10 bit vs 8 bit output

[James Houk]

Originally Posted by Paul Curtis

When it comes down to actual image quality i've seen nothing yet that really shows major differences. I am not comparing Slog or anything like that, just a base F3 vs FS100 in terms of sensor and the image quality you can get out of it. I think Alister himself states they are very close in his review.

But if you monitor externally and record externally - i've still seen nothing that really makes a case for a quality difference between them to such a degree. Purely from image quality - not handling or workflow.

The fact that both cameras share the same chip means there's going to be a similarity in look. But to get back to "8 bit vs 10 bit," you're not going to see a substantial visual difference in ungraded footage recorded 8 or 10 bit.

10 bit on the F3 (even without SLog) is going to give you more room to push the images in post.

Under ideal controlled lighting the differences between the images are going to be minimized. The F3 is going to shine under circumstances that aren't ideal.
In circumstances where you can control your lighting and

[Paul Curtis]

Thanks Doug and James,

The crux of my point is that if you remove the internal compression and do your post in 10+ bit depth are the images from both cameras really that different.

Doug, you have a super strong opinion on this and made some black and white statements around the highlight handling - are you aware of anywhere i can see this? Something like that *would* make a big difference to me.

I don't think anyone is arguing against the ergonomics or the feature set - the F3 is obviously better.

cheers
paul

[Brian Drysdale]

Unfortunately, you won't have the 10 bit images in the first place from the FS 100.

Although it's not the full F3 range, I don't know if anyone has tested Andy Shipsides set ups yet.
Sony NEX-FS100: Dynamic Range Test | CineTechnica

Of course, it's more than just numbers and if you look at the single sensor evaluation video The Great Camera Shootout 2011: SCCE ~ Episode One | Zacuto USA you can see cameras that have the same dynamic range handling highlights rather differently to each other.

[James Houk]

Originally Posted by Paul Curtis

The crux of my point is that if you remove the internal compression and do your post in 10+ bit depth are the images from both cameras really that different.

You can do "post processing" on the FS100 in 10 bit, or 12bit, or 32bit floating point. Ditto for the PMW-F3 or any camera, that comes down to your post hardware - and doing such can be useful if you're stacking a lot of filters or effects.

BUT. The FS100 only *outputs* an 8 bit feed via HDMI. Your source is never going to be better than bit on the FS100. On the PMW-F3 the SDI output is 10bit. That means 1024 graduations for each color channel instead of 256 (well, not quite, you can't use 0-15, but still, it's close).

[Alister Chapman]

Forget 8 bit, 10 bit, codecs, whatever.

The F3's output just looks better IMHO. It appears to have a richness, depth and realism to the image that just isn't quite there with the FS100. That's why I used the word verisimilitude in my first post in this thread. You look at the pictures and perhaps you can't quite pin down exactly what it is, but the F3 just plain and simple looks better. I believe it is down to improved micro contrast, less obvious processing and smoother highlight handling.

Don't get me wrong the FS100 is a great little camera and a great piece of kit for the money, but it's not in the same class as the F3 and no external recorder will ever make it into an F3. I just purchased a second F3, for the money I could have purchased two FS100's and some accessories, but that just would not have given me the image quality I get from the F3.

[Paul Curtis]

Originally Posted by Alister Chapman

Forget 8 bit, 10 bit, codecs, whatever.

The F3's output just looks better IMHO. It appears to have a richness, depth and realism to the image that just isn't quite there with the FS100. That's why I used the word verisimilitude in my first post in this thread. You look at the pictures and perhaps you can't quite pin down exactly what it is, but the F3 just plain and simple looks better. I believe it is down to improved micro contrast, less obvious processing and smoother highlight handling.

Okay, i get the idea :)

On your still frame examples did you use the same lens on both cameras, to eliminate that difference?

There is more detail in the F3. The detail circuits seem a little higher but there is more there. I would have thought it would have been the lens. It's true that the debayering could be different but why on earth would sony develop two debayering solutions for one sensor. I can understand cutting the outputs and more consumer compression - but at that deep firmware level why double their engineering effort...

Maybe i'll see if i can hire both and see for myself.

Cheers
paul

[Brian Drysdale]

The cameras are made by two different divisions of Sony, aimed at different markets. They're saving on the sensor development and production costs, but the rest is what each design department believe will work in their products for a particular selling price. The F3 customers are prepared to pay more for those extra percentage points of image quality, so it's worth Sony's while putting in the extra effort to use higher spec processing.

