Friez ML-3A Barograph Accuracy - 31 Day Chart
For the month of January 2017 I analyzed the accuracy of my second Friez ML-3A microbarograph by comparing its recorded sea level values with calculated sea level values from my Henry J Green observatory mercury barometer. The barograph has a 31 day battery-driven clock transplanted from an Oakton hygrothermograph and used Belfort #15621 31 day charts. The chart and the clock drive are about 5% out of sync, so I needed to calculate the chart time corresponding to the actual clock time when reading data off the chart. Values were determined once-per-day near midday.
Four statistics were calculated:
- The average error was calculated. It is the average difference calculated as the mercury barometer pressure minus the pressure read from the barograph.
- The standard deviation of the individual error values was calculated. By definition, 68% of the error values fall within +/- one standard deviation of the mercury barometer value. The average error can be reduced by recalibration of the barograph, while the standard deviation generally cannot, it is more inherent in the performance of the barograph.
- The average drift was calculated. This is simply the slope of a straight line fitted to the the scatter of error data points plotted against day of the month. It is an indicator of how the error changes with time.
- Finally, the average non-linearity was calculated. This is the slope of a straight line fitted to the scatter of error data points plotted against the mercury barometer sea level pressure, in mb per mb of barometric pressure. It is an indicator of how much the error changes on average with the local pressure.
For the ML-3A with 31 day charts, the calculated average error for January 2017 was only -0.038 mb, but the standard deviation of the error values was a huge 6.44 mb. There was significant drift in the error values, amounting to -2.06 mb/month. The calculated nonlinearity was a very large 0.329 mb per mb of barometric pressure. Plots of the error values showing the scatter, drift, and nonlinearity are presented below.
Drift
Data for January 2017
Nonlinearity
Data for January 2017
The calibration of this ML-3A appears reasonably good, it matches the mercury barometer near the middle of the barograph chart as seen in the nonlinearity graph. For a 31 day chart, the barometric pressure can change significantly in a small distance on the chart and a significant pressure error can occur for a small error in the time when the reading is made. I prefer to have this microbarograph record at 31 days because it presents a concise view of the passage of pressure systems for the month.
For the month of March 2017 I will disconnect the dashpots from the barograph mechanism to see if this improves the standard deviation and nonlinearity. Also, instead of correcting the chart reading time at the end of the month I will try to use a #15621 chart that is stretched horizontally 5.1% using Microsoft Publisher. This isn't a perfect solution as the time lines on the chart will be somewhat distorted by the stretching. will repeat this evaluation in April using the March data.
Go to Friez ML-3A microbarograph page.
Go to Home Page.
For the month of March 2017 I will disconnect the dashpots from the barograph mechanism to see if this improves the standard deviation and nonlinearity. Also, instead of correcting the chart reading time at the end of the month I will try to use a #15621 chart that is stretched horizontally 5.1% using Microsoft Publisher. This isn't a perfect solution as the time lines on the chart will be somewhat distorted by the stretching. will repeat this evaluation in April using the March data.
Go to Friez ML-3A microbarograph page.
Go to Home Page.
