Photo by Dan Bracaglia

The Sony Alpha 1 introduces the next generation of full-frame stacked CMOS from Sony Semiconductor. It’s the second chip of its type following the one in the a9, and with it comes improved readout speeds on the order of 5ms or faster, enabling flash sync speeds of 1/200s with electronic shutter as well as decreased rolling shutter and risk of banding under artificial light, and focus and exposure calculations at 120 times per second. This silent electronic shutter underpins the camera’s 20 and 30 frames-per-second (fps) burst modes.

With our in-depth review ongoing, we wanted to take a look at whether these sensor capabilities came at the cost of other imaging attributes of the a1’s sensor; particularly, its dynamic range capabilities.

Class-leading dynamic range

With the launch of the a9, we found a slightly unexpected decrease in dynamic range. We say ‘unexpected’ because we’ve grown accustomed to cameras using Sony Semiconductor’s sensors offering high dynamic range (i.e. low noise) and had gotten used to exploiting this in our photography.

It appeared the a9 traded off dynamic range – by dropping readout bit-depth we surmised – to achieve high read speeds (nearly 1/160s according to Jim Kasson). While the a9 II improved matters significantly, noise levels in deep shadows were never as low as those of Sony Semiconductor’s benchmark sensors. Our dynamic range tests showed that neither the a9 nor the a9 II achieved the noise-free shadows (i.e. high dynamic range) of Sony’s own a7 III, while Canon’s 1D X III only competed with noise reduction that significantly decreases detail.1

It seemed that you could only have high dynamic range and modest readout / shooting speeds, or high-speed sensors with accompanying noise penalties. Put more simply, you apparently couldn’t have your cake and eat it too. With the a1, you can.

You couldn’t have your cake and eat it too… with the a1 you can.

The sensor in the a1 proves that fast scan rates and high dynamic range need not be mutually exclusive. Compared to the chip in the a9 II, it offers a nearly 1 EV improvement in base ISO dynamic range2 while still offering at least a 20% increase in readout speed based on spec alone, and a 1.44x increase in linear resolution.

The a1 measures 13.71 EV base ISO dynamic range in all electronic shutter modes (including 20 and 30 fps continuous modes), compared to 12.96 EV for the a9 II in all its electronic shutter modes. This 0.75 EV improvement places the a1 firmly ahead of all professional sports-oriented cameras in low ISO dynamic range. High ISO dynamic range remains competitive, though lags slightly at very high ISOs due likely to a smaller pixel pitch and 1/3 EV lower dual gain step (ISO 500 vs. 640) compared to the a9 II.3

Note that all dynamic range comparisons are normalized to a common output or viewing size, for fair comparisons, removing disadvantages due simply to the presence of more pixels in higher resolution sensors.

The 0.75 EV improvement in e-shutter base ISO dynamic range (0.7 EV in mechanical shutter) over the previous generation of stacked CMOS places the a1 firmly ahead of not only Sony’s own prior attempts at pro-oriented cameras in this regard, but all other professional sports-oriented cameras as well:

Mechanical Electronic
Sony a1 13.9 EV 13.7 EV
Sony a9 II 13.2 EV 13.0 EV
Sony a9 12.6 EV 12.6 EV
Canon 1D X III

13.6 EV*

11.6 EV
Nikon D5 11.4 EV

* Canon 1D X III achieves this high dynamic range figure after noise reduction.

The a1 has 2.5 EV higher base ISO dynamic range than Nikon’s closest offering, while Canon’s 1D X III shows similar noise levels in base ISO Raws, but with considerably less detail due to the noise reduction that enables its low noise levels. Amongst its pro-oriented peers, the a1 sports class-leading dynamic range.

Not just good for a sports camera

The Sony Alpha 1’s dynamic range isn’t just good for a sports camera, though. It’s nearly a match for the very best full-frame cameras we’ve tested, despite the fast sensor scan rates that enable so many of its headline-grabbing capabilities. Have a look below at the a1’s dynamic range performance against the Sony a7R Mark IV (green), one of our current benchmarks for full-frame performance.

The base ISO dynamic range of the Sony a1 falls only 0.1 EV behind that of the class-leading Sony a7R IV in mechanical shutter mode, making it nearly class-leading in dynamic range compared to full-frame cameras. There is only at most a 0.2 EV dynamic range cost to e-shutter mode, which disappears at higher ISOs as amplification overcomes any extra read noise that accompanies the electronic shutter mode. High ISO dynamic range, which tends to vary with pixel size, the dual gain step, and upstream read noise, exceeds the a7R IV and compares favorably against other high-res full-frame cameras, albeit slightly less so against its lower-resolution pro- and sports-oriented peers.

The a1’s dynamic range stacks up well alongside its class-leading high-resolution full-frame peers: the Sony a7R IV (and for that matter, Nikon’s excellent Z7 II), being only a nearly inconsequential 0.1 EV behind either.4 But what’s particularly impressive is just how little of a dynamic range cost – just 0.2 EV – there is to the e-shutter mode that underpins the camera’s 20 and 30 fps burst modes. Until now we’ve typically seen a noise cost associated with fast readout speeds; the a1’s sensor retains nearly the same dynamic range in both of its shutter modes despite a readout rate that’s within 1ms of a traditional mechanical shutter.

