4K - Dynamic Range (CMV12000)

Does anyone know if BMD are using the 'Multiple Slope' feature of the sensor to maximise the Dynamic Range ?

Am thinking an 'Advanced Mode' where the user had the ability to tweak the kneepoints to suit their own purposes. Also there are a couple other settings that could be useful to expose to the user.

Would mean post processing would increase in complexity but is kinda the point.

Kneepoints

image.jpeg (62.56 KiB) Viewed 1459 times

Think of it like an in camera user configurable Compressor for video...

Can't speak specifically about BM's implementation but I know apertus was talking about a lot of readout tweaks they could do to supposedly get up to theoretical 15 stops off the CMV12000, potentially with some tradeoffs. If I remember correctly the biggest thing discussed was a dual-gain readout, ether a true parallel architecture like the ALEV III or else something a bit more hacky like dual-ISO on Canon DSLRs with Magic Lateran, which though with limitations can get something in the 14-stop range using ISO 100/800 or 100/1600 to lower the effective read noise floor by 2-21/2 stops by sacrificing a little resolution in the extreme highlights and shadows; it could even do it in raw video though I haven't heard of it being realistically usable without faster readout and write speeds on the Canon end of things.

Not exactly the same as what you have outlined but does have a very similar result :-

"The CMV12000 has the possibility to achieve a high optical dynamic range by using a multiple slope feature. This feature will partially reset those pixels which reach a programmable voltage, while leaving the other pixels untouched. This can be done 2 times within one exposure time to achieve a maximum of 3 exposure slopes."

Kinda on the path to virtualising the sensor, but will leave that for another day.

Sounds a bit like a very early/crude version of a prototype technology that, using advanced on chip circuitry, can reset photosites when they fill up and count the number of times that happened, which at least theoretically could break the highlight clipping barrier and lead to sensors with far higher DR than we see today. A lot of these sorts of things never make it out of the lab or have some caveats, but they actually did produce a working prototype. I can't find the link now though...

Is really interesting for anyone that follows IT history, as Motherboard manufacturers have gradually realised that it is better for them to expose as much of the possible configurable settings as possible....users love it.

This allows their customers to tweak the system to suit their own specific purposes and teenagers are only too happy to dive into the detail and engage at a skill set level that used to be considered something for a Systems Programmer.

I have been messing with Software Defined Networks on an Asus X99-E WS whilst overclocking an 18 Core E5-2699 V3......yeah I know I have voided my warranty as Xeons are not meant to be Overclocked....spare me.

The system is multi use and boots in multiple configurations, including as a render machine....which of course is lightening quick !

Kinda figure sooner or later camera companies such as Black Magic Design will realise to compete they have to do the same and open up the firmware configuration options to the user.

Geez, even my Yamaha Jet Ski has 2 modes.

Kinda figure sooner or later camera companies such as Black Magic Design will realise to compete they have to do the same and open up the firmware configuration options to the user.

I wonder if it will be other companies other than BM that will have to compete in that way first, as Blackmagic seem to be offering a quality level at a much more competitive price than the competition. The other guys gotta step it up before BM needs to.

What you've outlined is essentially the main selling point of the Axiom Beta/Gamma, openness and developer friendliness at every turn. Both literally and figuratively what you'd get if you'd combined BM with ML, since they are using the same CMV12000 and the ML team has officially committed to helping them; at the same time they are just as delayed with the Beta if not more so than BM is with the 4.6K, but on the other hand are the best in the business when it comes to communication and transparency of every stage in the process.

Although the basic sensor silicon is essentially the same, the off-die processing is highly customizable, which will hopefully allow them to implement an improved dual channel readout architecture similar but hopefully superior to what BM already has and closer to that of the Alexa, which essentially allows for the effect described by the OP by reading each photosite at two different gains simultaneously, and doing some HDR wizardry to allow for high DR with smooth highlight rolloff.