There are very few.... _very few_ projects out there that requires more than 11-bits resolution with an ADC. Most MCUs don't offer greater than that- you have to use a standalone ADC for just a few more bits. Or you can make your own. Making your own allows you to make as many bits as you wish (within reason, there are limits), and you can read them simultaneously (in parallel) with little effort.

Can you point me to a some reads about that?

 

"Also you can detect is the flash burst is only one pulse or various fast ones." What does that mean? Do you need to record the data from a single flash or are you intending to measure the data from multiple flashes, i.e. the average of multiple flashes?

To measure something to 12-bit accuracy means that your data is accurate within 1 part in 4096 parts. That means all components in the signal chain must be both accurate to that degree, stable to that degree, and with low enough noise that noise contributes less than 1 part in 4096 to the output result. I remain unconvinced that you can produce repeatable flashes with that degree of accuracy, and therefore any result that uses data from multiple flashes will be an average.

What you have described is very much like what a high-speed digital oscilloscope does: it performs sequential measurements at a very fast repetition rate and saves each result to memory. With very rare exceptions, those scopes measure with only 8-bit resolution and less than 8-bit accuracy.

Photographic flashes works in some different ways. They must shoot a only one pulse or some depends on the work mode:
-TTL produce 2 pulses. A first little pulse one to determine the power needed for a correct picture, and then the "good" pulse.
-HSS is a mode to overpass the sincro limit of the cameras, consisting in a train of pulses to take a picture.
-HyperSync is a technic consisting in make a long time pulse (is another way to overpass the sinchro limit of the cameras).

Sinchro limit mean the maximum time a camera take a shoot with the shotter full open, normally 1/250 to 1/500 seconds.
A shutter has 2 curtains. When you shoot, 1rst curtain start to move open the shutter and light start to impress the sensor. When time reachs the time selected, second curtain start to move and close the shutter. When you select a fast shutter speed, 1/1000, the 2nd curtain starts to move before 1rst ends her move (this is the sinchro limit) and only 1 pulse flash will produce a underexposed band in the picture. This is the explanation of HSS technic.

When you want to analize the shape of a flash burst, you dont know at wich power the flash will shoot. This is the problem, and for that you need to read a range so wide (maybe 1lx to 150000lx)

 

I wanted to respond earlier but had a meeting to go to...

The response of almost any photodetector to light intensity is wavelength dependent. That is especially true of photodiodes. What is the spectral purity of your flash, and most important, does it vary from flash to flash?

Your very precise intensity measurements could be distorted if the spectrum of emitted light changes. How is the flash produced?

Sorry, I dont understand your question.
Flashes used ara photograpic ones.

 

I just grabbed this off the internet:

Source: https://www.hamamatsu.com/resources/pdf/ssd/e02_handbook_si_photodiode.pdf

Take, for example, the S3759. The same amount of light (quanta) at 980 nm will give more response (higher quantum efficiency) than one producing light of 550 nm. In other words, the quantum efficiency of your photodetector varies with wavelength. If the spectrum of your flash is always the same, then the measurements will be comparable.

At greater and greater levels of precision, though, you might run into problems with reproduciblity. You haven't said much about how you are producing the flashes. At one extreme, if you are doing exploding wires of different metals, then I would expect the spectra to vary quite a bit. At the other extreme, a standardized flash bulb might give very consistent results.

 

I'm testing this circuit, and it seam to works.
I have a section of the light graph in each ADC, and the times measured are like espected ( t.5 duration of 1/20000~1/30000s for a flash shoot at 1/128 power (a speedlight flash).
I need to finetune the Rf values to have a lineal change of scale:
-ADC2 read > 0 at ADC1 reach 4000
-ADC3 read > 0 at ADC2 reach 4000

Thanks, friends, for all your help!