Page 2 of 2
Re: Accuracy of Detectors (Overall)
Posted: February 18th, 2020, 6:02 pm
by MTV<=>Operator
MoMoney$$$;)0) wrote: ↑February 18th, 2020, 5:49 pm
MTV<=>Operator wrote: ↑February 18th, 2020, 5:45 pm
MoMoney$$$;)0) wrote: ↑February 17th, 2020, 7:13 pm
It's quite easily actually once you know what you're doing, and if you're acquainted with the language you're working with. I'm pretty sure he's talking about using the Stienhart-Hart equation given with a thermistor, and then changing it with some testing.
Changing it as in adding a constant? I don't see how one could find new coefficients in the 20 minutes of allotted calibration time.
More than doable with the right tools. This can be like quickly using a good graphing calculator to take points with corresponding voltage and temperature. Then finding a line of best fit. Personally I've done it in a time crunch, with a little under 20 minutes. Definitely can happen.
Every time I have tried making a linear regression for voltage vs temperature, it ends up accurate for cooler temps and off by as much as 2 degrees for hotter temps. I was able to collect 5 data points in 20 minutes and each was in a different temperature range. When I collected 30 data points, I discovered that the slope for hotter temps is much greater than for cooler temps. Is there another type of regression that models this more accurately?
Re: Accuracy of Detectors (Overall)
Posted: February 18th, 2020, 6:26 pm
by MoMoney$$$;)0)
MTV<=>Operator wrote: ↑February 18th, 2020, 6:02 pm
MoMoney$$$;)0) wrote: ↑February 18th, 2020, 5:49 pm
MTV<=>Operator wrote: ↑February 18th, 2020, 5:45 pm
Changing it as in adding a constant? I don't see how one could find new coefficients in the 20 minutes of allotted calibration time.
More than doable with the right tools. This can be like quickly using a good graphing calculator to take points with corresponding voltage and temperature. Then finding a line of best fit. Personally I've done it in a time crunch, with a little under 20 minutes. Definitely can happen.
Every time I have tried making a linear regression for voltage vs temperature, it ends up accurate for cooler temps and off by as much as 2 degrees for hotter temps. I was able to collect 5 data points in 20 minutes and each was in a different temperature range. When I collected 30 data points, I discovered that the slope for hotter temps is much greater than for cooler temps. Is there another type of regression that models this more accurately?
A non-linear regression with multiple parts, with set ranges you already have set. I use a tinspire where I can sort through data, and then easily look through the data. Then you put it into data and statistics, go into regression and test all of the types of regression for the best fit. Personally, I have found using multiple parts with polynomial functions to work the best.
Re: Accuracy of Detectors (Overall)
Posted: March 2nd, 2020, 6:39 am
by lameloball
Has anyone else notices that pretty much the only thing that matters is how precise the calibration thermometers are?
Re: Accuracy of Detectors (Overall)
Posted: March 2nd, 2020, 6:50 am
by ericlepanda
lameloball wrote: ↑March 2nd, 2020, 6:39 am
Has anyone else notices that pretty much the only thing that matters is how precise the calibration thermometers are?
it's not really the only thing that matters -- usually, the thermometers are precise enough, and your device really does still have a huge impact on your results, but if a thermometer isn't precise it can indeed throw a team off by a considerable margin. It's just one of the many things that are wrong with this event.
Re: Accuracy of Detectors (Overall)
Posted: March 2nd, 2020, 11:37 am
by jinhusong
ericlepanda wrote: ↑March 2nd, 2020, 6:50 am
lameloball wrote: ↑March 2nd, 2020, 6:39 am
Has anyone else notices that pretty much the only thing that matters is how precise the calibration thermometers are?
it's not really the only thing that matters -- usually, the thermometers are precise enough, and your device really does still have a huge impact on your results, but if a thermometer isn't precise it can indeed throw a team off by a considerable margin. It's just one of the many things that are wrong with this event.
If it is regional, you bring your own thermometer and you have unlimited time to calibrate it before coming to competition. As long as the thermometer is repeatable, you should be able to calibrate your device to it.
Re: Accuracy of Detectors (Overall)
Posted: March 2nd, 2020, 1:07 pm
by MTV<=>Operator
Is there any reason why a team would not calibrate at all at a state competition? I recall seeing a top team not calibrate at an invitational and they got first place in the event. Also what is a sufficient number of values to collect when calibrating?
Re: Accuracy of Detectors (Overall)
Posted: March 2nd, 2020, 8:35 pm
by jinhusong
We did at invitation running state rule. 5 data points from scratch for thermistor, 0.1, 0.0, 0.1, 0.2.