Hellow from Poland (Minsk Mazowiecki)

[lukaszgryglicki]

Maybe also base64

[Andy Perrin]

12 minutes ago, ulf said:

Andy, how do you change the wavelength range of that simulator?

Export to Excel using the button on the top right, and then plot in Excel. I have already plotted above, though. There is a solar line or maybe many there. So it looks like it's probably not UV-B after all, since the sensor can only push things towards the long end, and there's no UV-B light to work with in the pass-band of the stack. Depending on where you put the UV-B cutoff of course, but still, it's so much more sensitive on the long side, that it's got to be mainly the UV-A leak. My bet is that the peak is right around 320-330nm. 

 

Here is the data, extracted from the original spreadsheet, for Global horizontal irradiance:

 

Solar Irradiation at Minsk Dec 21.xlsx

[ulf]

With different settings in the options you get a better graph on the web site directly.

Here I used 1nm steps.

I used the latitude of Mazowiecki (53°N) for this plot. 

It seams to be an increase of radiation of a factor of 3 for every 10nm

[Andy Perrin]

Ulf that is not what I get at all? 

- lon: 52.23

- lat: 21.02

- date: 12/21/22

- time: 1 pm

 

 

[lukaszgryglicki]

That would move peak to upper edge, still around 320 nm IMHO because after this filter stack curves go down very fast.

 

EDIT: (after seeing the final chart of global irradiance) I'm Ok with slope going down to almost 0 at 300nm (and rising starting from 310 to 320 and then 330) but I really really wonder why it rises again BELOW 300nm ?? see for example 290nm - this is almost like 330nm - totally don't understand it - it looks like a bug for me unless explained.

 

 

[ulf]

My mistake. Mixed up longditude and latitude

[Andy Perrin]

Nah, sensor curve goes up fast too, which we are not accounting for yet. It's going to be mostly in the 320's I think. 

Jonathan's data is here: https://jmcscientificconsulting.com/mini-uv-camera-matrix-vision-bluecougar-with-the-sony-imx487-sensor-spectral-sensitivity/

 

 

[lukaszgryglicki]

1 minute ago, ulf said:

My mistake. Mixed up longditude and latitude

Good catch, didn't noticed...

 

[ulf]

The 50% width of the combined filter is ca 17nm.

Difficult to see the slope of Andys diagram (Tou shalt not divide by zero), but if the slope is similar to what I saw that would make a gain difference of ca a factor of five between 300nm and 317nm.

With the additional slope of the sensor multiplied in  I would also say mostly light from around 320nm. 

 

To save the day there are different definitions of the borders between both UV-B , UV-A and VIS, so if we search and chose the one with the longest wavelength  It could still be UV-B-ish images  or at least an even mix of long UV-B and very short UV-A

 

Just as I expected this is a borderline case.

Still it is interesting to see what lurks down there.

 

Thank you for sharing Lukasz.

 

 

[Andy Perrin]

This is Jonathan's data on top of the solar data, both normalized to 1 at 400nm:

 

[lukaszgryglicki]

Any ideas why red curve goes to zero from 320->300 but then goes up?

[Andy Perrin]

3 minutes ago, lukaszgryglicki said:

Any ideas why red curve goes to zero from 320->300 but then goes up?

Because the sun has Fraunhofer lines (and also there are atmospheric absorption lines too)! There are probably one or more solar absorption lines or atmospheric lines in that region. Sometimes the lines are clumped together.

[lukaszgryglicki]

I need to research this, I wander what absorbs around 300-310 but not in 280-300 - don't know what Fraunhofer lines are - need to educate...

EDIT: OK Fraunhofer lines are just absorption lines, this is something I know - need to find/check what causes this absorption around 300-310...

 

[Andy Perrin]

Ozone does. It seems to be ozone mainly. But it's strange how it goes back UP after that. Compare to this graph:

From: http://tid.uio.no/ozone/about_uv_o3_cloud.html

 

Let's see if we can find another solar calculator and verify the <300nm side. 

 

You can se that the surface UV without ozone would be much higher - maybe the reason has to do with the estimated ozone layer thickness in your location vs. wherever this above graph was made (we don't know...)

[lukaszgryglicki]

Yeah, I was thinking that ozone absorption is mostly in UV-C region... and below 200nm ozone is actually generated by UV-C.

EDIT: I also know that ozone blocks all UV-C, some UV-B and passes UV-A - but then why curve goes up below 300nm ?

[Andy Perrin]

I am wondering about whether the model has an issue in that region. Remember this is generated by a model, so we should verify with some other source. It's suspicious how fast it jumps up at exactly 300nm.