[Alister Chapman]

Yes in my samples, both cameras used the same Nikon 50mm f1.8 lens.

[David Heath]

Originally Posted by Paul Curtis

.........but why on earth would sony develop two debayering solutions for one sensor.

The simple answer is to save money. Have one sensor for economies of scale, but develop two methods of reading the data and processing it. One doing it as well as possible (but complex and hence expensive) the other more simply - but at a far more affordable price.

I wouldn't be surprised if the FS100 didn't debayer as such - rather directly derived a RGB value for each photosite quartet:

RG
GB

And there's evidence that it may from some of the alias fundamental numbers. Such an approach wouldn't be expected to give as high resolution as a full debayer - but would be better in other respects such as price and (especially) power consumption. And just compare power consumption of the FS100 and the F3......

I don't find it at all surprising that the F3 looks better than an FS100 - it costs a lot more. That's not to say the F3 looks at all bad - far from it - but you get what you pay for.

[Paul Curtis]

Originally Posted by David Heath

The simple answer is to save money. Have one sensor for economies of scale, but develop two methods of reading the data and processing it. One doing it as well as possible (but complex and hence expensive) the other more simply - but at a far more affordable price.

I wouldn't be surprised if the FS100 didn't debayer as such - rather directly derived a RGB value for each photosite quartet:

RG
GB

And there's evidence that it may from some of the alias fundamental numbers. Such an approach wouldn't be expected to give as high resolution as a full debayer - but would be better in other respects such as price and (especially) power consumption. And just compare power consumption of the FS100 and the F3......

I don't find it at all surprising that the F3 looks better than an FS100 - it costs a lot more. That's not to say the F3 looks at all bad - far from it - but you get what you pay for.

That's an interesting thought.

I would have thought most of the cost in the firmware/hardware design is R&D. The actual manufacturing costs for F3 vs FS100 can't be that different.

So there's evidence of more aliasing in the FS100, can you point me anywhere to find out more?

Also, rather than start a new thread as there's some very knowledgable people here, can i used E mount lenses on the F3 fairly easily? I have a lot of weird and wonderful lenses and E mount adaptors here. Would be nice to know they could by used in some cases...

thanks
paul

[David Heath]

Originally Posted by Paul Curtis

So there's evidence of more aliasing in the FS100, can you point me anywhere to find out more?

No, not saying there is more aliasing, it actually looks pretty good for lack of it, but any camera will have some, and the pattern may be seen as a bit like a fingerprint, giving away facts about the sensor and the way it's read.

The intensity of the aliases will change with lens, focus etc etc but the shape of the patterns won't. It doesn't matter what lens you put on etc, the shape and patterns will remain the same. I haven't been able to see like by like comparisons of zone plates from the two, but I suspect the patterns differ between the two cameras. Which is interesting.

The real point is that the quality is not solely defined by the sensor - it's a combination of the sensor, the way it's read, and the way it's processed. Quality wise the FS100 is pretty good for the price - but it doesn't surprise me that the much more expensive F3 is better.

[Doug Jensen]

Originally Posted by Paul Curtis

The actual manufacturing costs for F3 vs FS100 can't be that different.

You obviously have not had your hands on both cameras.
Could you build a Mercedes for the same price as a Fiat after all the R&D was finished?
There's no way the manufacturing costs are anywhere near being close on these two cameras.

[Alister Chapman]

The alias patterns for both the F3 and FS100 are extremely similar, the only difference is the FS100 aliases are a little stronger. I've done side by side zone plates with the same lens on both cameras.

[David Heath]

Originally Posted by Alister Chapman

I've done side by side zone plates with the same lens on both cameras.

Which is what I've not seen, so I'll defer to you, Alister - I've been having to go on different tests done at different times. It does then leave me wondering why there is such a difference in power requirements between the two cameras?

Is there any chance of seeing the results you talk of?

[Paul Curtis]

Originally Posted by David Heath

Which is what I've not seen, so I'll defer to you, Alister - I've been having to go on different tests done at different times. It does then leave me wondering why there is such a difference in power requirements between the two cameras?

Perhaps the compression is a lot more power hungry. The FS100 would be using cheap and cheerful mass produced AVC compression hardware.

paul