Is DCGlare and annual DGP the same?

I am currently using VI-Suite for Blender, to conduct some testing regarding glare analysis. As i have come to learn, i can come here to ask any questions for the radiance engine the VI-Suite uses.

As my company would like to implement a beter way to control for annual glare, i have used the feature of DCGlare. However when making sure it is calculated in accordance with NS-EN 17037 Daylighting of Buildings (formula in the picture underneath), i am now a bit unsure if DGP is the same now as DCGlare. The reason i am asking is that BREEAM-NOR requires the DGP annual.

I looked into the link (underneath) and for me it seems as DCGlare is DGP?

Could someone with some more experience clarify this for me. If however it is a different metric, how well does it measure up against the DGP? Can we say for sure that the quality of DCGlare can be seen as DGP, and is there any argument listed that we can use it for BREEAM check.

I forgot to mention, i have read this paper:

However, i find it hard to interpret an English academical paper.

Before implementing I would recommend looking a paper from 2022 about “A critical comparison of annual glare simulation methods” :


The DC-glare method is referred as “imageless DGP”. If you look at RMSE, its performance is similar than just using Ev as glare predictor, which we know that it is not appropriate in many cases. Specifically there is a systematic under-prediction of glare caused by EC-glazing (for both Ev and DC-glare) , which I find quite critical especially if such a method is applied in consultancy projectscomparing different facade solutions or for certifications.

Here are the core comparisons of different annual glare calculation methods from that paper:

And as numbers:

Hi @olemarius,

You can find an explanation of the dcglare method in Fast Climate-Based Glare Analysis and Spatial Mapping. It does use the DGP formula, with a substitution described in Equation 4 of the paper that enables calculations based on the input matrices. The main limitation of this is that it does not account for specular reflections, so it should not be used in cases where there shiny surfaces. However, it does offer much more specificity than just using Ev as a glare predictor.

If you have a typical Radiance model with only plastic and glass material modifiers, dcglare should fit your needs just fine.

Hi Nathaniel,

Good to have the discussion :wink:.

In your paper you write at the end “.…it may not perceive the sun as a glare source through ectrochromic
glazing.” - which is exactly what we experienced as well. And EC glazing is modeled as glass material (so in that case it would actually not fit the purpose I guess :wink:)… Using glass material with very low transmittance values, as well as BRTDfunc material (e.g. to model fabric shades) we found a systematic underprediction of glare of more than one glare category (also on the 95% percentile).

We also found that using a high transmittance glass it didn’t show this bias, as well not for trans material (at least not when the material had no extreme peak ).

But I want to ask this question : When there is a glass with high light transmittance, for what reason do we need to run a sophisticated glare analysis? A simple check if the sun is in the field of view will tell you the same thing: The frequency how often glare occurs (because if the sun is visible, then there is glare). So in my point of view a annual glare calculation method should work properly with all typical shading options and not just for clear glass where we actually dont need to quantify glare. DGP was developed exactly for that purpose to differentiate between different shading solutions (and not to tell that there is glare when the sun can be see through a normal glass).

So the question is if there is any plan to modify DCGlare so that it works also for such systems/materials without a bias.

In any case, the raytraverse method seems to me to be the more comprehensive solution since it can consider any facade material than can be modeled in Radiance, plus also specular reflections that can be also more complex (e.g. from opposite buildings, opposite metals or PV-systems).

Hi @Jan_Wienold,

The basic idea of dcglare is to look for sky patches that could be glare sources (the contrast term in the DGP formula) in combination with Ev (the brightness term in the DGP formula). This is more advanced than a sun visibility test like ASE or a brightness only approach like DGPs, but less so than image-based DGP. Specular surfaces and BSDF-type materials don’t really work with this assumption because they change the direction of the ray and therefore break the connection between view direction and sky patch.

The issue with EC glazing is that in two-phase calculations, the sun is smeared over the entire sky patch, so it is not bright enough to appear as a glare source when viewed through EC glazing. A simple solution would be to use a much higher MF for the sky subdivision to constrain the sun to a smaller region and boost that region’s brightness. As I recall, dcglare allows up to MF 12, but I never had the desire to run simulations that fine.

At one point, I worked out the effort that would be involved to include specular and BSDF materials in the dcglare process. Essentially, it would require storing a matrix of received radiance from each viewing direction - basically what amounts to an image. This seemed to defeat the point of the imageless method, so I did not pursue it further.

So, if your model uses generic diffuse materials and high-VLT glazing, which I would argue is the case for over 90% of Radiance models, then imageless calculations should be sufficiently accurate. Otherwise, you can use one of your other methods.

Evening Jan and Nathanial. Thank you both for your answers.

