The weird thing is that no one that I deemed qualified (like Jim Kasson or Bcliff from Photonstophoto etc.) is testing A9III’s image quality so that I can just copy paste their results. So I guess lazy boy has to move his ass for once now.
I did a very simple test (since I obviously don’t have the camera, my capability is quite limited anyway) just to check the quantum efficiency of this camera:
1: Steal some RAW files from DPReview
2: Check the noise level.
If you just do some eyeballing, it looks like A9III is slightly noisier than A9II (this is also DPR’s conclusion).
But the numbers suggest otherwise. I picked 3 patches from the ISO6400 RAW files:
And here is what Rawdigger tells me:
Looks quite comparable. A9III is even slightly better in patch 1 (the brightest patch). However A9II does show some minor advantages in the darker patch which seems to implicate better read noise.
All in all it’s quite negligible. The low light performance looks OK to me. Dynamic range is another story for sure, given that the base ISO is 250. However, since this is a super fast global-shutter sensor, there is a ton of tricks that can be applied here to increase the dynamic range, as long as Sony wants it. In theory it can do bracket exposure faster than any consumer camera ever built. It’s pretty much the king of image stacking or computational photography.
(In reality though, Sony RX100VII, a camera also with super fast readout speed, only does bracket exposure at 10FPS, totally wasting the sensor’s potential, so I have to emphasize the part “as long as Sony wants it”.)
In the old days there is no cheat code for the low-light performance so it’s always more important than dynamic range. Nowadays Adobe’s AI denoise is doing some unrealistic stuff. Call me impressed and it kinda changes my way of shooting (I kinda don’t give a shit about ISO anymore…). Anyways I still believe that low-light is the core spec here so nice to see that A9III is not really falling behind. Also, for the target audiences (sport guys?) I guess they’re always shooting at high ISO anyways, so the ISO250 situation doesn’t hurt that much.
Let’s say I haven’t had an orgasm watching Sony announcements for many years, but oh boy this one is just too good.
Luckily Sony has published a few papers about a prototype with 6.9um pixels around 2018 (not far from A9III’s 6um pixel), otherwise we wouldn’t know how powerful this new architecture actually is. The funny thing is that you can no longer measure the readout speed of a global shutter sensor.
Traditional global image sensors first capture the image and then store the information in the analog domain, waiting for their turn to do AD-conversion one by one. The problem is that the sensor can still absorb some light even after the exposure is supposed to be finished, or let’s say they just can’t close their “shutter” completely. Therefore, they need some extra measures to shield the sensor from “parasitic light”, which reduces the image quality in one way or another.
Sony on the other hand, has built a tiny Single-Slope ADC into each pixel, which is fxxking crazy. So the sensor has two stacked layers, one with 24 million pixels and one with 24 million ADCs, logic, I/O, SRAM and some other stuff. There are also at least 48 million Cu-Cu connections between the two layers. The prototype shown below only has 1.46M pixels but the principle is the same:
So basically there are no more waiting in the queue. The signal from all pixels is converted to the digital domain at the same time. We can call it the “digital global shutter”.
On paper this basically get rid of all the issues of traditional “analog global shutters”. Well just on paper… there is obviously some issues (ISO250!) that we’ll address later.
The sensor is not just about global shutter though. Traditional column-parallel ADC sensors only has a few thousand ADCs. This thing has 24 fxxking million. Even though each one is tiny and slow, the sheer number has made sure that this sensor is just insanely fast. The prototype has a readout speed of 660FPS!
And this is actually limited by the IO speed. The ADC can complete a cycle in 0.39ms. So in theory the maximum speed can be as fast as 2400FPS!
Another wonderful thing is that there is no such thing as a resolution-speed trade-off anymore. In conventional column-parallel sensors more pixels in total = more pixels for each ADC to handle = slower speed. In pixel-parallel sensors more pixels = more ADCs = the same speed whatsoever. So I assume that A9III’s sensor is in the same 2500FPS ballpark or even better.
Let’s say with this architecture Sony would never have to worry about the speed again for the next decadeor two.
So speed is no longer the concern, but what about the resolution?
Well that’s the main challenge here. First of all you’ll have to squeeze a tiny ADC into the pixel. Single-slope ADC is probably the tiniest ADC structure due to a lot of resource sharing but there is still a comparator + a 14-bit counter and some other stuff. For smaller format it’s even worse. 6um pixel = 10MP for APS-C or 10 pixels for smartphones. So finally we have some new technology that discriminates against smaller sensors lol.
