F5/F55 Sensor resolution and bayer pattern

nyvz said:
Steve, at this point it is still not clear to me what you are saying, and it sounds like you may be further confusing yourself. What we care about is the number of photosites read-out to create the images that we record from the camera. There are many pixel counts, and the only one we care about is the one that is used for the recorded image area.

It sounds like you do understand that sensors often have more gross pixels, and that specs are often given for numbers of pixels windows in that gross area that are used for imaging or and for recording. So where is the confusion? Why are you then surprised that Sony says a sensor has 11.6MP and is only using a 8.9MP portion of it to create recorded images? Clearly it is common for the area used for recording to be inside of the sensor, with outer area used for lookaround and for calibration area.

Are you talking about measured resolution or something? If so, measured how? MTF50? No manufacturer ever mentions measured resolution except RED who does not advertise it but sometimes tells people guidelines for it which often confuse people since there are many ways to measure resolution and most people have nothing to compare it to since no other manufacturer ever talks about measured resolution for good reason.
Click to expand...

There is considerable confusion on many folks in this thread. In this thread, there is the erroneous defintion of 'effective resolution' that is also combined with further confusion about double-bayer technology. Granted, a lot of this confusion can be attributed to Sony themselves. Incosistent documentation and perhaps intentionally misleading.

Again, "the number of photosites read-out to create the image" is called active pixel resolution. It is not effective resolution. This is where I think you are confused.

Let's consider for a moment the definition of 'effect'. It is "something that is produced by an agency or cause; result; consequence". Effective resolution is the "resulting" resolution from the sensor. It can't get more literal than that.
 
Steve Castle said:
Sony's defintion of "true 4K" is 4:4:4 4K. To do this, they interpolate Q67->Q100. The logic is that 26.5MP makes photosites smaller harming DR, hence having zero interpolation on green pixels and interploation only on red/blue pixels resulting in f.6 of full RGB.

Also, F5/F55 would be based on 11.6MP, which would be closer to ~Q43. It would only Q33 like any other bayer sensor if its only using the exact 8.85MP area. Obviously Sony wants you to buy the F65 if you want what they consider to be "true 4K".
Click to expand...
Obviously, "True" means different things to different people.

With 4K, 26.5MP are required to produce what I will term "Full" 4K. Anything less involves interpolation, and interpolation is not, at least for me, "True" in any sense of the word.

According to that definition, for Full 4K:

The 20MP F65 would have 75% of the needed photosites.
The 11.6MP F5 would have 44% of the needed photosites.
The 4.6MP Alexa would have 17% of the needed photosites.

Similarly, for Full 1080 HD (6.2MP):

The 3.4MP F3 would have 50% of the needed photosites.
The 20MP F65 would have 322% of the needed photosites.
The 11.6MP F5 would have 186% of the needed photosites.
The 4.6MP Alexa would have 73% of the needed photosites.

Obviously, how the cameras go about producing their output has great effect, but this is at least method to compare theoretical sensor resolutions without all the hype.

Cheers.
 
Admit One said:
Obviously, "True" means different things to different people.

With 4K, 26.5MP are required to produce what I will term "Full" 4K. Anything less involves interpolation, and interpolation is not, at least for me, "True" in any sense of the word.

According to that definition, for Full 4K:

The 20MP F65 would have 75% of the needed photosites.
The 11.6MP F5 would have 44% of the needed photosites.
The 4.6MP Alexa would have 17% of the needed photosites.

Similarly, for Full 1080 HD (6.2MP):

The 3.4MP F3 would have 50% of the needed photosites.
The 20MP F65 would have 322% of the needed photosites.
The 11.6MP F5 would have 186% of the needed photosites.
The 4.6MP Alexa would have 73% of the needed photosites.

Obviously, how the cameras go about producing their output has great effect, but this is at least method to compare theoretical sensor resolutions without all the hype.

Cheers.
Click to expand...

Which is the reason I put "true" in quotation marks. You're right in that it really isn't "true", its Sony's definition of true.

I posted Sony's reasoning in a previous post within this thread:
http://www.dvxuser.com/V6/showthrea...attern&p=1986238443&viewfull=1#post1986238443

But I get Sony's logic. 26.5MP would be cumbersome and would mean higher pixel density resulting in reduced low-light capabilities and dynamic range. Not having to do any interpolation of green photosites should be a huge advantage.

