cpc

I rarely encounter banding on my A7s and A7sII. What were you shooting? How were you grading? I'd be interested to see examples of banding given I've not experienced it, yet you seem to find it so bad to the point of unusable...
As you've sort-of covered... it isn't overexposure necessarily. Knowing the curve and where it places its values is extremely important when shooting log. Moreso when shooting 8-bit log. Blanket over-exposure won't necessarily give you better results. As for Arri, I regularly rate them at 400. 800 may give you the greatest spread of dynamic range, and results are generally acceptable. I find rating at 400 better. Knowing the curve and placing your exposure properly will give you even better results.
Have fun with constantly shifting depth of field and days (and therefore money) in the colour suite trying to get everything to match....
 

@kidzrevil I'm pretty certain what @cpc meant wasn't that you shouldn't ETTR - He's saying in order to keep exposure consistent,  rather than just ETTR as much as possible before clipping every time, it's better to overexpose by the same amount every time by rating the film speed lower. I agree - makes batch grading an actual possibility.

Banding is a combination of bitdepth precision, chroma subsampling, compression and stretching the image in post. S-log3 is so flat (and doesn't even use the full 8-bit precision) that pretty much all grades qualify as aggressive tonal changes. S-log2 is a bit better, but still needs more exposure than nominal in most cases.
Actually, I can't think of any non-Arri cameras that don't need some amount of overexposure in log even at higher bitdepths. These curves are ISO rated for maximizing a technical notion of SNR which doesn't always (if ever) coincide with what we consider a clean image after tone mapping for display. That said, ETTR isn't usually the best way to expose log (or any curve): too much normalizing work in post on shot by shot basis. Better to re-rate the camera slower and expose consistently.
In the case of Sony A series it is probably best to just shoot one of the Cine curves. They have decent latitude without entirely butchering  mids density. Perhaps the only practical exception is shooting 4k and delivering 1080p, which restores a bit of density after the downscale.

Banding is a combination of bitdepth precision, chroma subsampling, compression and stretching the image in post. S-log3 is so flat (and doesn't even use the full 8-bit precision) that pretty much all grades qualify as aggressive tonal changes. S-log2 is a bit better, but still needs more exposure than nominal in most cases.
Actually, I can't think of any non-Arri cameras that don't need some amount of overexposure in log even at higher bitdepths. These curves are ISO rated for maximizing a technical notion of SNR which doesn't always (if ever) coincide with what we consider a clean image after tone mapping for display. That said, ETTR isn't usually the best way to expose log (or any curve): too much normalizing work in post on shot by shot basis. Better to re-rate the camera slower and expose consistently.
In the case of Sony A series it is probably best to just shoot one of the Cine curves. They have decent latitude without entirely butchering  mids density. Perhaps the only practical exception is shooting 4k and delivering 1080p, which restores a bit of density after the downscale.

You will still get the best HD/2K delivery quality from a 4K camera even if you never deliver in 4K. A Sony FS700 + Odyssey 7Q goes for $3-4k nowadays and can shoot great 4K 12-bit raw. And it can also do high fps for slow-mo, which may be appealing for your music video/slow movement scenes.

In the ASC mag interview Yedlin readily acknowledges there is nothing new in the video (for technically minded people). It is deliberately narrated using language targeting non-technical people. His agenda is that decision makers lack technical skills but impose technical decisions, easily buying into marketing talk. Also, in the video he explicitly acknowledges the benefits of high res capture for VFX and cropping.

In the ASC mag interview Yedlin readily acknowledges there is nothing new in the video (for technically minded people). It is deliberately narrated using language targeting non-technical people. His agenda is that decision makers lack technical skills but impose technical decisions, easily buying into marketing talk. Also, in the video he explicitly acknowledges the benefits of high res capture for VFX and cropping.

In the ASC mag interview Yedlin readily acknowledges there is nothing new in the video (for technically minded people). It is deliberately narrated using language targeting non-technical people. His agenda is that decision makers lack technical skills but impose technical decisions, easily buying into marketing talk. Also, in the video he explicitly acknowledges the benefits of high res capture for VFX and cropping.

