Help me decide: Canon C300 Mark III or Sony FX9

[kye]

5 hours ago, Jedi Master said:

I didn't realize professional recorders were that expensive. Am I correct in assuming that the higher cost is due to better analog preamps and ADCs, reliability, and ergonomics versus a prosumer recorder like a Zoom? I doubt it's due to the digital audio section as cheap, modern ICs can easily handle 24-bit/192 kHz, which is vast overkill for audio.

I'd imagine it is partly to do with the analog and ADCs, but don't forget that just because modern ICs are very capable doesn't mean the manufacturer will sell you a product for a song.  Also, digital processing is another step that the high-end products may well also have capabilities for.

[IronFilm]

17 hours ago, Jedi Master said:

I have the Sachtler FSB 6 MK II fluid head and the Sachtler 75/2 carbon fiber legs.

The head's specs say it'll support 17.6 lbs (8 kilos for the Imperially-challenged). The C300 Mk3 is about 6 lbs, and with a 2 lbs cine lens and 1 lbs battery, comes out to around 9 lbs, leaving a margin of around 8.5 lbs.

Ah, getting say an ARRI ALEXA Classic might be pushing it beyond the reasonable max for it. 

16 hours ago, Jedi Master said:

I didn't realize professional recorders were that expensive. Am I correct in assuming that the higher cost is due to better analog preamps and ADCs, reliability, and ergonomics versus a prosumer recorder like a Zoom? I doubt it's due to the digital audio section as cheap, modern ICs can easily handle 24-bit/192 kHz, which is vast overkill for audio.

Nobody can really accurately tell apart a Zoom F8 vs a Sound Devices 788T (or a F8n vs a 888) in a double blind test. 

So it's more about the feature set and workflows, you can just handle more with the Sound Devices 8 Series. 

(I wish it was also about reliability too... but I tend to feel the Zoom F8n has the slight edge in reliability over my Sound Devices 833! Although, I am recording to not one, not two, but three media with my 833!)

[Jedi Master]

After reading this article:

https://www.eoshd.com/news/is-n-raw-real-raw-nikon-z9-under-the-spotlight-at-eoshd/

I’m beginning to wonder if RAW is all it’s cracked up to be. Is there a similar analysis of Canon RAW Lite? That would be interesting to read.

How about a codec that doesn’t skimp on color sampling, such as ProRes 4444 or 4444XQ? Would that be roughly the equivalent of RAW without the hassles?

[PannySVHS]

You have read several compelling accounts about NRaw here in this thread. Slashcam tested it on several occasions with best in class latitude, bettering S1H internal which is bettering R8 Canon Raw. Color fidelity is impressive. Andrews article shows a different perspective iirc, about difficulties to work it in post for best effect iirc. Give us your summary of it, if you find time. It's been written very early after that brandnew format came out. @Jedi Master

[TomTheDP]

7 hours ago, Jedi Master said:

After reading this article:

https://www.eoshd.com/news/is-n-raw-real-raw-nikon-z9-under-the-spotlight-at-eoshd/

I’m beginning to wonder if RAW is all it’s cracked up to be. Is there a similar analysis of Canon RAW Lite? That would be interesting to read.

How about a codec that doesn’t skimp on color sampling, such as ProRes 4444 or 4444XQ? Would that be roughly the equivalent of RAW without the hassles?

Prores 444 and 4444xq are really nice codecs to work with. The issue is compressed RAW is often actually smaller in terms of file size compared to Prores 444. I shoot 2k Prores 444 which is about 500mbps, which is a reasonable size. 

To me 12 bit is important but it doesn't matter if it in the RAW format, Prores 444 or semiraw like BRAW. For me I notice the difference in shadows and skintones compared with 10 bit recording. The differences can be subtle though. 

This video demonstrates some differences you can see in a real world scenario. FX6 10 bit vs REDRAW
 

 

[mercer]

8 hours ago, Jedi Master said:

After reading this article:

https://www.eoshd.com/news/is-n-raw-real-raw-nikon-z9-under-the-spotlight-at-eoshd/

I’m beginning to wonder if RAW is all it’s cracked up to be. Is there a similar analysis of Canon RAW Lite? That would be interesting to read.

