My own personal res test.

evinsky

evinsky

Well-known member
I decided to do my own res test using the HVX in 1080 24P mode. I suspected the HVX might be resolving a little more than Adam Wilt gave it credit for, mostly because of the condition and speed of the test. I have a Century Optics res test, not quite as accurate as the Trumpets but good enough for my uses.
hvx.jpg

On this chart "E" is 576 Lph and "D" is 640 Lph "C" is 896 Lph.

From my observations the HVX easily resolves 600+ LPH and on fine (as opposed to acute) detail, like texture and skin, up to 640 where moire becomes visible. There may be even more fine detail rendition beyond 640 but I can't measure it on this chart.
If you understand optics then you realize the difference between acute resolution and fine detail resolution. You need a good balance of both. This camera probably resolves about twice as much as a DVX. A little more than half a Cinealta and about 3/4 a Varicam.

Details.
10mm, F4, Sharpness -7, 5 feet from target.
 
Good job.. We need more people doing these tests. I don't know about anybody else and I'm not out to insult Adam Wilt but I just had a nagging feeling it wasn't completely accurate. The more of these we get the better an idea we'll have of the numbers. I personaly don't care that much about the figures, I love the images the camera produces and I'm getting one ASAP but it would be nice just to know.
 
If I understand Adams testing methods, your chart supports his findings since you can begin to see moire in E.
 
Thanks for the test but I have a question/remark. To my eyes even the word "century" does not seem in focus, am I wrong????? Is it shot perpendicular to the rez chart ???? Maybe the rez is even better than the 600+ ???



evinsky said:
I decided to do my own res test using the HVX in 1080 24P mode. I suspected the HVX might be resolving a little more than Adam Wilt gave it credit for, mostly because of the condition and speed of the test. I have a Century Optics res test, not quite as accurate as the Trumpets but good enough for my uses.
hvx.jpg

On this chart "E" is 576 Lph and "D" is 640 Lph "C" is 896 Lph.

From my observations the HVX easily resolves 600+ LPH and on fine (as opposed to acute) detail, like texture and skin, up to 640 where moire becomes visible. There may be even more fine detail rendition beyond 640 but I can't measure it on this chart.
If you understand optics then you realize the difference between acute resolution and fine detail resolution. You need a good balance of both. This camera probably resolves about twice as much as a DVX. A little more than half a Cinealta and about 3/4 a Varicam.

Details.
10mm, F4, Sharpness -7, 5 feet from target.
Click to expand...
 
This is a 200% blow up from photoshop to illustrate the res, it's not going to look in focus at that magnification.
 
So there is alias and color artifacts in E which is below 600lines. I don't see anything in D besides alias. If you don't see the actual number of lines the pattern has, it's just alias and does not contain anything useful. If it was flat grey it would look better in any real world situation. It wouldn't give you problems shooting building or sharp parallel lines.

Charts with parallel lines are generally more accurate and easier to use.

From an engineering point of view, the ideal system would resolve a number of lines and start moving into blur without having alias and color artifacting. This requires very good filtering algorithms and we are not going to see that in this category of cameras.
 
limo991 said:
So there is alias and color artifacts in E which is below 600lines. I don't see anything in D besides alias. If you don't see the actual number of lines the pattern has, it's just alias and does not contain anything useful. If it was flat grey it would look better in any real world situation. It wouldn't give you problems shooting building or sharp parallel lines.

Charts with parallel lines are generally more accurate and easier to use.

From an engineering point of view, the ideal system would resolve a number of lines and start moving into blur without having alias and color artifacting. This requires very good filtering algorithms and we are not going to see that in this category of cameras.
Click to expand...

Maybe from an engineering POV, but from a filmaker's POV the detail beyond 600 can still be very usefull. Acute (contrasty Line) detail is only of use for buildings and such, fine textural detail does not suffer from the same problem of aliasing. So skin detail or other resolved textures can be clearly rendered beyond what a black on white line can.
 