Testing camera profiles comes with the job characteristic of being a videographer/cinematographer. The easiest (cheapest) way to get an idea of DR distribution is to shoot a chip chart at a few exposures with a constant increment and use these to make a DR curve. Here is an example I did years ago using a Canon DSLR:

 
Also, it is been mentioned already but log profiles do not introduce noise. They raise the blacks making the noise visible. If you see more noise in the graded image, you are not exposing the profile properly. Most log profiles need to be exposed slower than nominal because the nominal ISO of the profile is chosen to maximize some notion of  SNR, which is not necessarily the rating you want for a clean image (after grading). In fact, cameras do it all the time. Take a Sony A7 series camera, for example, and look through the minimal ISOs of their video profiles. See how the minimal ISO moves around between 100 and 3200 depending on the profile? That's because the camera is rated differently, depending on how it is supposed to redistribute tones according to the profile.

High detail--deep focus--bright will be the highest rate. Movement is irrelevant cause compression is intra-frame.

The human eye does not perceive bits so such a claim makes no sense by itself. The necessary precision depends on many factors. There are multiple experiments on the tonal resolution of the eye. Perhaps the most applicable are the ones done during the research and setup of the digital cinema specifications, which determined that for movie theater viewing conditions and for the typical tonal curves dialed in by film colorists during the DI process and for the encoding gamma of the digital cinema spec (2.6 power gamma), you'd need around 11 bits to avoid noticeable posterization/banding artifacts after discretization (seen in gradients in the darks, by the way). This was rounded up to 12 bits in the actual specification. In an office viewing environment (brighter environments) you can do fine with less precision.
This is about delivery though -- meant for images for display. You'd usually need more tonal precision when you capture an image, because it needs to withstand the abuse -- pushing and stretching -- of post production. The precision will mostly depend on transfer curves -- you need relatively more for linear images, and relatively less for logarithmic images. With today's DRs 8 bits is absolutely not enough anymore for log curves (and not even remotely close for linear images). It usually does fine for delivery for consumer devices (some of these displays are 8-bit, some are 6-bit; likely moving to 10 bits in the future).

The true power of raw shooting is that you don't need to fix anything in advance and you can raw develop, edit and color at the same time in a liquid, flexible and creative post workflow. At least in Resolve you can. Proxies shouldn't be necessary. I don't see how a proxy workflow can be simpler -- you add a step (possibly two steps, if you round trip).  

Just went and tested this out, 4,6k60p is available in lossless RAW as well as 3:1 and 4:1. The RAW/Prores compression ratios are completely separate from the frame rate/resolution.

The DSLR filmmaker community is funny. It's like "Gimme 8bit proxy, 10bit, RAW, Canon colour science, Log, 4K, IBIS, DPAF and HFR in one package, in A7S form factor, max 2500 bux. I need a silver bullet that covers all cinematography and videography scenarios but make it cheap because I'm not a pro and want to shoot flowers/cat in my garden with it. Also has to shoot on SD cards because Cfast is too expensive". 
There's no silver bullet for everything in cinematography. Every camera body and ecosystem has its compromises. You can hold out for the perfect dreamworld unicorn camera, or you can pick up one that fits your shooting style and then actually shoot something.
Also, the C200 looks awesome and exciting - and daaaymn, official 4K RAW on an affordable Canon frame, without the unreliability of hacking stuff. Had I not sprung for an UM4,6K recently I'd be throwing my money at the Canon.

Canon cinema cameras raw is logarithmic like ARRIRAW and REDRAW (which are both 12-bit). 12-bit is plenty for log raw. 10 bits is still pretty good with a log curve.

Worked for a german short film one day, directed by the then-unexperienced daughter of a famous TV producer. A friend of mine was the production designer/set dresser/prop maker (the first being her profession), and I helped her. Everbody got paid after profits, which means nobody. Regular medium sized crew, credits ran long, including a long list of sponsors (???). The regular TV cameraman had a RED, he also worked three 12-14 hour days without payment. Well-known TV and stage actors (in part "borrowed" from the father's TV shows). Everything looked promising, but at the premiere (free buffet with champagne for the crew) I found the result rather mediocre. Couldn't tell a moral from this.
My friend also worked for Cronenbergs A Dangerous Method, and out of curiosity I volunteered to help demount the studio sets. I liked this film very much, but the actual sets were really amazing, I expected the visuals to turn out much more spectacular than they eventually were. To be more precise, I expected a much higher production value. Again, I don't know what to think of that. Both experiences were inspiring.