This means that the a1 can offer dynamic range competitive with the best landscape cameras even when using the fast readout modes that allow sports camera performance. A table below summarizes the base ISO dynamic range numbers for the cameras, alongside the a7 Mark III referred to earlier:

D-Range (pixel level) D-Range (24 MP)
Sony a1 Mechanical 13.4 EV 13.9 EV
Sony a1 Electronic 13.2 EV 13.7 EV
Sony a7R IV Mechanical 13.4 EV 14 EV
Sony a7 III Mechanical 13.8 EV 13.8 EV

Furthermore, although 30 fps bursts require you switch to lossy compressed Raw, we measured no additional dynamic range cost (there may still be potentially distracting local compression artifacts around high contrast edges of deep shadows in low ISO files).

It seems that with the launch of the sensor in the a1, Sony shows us there needn’t be a dramatic tradeoff between speed and image quality, in terms of either dynamic range or resolution.

All-round performance

So far we’ve primarily looked at low ISO settings, since these are the ones that you’d use if you were concerned about dynamic range. However, as light levels drop, the a1 remains competitive, especially when you consider its high resolution. Noise levels in midtones are only slightly behind class leaders Nikon D5 and Canon 1D X III. We take the position that signal:noise ratio of midtones is more relevant at high ISO than dynamic range, but it’s worth noting that the a1 fares well in both regards, exceeding the dynamic range of the a9 II at intermediate ISOs, but eventually falling slightly behind the a9 II and other lower resolution chips at the highest ISOs. Pay close attention to the noise levels of pushed shadows of the high ISO crops in the links above.

These additional visual results compared to the a7R IV, a9 II and a7 III confirm a clear pattern: there’s a small high ISO dynamic range cost to be paid for high resolution sensors due to the increased cumulative read noise from the extra pixels which, with the greater amplification at the highest ISOs, becomes more noticeable as visible noise in deep shadows. It also seems that fast readout still does come at the cost of at least some, albeit now small, noise cost.5 That considered, the a1 does compare favorably in high ISO dynamic range next to its high-resolution peers.

A visual comparison

While a single number isn’t meant to sum up image quality by any means, when comparing across the same sensor format, base ISO dynamic range numbers give you a reasonable idea of how noise-free shadows of Raw files are (and for the sensor geeks amongst us, how little downstream read noise the sensor and camera electronics add to the signal). The cleaner those darker tones are, the more readily you can brighten them for incorporation (or tone-mapping) into your final output. And, indeed, our measured numbers align very well with the visual results our tests, below.

Take a look at the dynamic range capabilities of these cameras – and our previously mentioned high-resolution benchmarks – by examining shadow noise in our ISO invariance widget below. Our ISO invariance test looks at how much electronic noise a camera’s sensor and electronics add to an image, which provides a rough idea of the noise penalty you can expect if you were to reduce the ISO setting to protect highlights at the time of exposure – while still using the shutter speed and aperture of a high ISO exposure – compared to using the high ISO setting in-camera to obtain a ‘proper’ exposure (learn about ISO-invariance here).

After a +6 EV push, the a1 shows similar noise levels in shadows to the a7R IV, as our graphs and numbers predict. Electronic shutter, the mode required for the camera’s highest burst rates, shows only the tiniest increase in noise in the darkest black tones (to the right of the newspaper crop above). There’s significantly less noise and more detail compared to the a9 II, and especially in comparison to Sony’s first attempt at full-frame stacked CMOS, the a9.

Compared to ‘pro’ offerings from Canon and Nikon, the a1 shows significantly more detail and less noise, with the 1D X III competing on noise levels due only to noise reduction that reduces detail. Switch the Canon to its e-shutter mode and the difference in dynamic range is rather stark (we use the R6 as a proxy here as the sensor measurements are identical).

The Sony a1 has easily the highest dynamic range among pro-oriented sports cameras

Though measured base ISO dynamic range falls within 0.1 EV of Nikon’s Z7 II, engineering dynamic range does not account for the increased total light the Z7 II can tolerate at its lower base ISO of 64 (compared to 100 for the a1 and a7R IV), which does give the Nikon a leg up in shadow noise, provided the extra exposure is possible.

Conclusion

The sensor in the Sony a1 displays a marked improvement over the first-generation full-frame stacked CMOS seen in the a9 and a9 II, with increases in resolution, readout speed and dynamic range. Up until now it had seemed as though fast sensor readout and high dynamic range were mutually exclusive. The a1 has the fastest sensor scan rate of any camera we’ve measured, to our knowledge only roughly 1ms or ~25% slower than a traditional mechanical shutter.

