[photogerald]

It's been really great reading all the K-5 user reviews (though I'm sure I missed a few), I'm so envious of you folks!

I had a quick question - and my apologies if this has already been covered - but how is the live view display on the K-5 when zoomed in? As I understand it the K-7 was a great improvement over the K20D in this regard, but I remember hearing that the K-5 might've taken a step back here (more stair-stepping when zoomed-in), at least on the pre-production model that was tried? Can anyone set the story straight here?

The reason I ask is because I have some nice MF glass and I would like to use LV to assist with focusing these lenses (the K20D LV is too coarse, though it's better than nothing).

[Col]

No I not observed pixelation. Oh and in MF the K-5 has 10x magnification, rather than K20d's 8x.

[photogerald]

No I not observed pixelation. Oh and in MF the K-5 has 10x magnification, rather than K20d's 8x.

Thanks Col, this is great news!

[Col]

Thanks Col, this is great news!

No problem, here's a couple of quick shots of the LV.

As you can see OOF areas are blurred rather than chess boards.

... oh and also its not bright in my front room. 0.5sec at 100ISO f3.2!

[falconeye]

Col,

please focus on a line or border and turn the camera so the line is 45 degrees.
As I wrote, at least the Photokina K-5 showed a clear (and sharp) staircase then. Sharp staircase is better than soft upscale. This is why I would like to see an area in focus and in the image center.

[Col]

Col,

please focus on a line or border and turn the camera so the line is 45 degrees.
As I wrote, at least the Photokina K-5 showed a clear (and sharp) staircase then. Sharp staircase is better than soft upscale. This is why I would like to see an area in focus and in the image center.

I'll try, are these any better than you saw at the show? Got about as many diagonals as I could think off ...

Crop of same



What it looked like to the K20d in real life

[falconeye]

I wouldn't say better, just different (if your's are in focus actually).
I would say that the 10x LV shows less detail than the image when shot. Personally, I prefer to see the jaggies then. YMMV

Update:
In the last sample, the E looks considerably sharper. Maybe, an a bit more controlled focus is required. And the corresponding 100% crop of the same area, but of the shot when taken. Maybe on the disply as well, at 10x, so everything else is known to have stayed the same.

---

Last edited by falconeye; 10-26-2010 at 03:58 AM.

[Col]

Because of the K20d rarely used LV for focusing. Have to get used to the display lag when manual focusing. Also I guess LV suffered from the lack of DOF of the Tamron 90 being wide open at f2.8. (I was using the tripod for the K20d so the K-5 could slip on the table when I removed my hand.

[clark]

so as far as i can gather- the camera doesn't automatically zoom in when manual focussing right?

looks nicer than the k7 though, i'll be interested to try one

[Col]

No the zoom in shown in contrast AF.

[falconeye]

so as far as i can gather- the camera doesn't automatically zoom in when manual focussing right?

No, simply because the camera cannot know when you turn the focus wheel.

Except for some clever image processing where a change of contrast not due to motion would indicate that the focus distance is changing ... It could even display or visualize some numeric contrast value to aid manual focussing. Like two triangles approaching to a center position when the contrast increases, where the triangles touch an inner solid segment as soon as the contrast reaches the last maximum value before it was seeing decrease again (the triangles can continue to approach within the solid segment).

[photogerald]

Thanks Col for the detailed snaps - I really appreciate you taking the time to take and post these!

It looks like I have nothing to be worried about - the LV resolution of the K-5 looks just fine!

One other question - does the K-5 allow you to zoom beyond 10x? If so, does this show any added benefit, or do things start to become pixellated?

I'm still curious to see a direct comparison of the LV between K-5 and K-7.

By the way, these images courtesy of Gerry Winterbourne from the other forum nicely illustrate the improvement in LV between the K20D and K-7:

K20D:


K-7:

[photogerald]

Col,

please focus on a line or border and turn the camera so the line is 45 degrees.
As I wrote, at least the Photokina K-5 showed a clear (and sharp) staircase then. Sharp staircase is better than soft upscale. This is why I would like to see an area in focus and in the image center.

Hi Falk, nice of you to stop by! I remember now that it was your Photokina report that triggered this question.

[falconeye]

By the way, these images courtesy of Gerry Winterbourne from the other forum nicely illustrate the improvement in LV between the K20D and K-7:

Gerald, impressive comparison.

If you're really curious to know what's going on in LV, I recommend doing one of my standard tests.

1. Print my zone plate test chart (A4 B&W is ok), downloadable here:
   
   (make sure to download in original size -- please don't delete the (c) mark)
2. Zoom onto the large plate hanging at some distance, getting you an image like attachment #2 (K-7 in 10x)
3. Actually take a photo and crop to 100%, getting you an image like attachment #1 (K-7 100% crop)

Discussion:

1. 100% crop:

The printed zone plate has all rings printed correctly, with only minimal printing Moiré pattern. All Moiré you see in the 100% crop is from the demosaicing process. BTW, this would look more ugly for a camera w/o AA filter. The K-7 only produces neglegible Color Moiré near the Nyquist frequency.

At about half the Nyquist frequency, you see 4 pseudo rings (i.e., they are not on the printed chart!) which you can use to determine the exact pitch of the Bayer pattern (2x the pixel pitch). As you can see, the demosaicing manages to resolve beyond the Bayer pattern pitch almost up to the Nyquist freqency (LR2).


2. 10X Live View zoom:

The LV image shows 2 pseudo rings. It turns out they are at the same location as in the 100% crop, i.e., at 50% Nyquist frequency. One sees too that in the vertical direction, the resolving limit is at about half that which is 25% Nyquist frequency, in horizontal direction, it is better and it resolves up to 25% Nyquist frequency w/o the pseudo ring.

This means that in vertical direction, 10X LV uses line skipping (only every 2nd Bayer cell is read out -- pseudo rings are a sampling artifact) and in horizontal direction, every Bayer cell is read out and converted into a single rgb value. Proper demosaicing is skipped because some neighbors are missing.

Overall, this means that the effective K-7 LV resolution in 10X is 2336x776 or 1.8MP.

Some last note: The K-7 LV resolution improves at the 8x and 10x scale. Up to 6x, it uses line skipping and row skipping.


A similiar analysis for the K-5 would be interesting indeed.

[clark]

No, simply because the camera cannot know when you turn the focus wheel.

thanks
fwiw (i'm sure you know this but...)the gh1 senses when you're manually focussing and automatically zooms in, i'm 80% sure it still does this even when you're mounting an adapted lens