WimH

There seems to be a fix for this, see this thread: https://www.raspberrypi.org/forums/viewtopic.php?f=43&t=273358&p=1658103#p1658078 I received my camera + 16mm lens today. I did a fresh install of Raspbian on a spare RPi3b and played with CLI commands a bit, trying different exposures of 1s, 2s, 4s, 8s, 16s. Note that I manually fixed as many parameters impacting the exposure as possible to avoid the camera to take exposure related decisions for me. Everything below 10s worked fine but didn't get it to work just using e.g. following command for a 16s exposure. The embedded text in the image showed the exposure as 10.2s instead, same upper limit as you were hitting. raspistill -ISO 100 -ex verylong -ss 16000000 -a 16 -o test.jpg However, when I forced the sensor mode to mode 3 (see camera guide p.110) with below command, the image data did effectively show the captured image had a 16s exposure (15.999-something to be exact) and subjectively it indeed looked 1 stop more exposed than the 8s exposed image. raspistill -md3 -ISO 100 -ex verylong -ss 16000000 -a 16 -o test.jpg Note that it takes a long time to take a long exposure (approx. 5-6 times the exposure time). The reason why is explained at the bottom of the linked thread, still something to be fixed. Will play a bit more with this the coming days, including a proper recompile of the raspistill as mentioned on the linked forum thread.

I've been looking into this option as well to use 4x4 binning. However, if the datasheet is correct, the increased sensitivity would not do you much good if your longest possible exposure in this mode is 20ms (50fps).

Here's my attempt to link LSB and mV specs :-) Consider this as 1st-order approximation as it assumes e.g. ideal ADC being used etc. Both mV and LSB quantify the luminosity response of a pixel under specific conditions, when expressed in mV the measurement is before the ADC (analog domain), when expressed in LSB the measurement is after the ADC (digital domain). There is no real way of comparing them unless we know for instance more about the ADC specs and settings used. IMX477 has for instance 8, 10 and 12-bit ADC settings. When looking at the IMX477 datasheet, it mentions 250LSB sensitivity and 1023LSB saturated signal. This means 10-bit ADC setting was used for the sensitivity measurement. Furthermore, the datasheet mentions the sensor uses an analog supply voltage of 2.8V. Accepting that this is also the input signal range of the ADC (which is not 100% correct, but again, this is 1st order approximation), a 250LSB response at the output of the ADC would equate to 683mV on the input. If Sony is really talking about mV per pixel and not mV per um2, this looks like an acceptable number, considering the difference in pixel area between both sensors (assuming both sensors have similar quantum efficiency and charge conversion rate).

Long exposure support is also what held me off buying the RPi v2 camera. However, what I've been able to find back for the HQ camera so far (see the camera guide, p. 110: https://magazines-static.raspberrypi.org/books/full_pdfs/000/000/036/original/Camera-Guide.pdf?1588180275) is that there are 4 capture modes. Mode 3 (12MP) supports frame rates between 0.005fps to 10fps, i.e. shutter speeds of 0.1s to 200s. This could be more interesting. EDIT: and this is confirmed elsewhere in the document in the CLI description (p.115): Shutter speed --shutter or -ss Sets the shutter speed to the specified value (in microseconds). The upper limit is around 6000000 μs (6 s) for CM v1; 10000000 μs (10 s) for CM v2; 200000000 μs (200s) for HQ Camera.

Wouldn't that be exactly the same? Delta between F2.8 and F5.6 is 2 stops, delta between 1/120 and 1/30 shutterspeed is also 2 stops so in both test conditions, you have the same amount of light hitting the sensor during the time the shutter was open.