Quote "If your observation that the menu override worked on the old build is correct, then I'm pretty certain my build does not "fix" scripted overrides. They both use the same mechanism under the hood, so either there is some other variable at play, or one or more of the test results is in error."
This is very interesting. For your test, on previous builds your on-camera procedure worked exactly as you predicted. My own test results (the attachment with annotated -svs) is equally robust .. it worked immediately, and to be certain I tried it on both cameras. So I can confirm both results are reliable.
My quote
"What did you do? Please explain in as much detail as your pen has ink."
Quote "Code: [Select]
...
(by analogy to g12)"
Clear as mud .. words, not (ARM?) assemby code ... you know I don't have clue about it.
Quote from: reyalp on Today at 07:52:24
"Sometimes the higher ISO levels are handled differently in the Canon firmware.
Quote "To elaborate, the highest ISO setting is frequently done in software in the Canon jpeg process. You can check if this is the case by comparing raws. I would test with with CHDK DNG, since Canon might do equivalent trickery in their raw software. CHDK override will probably not be able to set ISO values that behave this way."
I am almost certain you must be referring to the faked very high ISOs that they advertise in some cameras. They do that by reducing resolution and binning pixels, in effect, summing the light flux over a larger sensing area. For example, with uniform photon flux falling on a sensor surface, if you double the sensing area electrical noise increases as a function of the ratio of sqrt(2) of the respective noise levels, whereas, signal increases by 2. The S90 does not have this "feature." So at 3200, it is definitely using a gain control signal in its CCD variable gain amplifier ... in my opinion the S90 design is at the verge of solid state physics for room temperature operation, and is part of the many electronic design aspects that make this great camera. This high ISO "feature" in other cameras has some relationship to "the more pixels the better" paradigm that has taken the general general public years to realize nothing is for free. To be certain, I did all the tests you suggest here, and indeed they confirm my earlier engineering evaluation of this hardware: your DNGs and Canon CR2s are EXIF-tagged at 3200 but exposed at 1600, behaving like JPGs. At shoot -sv values 1600 and lower, everything still works perfectly, so far. To have access to this fabulous S90 high 3200 ISO via shoot -sv= will be very useful if you can think it through and find a way to enable it. If possible, I would hugely appreciate it. On another note, it is obvious to me your first explanation was insufficient and could have resulted in misguided perceptions. Your elaboration was much better ... elaboration, thorough thought expression, and wordiness always helps, in my opinion // you can do it if you care to. I believe my careful explanations helped you find a difficult problem which led you to one line of code that solved it.
AN IMPORTANT and INTERESTING DISCOVERY that may impact the CHDK Community
Your test above was planned for later but your request yesterday accelerated it. It is known to most of us that Canon firmware corrects JPGs in many of its latest PowerShots to compensate for lens barrel distortion (straightens the curvatures in the photo). This is an area where digital photography beats film photography hands-down. It turns out that in the S90 (and likely in the S95) Canon native RAWs (CR2s) are also corrected. Hence the conclusion is that Canon S90 native RAWs are not true RAW images. This feature may help the average photographer, but carries with it the disadvantage that the original image space is re-mapped to another one, losing edge data. Worse, since pixels are moved around their values must be manipulated. Thus the original 12-bit re-mapped pixel values cannot be reliable either. In my lens-less application this is a very important consideration. On the other hand, CHDK DNGs do not manipulate image data and are true RAW in this respect. The difference is dramatic // see Fig 1 (Canon native CR2) and Fig 2 (CHDK DNG). There is more research that needs to be done // for example .. is full zoom on this lens (3.8x) enough for Canon disable the correction? Because compression advantage of Canon native RAWs for transfer and PC disk storage efficiency ... could CHDK find a way to disable the correction??