May i interoperate? In accordance with the second article provided by Nathanial i read from figure 3 that for DGP and image less DGP (DCGlare), the difference in percentage can yield as much as 4%. Did i understand the chart, and that imageless DGP is DCGlare from radiance engine?

I understand furthermore that figure 5 mentioning the increase of sky patches reduces the error of accuracy. Though based on figure 3 to achieve the 4% error, how many sky patches do you have to use? In accordance with the first article provided by Jan, the source of error seems for me to be as high as 9 percentage , or did i missunderstand the article? I have tried to interperate that article before, unfortionaly my English falls short here.

Also, Nathaniel, in your article, how many sky patches do you advice to use?

When it comes back to the metric of annual DGP_40% = 5 percentage , can one potential approce be that the 4 % error as previous mentioned be simplified in a way where you argue that 40 percent discomfort has a 40*(1-0,04) = 38 percentage to account for a potential error?

When it comes to limitations for high reflective surfaces. Will this one day be implemented? I have noticed now and then some buildings have either a exterior metal surface, or interior, which is metal and thus yields glare when walking by. To peforme such analysis to detect would come in good use im sure.

So as i understand i can say to my employer that yes DCGlare use the formula with some modification. The source of error can be ish 4%, we have to use x sky patches to do our simulation. Also the day light standard when it comes to glare is not fully developed as it derives from a thermal comfort standard, and may i assume ashrae?

Hi @olemarius,

I generally use MF 4, which seems to strike a good balance between speed and accuracy in most cases.

If I understand your question correctly, you are asking whether you should multiply the results of imageless DGP by some factor to calculate DGP. You should not do this. It is important to differentiate statistical correlations, which are what @Jan_Wienold and I examine in our articles, from conversion rates, which imply a constant relationship between two metrics.

You may notice from @Jan_Wienold’s paper that imageless DGP, calculated by dcglare, predicts glare with more accuracy than other approximations like ClimateStudio, which are accepted for regulating glare, as I understand. So if you are attempting to meet the regulations of EN 17037 in a location where you would be allowed to use ClimateStudio, then dcglare’s accuracy should also be acceptable.

For metal buildings that create specular reflections, dcglare will generally not identify them as glare source. However, the DGP metric is not intended for outdoor applications, so it would not be meaningful anyway. If you want to understand glare in those cases, you should be using an outdoor glare metric like the one described in Methodology to Assess Potential Glint and Glare Hazards From Concentrating Solar Power Plants.

Thank you for pur reply. At tr moment i am experimenting with VI Suite for
Blender. Climate studio seems however to yield a bether visual display.

For your other feedbacks much apritiated. Then I take it the feedback I can
give is that the accuracy is somewhere around 5
% and thus a meteic that can be applied, of course with the limitations you
ate mentioning.

MF 4 is then how many skypatches?

MF 4 is 2305 sky patches.

Answering over phone whilst beeing out seems to have given many misspellings on my last reply. My apologies. Never the less ut seems we were able to communicate rather well.

All in all i am greatful for your valuble feedback. If i may summerize and forward an answer to my employer, it can be this:

The use for Blender VI-Suite is using the equation from NS-EN 17037 Daylighting of Buildings. Thus DLCglare meets the requirements for annual DGP calculation. However, there is a difference in how DLCglare calculat the E_v component. As it uses a simplified imageless approach rather than a fish eye calculation, it yields a correlation which has an error rate around 5 percent compared to a image based calculation of E_v.

The applications for DCGlare only apply for indoor simulations as the source of glare is the skydome itself, and the sun. High reflective surfaces can not be detected with DLCGlare. In this instance a Fish Eye calculation most be used with a specific time calculated. Electocromatic window is also an application which DLCGlare can not take into consideration.

To achieve the small error of 5 % it is advisable to use MF4.

Have i understood your guidelines now?

Actually, your interpretations are not correct:

  1. The 5% relative deviation you get only for cases where you don’t need a glare calculation: a) Clear glass with high transmittance and sun is visible (=always glare larger than any threshold) b) Clear glazing and sky visible → DGP is always lower than any glare threshold c) Venetian blinds activated → Always meet the glare criterion according to EN17037 (explicitly mentioned in standard, therefor no calculation needed)
  2. For shading options that are customizable (fabrics with OF) or electrochromic glazing neither DCglare (in its current version) nor Ev are suitable glare calculation method since both methods significantly underestimate glare (negative bias). For such cases the deviations are much larger that 5% and might be even 2 glare categories.
    Shading devices based on fabrics have a market share of around 40% on the European shading market, therefore this is relevant.

In the next revision of EN17037 there will be very likely specifications on boundary conditions for the simulations to meet the standard, since this was requested in an inquiry. This will include the resolution needed for sampling the glare source (sun or a reflection of it) and which cases (=shadings and/or materials) a calculation method needs to be able to correctly represent (including fabric shadings and glass with low transmittance). In its current version and low patch resolution DCglare would not meet these specifications, as well as Ev or DGPs (that would never meet).

