The Danger of Calculated Glazing Data

The Danger of Calculated Glazing Data

Last Updated: 

Thursday, September 21, 2017

Software/Versions Affected:

  • IGDB›All

This often-misunderstood issue deserves its own page. Optics5 is a freely available program developed by LBNL; other programs exist with some of the same functionality. O5 can use optical data for glazing components to calculate the properties of permutations of that original glazing. This can save a lot of measurement and assist with prototyping. The question that has often been asked: "If the data in the IGDB is accurate and the algorithms of Optics5 are correct then why can we not use it to calculate the properties of new glazing materials and submit them to the IGDB just like measured data?" The response is "what do you mean when you say "accurate" data"?

How Accurate Does the Data Need to Be?

Every ASTM (and NFRC) standard is supposed to have a "statement of precision and accuracy". It is one of the most difficult things to do properly. For one thing, the level of accuracy that you shoot for is determined by the application. If you want to do better for another purpose, then you will have to modify your standard test procedure which may involve much more time and expense. For the original purposes of the IGDB, it was simply desired to calculate the properties of a glazing system (layers separated by air) given the properties of the individual layers in the IGDB. Window5/6 does a fine and mostly failsafe job of this. If your measurements are off by say 1% then the final result may be off by 1%. Something of this order is usually considered adequate for the purposes of calculating energy performance values. Using Inter-laboratory Comparisons (ILCs) we ensure that all data submitters can perform to at least this level of accuracy. Some can do better than that already.

O5, however, requires something very different. From radiometric data (R and T), O5 extracts more fundamental optical properties and then puts them back together in new configurations. This sometimes requires much higher accuracy on the component data. For failsafe operation the component data might typically need to be accurate to 0.1% or even better but only in some parts of the spectrum. A typical trouble spot is the UV where all components have near zero transmission. Our current ILC does not test submitting laboratories to this standard of performance.

NFRC Issues Related to Calculated Data

Lack of safeguards: Using Optics5 to try out new things can be very useful and we encourage that type of usage. If something goes wrong then you can try to refine your data with more careful measurement or by resetting some values that are causing trouble. Usually the problems can be fixed with one or two attempts. Maybe you will decide to give up. The worst that can happen is that you have to make more samples and measure them all. Unfortunately, the desire to take advantage of the power of O5 has led NFRC committees to approve procedures without what I consider proper safeguards.

Infrequent (quarterly) releases: There are six IGDB releases per year. Usually this is an acceptable interval. When there is a problem with data submission however even for a standard submission this can delay the release of data for another 2 months which is usually unacceptable to the submitter. At present there is no mechanism for intermediate or accelerated review. Perhaps a fee would be necessary for special services. An option that has always existed is to simply submit all measured data. Unfortunately NFRC now has at least one procedure for Applied Films that requires the uses of O5.

Lack of access to the user database: Another practical issue is that to test the results of an O5 calculation, the component data must be in the official IGDB first. There is no option for entering component data into the User Database for testing prior to submission. Obviously this will cause disappointment if the data does not work in O5 and it does not give the submitter a chance to recover from the problem for at least 2 months.

Embedded Laminates: This is an issue within an issue. The ability to modify laminates with embedded coatings is the most requested feature for Optics5. It is often assumed that Optics5 does not contain this feature, but the fact is that our IGDB "Checker" tool does not have the ability to make the proper links between files automatically. There are some existing embedded laminates in the IGDB because they were added manually which is a time consuming process. Modifications to the Checker rather than Optics5 would be needed to make this a standard process.

Possible Solutions

  • Forbid all submissions of component data.
  • Allow submission of component data, with no guarantee that they will work properly in O5.
  • "Fix" the data. This is not as bad as it sounds. Usually the operation will fail when a property gets very close to zero or one. Intelligent analysis of the situation often leads to the conclusion that the slightly noisy values can be reset to be exactly zero or 1. Not always however.
  • Provide more funding to LBNL to review each submission. This can sometimes turn into a small research project for each submission. This could be used either for additional review and testing or for reprogramming the "checker" tool to find some types of errors.
  • Ask submitters (or simulators) to verify that the data they provide is adequate by:
    • constructing a number test samples and using O5 to model them
    • learning more about the process so that they can spot problems in advance
  • Perform a new ILC and perhaps qualify submitters at two levels of performance.
  • Enable the testing of component data from the user database prior to entry in the public IGDB.
  • More frequent IGDB release or accelerated release for a fee.