endofoto: Medium formats have the lowest DR, and the worst high ISO performance. Which medium format camera are you referring to?
And I should add that the High ISO's good too with this moderm MF camera. The DxOMark Sports Score is usually (but not always) based on the highest ISO that gives an "acceptable" 30dB SNR.
The Sports Score of the 645Z is 4505 vs the D810's 2853.
Note: the "Sports Score" really doesn't indicate the suitability of a camera for sports. That's more about AF, burst speed and burst buffer depth, and the D810 is much better equipped for this role than the 645z. But with its much bigger sensor, the 645z is probably better in LL situations.
Modern MF cameras with Sony CMOS sensors have good DR. The sensor in the 645Z has a 14.7 stops DR, similar to the 14.8 stops of the D810 at ISO64.
Prognathous: With this kind of performance, Pentax should add an option to expose-to-the-left, basically aligning exposure to the left of the histogram. The pictures may look dark before processing, but at the same time will provide the fastest possible shutter speed (for a given aperture) - without loss of detail in dark areas and without blown highlights (possibly not even light sources).
In the interests of reducing shot noise, you should always maximise the number of photons captured during an exposure. So you should ETTR, but only at base ISO.
Once the exposure results in a capture which is too dim for your liking, you normally then boost the ISO. This increases the mapping of the number of photons to a digital value representing a pixel's brightness, so the image is now brighter. But it doesn't increase the number of captured photons, so the signal-to-shot-noise ratio stays the same.
Good "ISO Invarient" performance is due to very low ADC (later-stage) noise.
What Ricoh should be doing, is offering an "ISOless" (or whatever they call it) mode for raw shooters, where the ISO either stays at base ISO, or only increases a little e.g up to ISO400 for shots that normally use high-ISO, and boosting the review image's brightness, so what's seen on the back LCD is not too dim.
This boost value would also be stored in the metadata for "default" raw conversions.
TrueAmateur: I hope some knowledgeable readers can help me understand "pushing shadows". If I brighten the dark areas of an image without affecting the bright areas (using Shadow slider?), wouldn't the midtone range get squashed so that two adjacent midtone pixels that used to be slightly different now become identical? That is a loss of midtone detail. But if I brighten only the dark area, some distant pixels that should be different will become identical; the image would lose faithfulness and acquire that dreamy dizzy HDR look. As long as we have only 256 levels for each colour, pushing shadows exacts a cost in image quality. Am I missing something here? Thank you!
The 256 steps in 8-bits-per-colour JPEG are sufficient to prevent posterisation if you perform any brightness manipulation *before* the application of the Tone Response Curve and Gamma, which shoe-horns the 16K linear steps of 14-bit raw into the JPEG version. (Also, the JPEG almost always has lossy compression applied too. Although some JPEG versions offer lossless compression, these are not used much.)
So you boost levels and saturation in a raw editor, *and then* you save the final output as a much more constrained JPEG.
If you're heavily manipulating a JPEG you need to be careful as you've got a lot less data to work with.
princecody: What the main difference between this tech & Olympus high res mode?
No. Richoh tech is not necessarily better than Olympus. Olympus uses a smaller MFT sensor, which benefits from the greater NR of 8 shots vs 4 shots. The FF sensor has less need of NR from shot-stacking.
Also the 1/2-pixel shifts in Olympus creates Super Resolution. There's less need of this with the 36MP in the K-1 sensor.
But both the 4-full-pixel-shift & the 8-half-pixel-shift systems give you much better chroma resolution than a standard Bayer CFA.
disasterpiece: Is it just me or do the D810 images look better?
You prefer the D810 picture in this comparison? So you like the moire in the finer text?
Have a look at the mushiness of the edges of the coloured roundels in the 5 o'clock position of the tableaux. The red/blue transition is presented much better with pixel-shifting.
dosdan: Cary, I understand you only had a few hours to use the camera. But there's some nice features in Pentax cameras that make the shooter's life easier.
For example, in your "Pentax K-1 Pixel Shift Resolution demo" video, you zoomed in repeatedly to check the effect of the pixel-shifting. If this camera is like earlier Pentaxes, when the review image appears on the back LCD, you can just turn the rear dial (now used for zooming) 1 click to the right.