The good news is that you can always shrink your ADC with more advanced process node, and Sony still has a big room for improvement since the prototype only uses the 65nm process node for the logic chip:
Sony also published a detailed paper about A1’s sensor, which has a full-frame size 40nm chip stacked on the 90nm pixel layer. I assume that A9III is also using 40nm. Anyway, if they can make a 6.9um pixel with 65nm then I think 4.2um (50MP) with something like 28nm isn’t really a pipe dream. Sony was already doing 28nm chips for smartphone sensors back in 2016:
There is a second issue though. 24 million ADCs working at the same time is going to create a huge peak current, which could be an issue for the power supply. The total power consumption is not the main concern here. It’s like you’re trying to move 100 bricks from A to B. You can either do 100 runs, each run moving only 1 brick, or do only one run with all the bricks. The result is the same but it depends on whether you can handle the total weight of 100 bricks. And even if you’re operating at 1FPS the peak current is still there. Sony has made quite a lot of effort to reduce the current consumption for each ADC. The average consumption can be as low as 1.7nA for a 660FPS output speed. The funny thing is that due to their immense effort to drop the peak current, it looks like this sensor is actually more efficient than the A1 sensor. Based on the number provided in the paper, if you increase the resolution to 50MP and drop the speed to 44FPS (just like A1’s sensor), then the power consumption is going to be 1.73W, much lower than A1 (2.76W):
This is not reflected in the battery life of the final products though. First of all nobody is doing full speed output when you’re just looking at the viewfinder. And the sensor’s power consumption is just a small part of the total consumption. A1’s power consumption is actually 4.3W when using the EVF.
Anyway, it looks like the new technology isn’t that power hungry, which is good news. But you probably noticed that it has two power consumption number. Actually one is for the low power/high noise mode and one is for the low noise/high power mode. Now, here comes the bad news…
The Image Quality Sucks?
Even in the high-power mode the read noise is still 5 times larger than A1. So you have more ADCs now but it looks like they’re not as capable as those big column ADC that can utilize more area or more current.
But this is not even the worst part. The full well capacity is only 16.6k e-. A1’s FWC is 60k e- and the pixel size is much smaller! It’s basically an ISO1000 sensor vs an ISO100 sensor!
Sony explained that the FWC is low only because the DAC in the prototype isn’t top notch, it’s not the pixel’s fault blablabla. But as you can see… A9III has a base ISO of 250, and also no dual-gain. Ouch!
My guess is that the reason behind the low FWC is that the pixel design is actually quite different:
This part is difficult to explain to people who’s not familiar with analog circuit design but let’s say a small part of the ADC is actually in the pixel layer instead of the digital layer. The three transistors on the left are the input pair and the biasing current of a simple comparator. The other half of this comparator is in the digital layer, connected via two cu-cu nodes.
If you take a look at A1’s pixel design, which is your typical 4T structure, it’s much simpler (the red part):
These three extra transistors could also be very large due to the need to drop the current consumption. It’s difficult to tell because Sony is operating them in the sub-threshold region where everything changes exponentially. It’s possible that these transistors are taking too much space and drops the FWC. But at this point it’s really just my wild guess.
If my guess is correct, then first of all… ISO250 just sucks. It’s going to be a deal breaker for many people. I guess our digital global shutter sensor is just like their analog cousins that have poor image quality. The main difference is actually the speed. It also means shrinking the pixel size might be quite difficult. You’ll need better process technology for both the digital layer and the pixel layer, which kinda defeats the purpose of making a stacked sensor. As far as I know Sony’s pixel layer has been stuck with 90nm for many years even for smartphones.
A potential fix is to the ISO 250 issue is to use some on-chip multi-frame stacking to achieve higher dynamic range. As I said the ADC can be as fast as 2500FPS internally so if you have some integrated memories and processing circuit then you can finish a 4-frame integration in merely 1/600s. 1/600s is certainly not ideal for fast moving subject but probably good enough for most situations, better than nothing after all. Practically, for birding or sport shooting you can rarely achieve a high shutter speed + low ISO combination anyway. Sony has indeed introduced a new multi-frame stacking feature in A9III, with some interesting description like “possible to shoot wildlife at higher shutter speed”. But it’s not really on-chip (extra work flow with made in Japan software lol), and it remains to be seen how it works and what’s the integration time.
Anyway… even with such a huge flaw, let’s say global shutter is regarded as “the Holy Grail or photography” for a very good reason, and the speed aspect of the new technology is certainly appealing. But for now, I’d say the price is just too steep. I was expecting something like $4500. SonyAlphaRumors has destroyed my dream. If I had $6000 to burn on a camera then I’d rather buy an A1II. A1’s sensor is still a beast of its own even though it’s not a global shutter sensor. The major downside is that it doesn’t have the fancy 4-axis screen or all the latest software. Sony has basically given A9III everything they could, and I expect no less from A1II.
