Odd Si7021 humidity readings over 6 nodes in same jar

[WhiteHare]

I measured the standby current on the el cheapo si7021 breakout board (above), and it was around 8uA.  Looking closer, I noticed it had a voltage regulator on it.  After removing that and shorting across its Vin and Vout pads, I'm now measuring the roughly 60nA standby current that's in the datasheet.

[joelucid]

Yeah it's a very low energy sensor. Has been running fine for years with my coin cell motes. BTW, humidity about 90% outside today and the si7021 needs heater level 2 to keep humidity at the sensor <75%. That's 15mA! If I end up doing this I'd probably use 4x Eneloop to power so it last a couple of weeks.

My two new si7021 motes now both measure 43% RH here in my office after two days of conditioning. Four other si7021 nodes in the same place measure 62%, 58%, 53% and 52%. None of the nodes have recently experienced high humidity.

Now some of them I've treated pretty badly - reflowed a couple of times after having shorts underneath the chip etc. But the two 43%, the 58% and 52% weren't mistreated.

Sigh

[WhiteHare]

In comparing the el cheapo si7021 breakout board (BoB) to the el cheapo BME280 BoB, I notice that aside from the sensor chips themselves the board layout and other discrete components are nearly identical.  So, for comparison, I unsoldered the voltage regulator on the BME280 BoB and shorted across its Vin and Vout pads also, and then did a standby current measurement using the Dave Jones Microcurrent in combination with my Fluke 87V (the same instruments I used to measure standby current on the el cheapo si7021 BoB above).  Doing that, the standby current on the el cheapo BME280 BoB measured at 148na.  So, not as low as the si7021, but I have no complaints with that. 

[joelucid]

My two new si7021 motes now both measure 43% RH here in my office after two days of conditioning. Four other si7021 nodes in the same place measure 62%, 58%, 53% and 52%. None of the nodes have recently experienced high humidity.

Did a test over saturated nacl solution today and amazingly the two new si7021 motes are accurate. Everybody else is off by >10%. I could easily see this being the result of contamination by dust during their lifetime given I don't have covers on.

5 new Si7021 with covers arrived today. I'll replace the problematic chips and see if that leads to a stable th mote.

Tom, I think you mentioned your motes were stable. Do you have covers on?

Joe

[WhiteHare]

I wonder whether your new si7021 nodes with covers are also more resilient against high humidity?  e.g. imagine your particular dust/contamination is hydrophilic, so that once it absorbs moisture it resists letting it go.  If that dust/contamination is in contact with your sensor--because it lacked a cover--then maybe it skews high for a long time after high humidity exposure, which is what you originally observed.

[joelucid]

I wonder whether your new si7021 nodes with covers are also more resilient against high humidity?

It's possible. But for the outdoors installation according to datasheet I think one really should use the heater for prolonged rainy periods.

I've got the new sensors installed. They measure:

76.6%, 77.5%, 75.2%, 78.5%, 77.1% over NaCl (should be around 75%). So all but one are in spec. The 78.5% one measures 0.3 C lower in temp than the others so for dew point measurements it's closer.

I've looked at my long term charts and humidity has systematically been increasing since the sensors installed. Moral of story: you really do need a cover or a different type of dust protection.

Joe

[WhiteHare]

I recently built a little experimental wireless temp-rh node (see attachment).  This particular one has a detachable si7021 breakout board hovering over the RFM69HW, and on the back there's the usual atmega328p-au, an LED, a resistor, and a capacitor.

Originally, I hadn't really intended for the breakout board to be mounted "upside down."  However, thanks to this thread, I'm starting to think it might actually be better that way, because it seems that dust will be less prone to settling on the sensor cover over time.  Anyhow, thought I'd mention that in case there's merit to the idea, as it may turn out to be one of the  "give backs" in the spirit of open source hardware to help improve things for everyone going forward.

[CallOne_Phil]

I have not had a lot of time to do more testing, but you guys have been killing it with all these posts. Finally got the new BME280 sensors running with accurate results so far on an open air test of 66%(see attachment).  Didn't have much luck drying out the si7021 boards naturally in my sealed 60%-63% container and I'll need to look into using the build in heater when time allows.

