Davis Barometer Drift With Temperature

For those interested, I’m posting a study looking at Davis barometer drift as a function of temperature. See Power Point attachment. Let me know if you have questions or file format issues.

Thanks,
James
DW4536


Pressure Differences WRT Temperature.zip (368 KB)

So what’s your conclusion?

Hello Nico

The purpose of the study was just to determine the pressure drift WRT temperature, specifically pressure differences. That’s what the study did for a range of temperatures and local pressures. Many readers will draw different conclusions based on how they want to use this data. If the study results are meaningless to a reader then that reader should ignore the study.

For my purposes, the Mean Pressure Differences shown on each graph indicate that my Vue and ASI DBX2 are calibrated the best that one can do given a 0.1 mbar calibration step. If the magnitude of the Mean Pressure Difference was greater than 0.05 mbar then a 0.1 mbar calibration step should be made.
Since I send the Vue pressure to CWOP, I desire to keep the error in the reported pressure to 0.1 mbar or less. Others may have different error criteria.

I use my ASI DBX2 as a reference barometer for calibrating the Davis Vue console. By knowing the temperature of the Vue console, I can estimate the range of pressure differences that I could see between the calibrated DBX2 and the calibrated Vue.

I hope that answers your question,
Thanks,
James
DW4536

I see, can I extrapolate that to conclude that the Vue baro meets its specification?

Hello Nico

The Vue, GE Druck graph indicates that “my” Davis Vue console pressure transducer in operating well within the Davis specification of ±1.0 mbar. You can also conclude this from the Vue, ASI DBX2 graph Of course, the Vue and DBX2 barometers get periodically single-point calibrated (due to long-term drift) to insure they meet specifications. The Vue console is setup to send altimeter/ QNH pressure, so the data going to CWOP is correct.

I should say that the GE Druck Barometer introduced no significant error into the study.

Thanks,
James
DW4536

Interesting, thanks :slight_smile:

These measurements are all indoors - correct? How do you ensure that the air pressure in the test location isn’t affected by e.g. forced air heating/cooling, or wind and open windows/doors etc.? Does there need to be a correction because the air inside is locally at a different (higher or lower) temp than the outside air?

Hi Nico

Yes, all measurements are indoors.

The test was not affected by other factors because pressure differences were used.
Also, the pressure from home forced air handlers are measured in IWC, inches of water. atmospheric pressure is measured in inches of mercury. Big difference. Assuming your air handler has a return duct, the net interior pressure is a fraction of a inch of water. This is insignificant when compared to the pressure from 20 to 30 inches of mercury.

No correction needs to be made in the study to compensate for indoor/outdoor temperature differences.

Nico, I’m curious if you own and operate a personal weather station that reports to NWS through CWOP?

Thanks,
James
DW4536

Got it, poorly expressed. I guess I was looking past the differential sensor test to the final objective of demonstrating reporting atmospheric pressure to an accuracy of 0.1mb.

Also, the pressure from home forced air handlers are measured in IWC, inches of water. atmospheric pressure is measured in inches of mercury. Big difference. Assuming your air handler has a return duct, the net interior pressure is a fraction of a inch of water. This is insignificant when compared to the pressure from 20 to 30 inches of mercury.
Although the net interior pressure is very small is there not a potential issue with local pressure increases? Most systems that I have owned or seen have a single central return register. Surely there is a local variation in pressure in a closed, or semi-closed, room with only supply registers. After all the study is designed to support reporting to an accuracy of 0.1mb which is only 0.04 inches of water.
No correction needs to be made in the study to compensate for indoor/outdoor temperature differences.
In the differential study I agree. My question relates to effect on the accuracy of the reported pressure of air which is heated or cooled vs the outdoor air (since we are considering a very precise measurement).
Nico, I'm curious if you own and operate a personal weather station that reports to NWS through CWOP?
Own, yes, for many years, report to CWOP, no. I have an official airport station a couple of miles away. My data has always tracked has very closely with that station and I can't hope to compete with the accuracy of several $K of equipment with Davis equipment, so for geographical and technical reasons I don't see that sending it to CWOP would have any real value.

Hello Nico

“Although the net interior pressure is very small is there not a potential issue with local pressure increases? Most systems that I have owned or seen have a single central return register. Surely there is a local variation in pressure in a closed, or semi-closed, room with only supply registers. After all the study is designed to support reporting to an accuracy of 0.1mb which is only 0.04 inches of water.”

To the extent that air handler pressure effects barometer pressure, the study used pressure differences so those effects are cancelled. The degree that I attempt to calibrate my barometer was not the subject of the study and is not the subject of this post, but, to answer your question, I do not apply corrections for air handler, wind surface pressure on the structure, open or closed windows, return duct distance from the barometer, etc. Neither does the NWS or the FAA.

“In the differential study I agree. My question relates to effect on the accuracy of the reported pressure of air which is heated or cooled vs the outdoor air (since we are considering a very precise measurement).”
Again this is beyond the scope of this post, but, I’m sure there is a small effect. I believe the NWS and FAA systems consider air field temperature in their QNH reductions. As I understand it, if the pressure sensor is at the same altitude as the reference runway point, then air field temperature is not used and the sensor’s pressure is “standard atmosphere” reduced to calculate QNH. In other words QFE equals local measured pressure. If the pressure sensor is at a different altitude than the reference runway point then the altitude difference and the air field temperature are used to reduce the local measured pressure to a QFE pressure. Then this QFE pressure is “standard atmosphere” reduced to sea level for the final QNH value. I hope this helps.

Thanks,
James
DW4536

Thank you for your time :slight_smile: I understand very well the scope of your project now.

