Hi Paul, I was referring to two things by this wording:

First of all – the silicon is fabricated only with columns, and in a different layer a set of grids is made – like if you paint rows and columns on the opposite side of a paper – these never physically connect in order to make the pixels – only if you see through it, you can notice the imaginary squares. The only pixel-physical thing on the CCD is the color filter array and the micro lenses, but not the light sensitive pixels.

Secondly, you almost got it right – the pixels are created by means of an electric field in the multi-phase grid perpendicular to columns. These phases (or clocks) are the ones that virtually split the sensing area into rows while exposure happens. And during the readout, the phases are “walking” and pushing the charge towards the edge of the sensor when there is then another set of similar structure, to serialize the charges into a single signal (but real sensor has usually 2, 4 or more such taps, easily benefiting from the symmetry of the chip). Its not a simple pulse – but a state (or combination of those multi-phase clock lines) that make it to do something.

This charge pushing by electric field is somewhat similar to peristaltic pumps – in sense that you can see how analog and continuous the process is, not that there are some discrete cups to shift the data digitally around.

CCD and no pixels? Whilst they are read in columns and summed up that way (hence vertical streaks) nowhere have I ever read or been told that the CCD sensors are not ‘pixel’ based? Can you reference this – just curious? So there are no pixel micro lenses on a CCD? Or you’re just saying it’s a continual rail underneath the pixels where charge is summed and divided by a clock pulse? (Not that it’s important but I do still like the colours of a CCD!)

AFAIK most of the bigger camera manufacturers have licensed compressed RAW, there have been enough nods from the Red camp to suggest that.

So the choice not to is marketing and the fact that most people just don’t need it. Also the extra hardware to do real time compression inside small cameras is power hungry and a heat generator.

Mirrorless cameras just don’t really need it – most users are not plugging into a complete end to end RAW pipeline either.

The various log codecs these days are very good, robust and the software support is so much better than it was a few years ago

I was asking whether you had considered changing the way the waveform screenshot is shown because even if you have a really clean camera in shadows, because those stops are all log then you still couldn’t see. If you have a waveform that had exp applied to it to that each stop was the same size it might visually show more? Or at least expand the image under 256? It’s just that most arguing about range is really subjective based on the shadows. It would be a nice way to compare if that was visually obvious?

Also I’m undecided whether not include the reconstructed stops is fair, in real shooting scenes you do want highlights to white most of the time and they are included in real shots. But I also understand why you may not want to include them. My only observation is that you don’t know for sure whether other cameras are reconstructing in their SDKs, because they may choose to not go to white. In fact I think the older reds did try to maintain the colour gradient into the highlights. If a reconstruction did attempt to maintain the same colour gradient then your test would show as real stops despite being reconstructed. If that makes sense?

Great test as always..!

A mechanical shutter is not a global shutter… it’s a rolling shutter. So it can never be the best global shutter.

+3 stops like RED is saying. They merge 2 pics together and one of it is 3 stops underexposed.

Just look at the CVP video on Youtube. Right now the extended highlights can be used only for non moving subjects. RED itself stated that only “excessive movement” will lead to artifacts ;)