PDA

View Full Version : This may explain the FS-100's great low light performance.



David G. Smith
05-06-2011, 03:40 PM
Here is an interesting article, from Sony, on the ProVideo Coalitions website. Sony is claiming that their compressor encoders on the FS100 and the NX5U are not only "Content adaptive", but also "Operations-adaptive". This means that the camera settings, such as gain up, are sensed by the compression encoder and changes the way it works. Very interesting. Have a look.

http://provideocoalition.com/index.php/sony/story/avchd_with_a_twist/

Barry_Green
05-06-2011, 08:35 PM
That would be really cool, if it actually works. Funny thing though, I was interested in the performance of how Sony's NXCAM implementation would fare, so I got an NX5U for a day and chained it to an AVCCAM recording deck through SDI, so I could directly compare the two codecs side by side. And I tortured them. Super-highly-detailed scenes, zooms and rotation throughout, etc. Conclusion -- six of one, half a dozen of the other. There were instances where the Sony codec looked better and the Panasonic choked, and there were as many instances where the Sony codec choked and the Panasonic looked better. I came away from that test with the idea that they are comparable level codecs.

I didn't try applying gain, I don't know if that would make a difference or not.

David G. Smith
05-06-2011, 08:51 PM
This is far from my area of expertise, but I would imagine that if what they are asserting is true, it would be for "By the numbers" types of parameters, like gain. My fear is that under high gain settings other codec stressing variables might be compromised.

David G. Smith
05-06-2011, 09:01 PM
I have put my deposit in with AbelCine (NYC) and am pretty high in there pre-order batting order for delivery of a FS100 at the end of month or so. How do you suggest I set up a test to try and confirm this assertion of the "Operations-adaptive" compressor encoder? I would not have access to other cameras, but it might be worth a try.

nyvz
05-06-2011, 10:36 PM
Well it seems unlikely that much of the FS100's great low light performance comes from a particularly efficient compression scheme. It probably helps retain certain types of detail despite the low bitrate, but if the images look clean on the LCD or HDMI output before compression, which they do, its probably just the sensor. It's really not that surprising that it the sensor is so good in low light when you consider that the 5dmk2, which only actually collects light from ~1/8th of its sensor area, is really only collecting light from a sensor area about the same size as S16mm. The FS100, with its newer sensor and big photosites that do not throw out photons from the S35mm image area, has the potential to be about 2stops better in low light given similar photosite technology. And sensor technology could have improved in the past 3 years... Seems like there has always been a lot of room for gains in low light performance, it's just that no one went to the trouble of doing it properly and making it this affordable.

David G. Smith
05-06-2011, 11:09 PM
Well it seems unlikely that much of the FS100's great low light performance comes from a particularly efficient compression scheme. It probably helps retain certain types of detail despite the low bitrate, but if the images look clean on the LCD or HDMI output before compression, which they do, its probably just the sensor. It's really not that surprising that it the sensor is so good in low light when you consider that the 5dmk2, which only actually collects light from ~1/8th of its sensor area, is really only collecting light from a sensor area about the same size as S16mm. The FS100, with its newer sensor and big photosites that do not throw out photons from the S35mm image area, has the potential to be about 2stops better in low light given similar photosite technology. And sensor technology could have improved in the past 3 years... Seems like there has always been a lot of room for gains in low light performance, it's just that no one went to the trouble of doing it properly and making it this affordable.

Yeah, probably so. It is just that the recent low light tests that have been circulated show almost an eerily clean image at very high gain settings. I agree that the best part of the set up is the sensor design, but I think that Sony may very well be up to something very cool with their "Operations-adaptive" compression encoders. That they have announced this in regards to two of their low end pro-sumer cameras is very interesting, IMHO.

brunerww
05-07-2011, 04:03 AM
...almost an eerily clean image...

I like the way you put that, David. "Eerie" is a good word for it. The thing that has me most excited about this camera is how much money/hassle that filmmakers will save on lighting. If it performs as well initial tests seem to indicate, shooters can sell their old cameras and their old lights to buy it.

Congrats on getting on the list.

Bill

greymog
05-07-2011, 07:01 AM
I think someone should keep making this point, I know it's been made to death, but I've been finding out the hard way. RPO put it well in his thread when he said lighting is there for two reasons, one, to illuminate, and two, to create mood.

The idea of a highly sensitive camera that can deliver a clean image is great, but shouldn't have to mean 'doing away with light'. That said, it would be nicer to get on set with one or two soft bulbs and just focus on creatively lighting, as opposed to also worrying about sufficient illumination.

I been shooting a film recently, (after I sold my cameras) on whatever I can find that day. So far I've used a 60D a 550D a 7D a 5Dii and an iphone 4 and a 500d, along with a sony camcorder form radio shack. Then one day this guy from the newspaper opposite our office offered us his D3s. The first day I had it I dispensed with all the lights, the thing can see in the dark, granted, but the footage didn't look so nice for the film. I reshot, with the D3s, but realized that no matter how clean the hi iso gets, I had to consider light.

Just have to make this point. It would hurt if high isos meant getting rid of lights altogether.

Also, that said, that D3s is ridiculous.

Jason Allen
05-07-2011, 08:24 AM
I like the way you put that, David. "Eerie" is a good word for it. The thing that has me most excited about this camera is how much money/hassle that filmmakers will save on lighting. If it performs as well initial tests seem to indicate, shooters can sell their old cameras and their old lights to buy it.

Congrats on getting on the list.

