BM's Dual Gain Fairchild vs Sony

[TomTheDP]

I've seen many many times people talk about how the older Blackmagic cameras have a certain quality that the Pocket 4k and 6k seemed to have left behind. The OG Pocket, the BMCC 2.5K, or the URSA mini. The Fairchild sensors used in all the Blackmagic cameras other than the Pocket 4k and 6k were all dual gain. They process the image like the Arri Alexa. Taking two 11bit images and combining it into one 16 bit image that is then processed in camera into 12 bit LOG RAW. The Alexa is a bit different as its taking two 14 bit signals rather than 11. Its similar to an HDR image where its taking a higher and lower amplified image and combining for more dynamic range. One of the reasons why the Alexa has probably the best highlight and shadow range on the market. 

Anyways this info is all on Arri's website but where I am confused is if this pipeline also creates higher color depth. Two 14 bit images means 28 bits, which is then processed down to 16 bit(not sure how that works?) and then into 12 bit LOG. Does that mean the final 12 bit log RAW is similar to a 16 bit linear RAW? Does that mean the URSA mini in regards to color depth is no different than a RED camera doing 16 bit linear. Or the OG pocket for that matter. 

[kye]

In terms of combining the bits together, you would only add the bit depths if they didn't overlap in DR.

For example, if I took two 14-bit readouts, one a stop below the other, then 13 of the 14 bits from each read-out would be duplicate data, and my effective bit-depth is really only 15-bits.  So in order to understand the total bit-depth you'll need to know the overlap range in the sensor.  That would also be further complicated if they weren't offset by a whole number of stops, which would place the values of one readout between the values of the other, providing more bit-depth but less increase in DR.

If you really want to understand this, try modelling things in excel and graphing them.  That should give you a more intuitive sense of what is going on.

In terms of there being a certain quality in the older sensors, this guy has done lots of tests comparing the older BM cameras to the P4K and Ursa.

https://www.youtube.com/user/joelduarte27/videos

My overall impression is that most people don't utilise anything like the potential of their cameras, and that the difference between what images most people get and the images that you see from an Alexa or RED is more down to user skill (in terms of lighting, composition, camera operating, and the complete image pipeline in post) than it is about any camera limitations that might exist.

[Video Hummus]

1 hour ago, kye said:

My overall impression is that most people don't utilise anything like the potential of their cameras, and that the difference between what images most people get and the images that you see from an Alexa or RED is more down to user skill (in terms of lighting, composition, camera operating, and the complete image pipeline in post) than it is about any camera limitations that might exist.

Yep. Money well spent on lighting, lens, and the training or time to master those. Camera bodies really are the least important thing for most people.

[TomTheDP]

2 hours ago, kye said:

In terms of combining the bits together, you would only add the bit depths if they didn't overlap in DR.

For example, if I took two 14-bit readouts, one a stop below the other, then 13 of the 14 bits from each read-out would be duplicate data, and my effective bit-depth is really only 15-bits.  So in order to understand the total bit-depth you'll need to know the overlap range in the sensor.  That would also be further complicated if they weren't offset by a whole number of stops, which would place the values of one readout between the values of the other, providing more bit-depth but less increase in DR.

Interesting so the bit depth on the URSA really wouldn't be over 12. 

 

2 hours ago, kye said:

My overall impression is that most people don't utilise anything like the potential of their cameras, and that the difference between what images most people get and the images that you see from an Alexa or RED is more down to user skill (in terms of lighting, composition, camera operating, and the complete image pipeline in post) than it is about any camera limitations that might exist.

Definitely true for coloring in post. Dynamic range is pretty easy to break though thats becoming less of an issue with the latest batch of full frame cameras. Lighting is not always as much skill as it is time and money. 

[kye]

30 minutes ago, TomTheDP said:

Definitely true for coloring in post. Dynamic range is pretty easy to break though thats becoming less of an issue with the latest batch of full frame cameras. Lighting is not always as much skill as it is time and money. 