Canon cinema cameras raw is logarithmic like ARRIRAW and REDRAW (which are both 12-bit). 12-bit is plenty for log raw. 10 bits is still pretty good with a log curve.

No profile --flat or otherwise -- will be able to properly capture what you will be able to recover in post, so you might as well use it to help with other things. A punchy profile like Standard will help with manual focusing. Couple this with digic peaking + magic zoom, and you'll be surprised at what focusing precision is possible with just the camera display. Also, learning how the monitoring profile relates to the specific post workflow pays good dividends, especially if you are in the habit of exposing consistently. And if you know which profile value ends where in post, then all you need for exposure is the digital spotmeter (which is really the ultimate digital exposure tool anyway).
For B&W Monochrome might be useful in that you get to view the tones only, without color distractions, and this might help with lighting choices and judging separation. But this can also be a detriment since color cues can help with space orientation and framing if there is movement in the shot. And the profile's mapping to grey values will not necessarily match the channel mix you will do in post, so it might be misleading.

No profile --flat or otherwise -- will be able to properly capture what you will be able to recover in post, so you might as well use it to help with other things. A punchy profile like Standard will help with manual focusing. Couple this with digic peaking + magic zoom, and you'll be surprised at what focusing precision is possible with just the camera display. Also, learning how the monitoring profile relates to the specific post workflow pays good dividends, especially if you are in the habit of exposing consistently. And if you know which profile value ends where in post, then all you need for exposure is the digital spotmeter (which is really the ultimate digital exposure tool anyway).
For B&W Monochrome might be useful in that you get to view the tones only, without color distractions, and this might help with lighting choices and judging separation. But this can also be a detriment since color cues can help with space orientation and framing if there is movement in the shot. And the profile's mapping to grey values will not necessarily match the channel mix you will do in post, so it might be misleading.

No profile --flat or otherwise -- will be able to properly capture what you will be able to recover in post, so you might as well use it to help with other things. A punchy profile like Standard will help with manual focusing. Couple this with digic peaking + magic zoom, and you'll be surprised at what focusing precision is possible with just the camera display. Also, learning how the monitoring profile relates to the specific post workflow pays good dividends, especially if you are in the habit of exposing consistently. And if you know which profile value ends where in post, then all you need for exposure is the digital spotmeter (which is really the ultimate digital exposure tool anyway).
For B&W Monochrome might be useful in that you get to view the tones only, without color distractions, and this might help with lighting choices and judging separation. But this can also be a detriment since color cues can help with space orientation and framing if there is movement in the shot. And the profile's mapping to grey values will not necessarily match the channel mix you will do in post, so it might be misleading.

I've been using 5D3 Magic Lantern RAW for a few years now and it's my personal gold standard for image quality as I'm very familiar with it, can predict the results I'll get in various situations and have developed my own look that I like using in post. I do a lot of documentary and corporate stuff and wanted to do more handheld, so I got the GH5 for its IBIS, 10bit V-Log, 4K and general ease of use. Although the GH5's colour is an improvement on its predecessors, the 5D3 is nicer looking to me. So I wanted to see if I could tweak the GH5's VLog colour to be more similar to the colour I get from Magic Lantern Cinelog-C, processed via Adobe Camera Raw (other debayering workflows will give different results).
I shot a chart with both cameras, extracted the squares and put them side by side in Resolve. V-Log on the left, ML on the right:
 

 
I used Hue v Hue and Hue v Sat to line up the signal on the vectorscope:
 

 
Here you can see the effect of this correction.This is accurately white balanced V-Log (default colour) with a curve and saturation added:
 

 
And here's the same shot with the colour correction applied:
 

