Re: *ist D resolution (was:Soon to be new istD owners)
Thanks, John. Very interesting. Jostein Quoting John Francis [EMAIL PROTECTED]: John, (or Rob or anyone...) This may be a stupid question, but is the relationship between lp/mm and sensors as straightforward as two adjacent sensors equal to one line pair, irrespective of sensor colour? Not quite, because the sensor colour does enter into it. But if your stimulus pattern is black-and-white lines, two sensors do equal one line pair, pretty much, if the interpolation is done based just on intensity. If, however, the stimulus is black-and-red (or black- and-blue) lines, you only get half the spatial resolution, so you need four sensors to get a distinguishable line pair. Black-and-green falls somewhere in between. In real life, of course, most patterns are lower-contrast than all black/all white, so deliver lower than theoretical maximum resolution. But they tend to be closer to black-and- white than to single-sensor colours, so the right limit to use is the two-sensor-per-line-pair number. It all depends on the input colours, the contrast ratio, and most of all on the quality of the image reconstruction algorithms (commonly referred to as Bayer interpolation). This message was sent using IMP, the Internet Messaging Program.
Re: *ist D resolution (was:Soon to be new istD owners)
William Robb wrote: With digital capture, there is no resolution loss when subject contrast drops, unlike film.. John Francis wrote: In real life, of course, most patterns are lower-contrast than all black/all white, so deliver lower than theoretical maximum resolution. But they tend to be closer to black-and- white than to single-sensor colours, so the right limit to use is the two-sensor-per-line-pair number. Um... Now you got me a tad confused here, guys... Cheers, Jostein This message was sent using IMP, the Internet Messaging Program.
Re: *ist D resolution (was:Soon to be new istD owners)
William Robb wrote: With digital capture, there is no resolution loss when subject contrast drops, unlike film.. John Francis wrote: In real life, of course, most patterns are lower-contrast than all black/all white, so deliver lower than theoretical maximum resolution. But they tend to be closer to black-and- white than to single-sensor colours, so the right limit to use is the two-sensor-per-line-pair number. Um... Now you got me a tad confused here, guys... Mr. Robb is correct. There are losses due to low contrast in the later software processing, though; the end result is somewhat reduced resolution. I was over-simplifying.
RE: *ist D resolution (was:Soon to be new istD owners)
Hi Rob Hmmm. Not so sure about how to test lens resolution on film. I'm sure 100 linepairs pr. mm has been occationally achieved on film. In those days the lenses was often setting the limmits. As opposed to todays technology where the CCD are setting the limits. Hav you ever seen digital images blown up to - let's say - 1200 x 1800mm. Like in a slide show. Or a moovie - made with 35mm film. Simple math tels me that 3000 pixel covering 24mm gives 125 pixel/mm. But you need three to make a PAIR of lines. Gives you appr. 43 line pairs/mm, right. So, using a 6MP body is like using the poorest lens ever made by Pentax - resoluton wise, of cource. Like 100 lp/mm seemed to be the sound wall of analog photography, it seems appr. 5000 dpi is the sound wall of current digital photography. But I'm sure they'll break throug this sometime soon. SONY is now marketing a 8MP consumer camera - with a 2.0 Carl Zeiss lens - for appr. 1000$ (Sony DSC F-828). Maybe we'll get there earlier than we relly want. My old faithful PZ1 has served me well for 12 years. How long do you think my Pentax *ist D will keep up - 2 years, 3 years? By the way - did you ever think about this - the digital photographic technology pretty much works like the human eye? All the best Jens -Oprindelig meddelelse- Fra: Rob Studdert [mailto:[EMAIL PROTECTED] Sendt: 11. januar 2004 23:56 Til: [EMAIL PROTECTED] Emne: *ist D resolution (was:Soon to be new istD owners) On 11 Jan 2004 at 11:26, Herb Chong wrote: i don't know where you read that. there are a few sensors in use as digital backs that have that kind of dynamic range, but not the sensors in DSLRs. I'll eventually put up a page like I did for my Oly with my findings on this matter, see: http://members.ozemail.com.au/~distudio/contrast/ as for resolution, the conditions under which you can reliably achieve 100 line pairs per mm are when shooting resolution charts under high contrast conditions with very high resolution film on a tripod at high shutter speeds with a top lens at its optimum aperture and focus bracketed. It's not actually that difficult if you have gear that it well matched. the reason digital cameras can compete is that under real life conditions, few people can possibly exceed 40 line pairs per mm worth of resolution and that a digital SLR has negible amounts of noise compared to film even at higher ISO ratings. The *ist D sensor provides a theoretical 128 pixels per mm after demosaicing however in practical terms the body can only resolve between 45 and 50 lpmm, which is very close to what I expected. This figure I derived using bracketed focus on a technical target with optics that should provide well over 100lpmm on film. Cheers, Rob Studdert HURSTVILLE AUSTRALIA Tel +61-2-9554-4110 UTC(GMT) +10 Hours [EMAIL PROTECTED] http://members.ozemail.com.au/~distudio/publications/ Pentax user since 1986, PDMLer since 1998
RE: *ist D resolution (was:Soon to be new istD owners)