The sensor in the Sony a1 displays a marked improvement over the first-generation full-frame stacked CMOS seen in the a9 and a9 II

And yet despite the speedy readout and shooting rates, the a1’s base ISO dynamic range is class-leading compared its sport-oriented peers, and lies within our measurement error of its class-leading high-resolution full-frame peers. Meanwhile, high ISO dynamic range remains competitive, surpassing that of many of its high-resolution peers and falling only slightly behind lower resolution cameras that have less cumulative read noise due to fewer pixels.

Final Result | Original JPEG | Raw
Photo by Carey Rose

That means you won’t have to worry about trading off any image quality for the capabilities the a1 bring that hinge upon its fast readout. High contrast scenes such as the one above, and more challenging ones, can be handled with ease if you expose to retain highlights and tone-map underexposed tones to be visible in post-processing. And the dual gain design of the sensor ensures low noise levels as light levels drop. The Alpha 1 should offer a great deal of flexibility regardless of your shooting scenario.

Addendum: What about lossy and lossless compression?

We’ve written much about Sony’s lossy compression, mostly about how it should be modified to not include local compression artifacts, in addition to being offered alongside lossless compression as it is for most other brands. The destructive lossy compression was the only form of compression for Raw files offered up until now (remember when it was the only Raw option?), but that’s changed with the a1.

With the a1, you now have three options for Raws: uncompressed, lossless and lossy compressed.

Uncompressed and lossless compressed files retain the largest dynamic range, while lossy Raw files pay a slight 0.11 EV dynamic range cost. 0.11 EV is within our margin of error, so you’re more likely to notice the local compression artifacts around high contrast edges of deep shadows than you are to notice this slight drop in measured dynamic range. Furthermore, if you’re switching to lossy compressed Raw to achieve the camera’s 30 fps burst rate, there’s already a 0.2 EV drop in dynamic range due to the use of a fully electronic shutter, which makes this 0.11 EV cost typically irrelevant.


Footnotes:

1This link takes you to our ISO-invariance results of the Canon 1D X III where you’ll see the progressive loss of detail as the ISO drops from 6400 down to 100, due to increasing mandated noise reduction in Raw at lower ISOs. This is not a tradeoff you’ll have to pay with sensors that perform better in this regard, like the one found in the a1.

2We report the ‘engineering dynamic range’, which measures the range of tones between clipping (white) and point where the signal-to-noise ratio drops to 1. We then take the log2 of this range and report this value in EV.

3‘Dual gain’ or ‘Dual Native ISO’ sensors are now common across both stills and video cameras. The approach was first seen in Aptina’s ‘DR-Pix’ technology, which adds a dual conversion gain (DCG) transistor to each pixel to either connect or disconnect a physical capacitor from the floating diffusion (FD) node (or, for the purposes of our discussion, the ‘circuitry’). With the capacitor connected in the circuitry, the resulting large capacitance allows for a large full-well capacity (FWC), or a large signal to accumulate and be read at each pixel, enabling a low conversion gain (LCG) mode for low ISOs. In low light conditions the capacitor is disconnected from FD node (the ‘circuitry’), and the lower overall capacitance only allows for a smaller FWC and, therefore, a much higher conversion gain (HCG) mode that can be thought of as an extra analog gain inside the pixel.

The HCG mode has a far smaller total charge handling capacity (so bright signals clip more easily), but the higher sensitivity and reduced read noise lead to lower noise, particularly in darker tones or shadows. The ISO at which each pixel switches to HCG mode can be thought of as the camera’s ‘second base ISO’, and the higher the ISO at which this occurs, the higher the conversion gain of the HCG circuit, and the lower the resulting read noise, at least theoretically. This is likely to, at least in part, explain the greater high ISO dynamic range of the a1 over the a7R IV, whose dual gain step or ‘second base ISO’ of 500 will have a higher conversion gain than that of the a7R IV’s second base ISO of 320.

4Alongside the Sony a7R Mark IV (and Mark III) the Nikon Z7 II is one of our benchmarks for base ISO dynamic range amongst full-frame sensor cameras. The Z7 II actually has an edge in situations where a photographer can use longer exposure times or brighter apertures, since the lower base ISO of 64 allows for an extra ~2/3 EV exposure before clipping highlights relative to ISO 100. That extra exposure increases the signal:noise ratio of all tones in the entire image, leading to a cleaner, crisper image overall.

Such differences, along with small 0.1 – 0.2 EV differences between the a1 & the a7R IV do add up cumulatively, so it’s worth considering that when you compare the Sony a1 vs. the Sony a7R IV and the Nikon Z7 II (at ISO 100 and ISO 64) side-by-side, the Nikon Z7 II offers the highest image quality.

5The a1’s dip in high ISO dynamic range relative to lower resolution peers is also confirmed in engineering dynamic range measurements here. Aside from the decreased cumulative read noise of lower-res sensors, both the a7 III and a9 II sensors switch their pixels to high conversion gain (HCG) mode at ISO 640 as opposed to 500 for the a1 (indicated by the smaller circles in the graph linked above). The higher conversion gain for these HCG modes on the a7 III and a9 II may also contribute to the decreased upstream read noise and the increased high ISO dynamic range relative to the a1, though the pixel count is more likely the dominant factor.

Author: Go to Source