I'm actually not so sure about DR anymore.  I compared the HLG mode vs Cine-D (709 equivalent) modes on my GH5 and the HLG has a couple of stops more DR (IIRC) than the Cine-D, but in real-life the differences weren't that much, even in extreme situations.  The only time I missed some DR in my comparison shots was when the sun was in the shot, but even then it wasn't much.

I used to shoot and think about DR in terms of making sure nothing was clipped, and then grading it to control everything.  Now I realise that I don't care about things with that much DR.  If I'm shooting inside and the outside is blown out then I can choose which thing I expose on, and if I care about the relationship between both then typically I can get a silhouette, and times when a face is important on the person inside (maybe they're looking out) then their face will be much better exposed and I can get the outside and their face.  
I'm not saying there are no situations where extra DR matters, but I'm saying that with, say, the 9-10 stops of DR that most cameras have now in 709 modes, that's enough for most situations.
Also, the situations where 12 stops isn't enough, you might find that the 15 stops of high-end cameras is also not enough.  Things with high DR that are common like fire, welding, the sun, any night scene where there is no ambient artificial light (eg, moonlight with torches, or moonlight with headlights) etc will be more than 15 stops of DR, so there's no point lusting after an Alexa in those situations either.

In terms of lighting, I think maybe you're underestimating how much skill is involved in getting the highest quality shots that top-end shows and movies have.  Give someone who is clueless an unlimited lighting budget (and an Alexa and great cine glass) and you'll still get something that looks awful.  Lighting budgets only make a difference after you have someone with the skill to know what to do with them.

[TomTheDP]

2 minutes ago, kye said:

I'm actually not so sure about DR anymore.  I compared the HLG mode vs Cine-D (709 equivalent) modes on my GH5 and the HLG has a couple of stops more DR (IIRC) than the Cine-D, but in real-life the differences weren't that much, even in extreme situations.  The only time I missed some DR in my comparison shots was when the sun was in the shot, but even then it wasn't much.

I used to shoot and think about DR in terms of making sure nothing was clipped, and then grading it to control everything.  Now I realise that I don't care about things with that much DR.  If I'm shooting inside and the outside is blown out then I can choose which thing I expose on, and if I care about the relationship between both then typically I can get a silhouette, and times when a face is important on the person inside (maybe they're looking out) then their face will be much better exposed and I can get the outside and their face.  
I'm not saying there are no situations where extra DR matters, but I'm saying that with, say, the 9-10 stops of DR that most cameras have now in 709 modes, that's enough for most situations.
Also, the situations where 12 stops isn't enough, you might find that the 15 stops of high-end cameras is also not enough.  Things with high DR that are common like fire, welding, the sun, any night scene where there is no ambient artificial light (eg, moonlight with torches, or moonlight with headlights) etc will be more than 15 stops of DR, so there's no point lusting after an Alexa in those situations either.

In terms of lighting, I think maybe you're underestimating how much skill is involved in getting the highest quality shots that top-end shows and movies have.  Give someone who is clueless an unlimited lighting budget (and an Alexa and great cine glass) and you'll still get something that looks awful.  Lighting budgets only make a difference after you have someone with the skill to know what to do with them.

Really just depends on the shot in mind. If you have a cloudy day and want it to look sunny in a oner that involves three different rooms you have a budget issue. Of course you could also chalk that up to an unreasonable DP or Director. Even getting creative with cheap equipment takes time and time is money. 

More dynamic range also makes it easier to light a difficult dynamic range situation, when there is more room in the highlights and shadows you don't have to use as strong of light to get the balance you are looking for. I personally notice the dynamic range difference going from my GH5 to Panasonic S1.
 

There are a lot of workarounds of course, I've seen great stuff with a Panasonic G7 but when I am shooting for someone else and need to get the shots they desire in the amount of time they desire its a different story. 

[kye]

2 minutes ago, TomTheDP said:

Really just depends on the shot in mind. If you have a cloudy day and want it to look sunny in a oner that involves three different rooms you have a budget issue. Of course you could also chalk that up to an unreasonable DP or Director. Even getting creative with cheap equipment takes time and time is money. 