How about a codec that doesn’t skimp on color sampling, such as ProRes 4444 or 4444XQ? Would that be roughly the equivalent of RAW without the hassles?

I can't speak for Z9 raw as I have never used it, but I have been shooting raw video exclusively for the past 6 years... 1080p ML Raw with a 5D Mark iii and more recently 4K cDNG with a Sigma FP...

...and I can honestly say that I will never shoot with a heavily compressed codec again...

ProRes is still good, I just can't afford a camera that hasn't effed it up.

...that said, raw isn't for the faint of heart. It takes a lot of storage space, both in camera and in post, but the difference in IQ is instantly apparent. With my 5D3, you can see it on the LCD and smile, but when you get the footage into the computer... it's kinda mind blowing.

And with that said, I still prefer the look of film over digital, so your mileage may vary.

[kye]

2 hours ago, TomTheDP said:

This video demonstrates some differences you can see in a real world scenario. FX6 10 bit vs REDRAW

Interesting video.

First things first - could that guy be any cooler?  I'm pretty sure there isn't a single element in that video that isn't an automatic 10/10 for hipster chic.  Wow.  Talk about those people whose whole life is their own art project!

Secondly, his commentary is all over the place.

I see the differences he's talking about in the footage, plus a great many more that he probably sees but didn't mention - there are hue shifts and gamma shifts and subtractive colour operations and all sorts of wonderful things that are different between the two cameras.

However, he points to differences and says "see the difference with 10-bit?" whereas I think the differences are likely to be a mix of:

1. FX6 sensor read-out bit-depth
2. FX6 image processing
3. Sony RAW-LOG profile conversion (gamma, gamut, and bit-depth)
4. FX6 compression

I built a DCTL plugin for Resolve that reduces the bit-depth of the image, and to my surprise, you can reduce the bit-depth of rec 709 footage to 6-bits (and some shots to 5bit!) before there are visible changes.  

I'm not saying that there aren't any differences between 16-bit and 10-bit, because there are (however subtle they might be) but the things he was pointing at were definitely NOT all bit-depth related, and I'd suggest that most of them were processing/compression related actually.

If you want to understand what the differences are with one parameter, you can't change dozens of them at once and then just declare that all differences are due to the one variable you want to talk about.  You could take his whole video and just replace the phrase "10-bit" with "compression" and it would make just as much sense, despite having exactly the same examples.

So yeah, it's a great video to show an FX6 LOG vs Komodo X RAW comparison, but it isn't an isolated 10-bit vs 16-bit comparison.

[PannySVHS]

2 hours ago, kye said:

If you want to understand what the differences are with one parameter, you can't change dozens of them at once and then just declare that all differences are due to the one variable you want to talk about.

Easy, easy on us, it wasn't us who did this test.😊 Alexa with Prores HQ instead of the FX and that 10bit codec will shine with that gorgeous image. But yeah, the whole image pipeline is to be accounted.

[kye]

6 hours ago, PannySVHS said:

Easy, easy on us, it wasn't us who did this test.😊 Alexa with Prores HQ instead of the FX and that 10bit codec will shine with that gorgeous image. But yeah, the whole image pipeline is to be accounted.

But I hold you personally accountable for the entire content of the internet!  Well, except for when @BTM_Pix makes funny jokes or posts ironic pictures of spam - that's all him.

[TomTheDP]

10 hours ago, kye said:

Interesting video.

First things first - could that guy be any cooler?  I'm pretty sure there isn't a single element in that video that isn't an automatic 10/10 for hipster chic.  Wow.  Talk about those people whose whole life is their own art project!

Secondly, his commentary is all over the place.

I see the differences he's talking about in the footage, plus a great many more that he probably sees but didn't mention - there are hue shifts and gamma shifts and subtractive colour operations and all sorts of wonderful things that are different between the two cameras.

However, he points to differences and says "see the difference with 10-bit?" whereas I think the differences are likely to be a mix of:

1. FX6 sensor read-out bit-depth
2. FX6 image processing
3. Sony RAW-LOG profile conversion (gamma, gamut, and bit-depth)
4. FX6 compression

I built a DCTL plugin for Resolve that reduces the bit-depth of the image, and to my surprise, you can reduce the bit-depth of rec 709 footage to 6-bits (and some shots to 5bit!) before there are visible changes.  

