Talk:Steinhart–Hart equation

"Steinhart" vs "Steinhardt"

The correct name for the first author of the Steinhar[d]t-Hart equation is almost certainly Steinhart. I noticed that the name was not spelled consistently within Wikipedia. After considerable Google searching, it became evident that "Steinhart" is the correct spelling. I learned that "Hart" is Stanley R. Hart, still a Senior Scientist at Woods Hole Oceanographic Institute. Steinhart's full name, I believe, was John S. Steinhart, former Professor and Professor Emeritus at University of Wisconsin-Madison. See following links:

http://www.whoi.edu/science/GG/people/shart/publications.htm http://www.secfac.wisc.edu/senate/2004/0405/1775(mem_res).pdf

Coefficients

The coefficient D is referenced in the where clause, but is not present in the equation. Either the equation is incorrect or the where clause needs to be modified.

more data

I translated this article on the French Wiki, and put some calculation to use it. I let you import those equations from FR to EN (talking French is easier but not needed... it's math). (in French)Relation_de_Steinhart-Hart. Enjoy !

units

This equation doesn't make sense. You can't take the log of a resistance. You can only take the log of a dimensionless number. The datasheet I'm holding for some thermistors I'm using instead have R/R25 in the log - the ratio between the measured resistance and the resistance at 25 degrees celcius. The ratio makes it dimensionless, and without that it doesn't make sense - as it is in this article R cannot be resistance in Ohms. 118.139.33.247 (talk) 05:21, 15 December 2009 (UTC)[reply]

You are correct that the dimensions don't agree. This formula is only a curve-fitting approximation and not an absolutely-correct calculation. Morcheeba (talk) 18:21, 3 February 2010 (UTC)[reply]
- I would say that this formula is not even a good curve fitting approximation, as it adds 3-rd order term but misses 2nd order term or polinomial regression. Shcha (talk) 16:47, 9 April 2018 (UTC)[reply]
  - I agree, and is reflected in an updated I did in the article. Cstausland (talk) 18:29, 6 July 2019 (UTC)[reply]

Typical values and dimensions of coefficients

It might be useful to give some typical values of the coefficients A, B and C for a typical thermistor. Also, are the units of A, B and C simply Kelvin? — Preceding unsigned comment added by 130.246.121.103 (talk) 09:27, 29 July 2016 (UTC)[reply]

I agree, but could not find any proper reference for typical values. I wrote some Python code that calculates coefficients for 141 thermistors from TDK, which seem to support that A ~ 1E-3, B ~ 1E-4, C ~ 1E-6 and D ~ 1E-7. We need a reference, though. Cstausland (talk) 18:29, 6 July 2019 (UTC)[reply]

Credibility[edit]

This curve-fitting formula is basically a polynomial regression with missing 2nd order term. Omitting 2nd order term and adding 3rd order term makes no sense to me. I've tested this on a number or NTC sensors I work with, and they all consistently give better approximation with $[\ln(R)]^{2}$ instead of $[\ln(R)]^{3}$ , and even better of course when both are used. I would seriously question scientific credibility of this article and/or original paper. Also, I would question significance of this formula - this is just an inferior special case application of a well known approximation method. Shcha (talk) 16:49, 9 April 2018 (UTC)[reply]

In fact, many citations for this method say that it is not adequate. See this article, for example: "The results of this study indicated that the basic equation and Steinhart and Hart equations were not the adequate calibration equation for four table data of thermistor". Or this: "It was confirmed that the Steinhart–Hart equation shows a poor performance and should be replaced by the more suitable models recommended in White et al. (2014)." At very least I would like some clarification as to what advantages this formula has compared to more simple fitting curve ${1 \over T}=A+B\ln(R)+C[\ln(R)]^{2}$ ${1 \over T}=A+B\ln(R)+C[\ln(R)]^{2}$ ? Shcha (talk) 17:20, 9 April 2018 (UTC)[reply]
- I agree with this. I changed the article, including some nice references that agrees with this. One of them is the one you are referring to. Cstausland (talk) 18:29, 6 July 2019 (UTC)[reply]