Thanks for the head’s up on that.
this!
I got in the second rev of the PCB this weekend. Seems to work very well so far. It fixed a routing error in the power mosfet circuit. Also the board form factor was changed from 4x5 inches to 100x150mm (so it fits within one of the pricing rectangles that SeeedStudio charges when they build boards).
But the main change was in the lux/lumen light sensors. I am no longer using the Rohm BH1750FVI sensor chips, but am now using the same Taos TCS3210 color sensor chip/board that the color temperature sensor uses. There are two TCS3210 interfaces. The first is for the LUX meter function. The second is for the LUMEN/CCT sensor (usually mounted in an integrating sphere). I use the GREEN color channel in the chips are used for the LUX and LUMEN measurements. This change eliminates one sensor board and a couple of IIC interface chips from the design.
There were a couple of issues with the Rohm chip. First was that it is in a TINY chip scale package that is just about impossible for mortals to solder. Second was an issue with the wavelength response of the chip. It has a second peak in the RED part of the spectrum. This causes it to over-estimate the LUX and LUMEN readings if calibrated with cool white, high CCT lights, or underestimate it if calibrated with warm, low CCT lights.
The GREEN channel in the Taos chip is a much better match for the human eye response that LUX readings are based upon. The disadvantage of the Taos chip is it does not have the dynamic range of the Rohm chip. It saturates at around 50,000 lux. Another disadvantage is that the color filters on it have a significant response to infrared light and a smaller one in the ultraviolet. That makes it a poor sensor for incandescent lights and sunlight. One might be able to use an IR blocking filter for incandescent lights (like anybody uses tungsten bulbs around here…)
Cool to read some news here!
Also, you advertised your project as a “LED analyzer”, not a “light source analyzer”.
I might be interested in one, even without hotwire-support 
How many (copper) layers does your layout have? Is this a layout a hobbyist could etch at home?
100x150mm sounds perfect for euro-pcbs (100x160mm).
I could be wrong, and I'm sure texaspyro will confirm, but if memory serves he got the last set of boards from Oshpark which means they're two layer (double sided). I would guess the new revision to be the same. Some of the parts are fairly fine-pitch though, making them less home-etch friendly.
PPtk
The boards are 2 layer (OSHPARK can do 4 layer boards). I’ve kept the design to a minimum of 12 mil lines and spaces (most are at least 16 mil). Finest pitch devices are SOiC-8. It has plated through holes… not good for home-brew boards.
SOIC8 is not a problem. And “plated-through holes” become “holes with short wires soldered to both sides” when making a pcb at home When it is only double-sided, then the remaining problem that i can think of is vias underneath a part (when there is not enough place to solder such a wire in), i.e. thermal vias.
I don’t know whether i would actually do such a board myself, but if i decide that i want to, would you share your layout + part list? Maybe this is also interesting for others.
I did a test with the TAOS lux sensor chip and a halogen bulb. Without a filter, I read about 12 times too high. With an IR blocking filter it read about 20% high (using the calibration factor for LEDs ). If you had the hots for hot wires, it looks you could use a filter and adjust the calibration value and get decent numbers. The color temperature numbers with the halogen bulb and filter were wrong (said 5000K when the bulbs was around 3100K.
BTW, the IR filter that I used was around $30. Those cheap UV/IR camera filters (little more than plain glass) are useless here.
bump to this post. I ordered a CBA4, too 