I'm not saying that there aren't any differences between 16-bit and 10-bit, because there are (however subtle they might be) but the things he was pointing at were definitely NOT all bit-depth related, and I'd suggest that most of them were processing/compression related actually.

If you want to understand what the differences are with one parameter, you can't change dozens of them at once and then just declare that all differences are due to the one variable you want to talk about.  You could take his whole video and just replace the phrase "10-bit" with "compression" and it would make just as much sense, despite having exactly the same examples.

So yeah, it's a great video to show an FX6 LOG vs Komodo X RAW comparison, but it isn't an isolated 10-bit vs 16-bit comparison.

That is LA for you haha.

Yeah I think you are right. Too many variables in that kind of test.

This might be a better comparison. Same camera and Prores 422 HQ has less compression than what the FX6 is doing. I also imagine the readout is the same in prores as it is in RAW. Can't confirm that though. 

 

 

[TomTheDP]

Not a scientific test either but maybe more informative than the first one haha. 

I think each camera's processing is just as important as the codec itself. That is what RAW theoretically gets around, the processing. Though with BRAW it's not really getting around that as much as it isn't a true RAW format. 

[eatstoomuchjam]

4 hours ago, TomTheDP said:

I think each camera's processing is just as important as the codec itself. That is what RAW theoretically gets around, the processing. Though with BRAW it's not really getting around that as much as it isn't a true RAW format. 

Keep in mind that even a number of cameras still do some processing even on so-called "raw" formats.  If they didn't, any two cameras using the same sensor would look exactly the same, but they don't.  There's a reason that raw looks different from multiple manufacturers despite that each one is using Sony sensors (for those who do - Sony, Fuji, Z Cam, Panasonic (I think?), and Black Magic (also I think?)).

Unrelated to that...
As far as whether raw is worth it, it depends on a combination of how much work you want to do with it and how flexible you want the footage to be.  For what you're doing, if you have lots of time to set up and prefer a slower pace of shooting, raw is probably not too important.  You have plenty of time to make sure to "get it right" in camera.  On the other hand, if storage space isn't a concern, there's very little reason _not_ to shoot raw.  

[kye]

On 11/29/2023 at 11:53 PM, TomTheDP said:

Not a scientific test either but maybe more informative than the first one haha. 

I think each camera's processing is just as important as the codec itself. That is what RAW theoretically gets around, the processing. Though with BRAW it's not really getting around that as much as it isn't a true RAW format. 

Agreed that it's often about the processing.

I think there's a bunch of stuff going on and often people don't understand the variables, or aren't taking into account all the relevant ones.  Also, people forget that the main goal of any codec is getting nuanced skin tones in the mids of whatever display space you'll be outputting to.

In that context:

• RAW in 12-bit is Linear, which is only about the same quality as 10-bit Log
• LOG in 10-bit isn't significantly better than 709 in 8-bit when exposed well
• It's commonly believed that only 12+bit RAW let's you adjust exposure and WB in post, and that 10-bit LOG is required for professional work, but thanks to colour management we can adjust exposure and WB in any of these (obviously 12-bit RAW > 10-bit LOG > 8-bit 709 from most cameras in real life and for big changes but if you're only making small changes then the errors are often negligible)
• 8-bit LOG is much worse than 8-bit 709 unless you were delivering to HDR (because if you convert LOG to 709 then you're pulling the bits in the mids apart significantly, which is absolutely visible)
• HLG is sometimes better than LOG - despite "HLG" being a marketing phrase and not a standard (unlike rec2020 or rec2100) it typically has a rec709 curve up to the mids then has a more aggressive highlight rolloff above that to keep the whole DR of the sensor, and combines this with a rec709 level of saturation.  This is brilliant because it typically means that you get the benefits of having a 10-bit image with 709-levels of saturation and the full DR of the sensor.  This is superior to a 10-bit LOG profile from the same camera (unless you clip a colour channel) because there is greater bit-density in the mids for skin tones (roughly equivalent to 12-bit LOG) and you get to keep the whole DR.  You can also change exposure and WB in post with proper colour management.  The GH5 and iPhone implementations of HLG are like this.
• The pros require greater quality than consumers because they have to keep clients happy when viewing the images without any compression.  Much of the subtleties get crunched by streaming compression, and unless you're shooting in controlled conditions you can't really expect to keep the last levels of nuance right into the final delivery.