It seems to me that the most interesting upgrade on the new A7C duo is actually the focus bracketing, which is sad given that it’s a basic feature for the other brands. Other than that none of the new upgrades really interests me that much.
The saddest thing is the price of A7CR. After the release of A7RIII in 2017, all the sensors in the A7 line basically have similar output speed = resolution * readout speed. So you have 24MP * 20FPS, 33MP * 15FPS, 42MP * 12FPS, and then the IMX455, 60MP * 9.42FPS (listed in the datasheet). It’s so slow that you can’t even shoot 10FPS without sacrificing a little bit of the image quality (well on the A7CR you can’t even get 10FPS, finally revealing the sensor’s true color). So Sony’s mainstream sensors haven’t really improved in 6 years. Some people may argue that it’s actually 8 years because A7RII also has a 42MP sensor. It looks to me that the A7RII sensor is crippled somehow. 6 or 8, doesn’t really matter. The point is that the cost of these sensors should be very similar based on my experience in the semiconductor industry. So the 60MP sensor is not much more expensive than the 24MP one in the A7III.
But of course the price is not dictated by the BOM. Sony charges more money for the higher resolution bodies for sure. In the case of A7RV I’d say at least it’s really well equipped compared to the A7IV. You have a monstrous 0.9X 9MP EVF which is the best in the industry. A super fancy screen twisting mechanism. Two CFA compatible card slots instead of just one on the A7IV. USB 3.2. 2*2 MIMO. You get the point. So at least it’s a real deal premium camera. For A7CR there is absolutely nothing. It’s just A7CII with a different sensor and some AI framing shit. And suddenly Sony is charging $800 more. This is very unsettling.
Also you can understand why Sony’s A7 lines are all capped at turtle speed while Canon is already in the 20FPS era. If you push the sensor to the limit then A7IV can easily achieve 20FPS with slightly reduced image quality. Then people will start to think hmm maybe Sony’s super expensive 60MP turtle speed camera is actually not worth it?
So I was always saying that the 42MP sensor would be a better fit for the next A7C. Not only is it faster than 10FPS but also Sony wouldn’t dare to charge you $3000 for an old sensor.
The whole thing is just sad. And it gets even worse: the buffer size of A7RV is 583 compressed RAWs. On A7CR it’s reduced to 36. This is even worse than A7C (115 frames) even taking the resolution difference into consideration. And you don’t get the CFA slot to compensate for the limited buffer size (BTW CFA cards are super cheap in China now).
What can I say? At least 16-35GMII is really yummy.
A7RV is a well-polished camera with a slow sensor which wasn’t even good enough 3 years ago. Very sad situation. Sony kind of have this strange rule that every sensor has to be recycled at least once, with very few exceptions like the A7R. But I’ll not beat the dead horse again (already bashed the sensor when A7RIV was announced). There are plenty of new stuff on this camera:
4-Axis screen: I think the only shortcoming of Sony’s implementation is that you can’t quickly go to the vertical low/high-angle position unless the LCD was facing the camera body to begin with. Nikon Z9 is capable of doing that, but it can’t do selfies or narcissistic vlog. Overall I still like Sony’s idea better. Compromise has to be made. Another upside is that when you’re doing vlogging you can move the screen away from all the cables. More degrees of freedom are always good.
Bulb-Timer: you can shoot maximum 900s in the bulb mode now. Sony cameras were able to do it years ago (NEX era? Don’t remember, don’t care to find out) but somehow they removed it, and you can only do 30s automatically without a remote control.
Ultrasonic sensor cleaning: this is not the regular shake-the-fucking-sensor-out-of-the-body-cleaning. It’s a new cleaning method starting from A7SIII (could be wrong, maybe another one) and the reception is very good, although Sony somehow doesn’t talk about it. This is much better than just shield the sensor with the shutter blades. Love it.
Low-light AF: 1 EV better than the A7RIV and it’s confirmed by Sony Northrup’s test. Although still not as good as a Canon R5.
Focus bracketing: we still don’t have in-body stacking which means no preview. OK I guess you can’t expect that much from Sony, but what’s weird is that there is also no focus breathing compensation, which makes post-processing much easier and is already available for video shooting. Anyway, better than nothing.