Thank you for your comments, Nico.
James
DW4536

Based on an interest to see Davis Vantage Vue pressure druft data down to 68 deg. F, I’ve updated the 1st order Vue chart. See attachment.

Thanks,
James
DW4536


Hello all

I have been asked to clarify the drift. Based on the latest data, the barometer drift with respect to temperature are:

Davis Vantage Vue: -0.0216 mbar/deg. F
ASI DBX2: +0.0141 mbar/deg. F

Thanks,
James
DW4536

I ran an experiment and the worst case pressure increase (closed room at the start of the ductwork from an HVAC unit with two inlet registers and no return) was 0.2 to 0.3 hPa. So I finally wonder no more about this :slight_smile:

If you had a Vue, you could correct that by increasing the room temp 10-15 deg. F

During the winter heating season it could be self correcting :wink:

Hello All

Well, I’ve finished collecting data. The final pressure drift WRT temperature values are:

Davis Vantage Vue: -0.0212 mbar/ deg. F
ASI DBX2: +0.0134 mbar/ deg. F

Thanks,
James
DW4536

For some time now, since the graphs in this study were specific to my Davis Vantage Vue Console, I’ve wanted to answer your question, “How do I get a graph that is valid for my Vue Console.” For the remainder of the post please refer to the attached graph. The most important element is the Mean Pressure Difference Model. That model is what has to change to make it specific to another Davis Vantage Vue Console.

So, let’s start by a general discussion of the model so you know how to use it. This model is the average pressure reading difference between the Vue Console and the GE Druck lab. reference barometer as a function of inside/room temperature. So it is a average pressure error WRT temperature. Since the GE Druck pressure reference has a pressure drift WRT temperature much, much less than the Vue Console’s pressure sensor, we can assume that the average error is entirely due to the Vue Console pressure sensor. Therefore, it is the Vue console’s average error. From here on out, I will just refer to this as the average error model or average error.

The average error model is:
average error = C + D T, where:
C is a term related to your console’s calibration. This term is what you need to determine for your console.
(D T) is the product of D and T, where D is the pressure sensor’s error WRT temperature and T is the inside temperature shown on the Vue Console at the time you read the console’s pressure.
Note that we a talking about local station pressure not a reduced pressure like Altimeter, NOAA or SLP pressure. Also note that the D term is valid for all Vue Consoles using the same pressure sensor as is found in my console. I assume Davis is using the same pressure sensors in all their Vue Consoles.

The average error model shown in the attached graph is:
average error = 1.65608 - 0.021844 T, where T is inside temperature reading (deg. F) on the Vue Console, C =1.65608 and D = -0.021844.
The room that contains my Vue view console will vary in temperature between 70 and 80 deg. F during the year. Given that then the average errors can be calculated so follows:
At 70 deg. F, average error = 1.65608 - 0.021844 70 = +0.125 mbar
At 75 deg. F, average error = 1.65608 - 0.021844 75 = +0.0178 mbar
At 80 deg. F, the average error = 1.65608 - 0.021844 80 = -0.091 mbar
The 0 average error temperature = -C/D = (-1.65608)/(-0.021844) = 75.67 deg. F

To determine the model for your console you need to have it calibrated. A calibration lab will calibrate your pressure sensor and, if your ordered it, will also give you the calibration data. This data is usually a table over the calibrated pressures and contains the actual pressure, the vue console pressure, the difference in these 2 pressures or the error, the calibration adjustment made and the temperature during the calibration process. The calibration adjust is usually made to force the error to zero. For the remainder of this example, we will assume the adjustment cancelled the measured error and the after-calibration error is zero.

When you receive your console back after calibration, your need to compute your C term as follows:
Let’s assume that your console is calibrated at 67 deg. F and the calibration data indicates that the error after calibration is zero.
Then, C = -D T = 0.021844 67 = 1.463548, sorry about the number of decimal places.
The after-calibration model for your Vue Console is:
average error = 1.463548 - 0.021844 T
The after calibration errors for inside temperatures of 70, 75 and 80 deg. F are:
At 70, average error = 1463548 -0.021844 70 = -0.0655 mbar
At 75, average error = 1.463548 - 0.021844 75 = -0.175 mbar
At 80, average error = 1.463548 - 0.021844 80 = -0.284 mbar
A quick check for the 0 error temperature is: T = -C/D = (-1463548)/(-0.021844) = 67 deg. F

The last step is to drive the zero error point toward the middle of your yearly room temperature range. In our example the middle room temperature is 75 deg. F and the average error calculated from the after-calibration model at this temperature is -0.175 mbar. Rounding this error to 1 decimal place gives an error of -0.2 mbar. Your want to minimize the average error so the need to use the Vue console’s pressure calibration function and enter a +0.2 mbar calibration. Now your need to adjust your average error model C term by +0.2.
Your average error model is now equal to 1.663548 -0.021844 T
Recalculating the average errors at the 3 temperatures gives:
At 70, average error = 1.663548 - 0.021844 70 = +0.134 mbar
At 75, average error = 1.663548 -0.021844 75 = +0.0252 mbar
At 80, average error = 1.663548 - 0.021844 80 = -0.0840 mbar
Your 0 error temperature is: -C/D = (-1.663548)/(-0.021844) = 76.16 deg. F

Now to produce your graph, your just plot the average error model as a function of temperature: 1.663548 - 0.021844 T
To add the single sample 95% confidence bands, plot:
upper band = your average error model +0.1
lower band = your average error model - 0.1
Again, the 95% confidence band shows you where a single pressure measurement will fall 95% of the time. See the study. This band of possible errors is due to the uncertainty in the model due to sample size, sensor hysteresis errors, sensor repeatability errors and sensor linearity errors.

I hope this is of use,
Thanks,
James B
DW4536