On 11/30/2023 at 4:17 AM, eatstoomuchjam said:

Keep in mind that even a number of cameras still do some processing even on so-called "raw" formats.  If they didn't, any two cameras using the same sensor would look exactly the same, but they don't.  There's a reason that raw looks different from multiple manufacturers despite that each one is using Sony sensors (for those who do - Sony, Fuji, Z Cam, Panasonic (I think?), and Black Magic (also I think?)).

I'm keen to learn more about what happens at this stage of processing.  If you have resources on this, please share!

On 11/30/2023 at 4:17 AM, eatstoomuchjam said:

As far as whether raw is worth it, it depends on a combination of how much work you want to do with it and how flexible you want the footage to be.  For what you're doing, if you have lots of time to set up and prefer a slower pace of shooting, raw is probably not too important.  You have plenty of time to make sure to "get it right" in camera.  On the other hand, if storage space isn't a concern, there's very little reason _not_ to shoot raw.  

The other aspect to consider is that RAW and uncompressed aren't the same thing.

The megapixel race has obscured the fact that very close to 100% of material is now being shot at resolutions above the delivery resolution.  This is fine, and oversampling is great for image quality, but if you want to shoot RAW (and get the benefits of having no processing in-camera) and not have to deal with 5K / 6K / 8K source files when delivering 4K or even 1080p, then you have to deal with the sensor crop and having your whole lens package changed because of it.

The alternative to this, and what I think is the big advantage of Prores, is to have a full sensor readout downscaled in-camera, but unfortunately this typically means that you have some degree of loss of flexibility to the source image (either by compression, bit-depth reduction, or even straight-out over processing like iPhone 14 did).  The alternative is to downscale in-camera and then just save the images as uncompressed.  

There is nothing stopping manufacturers from implementing this.  The GH5 downscaled 5K in-camera and many cameras record Cinema DNG sequences in-camera - just combine the two and we're there.

[eatstoomuchjam]

22 minutes ago, kye said:

The megapixel race has obscured the fact that very close to 100% of material is now being shot at resolutions above the delivery resolution.  This is fine, and oversampling is great for image quality, but if you want to shoot RAW (and get the benefits of having no processing in-camera) and not have to deal with 5K / 6K / 8K source files when delivering 4K or even 1080p, then you have to deal with the sensor crop and having your whole lens package changed because of it.

The alternative to this, and what I think is the big advantage of Prores, is to have a full sensor readout downscaled in-camera, but unfortunately this typically means that you have some degree of loss of flexibility to the source image (either by compression, bit-depth reduction, or even straight-out over processing like iPhone 14 did).  The alternative is to downscale in-camera and then just save the images as uncompressed.  

There is nothing stopping manufacturers from implementing this.  The GH5 downscaled 5K in-camera and many cameras record Cinema DNG sequences in-camera - just combine the two and we're there.

Some of this is also being done at the sensor level instead of in the camera.  The GFX 100 series have a full sensor width 10-bit 4K mode (12-bit in raw).  They're not reading the entire sensor and downscaling.  They're using a 4K readout mode that's built into the sensor and when set for raw HDMI output, feeding that to the recorder.  It's "raw" in the sense that it's the exact data that the SOC receives from the sensor.  I assume that the implementation is similar in at least some other cameras.

27 minutes ago, kye said:

I'm keen to learn more about what happens at this stage of processing.  If you have resources on this, please share!