I’m really wondering for what reason the “90% of models” with only clear glazing need an annual glare calculation based on DGP? Areas in a floor plan that potentially have glare without any shading devices can be easily detected by sun-visibility checks or simply using Ev. You don’t gain any extra information calculating DGP for such configurations.
On the other hand, 100% of the non-residential buildings (in Europe) have shadings applied to prevent from overheating and glare (from which around 40% are fabrics) . The EN17037 (and actually also the EN12464 (light and lighting for workplaces) require on the workplace positions 95% of the usage time to prevent from glare (DGP smaller than 0.45 for the minimum glare category). In case a simulation is required, this means the façade and shading has to be modeled appropriately and DGP reliably calculated. For Venetian blinds that can block the visibility to the sun no detailed calculation is needed since it automatically fulfills the highest glare category. For the remaining shading systems neither DCglare nor Ev are able to properly predict glare, besides the fact that glare from opposite buildings or PV panels cannot be predicted.

I learned while working in another standardization group responsible for the characterization of shading devices that the shading manufacturers are very keen on methods that reliably and fair evaluate their products. I’m wondering how they would react when they realize that certain products are favored by simulation tools by underestimating the glare potential, especially if these are products from a competitor. So in my point of view we should avoid implementing methods that biases the evaluation of shading systems to have a fair competition. Further, by installing system that perform worse than the simulation predicted, the occupants of such buildings will suffer and it is our responsibility to avoid this.

Good morning, apologis for the late reply. It takes time to digest all the feedback received.

I agree with you that when giving our advice to clients, it should be in our best intrest to accurately give the best advice. And thus, i do hope to learn from all of you.

As it stands with our projects when we want to meet our clients wish for points for glare in accordance with BREEAM-NOR V6, our manual requires:

  1. The project assesses the likelihood of glare from daylight (see Definitions) in accordance with NS-EN 17037:2018 -A1:2021 Daylight in buildings for the building (see Method and Definitions).

  2. The project has drawn up and implemented a strategy for mixture control (see Definitions) for relevant areas of the building and where the probability of DGP (Daylight Glare Probability - see Definitions) exceeds 0.4 in more than 5% of the annual service life (see Definitions).).

  3. The chosen strategy for glare control is optimized so that daylight can enter when it is cloudy or when the sun does not shine on the facade. The strategy will help to reduce the unnecessary use of artificial energy lighting. The location and design of the solutions chosen must not conflict with lighting control systems.

In my hope i would like to avoid the simplified manual assessment you can use in NS-EN 17037:2018 -A1:2021 with its tabell E.2-E.6 To assess wall to window ratio, view parallell or perpendicular and distance of the Person to the window. This Methode surely most yields a much higher source of error than any simulation.

As i have come to learn from your reply, any shading system will be biased. As i understand it, the strategy for finding a good blind/screen is therefore point 7 in our BREEAM-NOR. I can read based on your feedback that no simulation per now can meet this requirement.

However, to map out area of a building which is problematic from glare point 6. and 7. is an assesment without any shading. As it stands i have to according to our BREEAM-NOR manual then either do a manual check, or i can use annual DGP. As i am reading the two of you, i must make a choice between plague or cholera (analogy of what to choose as Methode). I am still too unsure with the current feedback if i can meet 5. and 6. with DCGlare, or if i still must use the simplified tabell E.2-E.6. I would like to give a note that if still our current standards and simulation is not accurate enough i do belive you in that, and hopefully we can soon have more accurat tools.

as you are mentioning high transmittance, the normal glazing option in Norway seems to always lean towards LT70 window do to wish for both as many workstations in the space as possible, in combination with densification meaning more shadings. If LT70 is deemed as a high transmittance i am not sure, as this is the standard now here. I can understand that point 7. most then be met with another approach. I am also agreeing with you that most office building have blinds/screens to avoid overheating, however i am forced to use the BREEAM-NOR which requeirs more and more documentation which is not always a need for in reality as many things might be coverd with ex. thermal comfort studies.

I am happy to extend the talk, however the main point of discussion is (my fault for not including the BREEAM-NOR quote) what alternatives i have for DGP without blinds. I might need more time to digest all the feedback. I would however like to hear @Nathaniel_Jones remark of the side discussion i have and the main inquiry i have. It might be easy for experts to address a science report, however as a non-English speaker i unfortunately do find it difficult to read academical papers.

Hi @olemarius,

I’ll add one additional caveat to @Jan_Wienold’s, witch is that the substitution used by dcglare is in the contrast term, not in the brightness term of the DGP equation.

