2 New Nikon DSLR Rumours

If what you are saying is that 14-bit gives you a 2 stop improvement in ISO performance, as would be suggested by your explanation, then absolutely not. If that were the case, all we'd need for great high ISO would be 20 bit converters!

 

The output of the sensor, after amplification, is referenced to a set voltage which is the maximum voltage the ADC responds to. This is independent of the number of bits in the converter. The "extra" bits added in the 14 bit converter, therefore, aren't at the "top", they're at the "bottom".

Perhaps I didn't word that very well. I'm not saying that the 14-bit converter will give 2 stops better ISO performance, and I agree that the extra bits are at the "bottom" of the range. Let me try again with an example....

Set the 14-bit camera at ISO 100 and the 12-bit camera at ISO 400. If the exposure is correct for ISO 400, both cameras will give the same signal-to-noise raw outputs with the same number of useful quantization levels with the top two bit of the 14-bit camera always 00. However, the shot of the 14-bit camera will not be over exposed if the "correct" exposure happened to be ISO 100. So in effect you can shoot ISO 100-400 exposure at ISO 100 setting and get at least a good 12-bit output. If the exposure level is ISO 400, the 14-bit camera will not give a better output than the other one. If you happen to over expose what you thought was an ISO 400 exposure, you're protected by a couple of stops. I think the high dynamic range mode in the 1DIII is doing something like this.

What's more interesting is that you can think of exposing the bright parts of you image at ISO 100 ( the sunball for example ) and dark parts at ISO 400, and bring the ISO 400 parts up to ISO 100 in post processing. The signal-to-noise of the dark parts will be no better than an ISO 400 exposure, but it would not be limited by quantization problems.

The analog voltage representing 2 stops below maximum will be 1/4 of maximum voltage regardless of bit depth or ISO. If a sensor has less than 12 stops of range, then both the 12 and 14 bit converters will be digitizing noise in their lowest bits.

As I mentioned before even if the sensor has less than 12 stops of range as defined by the RMS noise level, there is still information below this level. It's better to record it than to cut it off, especially in cases where you have to bring up the darker levels for dynamic range reasons.

OK, I get that. I certainly wasn't thinking in those terms before!

What you say, though, is predicated on the sensor delivering meaningful extra information beyond the capabilities of a 12 bit ADC. While I agree that there's some value to the extra bits even when a full two stops isn't available, it's value is diminished.

After reading Clark again and thinking about it more, I would not be surprised if the 14 bit converters do offer meaningful value in the 40D. I believe the 40D should be able to exceed 12 stops at base ISO. I doubt it will above base ISO but it doesn't need to be worth it. Prior to that, I didn't see the 40D as being meaningfully better than the D200. We still need to see tests, but I can see the 40D being a better UW camera than the D200. If the D200 successor comes in at 12.5MP like expected, it's not clear to me whether the extra resolution or the extra stop is more valuable.

I read from a generally knoledgable source that today's DSLR 14 bit converters are limited to 78dB which would be about 13 bits. If that's true (which would be a bit surprising to me) then the value of 14 bits is only one stop over 12, not two like it should be. Any idea whether that's the case?

Well, I just heard that the official annoucnement will definitley be tomorrow (8/23rd). Little that I heard sounds promising. It won't compete with DsIII in certain aspects but otherwise it should hold up very well against anything else that Canon has to offer. It is a good thing I did not buy the 17-55mm yet either

I read from a generally knoledgable source that today's DSLR 14 bit converters are limited to 78dB which would be about 13 bits. If that's true (which would be a bit surprising to me) then the value of 14 bits is only one stop over 12, not two like it should be. Any idea whether that's the case?

That seems to be close to the state-of-the-art:

http://www.analog.com/IST/SelectionTable/?...on_table_id=203

The 12-bit A/D don't have 12 effect bits either, so the 14-bit ones are about 1.5 bits better.

Well, I just heard that the official annoucnement will definitley be tomorrow (8/23rd). Little that I heard sounds promising. It won't compete with DsIII in certain aspects but otherwise it should hold up very well against anything else that Canon has to offer. It is a good thing I did not buy the 17-55mm yet either

Has anybody checked amazon.com

I have seen specs as high as 78dB for 12 bit converters as well. That would exceed their resolution though.

Assuming 12 bit converters can do 74dB or better and 14 bit converters do 78dB, a 14 bit converter only offers 2/3 stop over 12 bit. Julia Borg, the source of this information, claims that an engineer gave her these numbers and it corresponds to her observation that the 1D3 only offers a 2/3 stop improvement over the 5D.

Perhaps the reason we haven't seen 14 bit converters before now is that they have been impractical as well as unjustified in these systems. I would think this would improve with time, but for now expecting a 2 stop improvement with 14 bit is, perhaps, too optimistic.