Bill

To quote Charleton Heston: "From my cold, dead hands!" No camera - regardless of how well it does in low light - will be my reason to sell my lights. The difference between a polished shot that I MADE and a guerilla shot that I CAPTURED is shaping the light in the scene! That said, my cameras are going on the auction block as soon as I wrap up my May freelance gigs. There is an FS100 in my future! :D

Rick Burnett
05-13-2011, 02:24 PM
For me I just want smaller more power efficient lights instead of the heaters I have now :) I can afford to cater the food now so we don't have to make grilled cheese under the lights on set! :) When we went from shooting the HVX200 to the 7D, one of the biggest things we liked was not requiring so much light. We still used a lot of light, but it seemed a lot easier and safer in certain scene setups to use less powerful options. We also used flashlights to do some really cool things, which was great, because we could hide them VERY well.

Justyn
05-13-2011, 08:50 PM
That's true and most of the best DPs I've worked with got better results without overcomepensating and having a ton of lights. I once worked on a shoot were a DP used 9 lights in a closet. Can you talk about grip jungle and then when it was done, it didn't look great at all....

That's what I like about the large format cameras.. We can do more with less, and in reality.. that's how life looks... natural.

Postmaster
05-18-2011, 07:29 AM
I think someone should keep making this point, I know it's been made to death, but I've been finding out the hard way. RPO put it well in his thread when he said lighting is there for two reasons, one, to illuminate, and two, to create mood. .

I second that.

Light (and shadows) is one of your best creative tools in the box. Throwing it away, just because you can get "such a clean picture" out of this cameras would be plain stupid.

What I found is, that most digital cameras are not very happy with low light temperatures, so daylight Kinos or small HMI lights are in order. But yeah, you gonna need less powerful lights.

Frank

David G. Smith
05-18-2011, 01:37 PM
I don't think that the issue here is that you can get rid of your light kit because the FS100 is so sensitive. That is of course crazy. What I found interesting is what appears to be the innovative way that Sony is claiming that they improved the sensitivity of this camera with the "Operational adaptive" encoding. Or actually, how they say that they increased the image quality at higher sensitivities.

That is interesting about the use of low color temp lights Frank. I have seen your posts about that before Frank. That just may another area where "Operational adaptive" encoding may help to clean up the image. It might be possible for the camera to sense the boost in the blue channel, and the associated increase in noise, and optimize the encoding to clean it up a bit. Does anybody have the the ear of a Sony camera design engineer?

nyvz
05-18-2011, 04:08 PM
Keep in mind that the only effect encoder improvements can have is to lessen the visible destructiveness of encoding compression. It has the potential to increase image quality of high-gain footage relative to the low-bitrate encoded footage processed without the optimization, but it certainly wont give you image quality over significantly higher-bitrate encoding schemes. If you plug your FS100 into an HDMI monitor and look at the live image, you are looking at the best possible image you can get out of the camera, running uncompressed over HDMI at ~800Mbps, and if you're lucky the 24Mbps data stream that is recorded to your SDHC cards looks pretty similar to that image despite reorganizing that image data (and removing some) so that it takes up ~33x less space. All the adaptive encoding is doing is adjusting things a little bit in various situations to make sure noisy images dont make that 33:1 compression really show up, it wont bring back that 97% of your data that is gone, it just tries to make sure high noise shooting doesnt make it look like you lost 98% of your data.

Also, you are always losing something increasing noise and detail in your images without changing the bitrate, if operational adaptive encoding tells the encoder to focus on dark areas to preserve unique noise detail, you'll risk more artifacting in midtones and highlights. If operation adaptive encoding tells it to avoid getting caught up wasting bits on compressing noise, you may get midtones and highlights that are better than without the optimization when shooting at high gain, but you'll be at much greater risk for artifacting and detail loss in shadow areas. In the end, higher gain means more noise, which means more unique detail that is hard to compress, and without changing the bitrate, you'll still lose some image quality shooting at high gain (higher noise) over not, its just that now Sony has chosen what part of your images should lose quality based on whatever their understanding of what is most easy to perceive. You'll still get better quality shooting at lower gain in the first place.

David G. Smith
05-18-2011, 04:27 PM
Keep in mind that the only effect encoder improvements can have is to lessen the visible destructiveness of encoding compression. It has the potential to increase image quality of high-gain footage relative to the low-bitrate encoded footage processed without the optimization, but it certainly wont give you image quality over significantly higher-bitrate encoding schemes. If you plug your FS100 into an HDMI monitor and look at the live image, you are looking at the best possible image you can get out of the camera, running uncompressed over HDMI at ~800Mbps, and if you're lucky the 24Mbps data stream that is recorded to your SDHC cards looks pretty similar to that image despite reorganizing that image data (and removing some) so that it takes up ~33x less space. All the adaptive encoding is doing is adjusting things a little bit in various situations to make sure noisy images dont make that 33:1 compression really show up, it wont bring back that 97% of your data that is gone, it just tries to make sure high noise shooting doesnt make it look like you lost 98% of your data.

Also, you are always losing something increasing noise and detail in your images without changing the bitrate, if operational adaptive encoding tells the encoder to focus on dark areas to preserve unique noise detail, you'll risk more artifacting in midtones and highlights. If operation adaptive encoding tells it to avoid getting caught up wasting bits on compressing noise, you may get midtones and highlights that are better than without the optimization when shooting at high gain, but you'll be at much greater risk for artifacting and detail loss in shadow areas. In the end, higher gain means more noise, which means more unique detail that is hard to compress, and without changing the bitrate, you'll still lose some image quality shooting at high gain (higher noise) over not, its just that now Sony has chosen where you will lose it and where you will gain it based on whatever their understanding of what is most easy to perceive. You'll still get better quality shooting at lower gain in the first place.

Oh yeah, of course. However, I am sure that Sony or the other camera manufactures will develop this innovation and apply it to cameras with more robust codecs in the future and adapt it to other camera parameters. Who knows, user control over the choices being made by the encoder optimization may be a must have feature set on future cameras.