the sensor really has a working range of 0 to 1.0 mPa and not 0 to 1.6 mPa. Of course I cannot rule out that the response becomes non-linear beyond the range that I was able to test, I just feel that it is less likely.
Considering that my present project will have pressures below 600 kPa, it would be a trivial matter to calibrate the sensor to my application using the average response line that I have produced and possibly obtain accuracies generally around 1 percent. The big BUT is that I am losing confidence in the integrity of this sensor. From other forum discussions I’ve found, the manufacturer seems to go silent when challenged with questions of bad batches and calibration discrepancies.
Your suggestion of substituting the sensor with a probable working unit is good and what I would normally try to do. However the thought of shipping these units back and forth in the face of uncertain performance specification would only be attractive to Australia Post.
That’s some good intel Andrew. Bear in mind that all we have to work with from the supplier is the FS error, which is only valid on the max reading. It’s likely there would be some non-linear response toward the top end where the 0.5-1% FS error was measured.
If a second gauge has a similar response curve then it would be of benefit for others that we include additional info beyond what the supplier has shared. We are forever updating product pages with info; please do share your findings and we can update the page.
Awesome work @Andrew38773.
It is sad when a wall of silent is hit when hard questions are asked of the manufacturer. Assuming both analog gauges are correct, it definitely looks like the sensor you have is reading wrong. Unlike institutions hobbyists do not have the resources to buy a batch of items to test; or have the ability to accurately test the calibration of a device.
Possibly others have put these devices into use not realising they are inaccurate. Or it could be a dump of a faulty batch. Either way it is not a good look for DFRobot. Up to now they have been a trusted supplier for me. Pity it is not a University that is using the device, they would have more sway with DFRobot.
Anyway, thanks for the posts and work you have done, buyer beware,
Fyi, I shared this topic with the supplier and they agreed there was an error with the multiplier for the sensor in their guide. An update should be rolling out in the next couple of weeks.
Until then, you ought to be fine to make that adjustment yourself based on the data you’ve collected.
That’s good to see that DFRobot is addressing this problem. It is to be hoped that it acknowledges that the issue is a bit deeper than just publishing a revised multiplier. I will make the following comments:
• If the device is non linear over the specified range then a (gradient) multiplier by itself is not meaningful.
• I’d venture to say that most people would be using these devices in the middle third of their specified pressure range and this is where you need to have some confidence in the accuracy / stability. Claimed accuracy at the extreme limit of the range may be of academic interest but it is not helpful particularly if the accuracy / stability is significantly worse in the range where the device is actually used in practice.
• If the device is non-linear and there is variability from unit to unit then this should be stated up front. The need to calibrate a non-linear device is not a deal breaker for most users of such devices, particularly in the hobby arena. Take thermistors for example which can be strongly non-linear but once calibrated over a range of interest can be very accurate and useful.
• If the claim is to be made that this device is in fact linear and that the published multiplier is incorrect then it would seem that the specified range limit of 1.6 mPa is also in need of revision.
It will be interesting to see how DFRobot promote this device in the future. Externally the device appears to be well made so it would be a shame if it was let down by inadequate documentation.
It’s unlikely to be non-linear after further comms with them.
Though at the time and assuming documentation was correct, then it was the only reasonable outcome based on a max rating and FS error (along with untested results for the second half of the range of operation).
Adjust the coefficient, and get that project completed
I see that the DFRobot application wiki for this product is now specifying a pressure range of 0 to 1.0 mPa rather than the previously understood range of 0 to 1.6 mPa. That document has also revised the gradient multiplier from 400 to 250. Core Electronics is currently still promoting this as a 1.6 mPa device. Core is in good company however, because all the other suppliers of the sensor throughout the world, including the DFRobot Sales site itself still promote it as a 1.6 mPa device. To be fair it will take some time before the world-wide distribution channels are all updated. The wiki also appears to be redefining the accuracy of the device to 0.5% of measured value.
The bad news is that the device has been derated from 1.6 mPa to 1.0 mPa. The possible good news is that the accuracy of the device has been increased. Previously the apparent error on a 200 kpa reading was claimed to be16 kPa. (1% of FS). The revised accuracy for the same reading would appear now to be 1 kPa (0.5% of reading). I’d wait for further confirmation before relying on this though.
I hope that DFRobot eventually issue an errata to all purchasers, past and present, of affected devices so they may review just how critical the revision might be to their designs.