 

Ozone absorption definitely explains the 320nm to 300nm decrease, however. 

--

UPDATE! I figured it out. The original paper for the model is here:

https://www.nrel.gov/docs/legosti/old/2436.pdf

 

If you look at their table of absorption coefficients for ozone, alpha_{o, lambda} is given starting at 300nm. So nothing past 300nm can be trusted here.

[lukaszgryglicki]

Totally agree.

[Andy Perrin]

Hopefully the final word on this whole thing (is it ever the final word on here?? ). Here are all the graphs and the combined graph, which shows that the peak (defined as 50% of the max value) went from 313nm to 322nm. Which is arguably UV-B.

 

All graphs are linear scale. This is not useful for detecting leaks, but I don't think there were any.

 

First the individual spectra. These have been interpolated with cubic splines. 

 

 

Then the combined spectrum, which is the product of the above spectra. The red lines are the 50% of peak levels and they are at 313nm and 322nm.

[lukaszgryglicki]

Thanks for compiling this, I only wonder about sensor sensitivity - your Nikon D850 plot vs. my monochrome quartz GFX 50R - in other thread I was informed that my sensor should be able to "see" 253.7nm mercury line, but you suggest that my sensor sensitivity drops to 0 at 300nm, I actually don't belive this single thing, others LGTM.

[ulf]

1 hour ago, lukaszgryglicki said:

Thanks for compiling this, I only wonder about sensor sensitivity - your Nikon D850 plot vs. my monochrome quartz GFX 50R - in other thread I was informed that my sensor should be able to "see" 253.7nm mercury line, but you suggest that my sensor sensitivity drops to 0 at 300nm, I actually don't belive this single thing, others LGTM.

Drop to zero is maybe not the correct expression.

 

However the GFX 50R do not have a BSI sensor as the D850 do.

There have been several indications that a BSI sensor have a deeper UV-reach and higher sensitivity in the UV range. 

 

Andy's findings above could be interpreted as: if you had used a Nikon-D850 for your images it would have recorded light from 313nm to 322nm.

We still do not know the sensitivity characteristics of your non-BSI sensor, but it is likely worse than the D850.

 

Your results are still quite interesting as they are from a wavelength range rarely seen in images.

The label UV-B might not be that important?

 

For me "drop to zero" mean that there are no possibility to detect anything.

That is when all imaging signals drown in background noise, regardless if incoming light.

You have that for silicon sensors somewhere above 1000nm where silicon becomes transparent.

I do not think that is the case for your camera at 300nm or shorter wavelengths.

 

Seeing light from a very strong 253.7nm mercury line directly from a light source is quite different from getting a picture of any scene illuminated with such light.

That might still be doable with your camera, if other wavelengths are rejected properly with very high OD  filtering.

The intensity at the light source is magnitudes more intense, compared to the reflected light from an object.

[lukaszgryglicki]

Yeah, I know that my sensor is FSI not BSI, but BSI one (from GFX 100 for example) also has IBIS which is something that Dan is not modifying for mono currently. Also it uses PDAF known to have other problems, while 50R uses CDAF (which is slower for normal photography, but this is not an issue for me).

 

[dabateman]

Lets not be too hard on the FSI sensor. 

Remember this comparison: 

https://www.ultravioletphotography.com/content/index.php?/topic/3730-multispectral-monochrome-nikon-d850-testing-and-shakedown/&do=findComment&comment=33297

 

Jonathan's D800 (imx094) is FSI and it's hindered with a 280nm blocking filter, but still seems to have similar sensitivity as the D850 in UVB range. The stop difference maybe unnoticeable below 330nm.

[lukaszgryglicki]

Yeah, I hope my sensor is still able to see at least that 254nm - this is the most available UV-C wavelength and the one I decided to focus on.

[ulf]

2 minutes ago, lukaszgryglicki said:

Yeah, I hope my sensor is still able to see at least that 254nm - this is the most available UV-C wavelength and the one I decided to focus on.

I hope so too for you as you have gambled quite a lot of money into that hope.

 

It is very dangerous to base too much conclusions on the data at the very extreme and low end of these measurement graphs.

 

[lukaszgryglicki]

Not really gambled, I just wanted broad band mono, the only "gamble" part is quartz glass (especially for UV-C) which is $500, all other "items" I wanted anyway. There is also UV-A, UV-B and IR which are possible with this setup. Also I have extra resolution in visible or whatever I choose due to Bayer matrix beeing off.