If you look at Sony's PDF linked on that post you'll also see a nice graph on 1080p/2K on the F65.
 
In the end all most people should care about is the NET RESOLUTION (Effective resolution is the "resulting" resolution from the sensor.) - not the quantity of RAW photosite pixel count as depending on various patterns or processing algorithms resulting resolvable resolution can actually vary.

The whole thread that Duke started was from when me and him were discussing the F5/F55 and that RED was only resolving around 3.2k, and my only argument is that the F5/F55 will be around the same. But I am talking about LINES OF RESOLUTION.

Forget the F65, 20MP/17.6MP / Q67 sensor since it has absolutely nothing to do with this thread.
 
Steve Castle said:
Again, "the number of photosites read-out to create the image" is called active pixel resolution. It is not effective resolution. This is where I think you are confused.
Click to expand...

So it sounds like you do mean to say you think effective resolution is expected to be measured resolution? Measured how? And why talk about it if no manufacturer ever talks about it or quotes it besides RED who has confused some people. Manufacturers only ever talk about the count of photosites used on their sensor since that is a simple number to count and compare. I know the number is often thrown around that 4k RED measures about 3.2k resolution, but do you actually know what that means or how that was measured? It is not as simple as a 4k bayer sensor with OLPF is equal in resolution to a 3.2k monochrome sensor with no OLPF... The 3.2k value is somewhat meaningless with nothing to compare it to, it really only exists to serve as a guideline that tells the average user not to be surprised when his 4k bayer images look approximately ~20% softer than expected when viewed as 4k RGB images. It is a good way to remind people they should be shooting in 5k bayer if they want their 4k RGB images to be sharper...

You guys do know that measured resolution in the real world is not as simple as pixel resolution, right? It is not so meaningful to give it a linear value like we do with pixel counts. Real measured resolution is rarely talked about and often misused because pixel resolution is so easy to talk about since it comes from a finite and counted number of samples. You guys might want to read up on MTF before talking about measured resolution. I think if you are expecting any manufacturer to talk about measured resolution then you are mistaken and do not understand the complexities of what measured resolution is. No digital camera manufacturer has ever quoted measured resolutions in their spec sheets or advertising that I know of. They just talk about the number of photosites on their sensors, since that is far more easily quantifiable.

Lenses often list MTF curves, which can be compared to some degree, but that is how measured resolution compares in the real world, its a bit harder to do between digital image sensors.
 
Last edited:
starcentral said:
In the end all most people should care about is the NET RESOLUTION (Effective resolution is the "resulting" resolution from the sensor.) - not the quantity of RAW photosite pixel count as depending on various patterns or processing algorithms resulting resolvable resolution can actually vary.

The whole thread that Duke started was from when me and him were discussing the F5/F55 and that RED was only resolving around 3.2k, and my only argument is that the F5/F55 will be around the same. But I am talking about LINES OF RESOLUTION.

Forget the F65, 20MP/17.6MP / Q67 sensor since it has absolutely nothing to do with this thread.
Click to expand...

nyvz said:
So it sounds like you do mean to say you think effective resolution is expected to be measured resolution? Measured how? I know the number is often thrown around that 4k RED measures about 3.2k resolution, but do you actually know what that means or how that was measured? It is not as simple as a 4k bayer sensor with OLPF is equal in resolution to a 3.2k monochrome sensor with no OLPF... The 3.2k value is somewhat meaningless with nothing to compare it to, it really only exists to serve as a guideline that tells the average user not to be surprised when his 4k bayer images look approximately ~20% softer than expected when viewed as 4k RGB images. It is a good way to remind people they should be shooting in 5k bayer if they want their 4k RGB images to be sharper...

You guys do know that measured resolution in the real world is not as simple as pixel resolution, right? It is not so meaningful to give it a linear value like we do with pixel counts. Real measured resolution is rarely talked about and often misused because pixel resolution is so easy to talk about since it comes from a finite and counted number of samples. You guys might want to read up on MTF before talking about measured resolution. I think if you are expecting any manufacturer to talk about measured resolution then you are mistaken and do not understand the complexities of what measured resolution is. No digital camera manufacturer has ever quoted measured resolutions in their spec sheets or advertising that I know of. They just talk about the number of photosites on their sensors, since that is far more easily quantifiable.