FYI: Julia Borg is a knowledgeable photographer and contributor at the dpreview forums. Her father, Iliah Borg, is the author of Raw Magick, a well respected raw converter for Windows.

FYI: Julia Borg is a knowledgeable photographer and contributor at the dpreview forums.

As a lurker over there, I found it hard to believe that you let yourself get dragged into this morass.

I mean, come on, you were talking with a guy who uses the handle TestEng, and we all know there are 10 types of people who understand binary.

I know and I'd successfully lurked there for years! The comments on Clark not knowing what he was talking about really set me off :-(

Here Julia says what I expect is true regarding the 40D: http://forums.dpreview.com/forums/read.asp...essage=24476549

Of course, just because the 14th bit isn't useful doesn't mean the 13th isn't a win!

I have seen specs as high as 78dB for 12 bit converters as well. That would exceed their resolution though.

 

Assuming 12 bit converters can do 74dB or better and 14 bit converters do 78dB, a 14 bit converter only offers 2/3 stop over 12 bit. Julia Borg, the source of this information, claims that an engineer gave her these numbers and it corresponds to her observation that the 1D3 only offers a 2/3 stop improvement over the 5D.

 

Perhaps the reason we haven't seen 14 bit converters before now is that they have been impractical as well as unjustified in these systems. I would think this would improve with time, but for now expecting a 2 stop improvement with 14 bit is, perhaps, too optimistic.

 

FYI: Julia Borg is a knowledgeable photographer and contributor at the dpreview forums. Her father, Iliah Borg, is the author of Raw Magick, a well respected raw converter for Windows.

The maximum dynamic range of a 12-bit A/D is 20*log(2^12) = 72db. State of the art for high speed ones seems to be around 70db.

http://www.analog.com/IST/SelectionTable/?...on_table_id=197

The 14 bit converters are probably ~1.5 bit better than the 12-bit ones.

The 5D cell area is about 30% bigger. The new generation sensor design from Canon may not have improve things that much.

Note the following:

http://www.maxim-ic.com/appnotes.cfm/appnote_number/641/

http://www.seis.com.au/TechNotes/TN200410A_SNR.html

Each of these sources say that 12 bits is equivalent to 74 dB. Where the 2dB difference is I don't know and don't really consider important. It's a constant in the equation though. Each bit is 6dB. The important thing is that 78dB means 12 2/3 stops, not 13.

Now, here is the 78dB link: http://forums.dpreview.com/forums/read.asp...essage=24467785

Note that Julia says that it was an ADC design engineer who told her and her father that. What exactly 78dB means in the context of a 12 bit converter I don't know. It would seem to me, like you, that 12 bit converters will all be below 74dB.

I really don't think we can assume anything about the effective difference between current 12 and 14 bit converters. Julia reports 2/3 stop improvement for the 1D3 over the 5D and Clark reports that the 5D achieves nearly 12 stops. If Julia is to be believed, the 1D3 is only achieving about 12.5 stops. If there is really a 1.5 bit advantage like you say, then the 1D3 should do better. Perhaps it does.

Interestingly, Sony's new CMOS DX sensor has many parallel ADCs. I wonder if that implementation results in a lower noise conversion since the ADCs are clocked much slower.

Each of these sources say that 12 bits is equivalent to 74 dB. Where the 2dB difference is I don't know and don't really consider important. It's a constant in the equation though. Each bit is 6dB. The important thing is that 78dB means 12 2/3 stops, not 13.

The 6 db/bit is the range. The difference is that to get to the maximum signal-to-noise, SNR, that most specs quote, you have to divide by the quantization noise which is a fraction of a bit, so that number is about 2db higher.

I really don't think we can assume anything about the effective difference between current 12 and 14 bit converters. Julia reports 2/3 stop improvement for the 1D3 over the 5D and Clark reports that the 5D achieves nearly 12 stops. If Julia is to be believed, the 1D3 is only achieving about 12.5 stops. If there is really a 1.5 bit advantage like you say, then the 1D3 should do better. Perhaps it does.

It could be limited by the sensor or the A/D. If a careful test gives 12.5 bits we still won't know which it is.

Interestingly, Sony's new CMOS DX sensor has many parallel ADCs. I wonder if that implementation results in a lower noise conversion since the ADCs are clocked much slower.

First, they are 12-bit converters, and the state-of-the-art in that is probably good enough to be not much of a factor. Also, multiple A/D's have threshold matching problems which can show up as noise or patterns. Since, Sony is sending this out they're likely to have a good handle on it, but it's not clear that is not a noise contributer.

http://www.photographybay.com/2007/08/22/p...-and-d300-leak/

Popular Science leaked this info this afternoon.