Camera Labs

Hello
I am new this forum and hope that the following subjects haven't been covered.
I have 2 subjects I would like to discuss.
1) Resolution in 'lines per picture height' (Lpph).
a) I am puzled by the numbers quoted in the reviews. As an example, in the review for the Canon 400D ( http://www.cameralabs.com/reviews/Canon ... ge4b.shtml ), the first chart shows the 400D with 2000lpph, 18-55mm EF-S at 35mm, f8, 100ISO.
The 400D has a sensor size of 22.2 x 14.8 mm, and a maximum picture size of 3888 x 2592 pixels. However, the 2000lpph over 14.8mm picture height implies a vertical resolution of at least 2000X2 pixels = 4000pixels. Since the sensor only has about 2393 pixels in it's height, how is this possible?
b) How is lpph converted to equivilant dpi. I have seen references to this as 2Xlpph - a line pair being one blacK and one white line. I have also seen references where they say that 1 pair is 2 sets of lines requiring 4 pixels each. Could you please explain the how to convert an lpph reading into an equiv dpi?

2) I think that your reviews with the test results is the best there is. Comparison of like cameras is great. even for the purposes of upgrading from all in one large cameras (such as the Fijifil S9500) to DSLRs. But I have a problem.
I have used a 35mm Canon AV1 with a Sigma 39-700mm zoom for years. My wife won a Ricoh Caplio G4 Wide about a year ago and stated using it. I is a versitile camera. 28mm wide angle, 3X zoom and small size. But, it has a couple of major problem. In auto mode without flash it is quite ready to jump to ISO800. With flash, which is only good to around 2.5 meters, after that it compensates again by increasing sensitivity up to ISO800. The trouble with this is that the pictures are so noisy that they are vertualy unusable. I certainly wouldn't want to say to anyone "look at this, I took this photo". I have read some good reports also about the Canon Ixus 850IS on other websites.
Now to my question. I would like to upgrade to a compact like the Canon 850IS, Fijifilm S9500, Sony Alpha A100 or Canon 400D. I am able to compare all of them except the Canon 850IS. In your reviews you mention upgrading, and at times you include the previous models like the Canon 350D in the test results but there never is a high end compact included in the test results. Maybe I am the only one who would like to upgrade from film or low end compact, but I don't think so. Would it be possible to include in your reviews, in the future, a high end compact in the test results such as the resolution and noise performance sections? I am not looking for full reviews for these compacts, I understand that you review high end cameras only.

Thanks a lot
Frank1

Hi Frank, welcome to the Cameralabs forums!

You make a very good point concerning resolution and limits imposed by Nyquist theorum. What you must remember with digital cameras though is the final image is the result of more than just the sensor resolution. The camera has to process the data, and interpolation can give the impression of higher resolutions or resolvable detail. It's often a hard one to call, so in addition to quoting the resolution measured on the chart samples, we also compare the detail on real life, outdoor samples.

As for a DPI conversion, it depends on what exactly it is you wish to measure. If it's dots on the sensor surface, then you can derive the spatial resolution from the physical dimensions and the number of photosites, but remember on all but Foveon sensors, these photosites record tones only, not colour. The vast majority of digital camera sensors employ a bayer pattern where groups of four photosites with coloured filters above them are used to generate a full colour image - generally with the same pixel numbers as the sensor's photosites. So the colour resolution is actually lower than the spatial resolution.

Obviously if you're just after a DPI figure at a given reproduction though, just divide the number of pixels on the image in one axis by the physical dimension in the same axis.

As for comparing compacts against DSLRs, we use the same tests, so you could just open two windows and put them alongside each other. You'll notice the compacts generally have much worse noise levels at high sensitivities, although normally far better macro facilities than a DSLR kit lens.

We're also preparing a feature comparing a high-end all-in-one against a budget DSLR which you might find interesting, and you'll no doubt have also seen our latest Canon A640 and Sony N2 compact reviews.

Hope that helps, and thanks for your kind words about our reviews!

Gordon

Hello Gordon
Thank you so much for your response.
I look forward to your reviews. I would especially like to see you review the Canon Ixus 850IS or include it in the Results section of some other review.
Now to the resolution. I followed what you were saying so looked into it some more. I found a website (http://www.sinepatterns.com/i_Stdrds.htm) selling the ISO 12233 Resolution Test chart (which looks like the one you are using). The picture of it can be clicked for a larger view and on the top margin it is stating "Values in 100X lines per picture height". My assumptions were wrong in my previous post saying that it may be Lines Pairs per Picture Height.
This now makes a big difference. The 400D having 2592 pixels in it's height resolves around 2100 lines. Now the question was, why only 2100 and not closer to 2592. This is where I looked at the Nyquist Theorem. Unfortunately, I couldn't find anything meaningful but did find something else along the same line. The Kell Factor (http://members.aol.com/ajaynejr/nyquist.htm) is used to prevent "pixel straddle" This is where lines fall between sensors. This is why the number of lines resolved cannot be as high as the number of elements. This Kell Factor is assumed to be around 0.7, but is debatable. Now, 2100/2592=0.81 which is close to the Kell Factor. The clever electronics interpolation and good lens may make the difference between the 0.7 and the 0.8. this does make sense to me.
Now, to display this resolution (printed on a 13X10" paper) how many DPI would I need. 2100/10 = 210 lines per inch. But these number of lines cannot be printed with 210 DPI because of the Kell Factor (0.7). So deviding the 210 by 0.7 we get 300 DPI required to resolve our 2100 lines per picture height on 13X10" of paper. I hope that this is correct.
Thanks for putting me in the right direction.
Frank1

Hi Frank, I'm very familiar with Sine Patterns, as they produce the charts we use, and are sold via Applied Image - you can read more in the About Us pages.

Just a quick note concerning resolution again - the lens obviously plays a massive role in resolving power, but don't forget the effect of digital processing, especially the process of turning filtered photosites into full colour pixels!

Gordon