More dynamic range also makes it easier to light a difficult dynamic range situation, when there is more room in the highlights and shadows you don't have to use as strong of light to get the balance you are looking for. I personally notice the dynamic range difference going from my GH5 to Panasonic S1.
 

There are a lot of workarounds of course, I've seen great stuff with a Panasonic G7 but when I am shooting for someone else and need to get the shots they desire in the amount of time they desire its a different story. 

Sounds like you're talking about situations where everyone is paid and everything is at or above standards, and of course, at that point it's well worth spending money on equipment as it pays for itself in lost time.  Having a dozen people or more plus equipment on set is expensive.

In terms of the whole of film-making though, lots of people are making content where they are time rich but cash poor, which is more where I was talking about.  I think it's easy to forget how broad a range of film-making there is going on - everything from YouTubers with a phone and (maybe) and LED light and a lav, people making features by working part-time and maxing out a couple of credit cards, folks doing weddings or corporates, people on low budget but industry rate sets, through to productions where there are people above the line and the daily rental on the trucks alone would make the credit-card film-maker cry.

I try to keep my comments generic and in the context of everyone.  Plus, these forums seem to be more frequented by people at the lower end of the scale than at the higher end.

[TomTheDP]

7 minutes ago, kye said:

Sounds like you're talking about situations where everyone is paid and everything is at or above standards, and of course, at that point it's well worth spending money on equipment as it pays for itself in lost time.  Having a dozen people or more plus equipment on set is expensive.

In terms of the whole of film-making though, lots of people are making content where they are time rich but cash poor, which is more where I was talking about.  I think it's easy to forget how broad a range of film-making there is going on - everything from YouTubers with a phone and (maybe) and LED light and a lav, people making features by working part-time and maxing out a couple of credit cards, folks doing weddings or corporates, people on low budget but industry rate sets, through to productions where there are people above the line and the daily rental on the trucks alone would make the credit-card film-maker cry.

I try to keep my comments generic and in the context of everyone.  Plus, these forums seem to be more frequented by people at the lower end of the scale than at the higher end.

I mean I work often on low budget indie projects, often where its 100% volunteer, with a very low budget for anything that requires $$$. Its always a race against time. At least in my experience. 

[TomTheDP]

30 minutes ago, kye said:

Also, the situations where 12 stops isn't enough, you might find that the 15 stops of high-end cameras is also not enough.  Things with high DR that are common like fire, welding, the sun, any night scene where there is no ambient artificial light (eg, moonlight with torches, or moonlight with headlights) etc will be more than 15 stops of DR, so there's no point lusting after an Alexa in those situations either.

 

I do think the differences are becoming negligible. Which is partly why I haven't made it a goal to buy an Alexa or RED as I don't think they'd really do anything for me besides get me more work LOL. 

But yeah the Alexa is 14 stops while the S1 is 12.5. That's only a 1.5 stop difference. Compare the S1 to the C500 MK2 which is $15,000 and there is only a half stop difference. Of course the C500 has a lot of features the Panasonic doesn't but in terms of pure IQ there really isn't much of a difference. Trying to get a Panasonic to look like an Alexa color wise is not an easy task though, or at least not a task for someone who isn't an experienced colorist. 

But yeah deep down I don't like to spend my money and am really happy with what the current cameras I own can deliver. 

[kye]

I think the hidden 'hack' for good IQ is to go back to 1080p.  The GH5 has great 1080p modes, but even if the camera you have doesn't, if you shoot in the typical 4K modes then put them on a 1080p timeline then they're significantly better due to downsampling.

I look at all the latest camera releases and don't really see much that I would want over what I have, from a 1080p perspective.  Sure, if you're doing commercial gigs where people are paying for the spec, or recording stock footage or whatever, then sure, go for it.  

But any time you see a great image come up on your phone, you're not looking at something that's great because of the resolution, it will be the composition, lighting, colour, etc.  That's where I'm spending my time and energy now.

