Sylvain has forwarded this article to me:[size="2"][url="http://www.lenstip.com/index.php?art=127"]http://www.lenstip.c...dex.php?art=127[/url][/size]
I reckon there's a need to comment on this:
If you folks removed the AA filter the results would be completely different. So far for "true measuring". Have you folks noticed that AA filters have different strengths ? Are you aware that this causes a variable amount of de-sharpening depending on the specific camera ?
A lens test which does not try to reverse (to be reprecise "recover") the effect of an AA filter has little meaning. Optical systems are linear. The USM sharpening procedure is linear as well. We try to "normalize" the sharpening by reaching max. LW/PHs near the sensor limit (unless it exceeds default USM settings). Whether this is too much or too little may be debatable but applying none at all is invalid. Regarding the recent discussions about the K5 a fellow (PhD in Physics) has actually proofed mathematically that a sharpening of 100%, 0.5px is actually lossless on the K5 in terms of effective resolution or in other words - only if you apply some sharpening you will approach the effective limit. If you measure K5 data without sharpening at all you actually wouldn't measure the amount of "existing details" but something much lower that that.
Just to mention Imatest again: "[size="2"]A camera with little sharpening will have an MTF50 that doesn't indicate its potential."[/size]
You mention that Nyquist is the limit by citing "Sensor response above the Nyquist frequency is garbage. It can cause aliasing, visible as Moire patterns of low spatial frequency.". Yes, this is true. This is about the -visible- response on a real sensor. So ? We do not oppose this. We just cite that Imatest can MEASURE a bit beyond Nyquist: "this allows analysis of spacial frequencies beyond the normal Nyquist frequency". You may note the difference between "sensor response" vs "analysis" here ?
[size="4"] [/size]
I reckon there's a need to comment on this:
If you folks removed the AA filter the results would be completely different. So far for "true measuring". Have you folks noticed that AA filters have different strengths ? Are you aware that this causes a variable amount of de-sharpening depending on the specific camera ?
A lens test which does not try to reverse (to be reprecise "recover") the effect of an AA filter has little meaning. Optical systems are linear. The USM sharpening procedure is linear as well. We try to "normalize" the sharpening by reaching max. LW/PHs near the sensor limit (unless it exceeds default USM settings). Whether this is too much or too little may be debatable but applying none at all is invalid. Regarding the recent discussions about the K5 a fellow (PhD in Physics) has actually proofed mathematically that a sharpening of 100%, 0.5px is actually lossless on the K5 in terms of effective resolution or in other words - only if you apply some sharpening you will approach the effective limit. If you measure K5 data without sharpening at all you actually wouldn't measure the amount of "existing details" but something much lower that that.
Just to mention Imatest again: "[size="2"]A camera with little sharpening will have an MTF50 that doesn't indicate its potential."[/size]
You mention that Nyquist is the limit by citing "Sensor response above the Nyquist frequency is garbage. It can cause aliasing, visible as Moire patterns of low spatial frequency.". Yes, this is true. This is about the -visible- response on a real sensor. So ? We do not oppose this. We just cite that Imatest can MEASURE a bit beyond Nyquist: "this allows analysis of spacial frequencies beyond the normal Nyquist frequency". You may note the difference between "sensor response" vs "analysis" here ?
[size="4"] [/size]