Rather than going 2x, 2.8x, 4x ... 16x to check the focus, you can preset the first-click amount. On my K-3, I've set it to 8.3x (100%). So I get a very convenient jump to 100% pixel-peeping for focusing accuracy. Once there, left-turning the dial zooms out (8x, 5.6x etc), while right-turning zooms in (11x, 16x), and pressing the OK button on the back returns you to 1x.
Your current zoom level & panning position (e.g. lower-right corner) holds as you use the front dial to move through reviewing shots.
All very sensible.
@Matthew Saville "I know Pentaxians use that dial to zoom, but for both Canon and Nikon shooters it is common to have 1-click 100% zoom assigned to a rear button, which IMO requires less effort than making a single dial click."
What does a 2nd press of this customised zoom button in Canon/Nikon do? Zoom in? Zoom Out? 1x?
Is the zoom level/position maintained when you jump to the next shot, or do you need to touch the button again? I would think a maintained zoom level/position would be helpful when comparing multiple shots of the same scene.
Cary, I understand you only had a few hours to use the camera. But there's some nice features in Pentax cameras that make the shooter's life easier.
I know DPR's policy is to use ACR because of its market dominance, but doesn't ACR incorporate undisclosed "optimisations" that vary with the ISO? Because if it does, then the results of this test will be affected, perhaps significantly.
A test of ISO100 vs ISO1600 (4stops) should be performed using both ACR and RawTherapee (an open-source raw converter whose neutral processing behaviour is verifiable). If there is no difference, fine.
I'm not against raw converter optimisations if they can be disabled when required. But if they are present in ACR and can't be disabled, then this is not a test of "true" ISO invariance, but of "real world" behaviour i.e. what ACR has been specifically designed for, with ISO100 shots expected to need less noise optimisation than ISO1600+ shots.
nikheat: Yes, but still no mirrorless A-mount!!!!!!!!!!!!!!!!!!
Neccessary343, upping the ISO does not effect AF performance. AF performance in LL conditions is what's being examined here.
Why were the streets so bare of people and the clothing factory so empty of workers?
forpetessake: Sony is soon going to make Canon a third party lens manufacturer.
ttran88: "A $3200 camera will hardly give you a 1% increase in market share. Troll better"
True, but the tech in this will filter down to the cheaper cameras in the Sony range.
It's good to see that Canon will still continue as a going concern, and be able to play a niche role in the photographic industry, as the innovation in their DSLR range decreases :-)
Lee Jay: No AA filter = no sale.No OVF = no sale.I suspect I'll never own a Sony camera since they'll never produce something I'd buy.
Lee Jay: "And AA filters are so important that I wouldn't buy a camera without one unless it were in the gigapixel range."
You do release that the sensor pixel size + AA filter + lens are effectively a series-connected optical low-pass-filter? Just how much resolution do the lenses you use have? John Sheehy's calculations indicate that 150MP on a FF should be sufficient with even the finest lenses.
If you're so worried about moiré in these current "low-res" sensors without AA filters, there is a solution. Since Sony uses IBIS, they could have a switchable AA simulator if they license the tech from Ricoh/Pentax. Just involves oscillating the sensor up to 1px of rotary movement via the IBIS .
FocusBogus: Still not fully electronic shutters on mirrorless cameras in 2015? Sounds like a conspiracy: camera industry can sell some generations of cameras with mechanical shutters and later generations without them.
frzRom: "But BSI + on chip phase AF + on chip Dram ... Not sur it is possible to do it."
With stacked BSI tech (Exmor RS), pixel memory for a global shutter without any reduction in fill-factor may well be possible.
For more info on the problems involved with implementing a global shutter see:
Androole: I think the SuperRAW is obviously the most (I would argue really the only) significant element of this announcement.
However, it does make you feel like there's a bit of an agenda given that they carefully measured the performance of the SuperRAW in absolutely ideal test conditions, yet they decided that it wasn't worth it for them to measure the E-M5 II's High Res Mode when they reviewed it just 2 months ago. Hmmm...