On 12 Jan 2004 at 22:28, Jens Bladt wrote: Hi Rob Hmmm. Not so sure about how to test lens resolution on film. I'm sure 100 linepairs pr. mm has been occationally achieved on film. Hi Jens, Plenty of resources here: http://members.ozemail.com.au/~distudio/lenstesting/ Simple math tels me that 3000 pixel covering 24mm gives 125 pixel/mm. But you need three to make a PAIR of lines. Gives you appr. 43 line pairs/mm, right. So, using a 6MP body is like using the poorest lens ever made by Pentax - resoluton wise, of cource. Sure, the Kell factor in this case appears to be about 0.7 therefore the resolution in lpmm can effectively be calculated as 3008pixels/23.5mm/2pixels*0.7kell factor=44.8lpmm. I used a conventional test chart in conjunction with a very high resolution lens in order to reduce it's effect on the measurement and I calculated an optical resolution of 44.6lpmm and this was before I made any theoretical calculations. Like 100 lp/mm seemed to be the sound wall of analog photography, it seems appr. 5000 dpi is the sound wall of current digital photography. But I'm sure they'll break throug this sometime soon. SONY is now marketing a 8MP consumer camera - with a 2.0 Carl Zeiss lens - for appr. 1000$ (Sony DSC F-828). Maybe we'll get there earlier than we relly want. I don't think so, I doubt there will be that much to gain by making smaller pixels, look at the noise generated by the Pentax *istD at higher ISO already. Smaller pixels will reduce the effective exposure latitude, colour accuracy and increase the noise floor. By many accounts the F-828 offers more promises than performance. My old faithful PZ1 has served me well for 12 years. How long do you think my Pentax *ist D will keep up - 2 years, 3 years? You are probably right, 2-3 years, the *ist D isn't an ideal solution, nor is the Canon 1Ds, the Kodak comes closest at 14MP as it offers the same effective lpmm as the Pentax but using the full frame. Unfortunately it was before it's time technology wise however I am sure that it would be all that most photographers would ever require resolution wise. I don't think we'll see 24x36mm sensors with higher densities any time soon. By the way - did you ever think about this - the digital photographic technology pretty much works like the human eye? I think it may be a while before we are doing image capture Matrix style :-) Cheers, Rob Studdert HURSTVILLE AUSTRALIA Tel +61-2-9554-4110 UTC(GMT) +10 Hours [EMAIL PROTECTED] http://members.ozemail.com.au/~distudio/publications/ Pentax user since 1986, PDMLer since 1998
Re: *ist D resolution (was:Soon to be new istD owners)
Simple math tels me that 3000 pixel covering 24mm gives 125 pixel/mm. But you need three to make a PAIR of lines. Sigh. Here we go again. You need two pixels (one black, one white) to make a pair of lines (also one black, one white). So 3000 pixels in 24mm is 62.5 lp/mm.
Re: *ist D resolution (was:Soon to be new istD owners)
- Original Message - From: John Francis replying to Jens Bladt Simple math tels me that 3000 pixel covering 24mm gives 125 pixel/mm. But you need three to make a PAIR of lines. Sigh. Here we go again. You need two pixels (one black, one white) to make a pair of lines (also one black, one white). So 3000 pixels in 24mm is 62.5 lp/mm. Jens, I take it you think of three adjacent sensors in RGB. Before you measure lp/mm, you have to interpolate (Bayer) the data for each pixel with data from the adjacent sensors. John, (or Rob or anyone...) This may be a stupid question, but is the relationship between lp/mm and sensors as straightforward as two adjacent sensors equal to one line pair, irrespective of sensor colour? cheers, Jostein
Re: *ist D resolution (was:Soon to be new istD owners)
- Original Message - From: Jens Bladt Subject: RE: *ist D resolution (was:Soon to be new istD owners) with 35mm film. Simple math tels me that 3000 pixel covering 24mm gives 125 pixel/mm. But you need three to make a PAIR of lines. Gives you appr. 43 line pairs/mm, right. So, using a 6MP body is like using the poorest lens ever made by Pentax - resoluton wise, of cource. You show me any lens anywhere that can do significantly better than 43lppm under real world (not high res test target) situations. With digital capture, there is no resolution loss when subject contrast drops, unlike film.. The thing about film and lenses is that they increase in resolving power as contrast increases. Thats why people like to look at the 1000:1 TOC numbers. In the real world, a TOC of 1.6:1 is considered closer to normal, and 43 lppm is very good indeed. William Robb
Re: *ist D resolution (was:Soon to be new istD owners)
John, (or Rob or anyone...) This may be a stupid question, but is the relationship between lp/mm and sensors as straightforward as two adjacent sensors equal to one line pair, irrespective of sensor colour? Not quite, because the sensor colour does enter into it. But if your stimulus pattern is black-and-white lines, two sensors do equal one line pair, pretty much, if the interpolation is done based just on intensity. If, however, the stimulus is black-and-red (or black- and-blue) lines, you only get half the spatial resolution, so you need four sensors to get a distinguishable line pair. Black-and-green falls somewhere in between. In real life, of course, most patterns are lower-contrast than all black/all white, so deliver lower than theoretical maximum resolution. But they tend to be closer to black-and- white than to single-sensor colours, so the right limit to use is the two-sensor-per-line-pair number. It all depends on the input colours, the contrast ratio, and most of all on the quality of the image reconstruction algorithms (commonly referred to as Bayer interpolation).