Lenses often list MTF curves, which can be compared to some degree, but that is how measured resolution compares in the real world, its a bit harder to do between digital image sensors.
Click to expand...

That's the thing, we really don't know how many photosites that the F5/F55 is actually sampling. If it is only 4K exactly, then it would be resolving around 3.2K, if its actually sampling a lot larger area (closer to 11.6MP) it would be more than that.

To clarify confusion, let's make a table:


F65:
Gross pixels : 20.4MP
Active Pixel Resolution : 18.7MP (From previous post, "18.7MP used for imaging. The others are for black balance, image correction, and output functions")
Effective Resolution : 17.7MP (resolution of the F65RAW file)


F5/F55:
Gross pixels : 11.6MP
Active Pixel Resolution : ?
Effective Resolution : 8.85MP (resulting resolution you are getting from the camera, this is what you work with)


We don't know what active pixel resolution is for the F5/F55. This is the where the confusion is coming from as defintions have become jumbled up to mean "active pixel resolution" ="effective resolution". It can be for bayer, but it also may not be. It could be that the sensor is only sampling 8.85MP and discarding the rest. But like I said previously, you're discarding almost as much as the entire F3 sensor. It also could mean that sensor is sampling a larger area afforded to it, in which cases it won't be ~3.2K after the bayer penalty is calculated.

So far, Sony is not talking. But we should not immediately assume that because the effective resolution is 4K/8.85MP that it can only resolve ~3.2K. It may not be the case.
 
Last edited:
Pretty sure it is quite clear that the F5/F55 Active Pixel Resolution is 8.9MP...

I think you are confused, 4k RAW bayer means it is using 4096x2160 photosites (active area) from the sensor and recording that data directly in their 4k RAW format with no scaling, no binning, etc. All the numbers you are listing as "effective resolution" numbers are actually "active pixel resolution" numbers. I can understand why you would be hoping for oversampling though, but RAW is not supposed to be oversampled/downscaled, its straight from the sensor.

No manufacturer has ever oversampled RAW recordings by anything other than 2x, 3x, or 4x. There is no way to properly downsample 11.6MP and still record it in a 8.9MP RAW format, it has never been done, and there is no reason to do it and no way to do it well that has been seen or talked about.

I think you are getting confused by Sony's F65 numbers which are so close to one another they are probably just roundings or typos or just slightly different specs (including or lookaround or black calibration area or not).
 
Steve Castle said:
So far, Sony is not talking. But we should not immediately assume that because the effective resolution is 4K/8.85MP that it can only resolve ~3.2K. It may not be the case.
Click to expand...
Clearly we can't talk about what it will be, but only the limit of what it could be given the sensors fixed photosite count.

One of the drawbacks to using a Bayer pattern is that the green resolution is not the same red and blue. As a result, the choice of OLPF strength (resolution vs. aliasing) becomes crucial, and Sony's choice for the F5 may not be the same as for the F55.

Although few seem to take the time to be accurate, any resolution measurement for a Bayer pattern sensor should be done separately for each channel.

Cheers.
 
nyvz said:
Pretty sure it is quite clear that the F5/F55 Active Pixel Resolution is 8.9MP...

I think you are confused, 4k RAW bayer means it is using 4096x2160 photosites (active area) from the sensor and recording that data directly in their 4k RAW format with no scaling, no binning, etc. All the numbers you are listing as "effective resolution" numbers are actually "active pixel resolution" numbers. I can understand why you would be hoping for oversampling though, but RAW is not supposed to be oversampled/downscaled, its straight from the sensor.

No manufacturer has ever oversampled RAW recordings by anything other than 2x, 3x, or 4x. There is no way to properly downsample 11.6MP and still record it in a 8.9MP RAW format, it has never been done, and there is no reason to do it and no way to do it well that has been seen or talked about.

I think you are getting confused by Sony's F65 numbers which are so close to one another they are probably just roundings or typos or just slightly different specs (including or lookaround or black calibration area or not).
Click to expand...

