Calibration of an SQM

[Stub Mandrel]

I've made a basic SQM using a TSL2591 and an arduino nano. I'm aware of its limitations but I'm not too concerned if it looses precision at the ~0.002lux end of the scale - if the sky is anywhere near that end of the scale i won't have anything to worry about!

At the moment it is just reading Lux on the longest integration time/highest gain using the Adafruit library.

I plan to take dark readings for both photodiodes and subtract these before calculating lux using the Adafruit library routine.

I realise that calibration is not particularly important except for comparing with other people's readings.

The obvious route is to get a reading against an existing meter or at a location with a good estimate using Clearoutside or similar, but the latter requires a good clear night with astrodark....

I was wondering if there is an alternative way of calibrating, such as using a camera to take pictures of a plain wall in a dark room and calculate the from the image data while taking simultaneous readings with the meter.

Has anyone tried anything similar and can they give advice of calculating magnitude/arcsecond squared with a camera?

This is a lash up - the lens mount (its an ex-binocular eyepiece) is too long and being reprinted, followed by a better version of that light shield.

 

P.S. I'm probably going to add temperature compensation later, but that's down the road...

[pete_l]

Have a look at Samir Kharusi's work in this area.

Alternatively. there may be a way to use a full moon (like yesterday's) as a sort of "standard candle". The amount of illumination it supplies will wash out most other sources of light pollution. You'd need someone close by with an SQM to help/

[vlaiv]

I think that easiest way and probably most precise is to use telescope and camera as calibration source.

Procedure is fairly simple - just shoot piece of the sky - preferably away from the milky way with a bright(ish) star in FOV. Star should be bright enough to be in catalog - with known magnitude but not too bright to saturate sensor.

Then it is simple matter of doing photometry on that star and measuring background value / dividing with sampling rate so that you get photon count per arc second squared (or just ADU per that unit) - ratio of the two, star ADU count and background ADU count will give you basis for magnitude calculations - take star mag and subtract log of ratio to get sky mag (per arc second squared).

Compare that with your device reading in the same region to get one calibration point. Do multiple parts of the sky (different sky brightness values) to get good calibration curve.

[Stub Mandrel]

That method makes sense, but not easy at the moment.

One way of calibrating could be to use a light source with known luminosity. The problem is standard LEDs have all sorts of variation, but it must be possible to make one...

The sensor library calculates a reading in Lux and as 1 lux  = 1 cd/square metre, so you can calculate magnitudes/square arcsecond directly from this.

 The formula from unihedron's website  works (e.g. a reading of 0.0002 lux gives 19.83).

While this is uncalibrated the sensor goes past mag 22 before it reads 'infinity'.

I noted a few things last night:

You had isn't opaque enough to block all light!

If you do block all light successfully you get readings of mostly 0 with an occasional 1 for both IR and Full spectrum. This is impressive as the data sheet says it could be up to 20. I've added potential for dark subtraction but set the numbers to zero.

The lens appears to have a real world FOV of about 30 degrees (measured by seeing when and obstruction reduces the reading)

The lens has about 0.4 times the response compared to a tube giving about an 80-degree FOV.

This is quite a lot more presumably than you would expect with a smaller field of view.

Even with a fairly bright but starry sky last night (about 11:00pm so two hours before the moon was at zenith, pointed at the Plough area) it was reading about 19.8, so realistically under-reading by a few magnitudes.

The readings definitely show variations. Assuming these are essentially random (due to thermal noise, motion of the sensor etc.) this opens the possibility fo using over-sampling to improve resolution at low levels.

For a trial I'm using 4x oversampling and in conjunction with an estimated correction to the Lux reading.

 It all works and gives believable readings but I need to calibrate properly.

 

 

 

 

[vlaiv]

11 minutes ago, Stub Mandrel said:

One way of calibrating could be to use a light source with known luminosity. The problem is standard LEDs have all sorts of variation, but it must be possible to make one...

Do you have a camera and lens combination that has known QE / attenuation and Gain in terms of e/ADU (although that part you can determine)?

Simple incandescent bulb with a diffusing panel could be used than - or flat panel for that matter.

[Stub Mandrel]

I can probably set something up with the ASI1600 but I don't have a calibrated lens combination. I could try and calibrate from a sub with eth 130P-DS or 150PL but that's abit clumsy to move to the living room...

I think I may just be patient and wait for the next dark evening.

 

[Stub Mandrel]

I've found this, which will let me get a value using a DSLR:

https://www.translatorscafe.com/unit-converter/en-US/illumination/1-7/lux-candela steradian/meter²/

Reflected light

Determining exposure by means of measuring light reflected from the object using a light meter

For light meters that measure reflected light, the camera settings (f-number and shutter speed) are related to subject luminance and ISO speed by the following incident light exposure equation:

N²/t = LS/K,   (2)

where

• N is the relative aperture or f-number
• t is the shutter speed or exposure time in seconds
• L is the average luminance of the scene measured in cd/m²
• S is the ISO speed (100, 200, 400, etc.)
• K is the reflected light meter calibration constant; Canon and Nikon use K = 12.5.

 

So I can let my DSLR calculate exposure for a dim, evenly illuminated surface and calculate the luminance with:

L = N²/t . K/S

 

 

 

[Stub Mandrel]

That's interesting!

A rough and ready try with the camera gives me 0.127 cd/m^2 for the dark inside of the bottom shelf of my 3D printer stand.

The SQM gives me 0.39 Lux with the lens, but that's with a 'correction factor' of four so the actual reading (it bounces around a bit) is 0.10.

Without the lens it's reading 0.89 Lux, so 0.22 uncorrected.

Clearly 0.127, 0.10 and 0.22 are all in the same ballpark, the difference between 0.127 and 0.100 is 0.3 magnitudes.

It looks like spending some time tonight when I can get a more consistently illuminated surface sorted out it will do as an initial calibration.

 

[Stub Mandrel]

In a very dark room, white wall lit from a shaded lamp (with a hoodie over the top!) 4m away:

Camera 6 seconds, f3.5, 800 iso. Formula gives a value of 0.0319.

Meter varying between 0.0133 and 0.0128, but pretty clearly centred on 0.0130 Lux (correction factor set to 1.0).

Comparing the first figures were 0.39 and 0.22, ratio 1.773.

Second figures 0.0319 and  0.0130, ratio 2.454.

That's a significant difference BUT the second experiment was done under much more consistent conditions.

I'll put the 2.45 ratio into the device and then see how it measures up against real skies in due course.

 

 

[Stub Mandrel]

Did a test last nigh. At 11:45 it read 19.22 and at 1:15 it read 19.40, aimed at the darkest patch of sky (CO suggests I am 19.66, but that would be during astro dark).

Without access to the exact formula for calculating Lux, I can't be sure of the resolution but after changing my code to take four readings and average them the final SQM readings seem to have a resolution of about 0.02 right down to well past 22.