No links to share just now - just stuff I've read over the years as well as applying logic that if two cameras have the same sensor, but non-matching raw output, they must be doing some processing.  Not all are confirmed, but Sigma fp/Sony A7 III, Sigma FP-L/Sony A7R IV, GFX 100/H2D, Z Cam E2/Panasonic GH5S/BMPCC4K, etc etc etc.  If zero processing is applied, any of those cameras should look exactly the same when shooting raw...  unless the colors are determined by the raw decoder software - and if that's the case, it should be possible for the vendors to easily transform any of those cameras to any of the others.  🙂

[kye]

22 minutes ago, eatstoomuchjam said:

Some of this is also being done at the sensor level instead of in the camera.  The GFX 100 series have a full sensor width 10-bit 4K mode (12-bit in raw).  They're not reading the entire sensor and downscaling.  They're using a 4K readout mode that's built into the sensor and when set for raw HDMI output, feeding that to the recorder.  It's "raw" in the sense that it's the exact data that the SOC receives from the sensor.  I assume that the implementation is similar in at least some other cameras.

No links to share just now - just stuff I've read over the years as well as applying logic that if two cameras have the same sensor, but non-matching raw output, they must be doing some processing.  Not all are confirmed, but Sigma fp/Sony A7 III, Sigma FP-L/Sony A7R IV, GFX 100/H2D, Z Cam E2/Panasonic GH5S/BMPCC4K, etc etc etc.  If zero processing is applied, any of those cameras should look exactly the same when shooting raw...  unless the colors are determined by the raw decoder software - and if that's the case, it should be possible for the vendors to easily transform any of those cameras to any of the others.  🙂

Yes, I understand the logic.

One element of colour science often forgotten is the choice of RGB filters for the filter array, but I'm not sure if that would also be the same between these cameras, or if they'd just be supplied by Sony and therefore be identical between brands?  In theory this would create differences due to the corners of the gamut, but once transformed to XYZ differences between two different filters on the same sensor would be rounding errors at best.

The best discussion I've seen on digital processing is the discussion of the Alexa 35 - page 52 onwards.

https://www.fdtimes.com/pdfs/free/115FDTimes-June2022-2.04-150.pdf 

As you say, much of what is going on might well be things that occur on the sensor.

The other component that is worth looking at between RAW sources is the de-bayer algorithms between cameras.  AFAIK you can't choose which de-bayer algorithm gets used on RAW footage, so there might be differences between the manufacturers algorithms contributing to the differences we see in real life.

[Jedi Master]

52 minutes ago, kye said:

The best discussion I've seen on digital processing is the discussion of the Alexa 35 - page 52 onwards.

Interesting. That’ll make for good bedtime reading. Thanks.

[eatstoomuchjam]

On 11/30/2023 at 11:33 PM, kye said:

One element of colour science often forgotten is the choice of RGB filters for the filter array, but I'm not sure if that would also be the same between these cameras, or if they'd just be supplied by Sony and therefore be identical between brands?  In theory this would create differences due to the corners of the gamut, but once transformed to XYZ differences between two different filters on the same sensor would be rounding errors at best.

Sure, RGB microlenses on the CFA could cause some variance - and I certainly don't have a lot of insight to that.  I've generally assumed that they're supplied by Sony since they sell both "color" and "monochrome" versions of the same sensors - and largely, the only difference in a color and monochrome sensor is the CFA.  😃
 

On 11/30/2023 at 11:33 PM, kye said:

The other component that is worth looking at between RAW sources is the de-bayer algorithms between cameras.  AFAIK you can't choose which de-bayer algorithm gets used on RAW footage, so there might be differences between the manufacturers algorithms contributing to the differences we see in real life.

That's definitely relevant to ProRes/H.265 footage and somewhat with BM Raw, but not for cDNG and ProRes RAW.  Perhaps there's metadata embedded in the container that instructs the editor/convertor for which color values to use when debayering?  And if so, could the writers of the software allow people to override it?

[A_Urquhart]

To the OP, I’d seriously consider the FX6. Having used both the FX6 and FX9 i find the FX6 slightly more versatile. 
The FX6 is a great camera that will handle any level of work you need it to. You will not get any complaints from professionals in regards to the footage it produces despite what you may read here. 

[thecouchguy]

C70 with speed booster. This way you can get some RF 24-70mm & RF 70-200mm and have class leading dynamic range and versatility. Then if you need more low light you can chuck on the focal reducer for effective full frame. Or as others have said, fx6.

[Jedi Master]

I’m not interested in RF lenses. I plan to get a set of prime cine lenses in PL mount.