We don't know this, and we shouldn't immediately assume that 4096x2160 is the active area, we just know that it is the resulting resolution. We actually don't know what the RAW file resolution is, the F5 won't even be able to output the RAW file in-camera either way. In this case, Sony is emphasizing 4K XAVC, hence that is what you are getting out of the camera, its effectively what resolution you will be working with. If the RAW file is 4096x2160 then we should assume 4096x2160 is also the active pixel area. As Sony has been fairly tight-lipped about it we shouldn't jump to conclusions. And again, oversampling is not uncommon. The F3 is an example.

Admit One said:
Clearly we can't talk about what it will be, but only the limit of what it could be given the sensors fixed photosite count.

One of the drawbacks to using a Bayer pattern is that the green resolution is not the same red and blue. As a result, the choice of OLPF strength (resolution vs. aliasing) becomes crucial, and Sony's choice for the F5 may not be the same as for the F55.

Although few seem to take the time to be accurate, any resolution measurement for a Bayer pattern sensor should be done separately for each channel.

Cheers.
Click to expand...

The whole 4K->3.2K seems to orginate from the original R1s, which I think we're projecting on this camera. You're absolutely right, OLPF would have a significant impact, and it may be different between F5 & F55. They are different sensors afterall.

In this thread we seem to be jumping to a lot of conclusions .
 
Last edited:
Steve Castle said:
In this thread we seem to be jumping to a lot of conclusions .
Click to expand...
Quite true. But it can also be quite confusing.

John Galt at Panvision has been harping about this for years. Here's a recent interview: http://magazine.creativecow.net/article/the-truth-about-2k-4k-the-future-of-pixels

One big question for me will be how the Sony's 16bit linear version of RAW compares with log versions.

Both Alexa and Blackmagic are using 12bit log, and the C500 is using 10bit log.

Cheers.
 
nyvz said:
No digital camera manufacturer has ever quoted measured resolutions in their spec sheets or advertising that I know of. They just talk about the number of photosites on their sensors, since that is far more easily quantifiable.
Click to expand...
Canon specs the C300 to resolve "over 1000 horizontal TV lines": http://www.learn.usa.canon.com/resources/misc/cinemaEOS_faq.shtml

I remember Sony use to spec the resolved TV lines on camera heads and Betacams in all their brochures.

To me it seems unlikely that Sony would make a sensor with almost 25% more pixels than its using. They must be doing something we don't understand.
 
OK. Confirmed 4096x2160 Bayer for both cameras as suspected.

I will buy myself a beer for correctly guessing earlier. But it wasn't that tough a guess really.

NOW what I want to know is precisely how the 2K raw and xavc high speed modes work - and if it has similar resolution / color issues to FS700 high speed HD mode.

Bruce Allen
www.boacinema.com
 
Steve Castle said:
We don't know this, and we shouldn't immediately assume that 4096x2160 is the active area, we just know that it is the resulting resolution. We actually don't know what the RAW file resolution is, the F5 won't even be able to output the RAW file in-camera either way. In this case, Sony is emphasizing 4K XAVC, hence that is what you are getting out of the camera, its effectively what resolution you will be working with. If the RAW file is 4096x2160 then we should assume 4096x2160 is also the active pixel area. As Sony has been fairly tight-lipped about it we shouldn't jump to conclusions. And again, oversampling is not uncommon. The F3 is an example.
Click to expand...

4k is 4096x2160, F55/F5 with the R5 record 4k RAW.

Sony also states:
"Both cameras have a Super 35mm 4,096 x 2,160 single CMOS
sensor (11.6M total photosites, 8.9M effective)"

8.9M effective photosites is the active area, and that is recorded as 4k RAW. I agree that there is a 0.01% chance I am misinterpreting Sony despite them using the same terms they and other manufacturers use to describe many other cameras specs.

I agree we should not jump to conclusions, but this is not a jump, this is quite clearly worded by Sony.

If that does not convince you, then I suppose we can agree to disagree, but please try not to confuse people by saying any "effective pixels" spec is not an active photosite count or that it is possible to oversample or scale RAW bayer images.
 
Piolet said:
Canon specs the C300 to resolve "over 1000 horizontal TV lines": http://www.learn.usa.canon.com/resources/misc/cinemaEOS_faq.shtml

I remember Sony use to spec the resolved TV lines on camera heads and Betacams in all their brochures.