[zerocool22]

6 hours ago, kye said:

 

My overall impression is that most people don't utilise anything like the potential of their cameras, and that the difference between what images most people get and the images that you see from an Alexa or RED is more down to user skill (in terms of lighting, composition, camera operating, and the complete image pipeline in post) than it is about any camera limitations that might exist.

Yeah this is very true, to some extent. Most of us can benefit more from lighting/grading. But at a certain level, just everything needs to be top notch. I feel camera, lenses, lighting, grading, sound, acting, set design, directing are all important, if one of those is off the film/scene is off. Cameras do matter though if you look at shootouts between cameras in same lighting setup, not only DR stands out but color. And yes I have heard numerous times if you shoot 10 bit(or raw) you can get the same results but to me it looks like almost nobody can pull this off. I mean I can easily spot the blackmagic footage in mad max fury road. And if they cant even match it, how should us mortal folks do it.

 

[kye]

11 minutes ago, zerocool22 said:

Yeah this is very true, to some extent. Most of us can benefit more from lighting/grading. But at a certain level, just everything needs to be top notch. I feel camera, lenses, lighting, grading, sound, acting, set design, directing are all important, if one of those is off the film/scene is off. Cameras do matter though if you look at shootouts between cameras in same lighting setup, not only DR stands out but color. And yes I have heard numerous times if you shoot 10 bit(or raw) you can get the same results but to me it looks like almost nobody can pull this off. I mean I can easily spot the blackmagic footage in mad max fury road. And if they cant even match it, how should us mortal folks do it.

You're have a good point, but I don't think this is what matters.

What does matter is the difference between what a camera is capable of and what most people get out of it.  If someone is getting half of the potential of a P4K then giving them an Alexa isn't going to give much of an advantage, because the same limitations that prevented them from getting even close to the potential of the P4K will also prevent them from getting the most out of the Alexa.

There's a saying about continuity - "if people notice continuity problems then your film is crap".  

I think colour is kind of the same in many ways.  As much as I love it, a great film with BM colour, GH5 colour, or even Sony colour, is still a great film.  There have been many reports of people that don't know what they're doing using an Alexa, and the results are reported to look like a home video.

[TomTheDP]

9 hours ago, zerocool22 said:

Yeah this is very true, to some extent. Most of us can benefit more from lighting/grading. But at a certain level, just everything needs to be top notch. I feel camera, lenses, lighting, grading, sound, acting, set design, directing are all important, if one of those is off the film/scene is off. Cameras do matter though if you look at shootouts between cameras in same lighting setup, not only DR stands out but color. And yes I have heard numerous times if you shoot 10 bit(or raw) you can get the same results but to me it looks like almost nobody can pull this off. I mean I can easily spot the blackmagic footage in mad max fury road. And if they cant even match it, how should us mortal folks do it.

 

Interesting I've never noticed a difference when watching Mad Max but I guess I wasn't really looking for it either.

 

9 hours ago, kye said:

You're have a good point, but I don't think this is what matters.

What does matter is the difference between what a camera is capable of and what most people get out of it.  If someone is getting half of the potential of a P4K then giving them an Alexa isn't going to give much of an advantage, because the same limitations that prevented them from getting even close to the potential of the P4K will also prevent them from getting the most out of the Alexa.

There's a saying about continuity - "if people notice continuity problems then your film is crap".  

I think colour is kind of the same in many ways.  As much as I love it, a great film with BM colour, GH5 colour, or even Sony colour, is still a great film.  There have been many reports of people that don't know what they're doing using an Alexa, and the results are reported to look like a home video.

I think where a lot of people go wrong is trying to color grade their footage or manually grade log footage. Most would be better off just using the provided REC709 lut. As long as they white balanced correctly the footage will look decent. Of course an Alexa with just the 709 lut will look more pleasing and accurate in almost any situation compared to a BM, or Panasonic or insert camera here. I do agree story is paramount as long as the color is decent enough to where its not distracting the average viewer. Hell the average viewer might think log footage looks normal until they see a side by side. Look at anything long enough and it starts to look normal. 

