Welcome to our community

Be a part of something great, join today!

Log in
  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

Red One Measured Resolution

D

Dave Blackham

Well-known member
Joined
Oct 14, 2007
Messages
2,543
Reaction score
0
Points
0
Are there any figures for the measured resolution figures for Lines per image height and horizontal resolution for a Red One Camera. Assuming a properly collimated and very sharp lens is used so the optics exceed the sensor performance.

Thanks,

Dave Blackham
UK
 
On 4k (4096*2304) recording I can see about 3.2k measured resolution, or 1800 l/ph. Someone on CML reported similarly.

Graeme
 
I am confused why you don't resolve 2304 (vertical axis) lines at 4K? If you divide the horizontal number of photosites (4096), by the horizontal size of the sensor (24.4mm), you get 167.9 pixels (or lines) per mm. Would this be the maximum resolution for this sensor? Am I doing this right? I wonder how this compares to film and how it could help specify lens requirements, if at all?

Just my inquisitive thoughts..
 
Vertical l/ph is bit confusing for digital Bayer sensors.
It comes from TV cameras world.
4K format is 4096 pixels horizontal.
Vertical pixels are different for different formats.
2:1 format will have 2048 vertical pixels and 2.39:1 will have 1714 vertical pixels.

RED CMOS sensor has 5.5 micron pixel size, so using simple mathematics you should have, to be on safe side, lenses that deliver 200lp/pmm at 50% MTF
 
It was measured with linear sinusoidal 4k zone plates.

The reason you don't get the measured resolution as the same as the sensor resolution is that if you did, you'd get massive aliasing and a nasty image. How does it compare to 35mm film? Easy - superior.

Graeme
 
Joofa, Kell factor is a display phenomena, where the perceived display resolution is less, some factor usually around 0.7 of the actual resolution. It is effected by the type of display, like the CRT's gaussian spot distribution, or the pixels in an LCD display.

Graeme
 
Joofa said:
Was that for a perfect registration of line pairs with photosites -- I mean to say was Kell factor taken into account?
Click to expand...

Are you talking about resolution extraction with perfect registration without a lens or OLPF?

These tests were with a lens and OLPF as is standard on the camera with no attempt made at aligning the chart.
 
Graeme Nattress said:
It was measured with linear sinusoidal 4k zone plates.

The reason you don't get the measured resolution as the same as the sensor resolution is that if you did, you'd get massive aliasing and a nasty image. How does it compare to 35mm film? Easy - superior.

Graeme
Click to expand...

So what would be the maximum potential resolution of something recorded 4:4:4 HDCAM-SR? Would it matter what type of sensor was being used? For example, the 12MP RGB striped sensor of the Genesis versus three 2MP HD sensors in the F23? Does "oversampling" in regards to using bigger sensors help improve resolution even when recording to 1920 x 1080 RGB? Or do you always hit the same limit? Or is the advantage in that case, larger sensors with more photosites, more in terms of reducing aliasing? Or can you use a less strong OLPF with a bigger sensor with more photosites? Or do you need a stronger one?
 
It's like this....

You need an OLPF to avoid aliasing, as aliasing is an unwanted artifact.
OLPF's are slow filters, in that to get serious attenuation at high frequencies, you've got to start them early.
Electronic filters can be as fast or slow as you want - but you trade factors of resolution, aliasing, ringing and edge sharpness.

So, if you have a camera, and tell me it's sensor is 1920x1080, and you can measure good MTF out at 1920x1080 resolution, it must, by definition, have significant aliasing out at that resolution. That's not good.

So, what's the "solution".

Best is to use good optical filters and have more that enough resolution, so even though you loose some through the filtering, you've still got more than enough. That's the RED "solution". You can always downsample, and choose that downsample filter to go to a lower resolution with as much or as little detail in that resolution as you want, trading the factors I mentioned above.

If you start with a lower resolution, you're going to want to keep as much of that as possible, and not be as aggressive on the OLPF, and let in some aliasing. But you don't have enough resolution to do much downsampling.

The smaller the pixels, the weaker the OLPF you need as the MTF of the lens will be much lower at that higher lp/mm. If you can get the resolution high enough so that there's little or no MTF at that lp/mm, you don't need an OLPF. We're not close enough to that yet....