Thanks for that test, evinsky... cool stuff.

I think what would interest me more, however, is to see a res test shot at factory default settings... no mods to the detail setting. So far, all the tests we've seen have been with user adjusted settings, which of course most would probably use, but may be another factor contributing to resolution loss (a little grain is ok for tests!).

-pastywhiteboy
 
Edge enhancement only gives the impression of a sharper image, it does not actually increase recorded detail. I did do this test with both -7 and 0 image detail settings, there was no visable difference in resolution. Although the "0" looked sharper before the blowup.
 
Thanks Evinsky... I guess I'm still trying to figure out why they didn't set "no edge enhancement" at the 0 mark... why does 0 have significant enhancing? Odd. Thanks for the info Do you happen to have any grabs from that 0 detail test? It'd still be interesting to take a look at.

-pastywhiteboy
 
This might now show in single frames, but the alias will behave terribly with motion. It's not detail. It is an artifact of detail tha was above nyquist and wrapped arround. It's not natural and even inexperienced viewers notice it. If the detail frequency frequency increases, the artifact frequency will decrease. It would be better if you had nothing in that area which will not look unnatural.

Lets say that you whistle while increasing frequency continuously. When your harmonics hit the resolution limit they will start appearing as alias that is wrong frequency and is going down instead of up. This is very unatural behaviour. Video has a similar phenomenon.

If there is no high resolution detail in that area how could alias appeat to "improve" the texure? Alias comes only if level beyond nyquist is present. I can't think of a creative use of alias in any type of signal. If we entirely remove antialias filters, you will have good detail up to a much lower resolution compaired to a properly antialiased signal. And in the aliased case, you can't use sharpening to improve the frequency response (contrast near the limit), the image will break and only look worse.

One possible explanation for what you are refering to is the higher visible contrast in detail near (but below) frequency limit in a signal that is not filtered for antialiasing. Is this what you mean? You still have alias on top of this area though. It is a sacrifice.

I'm working on software for an HD uncompressed camera and can't help relating everything to the design process. Sorry::engel017:
 
evinsky said:
This camera probably resolves about twice as much as a DVX. A little more than half a Cinealta and about 3/4 a Varicam.
Click to expand...

You're comparing the 720p Varicam to 1080 HVX and saying it still is only 3/4 the rez?

Yikes, what does that mean for 720p performance of the HVX? So far, I've characterized the HVX as an amazing SD camera, an excellent 720p camera and a lackluster 1080 one.

Now I'm beginning to wonder if even HVX at 720p will pack the DVCPRO codec with as much resolution as it can swallow. I do not want to record 1080 HVX and then software downsample to 720p just to get "good" 720p.

Jim Arthurs
 
There are two "hidden" advantages in 1080p. Lets asume that the system is limited to 600horizontal and vertical tv lines in 16:9. that is 600x16/9=1067 horizontal pixels and 600vertical pixels.

1) 1080p is recording 640x1080 chroma and with the resolution limit that will give you 640x600 chroma. 720p is recording 480x720 chroma and with the resolution limit that is 480x600 chroma. And, 640 is better than 480.

2) The 720p sometimes has to be scaled to 1080p or 2k or whatever you need as an output. 1080p will needs very little or no scaling. 720 will need scaling and that comes with artifacts. If you scale 50% or more you don't only get softness, you get some blockiness in lines of small angles close to vertical or horizontal. Not having to scale is a good thing even if the actual resolution is the same.

On the HDV cameras 1080 mode has even greater advantages since 720p is only recording 360 vertical chroma pixels and 1080p is recording 540.
 
Limo, your getting a little beyond me here. All I can tell you is that there is a greater ability to render texture and fine detail than the "Lines" on a chart can tell me.
I can see it in the scenes I'm shooting. The rez chart is useful but I'm over it and on to real subjects.
 
You must log in or register to reply here.
Back
Top