Higher dynamic range than a TSL2561 light sensor.

[ziplock9000]

I'm currently using a TSL2561 based module to track ambient light levels outside. The problem is that it becomes saturated (65535) when in normal daylight and registers zero when it's still light enough to read. I've tried all of the gain values and none of them work. Yes I know the human eye's logarithmic dynamic range is wonderful etc. But does anyone know of a sensor (actually a complete module) that has a better range in a similar price range?
Ideally it would not clip when at the brightest of UK weather nor in the darkest of UK winter nights.
FYI the TSL2561 is  0.1 - 40,000+ Lux range, but I'm not sure if that's for a single gain value or all of them.

Thanks.

EDIT: I found the TSL2591 with a range of 188uLux to 88,000 which is much better, but would still be clipped in daylight.

[DaJMasta]

If the dynamic range is sufficient of the replacement you've found, then you can always put a filter in front of the sensor to reduce the light slightly to fit the range of the sensor.  From wikipedia, the 188uLux is roughly equivalent of starlight on an overcast night, is that dark enough?  Direct sun seems to go a tad higher, but if you could attenuate the light coming in even just slightly, it would probably fit your range.  You can also go the multiple path sensor route for way extra dynamic range, though it's added complexity.


For attenuation, you could even mount your sensor a set distance from a window, then lower its mounting to be farther from the sensor, letting less light in through the window, or you could mount it so that it's showing outside but not into direct sun (glare could make it clip, but you can always just throw out short term clipping results in software.

[ziplock9000]

Good ideas thanks.
Yeah I was thinking the same thing re: the filter, trading the bottom end for the top end on a TSL2591. I could use the TSL2561 to partially re-calibrate the TSL2591 once a filter has been applied. Obviously that would assume that the shape of the response curve is the same, just shifted which may not be true, but good enough for my needs.

Or as you mentioned I could use the TSL2591 for the low to medium ranges and the TSL2561 with a filter just for the high ranges. I could use the same re-calibration trick too on the TSL2561 once the filter was added.

Anyway, TSL2591 has been ordered so I'll see the light soon. 

[Twoflower]

If you ain't afraid of the small package size you might have a look into this little one: VEML6030. They claim to go up to 120kLux. I have the RGBW version (VEML6040) of it and it seems to work nice so far.

[ziplock9000]

Very interesting thanks. Do you have a support circuit schematic to share?

[Twoflower]

The datasheet has a circuit diagram. It needs only VCC, GND and I2C. the VCC needs to be reasonably clean (some caps for filtering). Nothing else.

But be ware of the size. I showed the datasheet to a colleague who's very experienced and he said "No problem, you can do that. And by the way order one for me too." After I handed the device over he changed his mind :-) Still I managed to solder it on top of wires.

[ziplock9000]

2mm x 2mm  Oh My!

I wonder if I can get them in module form with caps already added and connections that don't make atoms look big 

Vishay have a sensor board for $37.56 after shipping and VAT. Just not worth it. It looks like there's no Chinese module either.

I can get the sensor for £3.49 GBP on eBay and I have the other components.

[Twoflower]

I've warned you about the size :-)

Mouser sells the Vishay dev board for 11.05€. And unfortunately this one is sold out: http://store.linksprite.com/linker-veml6030/ Maybe contact them to see if if and when new boards are available.

But there are other similar sensors out there. Just check if the OPT3001 suits your needs (e.g. if 0.01Lux...83kLux is sufficient). It seems that there is an Arduino board available.

[Twoflower]

One more idea. Since the sensors are damn cheap. Do a dual channel and use a neutral-density filter in front of one of the sensors. This way you get the best resolution at low light conditions and the non clipping at high luminosity situations. Probably go for an ND3.0 (0.1% of light passes through) as this might give enough head-room and enough overlap between both channels.

If you're lucky you might find a cheap one at a photo shop or photo rental shops. You probably can live with a partly defective filter.

If you can't get a photo-filter you might use a piece of foils used for car windows. And compare both sensors (read out at the same time) to identify the filter-factor.

[Twoflower]

Just for your estimation how this sensors react on daylight.

As mentioned I use the RGBW device from Visay and finally I managed to run the thing during the morning. My setup is the sensor 'looks' at a white paper card which faces the sun through a window. And the white channel saturates in the shortest exposure time at about 9:30 mid of October (location close to Hanover, Germany). And in the night it is close to zero. It does not reach zero because some streetlight (about 0.13Lux on the white-channel).

I hope that helps for your project.