I’m not an expert in BREEM, but I think what Jan is saying is that in cases without shading devices, a direct view to the sun is the main cause of glare, and evaluation of the contrast term may be unnecessary. That said, there doesn’t seem to be a widely accepted metric for evaluating indoor view-to-sun glare (except perhaps for ASE), and BREEM seems to require use of the DGP equation, regardless. (And, I would argue, dcglare does use this equation, so it meets the criterion that you’ve laid out.)

As to fair competition, all of the methods we are discussing are free (both gratis and libre), so anyone complaining that their competitor is using a more advantageous method is free to use that method as well. That said, I’m not sure why a manufacturer would be using these methods. Just like with UGR and electric lighting, the glare score of a product in a lab has little relation to the glare experienced in a real use case. What manufacturers should insist is that designers use consistent methodology when comparing products.

What I’m curious about, @Jan_Wienold, is if Ev and DGPs aren’t accurate enough to predict glare occurrence in simulation, then do you think DGP is accurate enough? If we’re talking about photographs, where the luminance field is known with certainty, then sure, we can use DGP to assign a glare category, but rendered images just don’t have sufficient accuracy for this. Unfortunately, both your paper and mine use DGP as a baseline for comparison instead of survey data, which is fine for academic discussion, but doesn’t answer whether the results are accurate enough for real use. Saying that DGP is 100% accurate compared to itself isn’t that impressive…

@J_Alstan_Jakubiec found in a long-term study that DGP can misidentify the glare category by two categories, with what appears (from my reading of the figures) to be a bias for being too high. My field work comparing HDR photography to several different simulation methods found that when DGP is near the transitional 35-45% range, it’s not uncommon for simulations to be off by 5%.

Keep in mind that I was using expensive equipment to precisely measure material reflectance and sky conditions. In predictive simulation, that information is not available, which will result in larger errors. Even using a database of measured values like spectraldb, a search for “white paint” results in a reflectance range of 30%, which is certainly enough to throw off DGP calculations. A thread on the accuracy of Radiance simulations lists a number of validation studies, through which I’ve found that typical expectations of Radiance are that it produces results within 20% of measured illuminance values. An error of 20% in Ev is enough to reduce DGP from 45% (intolerable) to 39% (perceptible). Again, these studies mostly used measured reflectance values. For simulations using LM-83 recommended Lambertian reflectance values (20% for floors, 50% for walls, 70% for ceilings), the error is probably greater, and those are the “90% of models” that I referred to earlier.

So that is your choice between plague or cholera (which is my new favorite metaphor). None of these methods are certain, but each is useful in some cases. Standards like BREEM are problematic because they set specific targets, where your results will be influenced mainly by the accuracy of your representation of materials, and also to some extent by your choice of algorithm. In my opinion, the best you can do is to make sure you follow the wording of the standard. This is why I prefer simulation to be used for comparisons, and why I emphasize the limitation of dcglare for specular surfaces (because with the right settings, dcglare can still help you rank the relative glare risk of open-weave shades and EC glazing, but it will not correctly rank different specular materials). Of course, I also contend that standards should require reporting of simulation error margins, so as you can see, no one is listening to me anyway :wink: .

1 Like

Thank you for your latest feedback @Nathaniel_Jones . I have been reading your comment with great intrest and had a need to digest your answer as well.

Glad to hear your found that metaphor humoring! It is a Norwegian expression when you are forced to make a choice where you can see that either or of your choice is bad.

I am leaning towards your though that simulation is comparable, and far beter than the manual check in the standard. I will forward a reply based on the discussion here to my supervisor and i hope for an approval as i see the benefit of using the DCGlare module.

I would like to address an apritiation for both of you to participate in a discussion with me. It is helpful to observe the community is active and have different opinions.

May i ask a last question Nathaniel?

I have noticed your AcceleradRT has some kinda support for Rhino so it is quick to see effects of objects you hide/unhide. Would that feature be hard to one day implement in Blender, as the BlenderBIM community is active though at the moment there is few environmental tools.

Hi @olemarius,

Glad to be of help. The Blender question is probably something to move to a new thread. Briefly, AcceleradRT just watches for the octree file to be updated and reloads it when this is detected. It’s not a particularly elegant method. If you can get Blender to overwrite the octree file, then the same will work there, too.

1 Like

Understood. I have received positive feedback from my supervisor, and we will perform after xmass a test of DCGlare for a ongoing project we have. I will look into how i can somehow make a fitting for general blinds to see how this will impact the annual results. As far as i can understand i will therefore have to investigate how both plastic and translusent material is being processed and how the different setting impacts a results. If you have a good paper or know a old tread which covers this please direct me there, i have tried a little to see what i can find with out much results.

With regards to your comment on the octree and Blender, i will follow your inquiry for a new tread if this is something i am to pursue it further.