what do people use to sense light these days

[SiliconWizard]

Depends on the application really - but these days I tend to use dedicated light sensors, possibly with a digital interface (if I'm going to interface them with some digital circuits anyway). You can find them in various packages, with various spectral responses.

Even if you just need a cheap way of sensing ambient light with no accuracy requirement, you can find digital light sensors for dirt cheap these days, as those are now ubiquitous in mobile devices. So I wouldn't bother with an analog solution, unless I had an extremely good reason for doing so.

[Simon]

So I wouldn't bother with an analog solution, unless I had an extremely good reason for doing so.

EMC!

[SiliconWizard]

So I wouldn't bother with an analog solution, unless I had an extremely good reason for doing so.

EMC!

Unless you had extremely strict requirements (far exceeding the levels allowed in current standards), EMC would be absolutely a non-issue here. Or we would never design any digital circuit.
If you're particularly concerned, just choose one with an I2C interface and configure I2C for low freq (100 kHz or slower) and controlled slew rate (many MCUs have this feature). EMI in this case would be completely negligible, and if you're going to use digital circuits in your design anyway (even just an MCU), this will then even be the least of your concerns.

Now obviously, a good reason would be if your design doesn't include any MCU or programmable logic able to deal with a digital sensor (at least in a sensible way). That's another story.

[Simon]

EMC is kinda critical, an un-shielded and unfiltered AVR held in reset fails this one.

[SiliconWizard]

EMC is kinda critical, an un-shielded and unfiltered AVR held in reset fails this one.

I would definitely question the design of your PCB and possibly of your power supplies then.

By "failing EMC", I suppose it failed meeting one of the applicable EMC standards? Certainly an MCU, all the more held in reset, shouldn't need shielding. Looks like something is seriously wrong.

[Simon]

EMC is kinda critical, an un-shielded and unfiltered AVR held in reset fails this one.

I would definitely question the design of your PCB and possibly of your power supplies then.

By "failing EMC", I suppose it failed meeting one of the applicable EMC standards? Certainly an MCU, all the more held in reset, shouldn't need shielding. Looks like something is seriously wrong.

Yes with your understanding of how sensitive some tests are....

[bson]

You don't have to compete with the sunlight that illuminates outdoors, only the portion that reflects into your eyes - assuming you're not staring into the sun.  But the user is looking directly into your device's illumination, which once again only has to compete with reflected light.  Lux also isn't the unit of interest since you don't care about illumination per area; you care about the amount of light emitted by both the device and its ambient surroundings.  I'd start by turning the panel light (or whatever it is) up all the way and expose is to sunlight.  See in what ambient lighting it tops out.  Take the light sensor and add enough ND filtering to bring just below saturation.  There's no point extending the scale beyond this when your light can't get any brighter.  Then take it to a dark environment and see at what level you can get a useable signal.  Low pass filter it to reduce the bandwidth and noise.

Did you do this?  What were the results (quantitatively, not blatheratively).

[Simon]

Did you do this?  What were the results (quantitatively, not blatheratively).

No, because until I complete the PCB I can't complete the housing and by the time I have a PCB, housing and decal i will be shipping it.... So i have taken the most pragmatic and simple solution of being prepared for anything. Whatever I need to do i can do as I will have the brightest LED's I can get and a sensor that will cover the entire possible range. Why have i got to guess and do tests that are not representative when I don't have to as I have a solution?

[pwlps]

Many years ago I was using a sensor from TI (I forgot the part number) featuring an integrated photocurrent-to-frequency converter, so that it could be interfaced directly to a microcontroller without any other components. And it had a fantastic dynamic range (frequency changing from Hz to ~MHz).  I don't find anything like this in the TI catalog anymore but there are other brands e.g. https://ams.com/light-to-frequency, also  found plenty of these here https://eu.mouser.com/Sensors/Optical-Sensors/Light-To-Frequency-Light-To-Voltage/_/N-8zn66.   If you want high resolution especially at low light levels these are probably hard to beat. 
 

[Kjelt]

If I may be so blunt, DON'T!
If there is something personal taste it is lighting level, people over 50 have far more need for light than young twentiers.
Don't decide for the operators which light level they need, let them please decide it for them selves and give them a manual potmeter or something they can regulate it themselves.
Even carmanufacturers reached this conclusion years ago that the day/night dashboard lighting should always be manually adjustable. 

[RJSV]

I like to take apart various TOYS, often drug store toy shelf has most often using a photo-transistor (3 wires,) but also with attached 'smokey' tinted plastic window cover. So, as per other folk's posts here, I'm not sure about UV issues and nighttime issues. (I need to read and study those posts, in more depth.)

Also I still see LDR type Cds (cadmium construction) and thought there was EU ban in place (toxic). BUT, the toy store still carries one toy type, a motion sending toy singing bird.  It appears some clever software involved, as the TOY senses motion in ways that seem erratic / unpredictable, but other times the TOY seems to work well, using the CDs type.