Does dynamic range change with choice of format?

S

scenedirector

Well-known member
Hey guys, been watching this forum for a couple of months. I REALLY hope I am not asking something that has already been answered. Please forgive me if I have.

With this camera and its ability to shoot a TON of different formats (i.e. 1080i/p,720p,480i/p,60i/p,30p,24p,4:2:2,4:1:1,etc. etc.) does the dynamic range change?
I have been hearing that even though these new generation of sub-$10K HD cams give off a great image for the buck, you do lose some dynamic range especially in low light. Trying to squeeze a high def image on a 1/3" chip does put limitations on some aspects of the image, obviously.

I ask this because I plan on buying this cam and doing event videograghy (weddings and such)with the MiniDv option because there isn't a huge demand from brides to see their day in HD....yet, plus P2 isn't large enough and affordable enough to invest in at the moment for live event, but plan on using the P2 card option for my independent feature filmmaking.
Filmmaking can be lit well, weddings, especially at the receptions can NOT.

Will the dynamic range be better in SD, DV for that matter, rather than in the HD mode? If any one can throw a bone on this one, it would help a lot. Thanks!
 
DV should be the worst of the bunch if theres going to be a difference. DVCPRO HD is 8 bit and so is DV though so you may see more tonal detail with DVCPRO 50/HD, but the range should be the same. although its hard to say whats what till we have camera in hand.. the proof is in the pudding.
 
I could be very wrong on this, but I'd say it does as a matter of pure physics. The wafer surface determines how many photons are "captured" in a given amount of time. If we use the same same chip for HD as SD then I'm guessing some sort of grid is used, for instance if we take 4 wafers of equal size and each is a pixel in HD, but in SD the 4 wafers make a single pixel then I have four times as many photons of light to extract the information from. But again I'm making a bunch of assumpations, and we all know what that means. Really some one who knows the actual chip properties would need to answer this.
 
the dynamic range of the CCD itslef will be the same whether its later sampled to SD or HD. the real question of dynamic range would be which compression scheme/codec holds more dynamic range if either at all given the image its presented from the CCD.
 
Dynamic range of the luminance channel would seem to be limited by the bit depth of the storage system. If it is an 8 bit storage system, then the dynamic range would be 2^8 or about 48 dB (8 stops). If it is a 10 bit system, then it would be 2^10 or about 60 dB (10 stops). The dynamic range of the detectors (CCDs) and quantizers (A>Ds) have limitations based upon the physical properties, but given modern devices, should easily exceed the capabilities of the format. DSP processing can either compress or expand the captured information, but the format determines what can be delivered to the recording media.
 
khmuse said:
The dynamic range of the detectors (CCDs) and quantizers (A>Ds) have limitations based upon the physical properties, but given modern devices, should easily exceed the capabilities of the format. DSP processing can either compress or expand the captured information, but the format determines what can be delivered to the recording media.
Click to expand...

If the format is the limiting factor then why would there be any difference in a 2/3" chip vs a 1/3"?
 
..format isnt always limiting factor.. could be DSP-or certainly could be the CCD
in the case of hvx vs varicam .. i would assume it to be CCD for the majority contributor.
 
My GUESS is that the SD image would be a downsampled HD image. However, it would be cool if they used those extra pixels to increase latitude when shooting SD. It COULD use some sort of double sampling, one for highlights and one for mids and blacks.

Another interesting idea would be to use the 60Hz sampling rate of the CCD to increase latitude when shooting 30/25/24p. Again it would take some kind of double sensitivity exposure. Is that possible?????
 
At higher illumination levels, an 8 bit system will not challenge the typical dynamic range of either a 1/3" or a 2/3" sensor. One area where a 2/3" system can have a perceived dynamic range advantage is at lower illumination levels. All things being equal (and they aren't) a 2/3" sensor has a 4:1 advantage (12 dB) as its sensor is twice the size in both the x & y axis's. This of course completely discounts the other advantages of a 2/3" sensor, but with an 8 bit delivery system, and at illumination levels well above the minimum both will have similar dynamic ranges. At lower levels, the noise contribution and S/N ratio disadvantage of the smaller sensor can be within the quantizers LSB (least significant bit) threshold and will degrade the ENB (effective number of bits) that are available. With higher resolution systems (those with greater than 8 bits) the 2/3" sensor will show greater advantages.

A DSP (or any non linear dynamic multiplier) can be used to represent a larger input range than the output path would normally allow, however, the number of discrete steps is always limited to the size of the data path. Consider how you can 'map' a 10 bit sample (1024 discrete levels) into an 8 bit storage (256 levels). Obviously, the upper and lower limits can be made to fit (by applying multiplying each sample by a non linear coefficient) but the total resolution is still limited to 256 levels. This is a resolution reduction of 4:1 in order to increase the 'latitude' by a similar 1:4 ratio. In the end, its a trade off, bound by the limits of the storage media.

These constraints are not unique to video, all sampled systems are bound by the same considerations. The only way around it is to utilize entropy or similar compression schemes, however, these are not lossless and introduce their own problems.

I'll certainly take a 2/3" sensor over a 1/3" anytime as it has considerable advantages, but with the same data path, dynamic range isn't likely to be one of them.
 
I agree that bit depth is the biggest factor in dynamic range, but

khmuse said:
At higher illumination levels, an 8 bit system will not challenge the typical dynamic range of either a 1/3" or a 2/3" sensor. One area where a 2/3" system can have a perceived dynamic range advantage is at lower illumination levels.
Click to expand...

Since dynamic range is the difference from white to black then what your saying above means that biggger area on CCD produces more detail in the near black area and greater dynamic range? Either that or I don't understand dynamic range, which is very possible.
 
Let me try to explain my comment in a little more detail. Your understanding of the definition of dynamic range is correct, it is the absolute difference between the minimum to maximum.

In a low light situation, your statement:

"... means that bigger area on CCD produces more detail in the near black area..."

could be true if the illumination levels are low enough that the S/N ratio disadvantage of the smaller sensor falls within the range of the LSB (darkest level that isn't absolute black). However, if the minimum level is high enough, then this disadvantage disappears, either sensor size would perform the same (at least as far as dynamic range is concerned).

A higher bit depth system (say 10 bit) would have 4 times more illumination levels available and this could increase the perceived quality in the near black region by a similar factor. Even though 8 bit systems gives you 256 discrete levels, in the near black region, there may be but a handful of levels used that are just above the minimum black. If, in a given situation, you had say only 10 levels used in your shadow detail, then going to a 10 bit system this would increase to 40 levels and this could easily result in something that could be judged to have "...more detail...", but not more dynamic range, at least based upon the sensor size alone.

Again, a 2/3" sensor is a much better device for many reasons. A 10 bit system is also better, but for different reasons.
 
mmm said:
Another interesting idea would be to use the 60Hz sampling rate of the CCD to increase latitude when shooting 30/25/24p. Again it would take some kind of double sensitivity exposure. Is that possible?????
Click to expand...
I asked one of the camera engineers specifically for that. They took it under advisement, but the initial reaction was "it'll look interlaced" because the two different "combined" frames would have been taken from different points in time.

So it's probably not practical.
 
Yeah..I guess...maybe a bit more like the HD100's motion smoothing... I wonder how quick the CCD can cycle, the 60Hz is only the DVCPRO format limitation. If it could bang the 2 cycle out very quickly and combine them, it would just lead to a slight blurring on movement, but would be fine for more stationary footage.
 
You must log in or register to reply here.
Back
Top