And I bet they won't test the Pentax K-3 II's multi-exposure mode, either...
But DxO only deals with raw files. If the stacked mode is only available for JPEGs, then DxO won't measure it. PhotoAcute, and now DxO, offers the result of stacking in a DNG output format.
BTW, Pentax offers a multi-raw shot HDR file in PEF format, but this is really a container format than contains multiple separate raw shots, each with different exposure parameters, rather than one HDR shot with combined raw data.
The comparison of the noise of the D810 and the A7S in the 2nd page of the article was not a good choice, as the A7S is likely to be using Aptina's (now ON Semiconductors) DR-Pix technology.
There is a step in the read noise between ISO1600 and IS2000+, with the higher ISOs show less RN than expected. DR-Pix technology allows the changing of the gain in the sensor, so the sensor designer's trade-off between Full Well Capacity, Conversion Gain & Read Noise can be altered in-camera.
A better pixel-size comparison choice would have been between the Pentax K-5 (16MP) & K-3 (24MP), both APS-C cameras.
On the DR graph you can use the Screen tab to seen the pixel-level performance, while the Print tab shows the result using the same output size e.g. same print size.
coyot3: I tested on my nikon d3100 and the results are at iso 100 1" f4 underexposing 4 ev and later recovering in Darktable(linux) give less noise than iso 3200 1/30 f4 no exposure changes.
I dont know if i understand right the article xD im going to re read ._.
¿Some thoughts on my results?
You'll find an ISO100 + 4-stops boost VS ISO1600, using a K-5, here:
The exposure and sensor was the same in both cases: same scene luminance, same shutter speed & same f-stop. So the shots have the same amount of shot noise. But the ISO100 image was sent to the ADC at 1/16th of the level of the ISO1600 shot, so it was "under-brightened" in the stored image file.
DuncanDovovan: Richard / Rishi: What I would really love to see is a shot noise simulation. Rishi looks to be the guy who can calculate the number of photons that a pixel gets.
From the number of photons caught, couldn't you calculate the S/R ratio?
And from the S/N ratio, couldn't you visualise/simulate the shot noise how it would look like in a 100% crop (pixel level) and a crop that would show the scene at the same scale (more pixels scaled down to the same magnification)?
Wouldn't it be possible to visualise the shot noise, other upstream noise and downstream noise separately that way?
I would love to see (as in picture crops / not tables) and compare the shot noise, upstream and downstream noise components under several conditions. Like higher vs lower ISO. Shorter vs longer exposures. Overexposure and compensating, underexposure and compensating vs normal exposure.
And all of that for 5 different tonal values from dark to bright and in different situations from cloudy to bright daylight.
"I would expect the bright areas to have less noise."
No, the bright areas have more shot noise because they are the result of more photons hitting the sensor in this area. (Shot noise increases at the sq root of the number of photons.) It's just that the Signal-to-Noise Ratio will be better in these areas, so you probably won't notice the extra noise.
Two examples:A: 1,000,000 photons. B: 10,000 photons.
A. Shot noise = sq-root(1,000,000) = 1,000.SNR = 1,000,000/1,0000 = 1,000:1
B: Shot noise = sq-root(10,000) = 100.SNR= 10,000/100 = 100:1
BTW, SNR is sq-root(number of photons), but I've worked it out here to show the amount of the shot noise.
www.sensorgen.info has been mentioned a few times in these comments. It is a site run by Prof. Bob Newman, which shows the Full-Well Capacity of sensor pixels and Total Read Noise, as calculated from DxOMark measurements.
I'm interested in examining the Total Read Noise, input-referenced in -e (photo-electron charges), curve-fitted to estimate the Sensor read noise (upstream of the PGA) and the ADC noise (downstream).
Here's a sheet that shows a few cameras.
There are more cameras on the RN (Read Noise) Contributions tab, and it's easy to use it as a template to add more cameras.
SNR (dB) is my calculated value. You can compare it against SNR (DxO) to see how well the curve-fit performs. On the RN Contributions page you also see this in the "Diff square" value.
The 16x Ratio compares the Total RN at base ISO & 16x base ISO. The closer this is to 1, the more "ISOless" the camera.