That thread is now in this forum:
No, TDCI is simpler but computationally scarier. Ideally, it's a processor per pixel, each independently and asynchronously timing how long it takes to accumulate the charge threshold sufficient to get an accurate reading (large enough to overcome photon shot noise) with a timing accuracy of about 1us. My goal is a 500MP large-format *4x5) sensor generating at least 16Ev HDR streams at over 1000FPS equivalent continuous rate (1TB/s if done as conventional raw, but it will be lossless compressed in the time domain)....
To be blunt/direct - I do actually think that some, if not quite a bit, of your argument is somewhat "off base" ...
See Human Target's response above. It's spot-on.
You can leave it there. The investment in variable gain amplifiers is tiny. In most cases it's simply a minor feature of the output amplifiers on the sensor chip. In earlier times, it was a standard feature of the off-the-shelf analog front ends that they bought. The true reason that they use variable gain is that it allows them to spend less money on the A to D converters.
The reason REI was adopted is that the purpose of multi pattern metering is precisely to produce a different result than you'd get from simple metering calibrated for 18% grey, and it's impossible to qualify what the different effect should be, so in the end ISO gave up and said 'do whatever you think is best'. The outcome of it is, that 'ISO' really means nothing at all, except an expression of what a manufacturer thinks should be the output brightness for a given exposure.
These excerpts are from the Wikipedia page for Film Speed (and apparently extracted verbatim from ISO 12232:2006):
My goal is to have something layman-accessible--it seems that those of us technically well-versed in the issues are pretty much already know what's going on.
I do appreciate your feedback; it gives me some insight on where I need to refine the paper.
Erhm, I've already mentioned that Nikon has recently introduced a highlights-weighted meter but there doesn't seem to be much interest in it, not even to check out whether it does act as auto-ETTR or not
No it's not at all.
Perhaps 'millions' is something of an exaggeration, at least regarding the singular specific feature we are referring to.
No, that is wrong.
That's pretty much what I just said - but, remember that 'pattern/evaluative' metering is not the only metering method a camera has.
I think that is a completely wrong deduction.
...is in my view, an assertion that is really without foundation - at the very least misleading/mistaken, and in fact very likely to be plain and simply wrong.
The difference between 18% and 12% (or anything between) is rather 'academic'.
With respect - I really don't think you are succeeding.
Judging by some of the responses/reactions I've received in this thread, as I see it, that really does not appear to be the case at all.
Another way of stating this is that the Signal-to-Noise Ratio (SNR) degrades by 41% with each increasing stop of ISO.
No - 'dosdan' is/was correct - but you however, are confused/not correct...
I hope you meant 'gollywop' here, but really, you don't have to put my name in quotes. My mother never did.
Yes, I should/meant to say... "...'gollywop' is/was correct - 'dosdan' is confused/incorrect..."
At the risk of sounding insensitive/politically incorrect - did your mother actually name or call you gollywop?!
I can only side with some of the viewer responses there...Somehow I've been finding this video to be increasingly relevant these days.
He posted the wrong answer/calculation.
The TDCI (Time Domain Continuous Imaging) research I've been doing aims to produce sensors that essentially have that property -- but they don't do it that way. Instead, they do it by allowing each pixel to be sampled at its own rate. See the slides for my SPIE Electronic Imaging 2014 talk: Frameless, time domain continuous image capture.
i wouldn't expect you to agree, but it was precisely right.
The 'much wider' sense is irrelevant to the question of the design costs of variable gain amplifiers. Their use is a design choice which comes purely to avoid using better specified, quieter ADCs
it is right.
not just higher bit depth, but ADC's with a higher effective number of bits. At the bandwidth modern cameras require, getting an ADC with 14 or more ENOB is possible, but expensive.
The ADC is the main contributor to base ISO read noise in many cameras.
That all depends on what A and B are, doesn't it? Maybe you need to think about that a little longer.
How would ISO change according to which metering mode you used?
Would you dare to question her?
Funny is not the issue; relevance is. I never found it very funny either, even when I first saw it many, many years ago. But somehow, at times like this, it keeps coming back to me as pretty much sizing up what's needed. Perhaps its lack of funniness is part of the reason.I can only side with some of the viewer responses there...Somehow I've been finding this video to be increasingly relevant these days.
> "I suppose that somewhere, this is considered funny"
>> "No, it's really not"
...these responses are a lot wittier/funnier than the video (which just isn't) - as is usually the case for the 99.999% of garbage and dross that makes up 'youtube'.
You're absolutely right.
You're absolutely right.
You're absolutely right.
Yes, and that was very good of you – a bit overly pedantic, but very good of you.
Yep. I'm willing to bet that's a good bet.