To me it seems unlikely that Sony would make a sensor with almost 25% more pixels than its using. They must be doing something we don't understand.
Click to expand...

Ah yes, thats true. Forgot about that one, haven't dealt with cameras that list TVL resolution specs in a while. But I suppose my point as it is relevant in this situtation is that there isn't precedent that I know of for camera manufacturers list measured resolution specs in pixels or megapixels.
 
nyvz said:
4k is 4096x2160, F55/F5 with the R5 record 4k RAW.

Sony also states:
"Both cameras have a Super 35mm 4,096 x 2,160 single CMOS
sensor (11.6M total photosites, 8.9M effective)"

8.9M effective photosites is the active area, and that is recorded as 4k RAW. I agree that there is a 0.01% chance I am misinterpreting Sony despite them using the same terms they and other manufacturers use to describe many other cameras specs.

I agree we should not jump to conclusions, but this is not a jump, this is quite clearly worded by Sony.

If that does not convince you, then I suppose we can agree to disagree, but please try not to confuse people by saying any "effective pixels" spec is not an active photosite count or that it is possible to oversample or scale RAW bayer images.
Click to expand...

Again, we're getting into semantics and jumping to conclusions; fabricating facts from information that really isn't there.

And again, effective resolution is not active pixel resolution, which continually insist that it is regardless of my best efforts to convince you otherwise. The F5/F55 outputs an effective resolution 4,096 x 2,160, that is all that its saying.

Now, I'm not saying its 100% oversampling, or its 100% just taking the 4,096 x 2,160 portion of the sensor (which btw would also increase the crop-factor). I'm saying that there is not enough information to decide from the information released so far.


But let's agree to disagree. There is no point in continuing this discussion.
 
nyvz said:
Did you read the 96page PDF? It clearly says the sensor is 4096x2160, that must clear it up for you, no?
Click to expand...

We always knew the sensor outputted 4096x2160, this was in the original press release. We didn't know whether this camera was oversampling from a 11.6MP sensor or just taking 4096x2160 crop from the sensor.

Looking at the PDF, it specifically mentions debayering and demosaicing of the F55RAW, which would suggest its a classic bayer as Mitch Gross has said and not Q67 double-bayer. Secondly, Sony also calls their F65RAW 16-bit 4K, so that in itself isn't indicative of the sensor output. But what we do see is that the RAW file itself is 4096x2160 or 2048x1080. So 'output' isn't as important as what he actual RAW data is, as this is what the sensor is outputting and not the camera. So yes, I would agree that F55 seems to merely be taking 4096x2160 from the center of the sensor, and this camera isn't going to resolve a true 4K.
 
Steve Castle said:
So yes, I would agree that F55 seems to merely be taking 4096x2160 from the center of the sensor, and this camera isn't going to resolve a true 4K.
Click to expand...

Hallelujah.


Admit One said:
One big question for me will be how the Sony's 16bit linear version of RAW compares with log versions.

Both Alexa and Blackmagic are using 12bit log, and the C500 is using 10bit log.
Click to expand...

LOG is a gamma curve which increases the "captured" dynamic range by compressing the highlights. It's characteristic is that it exposes much more similar to film when it comes to highlights which tend to have smoother highlights. LOG can be over-exposed slightly, but it is usually best to underexpose is because the highend is so heavily compressed and has much less data available in it for grading and pulling detail back out. The downside of LOG is you really want to shoot it in the highest bit-depth possible because each stop is not getting a uniform amount of data for it ie. remember it is a curve that flattens out near the top in the highlights which as mentioned are "compressed" and close together.

With linear RAW however each stop of exposure would get the same amount of data giving you a lot of flexibility in post for grading any aspect of the full tonal range. ie. With 16-bit-depth you get approx. 65,000 steps ÷ 14 stops = 4642 steps per stop of data BUT EQUAL THROUGHOUT THE ENTIRE RANGE of your image.

To me, in my opinion RAW is way more flexible to shoot in terms of exposure whereas with SLOG I you need to nail it within a stop and use extremely high bit-rates to get the most out of it. Even 10-bit SLOG which the F3 lets you do is not enough in my opinion as it only yields 73 steps per stop on average (again not equal per stop since its a curve).
 
You must log in or register to reply here.
Back
Top