 
Skin goes from green-ish to pink-ish. Reds become more saturated, blues are pulled back. Foliage separates out into varying shades of yellow to green to blue-sh green, rather than being one big block of pure green. So next, I wanted to test this by comparing the same shots to Magic Lantern RAW. Would this correction really turn my GH5 into a handheld and more usable version of my 5D3?
I stuck the GH5 on top of the 5D3. I shot the 5D3 at 3520 x 1320 (the maximum resolution I can get that's both continuous and at a sensible aspect ratio) - this has a crop factor of 1.63x. ISO was 100, and aperture at f5.6 on the Canon 24-105. The GH5 was 10 bit V-Log, Cinema 4K, ISO 400, aperture f4 (roughly equivalent) on the Leica 12-60. I used shutter speed to control exposure, so motion blur is different between shots. It seems like the GH5 has around 0.66 stops more info in the highlights at the same exposure as the 5D, so I shot it one stop over to maximise DR.
 
 
First I corrected the 5D shots to the way I wanted them to look. Then I tried to get the GH5 to match. It's not a perfect match by any means, there are individual hues that tend to go awry a little, especially in the first shot. But for me the exercise is a success: the GH5 footage looks more like Canon Magic Lantern than it would straight out of the camera. And I would be happy to intercut the two, or to use the GH5 in situations that are more suited to its features. The take home message is that V-Log has a lot of grading potential - you're not necessarily stuck with Panasonic's colours on this camera.
A nice surprise for me was how much detail is in the 3.5K 5D files compared to Cinema 4K GH5, especially when sharpened. However, it's not very practical to shoot in its high resolution modes at the moment, due to the slow refresh rate of the LCD preview.
If you want to check out the files yourself, here are a series of matched pairs of 5D DNGs, and GH5 V-Log TIFFS.
FOLDER:
https://drive.google.com/open?id=0B1exEpCRAfgFdi1FZ3hma09YZms
ALL FILES ZIPPED:
https://drive.google.com/open?id=0BwzsbjMgrAwzTFJZby0xSTV6VmM

At 2K you'll be giving up the full frame benefits for a miniscule increase of pixel count (I deliberately don't say "resolution", it is likely you won't get true higher res at 1:1 2K as 1:1 puts much higher requirements on lens sharpness). It is almost certainly better to just upscale 1080p to 2K.

At 2K you'll be giving up the full frame benefits for a miniscule increase of pixel count (I deliberately don't say "resolution", it is likely you won't get true higher res at 1:1 2K as 1:1 puts much higher requirements on lens sharpness). It is almost certainly better to just upscale 1080p to 2K.

Tested on the Win10Pro 6950X 64GB GTX 1080 machine: easily ran real-time with both C300 II and 1DX II 4K footage, along with transitions and some titles. Added 4 OpenFX effects to a C300 II 4K clip: Light Rays, Glow, Gaussian Blur, and Prism Blur (in addition to ARRI 3D LUT and basic curves/grading)- still ran easily in real-time (as one would expect with efficiently implemented GPU acceleration; CPU was at 23%). Rendering speed: 1080p render from mixed 1080 4K timeline 1m:40s long finished in 38s, about 2.6x real-time, which is pretty good. Similar test on OSX with 12-Core and GTX 980ti in PP CC was about 2x real-time for 1080p output.
Would be nice to know how to replicate Lumetri effects in Resolve (really fast and easy to quickly grade footage using Lumetri).
Audio isn't quite done:
As noted by someone else in this thread, I had to set my default/system audio device sample rate to 48KHz otherwise no audio output (including strangely the rendered output). Where are all the audio effects hidden? Poked around briefly and only found EQ and Dynamics (which doesn't appear to actually work (correctly) yet). Also no way to select audio device (checking/unchecking Use System Audio didn't do anything). Tested again on the 2010 MacPro 12-Core (3GHz) GTX 980ti in bootcamp/Win10Pro. Can handle C300 II 4K 10-bit up to about 3 tracks at once (2 50% opacity tracks on top of a base track). Only 2 tracks for real-time, so transitions are good. At 3 tracks, it's not real-time and audio starts to sound shabby. Thus it's clear in this case that the OSX version is slower in software/drivers, since it runs faster in Windows on the same hardware.
Overall excellent progress!

bought one.