All-time DMF: reserved feature on some of Sony’s $$$$ GM lenses but now it’s available for all Sony lenses (not sure about the 3rd party lenses). Nice touch.
Pixel-shifting movement removal: in previous Sony camera it’s basically unusable. You always have micro movement that ruined the shots. What’s weird is that this seems to be an improvement to Sony’s desktop software instead of the camera itself. So basically all Sony cameras that have pixel-shift mode should be able to benefit from it. But AFAIK this is going to be an A7RV exclusive feature. Dick move.
IBIS: seems to be better now but I haven’t seen anyone thoroughly tested it yet. We do have someone who claimed to have handheld a 1.6s exposure here. On the other hand the walking around vlog test doesn’t look impressive at all. It’s still better than the previous Sony cameras so if your standard is low then it’s usable I guess. Another test can be seen here but it’s easy to tell that she’s being cautious to not make Sony look bad.
AI AF: Sony’s main selling point but frankly there is nothing to talk about. It’s quite similar to what EOS R5 already has. Interestingly it can auto switch between other animals/birds, but not the other stuff. Hopefully one day it can be more intelligent so you don’t have to switch at all. A feature that I want is “ignore that fucking fence”. Would be very useful in a zoo. Shouldn’t be that difficult? A focus range limiter could also do the job.
All in all I think Sony has done a very good job fixing all their flaws. There are still rooms for improvement (smooth reflection or more advanced stuff) but I consider it “good enough”, meaning it’s not perfect but there is no dealbreaker compared to the other brand now. The only problem here is the sensor. Funny that years ago Sony cameras are basically “super sensor in a barely usable body” but not it’s the opposite. I guess we’ll have to wait for the ultimate camera, again.
After the announcement of 16-35G there are 3 lenses that have been completely purged from Sony’s lineup: 35ZA (replaced by the stellar 35GM), 70-200GM and 16-35ZA. And it’s basically confirmed that the next one is going to be 24-70GM.
There is also 50ZA vs 50GM, but 50ZA is much cheaper, which is a good reason for it to stay on the market (1400 EUR vs 2200 EUR).
What’s funny about these four purged lenses (24-70GM included) is that they’re more or less infamous for the quality control issue, or the so-called “lottery program”. For Sony lenses it almost always happened at the tele-end of a zoom lens, with 35ZA being the sole exception, a prime lens that can’t pull its shit together.
Most people think that quality issues are caused by the manufacturing capability, but that’s not always the case. At least for microelectronics it’s often design-related. Some design tends to amplify all the imperfection of the manufacturing process and gives you worse yields than the others. It seems to me that Sony was probably using some software/algorithm/design methodology in the past that consistently pushed out bad designs that’s difficult to be perfectly assembled. This could lead to a series of consequences: higher manufacturing cost, unhappy customers trashing the brand online and returning all the bad copies, and of course Sony has to bear the cost of re-testing/re-packaging. Very bad for Sony’s business. That’s probably why the 35GM is actually cheaper than the disastrous 35ZA.
It looks like Sony has fixed all these shit after the debut of 24GM. At least I haven’t see the same complaints on 70-200GMII or 12-28GM so far (correct me if I’m wrong). So Sony is probably cleaning the house now because these lottery lenses are not that profitable. Otherwise I wouldn’t say 16-35F4 is a spec that desperately needs any upgrade right now.
So who’s the next candidate? Well here are the “lottery lenses” that I know about…
24-70GM, 16-35GM, 24-70ZA, 70-300G
There are probably others but I think these are the lenses that are more frequently trashed online due to the inconsistent performance at the tele-end, some more than the others. I think 12-24G also has the same problem (at least my copy wasn’t so good) but it was less talked about. It’s especially bad for the 70-300G because most people are paying for the 300mm end instead of the 70mm.
Of course these are just my wild guess so don’t take it too seriously. There is also one lens that doesn’t have the QA issue, yet I still want Sony to upgrade it: the 85GM. It’s not the sharpest lens for sure and the focusing motor is a bit out-fashioned now.
16-35G POWER ZOOM
As impressive as this lens is (just like most of Sony’s recent lenses, minus the 24/40/50 trio), it’s kind of boring given that Canikon are all trying new stuffs like 15-35 or 14-30 blablabla. Meanwhile Tamron goes the opposite way and gave us an unbeatable F2.8 UWA lens with slightly compromised focal range. If you ask me which is the better landscape lens then I’d say Tamron is still a no-brainer (for video makers it’s another story for sure). However there is something interesting with this lens: it’s a power zoom lens.