Time will tell if the BME280 sensors have the same drifting issue as we are seeing with the si7021, based on what the datasheet says I am thinking it is possible.

[Felix]

Finally got the new BME280 sensors running with accurate results so far on an open air test of 66%(see attachment).

Phil,
Thanks for the follow up!
Very encouraging results from the BME280, I feel quite good about it 

[WhiteHare]

@CallOne_Phil
That does seem like very good agreement, especially for humidity sensors.  Are you seeing similar close agreement on temperature and pressure?

[joelucid]

I asked the guys at silabs about the max humidity level beyond which to switch on the heater of the si7021. Here's the response:

Typically for relative humidity sensing, it is recommended to use the heater with 80%RH or greater. However if the only point of interest is the dew point and if the power budget allows for it, running the heater with about 5C of heating at all times is recommended. The heater does not affect the dew point calculation in a significant way. Please refer to App Note AN607 for more details on the heater in high RH environments and dew point calculation.

Joe

[WhiteHare]

I asked the guys at silabs about the max humidity level beyond which to switch on the heater of the si7021. Here's the response:

Typically for relative humidity sensing, it is recommended to use the heater with 80%RH or greater. However if the only point of interest is the dew point and if the power budget allows for it, running the heater with about 5C of heating at all times is recommended. The heater does not affect the dew point calculation in a significant way. Please refer to App Note AN607 for more details on the heater in high RH environments and dew point calculation.

Joe

I guess "5C of heating" means heat until it's 5 degrees C above ambient temperature?  About how much current is that? 

It might be worth asking as a follow-up what the benefit of the constant 5C heating is, or equivalently, what happens if you don't constantly heat 5C.  I'm guessing sensor drift, but it would be good to know what the magnitude is both with and without the constant 5C heating, because then one can rationally decide whether it's worth doing or not.


[Edit: Actually, though, reading the apnote (https://www.silabs.com/Support%20Documents/TechnicalDocs/AN607.pdf) it may not be that.  It looks like the main concern is whether the sensor has a lower temperature than the dewpoint (at which point dew will condense on it and thus foil the humidity measurement).  Maybe simply keeping 5C above ambient mitigates against that happening (?).  So, aside from the 80%+ RH case, maybe you can limit yourself to doing the 5C heating only when the sensor temperature is close to the dewpoint--extremely unlikely indoors in a habited space, for instance.]

[joelucid]

My take is that even high humidity levels below 80% will cause shifts in the long term. Increasing temp by 5 c will decrease rh by about 25%, keeping rh readings always below 75%. That appears to be unproblematic.

[WhiteHare]

It's just amazing how much due diligence is apparently required before buying even a simple humidity sensor these days.  I think this sort of thing should be unambiguously and prominently disclosed in the SiLabs datasheet, in bold red and large font, not merely hinted at or left to us to discover by other means.

I guess whatever sensor I finally settle on, I'm going to make sure it's detachable so that it's easily replaceable and/or future upgradeable to some other I2C sensor.  I hope the BME280 is more forgiving, but now I'll always have a nagging thought (regardless of the sensor type or the make or model) that there's a gotcha that I won't find out about it until after it's deployed.

[Felix]

It's just amazing how much due diligence is apparently required before buying even a simple humidity sensor these days.  I think this sort of thing should be unambiguously and prominently disclosed in the SiLabs datasheet, in bold red and large font, not merely hinted at or left to us to discover by other means.

I guess whatever sensor I finally settle on, I'm going to make sure it's detachable so that it's easily replaceable and/or future upgradeable to some other I2C sensor.  I hope the BME280 is more forgiving, but now I'll always have a nagging thought (regardless of the sensor type or the make or model) that there's a gotcha that I won't find out about it until after it's deployed.

I'm with you. Imagine how I feel after manufacturing with a sensor that has such a gotcha  .
But by default (assume you have no clue about this gotcha and are shopping for a nice TH sensor) would you rather trust SiLabs or a noname chinese boat fab that makes the DHT11?