[Laurier]

I had the original 4.6k ( so with a Dual Gain Fairchild sensor) and a a7rii at the same time few years ago, the a7rii actually had more dynamic in the highlight in slog2. 

Of course the codecs on the 4.6k were a lot more solid, but nowadays with 10 bits 4.2.2 in most cameras, something like a a7siii do clearly better than a ursa mini pro.

Out of the camera, the images from Blackmagic are more pleasing because of the colour science , but purely in term of data quality Sony sensors are doing much better.

If you know what you are doing in post you can make very pleasing images out of Sony cameras, the problem with Sony is that their official LUT are just crap overall ( the most usable one is the typeA IMO), and most of the profiles have way to much contrast making the images look very harsh. 

The real value in an Alexa for me is that you can put whatever lens on it, shoot log C, just slap the official lut and it look great, you pay more for the camera, but you could technically pay less in post.

[TomTheDP]

11 minutes ago, Laurier said:

The real value in an Alexa for me is that you can put whatever lens on it, shoot log C, just slap the official lut and it look great, you pay more for the camera, but you could technically pay less in post.

Exactly. I honestly think Arri is the only company that has made a good digital camera that can compete with film in terms of color. Their domination of anything high end attests to this. 

11 minutes ago, Laurier said:

I had the original 4.6k ( so with a Dual Gain Fairchild sensor) and a a7rii at the same time few years ago, the a7rii actually had more dynamic in the highlight in slog2. 

Of course the codecs on the 4.6k were a lot more solid, but nowadays with 10 bits 4.2.2 in most cameras, something like a a7siii do clearly better than a ursa mini pro.

Out of the camera, the images from Blackmagic are more pleasing because of the colour science , but purely in term of data quality Sony sensors are doing much better.

This is why I sold my URSA. I will say the URSA in RAW could better bring down highlights without effecting the rest of the image, better than any other camera I've seen including the Pocket 6k. Data in terms of dynamic range, shadow recovery, and ISO performance the A7S3 or Pana S1 is superior.

In terms of actual data of course Prores is going to be far superior to codecs on mirrorless cameras. 4k Prores HQ is something like 800mbps I believe where as the highest bitrate on the A7s3 is 300mbps and for the S1H 400mbps, or 150mbps on the S1. You can record Prores externally though which kind of closes the gap, though its not the most ergonomical choice. 

For indepth grading 12 bit is far superior to 10 bit. If you want to isolate skintones and grade them 12 bit makes a big difference. For anything else not so much. Of course you can now record 12 bit externally on pretty much every camera with an Atomos ninja. 

I really thought the new Ursa was going to be using a full frame Sony sensor. I honestly think they would have sold more cameras if they went that route. Internal full frame BRAW, with the URSA form factor and ND's. 

[Video Hummus]

11 hours ago, TomTheDP said:

really thought the new Ursa was going to be using a full frame Sony sensor. I honestly think they would have sold more cameras if they went that route.

I think their next pocket camera will be FF with internal ND for the typical bargain price. I hope they do a more sensible design with better QC. 

[maxmizer]

14 hours ago, TomTheDP said:

Interesting I've never noticed a difference when watching Mad Max but I guess I wasn't really looking for it either.

Mad Max: Fury Road (2015)

Technical Specifications Camera:

Arri Alexa M, Panavision Primo, Ultra Speed MKII and Angenieux Optimo Lenses
Arri Alexa Plus, Panavision Primo, Ultra Speed MKII and Angenieux Optimo Lenses
Blackmagic Cinema Camera (some shots)
Canon EOS 5D Mark II (some shots)
Nikon D800 (some shots)
Olympus OM-D EM-5 (some shots)

13 hours ago, TomTheDP said:

I really thought the new Ursa was going to be using a full frame Sony sensor. I honestly think they would have sold more cameras if they went that route. Internal full frame BRAW, with the URSA form factor and ND's. 

They are selling like crazy ... S35 is the right sensor because there are no lens limitations