Problem with validation of Hyperspectral rendering

radiance-general
1

Hello everyone,

I am aware that the spectral simulation capability is still under development; however, I am trying to use it in my project. I built a scaled room 743 (1/10) and conducted spectral measurements (location: 35°47’49.4"N 51°24’48.7"E) at three depths with several different glasses using a spectrometer.

section
section3000×1700 159 KB

Here, I have provided the transmission data for three of the glasses that I measured with the spectrophotometer .

glass
glass1652×993 180 KB

I performed spectral simulation based on the material information of the model (rad and mat files here) in the ALFA plugin and Radiance 6.0a, but the results are not encouraging :frowning:

In order to compare the spectral output of Radiance with ALFA, given that simulating 81 channels is currently not possible with -co as I understand, I modified one of the components of Lark plugin that was used to convert 9 channels into 400 channels so that I could also convert the output of 20 or 24 channels to 400 channels. (CSV file 380-780 5nm Exp,ALFA and Rad6.0)

chart1
chart14957×2975 858 KB

At shallow depths, the performance of Radiance is close to the measurements I took, and at greater depths, the performance of ALFA is also close. I changed every parameter that I thought relevant, but I have not figured out where I went wrong (ipynb file)

Any suggestions would be greatly appreciated.

1 Like
2

For deeper position try something like this, adjust ab, ad and lw until you see the values stabilize.

tr_params = pr.SamplingParameters(
    ab = 8,
    aa = 0,
    ad = 16384,
    lw = 0.00001,
    cs = 20,
    co = True,
)

If you have sky measurements, we recently introduced -L dirnorm difhor option to genssky, where dirnorm and difhor are for direct normal and diffuse horizontal illuminance. Running with this will give you a better integrated values. This option is not in pyradiance yet, but will be soon.

3 Likes
3

Hi Taoning,

Thank you
I have also performed simulations with similar and even higher settings, but there is still a significant difference compared to ALFA and Exp.

Has there been any comparison between the Radiance spectral simulation and other tools or exprimental measurements up to this point? ( except this [article] (https://publications.ibpsa.org/proceedings/bs/2021/papers/bs2021_30583.pdf))

I wasn’t aware of the new genssky feature. Thank you for letting me know! I’ll test it.

4

I am expecting quite a bit difference (>50% maybe) between your original one and the one I shared above. If you are not seeing any difference then there is something wrong with the process.

1 Like
5

Make sure you set -lw -12 or so. Positive values of -lw are associated with low accuracy when -aa is 0

1 Like