I'm not, personally, keen on an RGB stripe pattern. If you're going to a single sensor approach, the bayer pattern has proven itself time and again. If you're resigned to downsampling to 1920x1080 anyway, I'd do a bayer pattern, demosaic to the 12mp resolution, then downsample with user definable downsample filter to the 1920x1080. But that requires serious horse power to do properly. But that would give you the best result from the conditions of a single 12mp sensor with camera output fixed at HD rez.

Graeme
 
The samples also need to be individually addressable by the cameras processing to take advantage of the increased sampling. If they're binned on the sensor then you end up with a single sample from effectively two photosites.
 
Deanan said:
Are you talking about resolution extraction with perfect registration without a lens or OLPF?

These tests were with a lens and OLPF as is standard on the camera with no attempt made at aligning the chart.
Click to expand...

Oh okay, thanks for the information.

Did you try gray scale charts, or red and blue colored charts were also tried? If it is not considered internal or private information, can you kindly inform what lens was used.
 
Graeme Nattress said:
Joofa, Kell factor is a display phenomena, where the perceived display resolution is less, some factor usually around 0.7 of the actual resolution. It is effected by the type of display, like the CRT's gaussian spot distribution, or the pixels in an LCD display.

Graeme
Click to expand...

I though Kell factor is present in both capturing and display device. I have not come across recent papers that talk about measuring Kell factor for digital sensors, (though a factor of 0.9 is reported sometimes). However, for analog TV capturing, several good papers such as Mertz and Gray and other authors exist. A really good one is below where an analytical treatment is given on the capturing side:

H. A Wheeler and Arthur V. Loughren, "The Fine Structure of Television Images", Proc. IRE, Vol. 26, No. 5, pp. 540-575, 1938.

Yes, from 1938! But a really good paper and worth reading in all respects.
 
Until the RED primes are finished, we use Cooke primes for our testing.

Jim
 
The problem the Kell Factor is very poorly defined, so a lot of different people have different ideas of what it is..... Doing some further reading, it looks like Kell was trying to determine how much resolution was lost from a theoretical (sampling theory based) camera + display system. The numbers he got changed over time and with methodology of testing too. I think my confusion comes from that the spot size and distribution alters the factor, and I forgot that this refers to both the CRT display, and the scanning tube based camera :-)

Interesting though, Kell's experiments were on progressive video! And interlace adds further confusion due to the line pair summation used to avoid interlace twitter.

If we're going to get into talking just about sensors, I don't think the Kell factor is an appropriate name any more. But the fact remains that no matter what method you use to capture resolution, you either can't capture it all, or you capture aliasing. I'd like a kind of scoring system for resolution where measured resolution is rated +ve, and aliases are -ve * 2 or something, as they are a sin of commission rather than a sin of omission.

So, I do think it's good to start with a high number of samples, to allow you to use proper filtering, and go for a smooth, continuous, nice image reconstruction, and not worry too much about absolute resolution, because you started with enough. It's when you start with what you want to try to end up with, that you get "into trouble".

Graeme
 
Graeme Nattress said:
...If we're going to get into talking just about sensors, I don't think the Kell factor is an appropriate name any more. But the fact remains that no matter what method you use to capture resolution, you either can't capture it all, or you capture aliasing. I'd like a kind of scoring system for resolution where measured resolution is rated +ve, and aliases are -ve * 2 or something, as they are a sin of commission rather than a sin of omission.

So, I do think it's good to start with a high number of samples, to allow you to use proper filtering, and go for a smooth, continuous, nice image reconstruction, and not worry too much about absolute resolution, because you started with enough. It's when you start with what you want to try to end up with, that you get "into trouble".

Graeme
Click to expand...


There is NO doubt in my mind that RED image is unique in that it pushes more than enough resolution without the usual "electronic" look.
...and more important, it looks "naturally defined" i'd say.

Now, i'm completely lost when i try to understand how it preserves this silky
look even when downsampled to 1920X1080. It's beautifull really. Beautifull in 4K, and beautifull in HD.

I stay away from adding too much sharpening, it looks too "square".
I agree 100% with Michael Lindsay.
 
Oversampling, in general, is good. But it all comes to choice of the downsampling filter. But I do like 4k at 4k, as in no up nor downsampling, too.

Graeme
 
RED has the best damn edges of any camera I've seen! That means not crazy sharpened ringing and aliasing but very nice transitions across edges.
 
You must log in or register to reply here.
Back
Top