I know I know, photographers hate power zooms: “it’s for the evil video empire…” My biggest complaint about my RX10M4 is also the PZ issue. But this lens is quite different: it uses four XD motors just for zooming, so the mechanism is exactly the same as focusing now.Just think about how fast a Sony lens can focus, this should also applied to zooming now, unless Sony set some speed limitation on purpose. If we can get rid of the biggest cons then you’re basically getting a video-friendly lens without any extra cost, so why not? And there are definitely some potentials for photographers too, like automatic zooming during a time-lapse shot. In theory it’s also possible to fix the focus breathing by zooming and focusing at the same time, which is good for focus stacking. The 16-35PZ does have extremely good focus breathing control. I’m not sure if it’s due to the optical design or Sony is already using the power zooming to compensate for it.
It’s all just empty talk right now given that Sony cameras don’t even support focus stacking at all (what a sad story), but I do think this is good for the long run. There is still one stupid issue though: Gerald Undone has confirmed that the lens reset the zoom position when power off, which is a strict no go for some serious landscape photographers. How hard can it be for the camera to remember the zoom position?
There are also some challenges for the lens designer. As I explained before to achieve fast autofocus you need a lightweight internal focus group. This is the same for fast zooming: a fast power zoom must be an internal zooming lens. If you have to move the heavy front group then there is no way that it’s going to be super fast. Given that almost all the wide-angle lenses are external zooming lens I guess it’s not the easiest thing to do, but it looks like Sony has done a great job with the new 16-35G PZ.
It could be more challenging for those telephoto lenses. A good example is 200-600G, which is obviously longer than its peers. Losing the 150-200mm range is also a pain in the ass. But I think power zoom is probably not the best idea for a telephoto lens anyway.
On the other hand it makes all the sense for a UWA lens. First of all it’s more widely used for videos. Second the difference of the optical length between the wide-end and the tele-end is not that large, meaning it should be much easier to design an internal zooming lens than say, a 200-600. So yeah… I think this new power zoom lens is very interesting. And since 16-35GM also belongs to the notorious lottery club, hopefully we’ll see an updated PZ version in the near future. I’m definitely intrigued. I love my Tamron for sure but 3rd party lens always leaves a void in my heart that can’t be filled.
Actually it’s still a very decent camera for most people. I didn’t have many complaints until my son was born. He brought one thing that completely ruined A7RIII’s eye AF algorithm:
Narrator: but your son is 20 months old, he should be getting rid of the fake titty now…
OK life is not always as planned. But I do notice that my RX100M7 can do a much better job recognizing his eyes. It’s not 100% for sure, but when it fails you can simply touch the screen to initiate the real-time tracking, which takes care of the rest.
The next problem is that I desperately need a flip screen for low angle vertical shots, because that’s how you photograph a baby. Actually low angle vertical is also a popular technique for landscape photography, so even before my son was born I was already bitching about this issue non-stop(you remember the 宾得大护法 episode?) . With my son it’s only getting worse. Stupid tilt screen cameras kill my fucking back. Even without the fake titty issue I still need a new camera.
And as usual I always appreciate high resolution and high speed. The most perfect camera out there that covers all my needs is actually EOS R5. But I need a good 35mm lens. I can even settle with something as wide as 24mm. But all Canon has from 24-35mm is that shitty 35F1.8STM that vibrates like a dildo when the motor is trying to push its little tip forward, and the optical quality is meh meh. Or you can buy some dinosaur lenses from the DSLR era that’s humiliated by the stellar 35GM. So not today, not today. Previously I was using Sigma 56 as my main portrait lens so a 24MP FF camera is really a hard sell for me. But now I’m using 35GM so I can loose the constraints a little bit. After some compromises here and there, there are only two cameras left on this planet that fit my needs: A7C and A7IV.
I wouldn’t say A7C is a no brainer between the two. It’s obviously a flawed camera and I can see a lot of reasons to pay 1000+ more euros to get the boring camera for better videos and user experience. But here is the next thing:
I don’t shoot videos with my cameras now
I was planning to use RX100M7 & A7RIII to shoot baby videos, one provides 200mm and the other provides bokeh. But after my son was born I’ve realized that it’s a stupid plan. First of all you’re not going to take videos of your newborn from 100 meters away. All you need is a ~24mm lens. And nobody gives a shit about bokeh for baby videos, or video quality or color grading or whatever. It’s about the content. And you don’t have time to process all those videos when your life is changing rapidly. So the best tool is actually your smartphone, it’s low profile (very important when your kid is playing in a public place), easier to carry, to edit, to share, and the stabilization is better than any shitty camera that Sony has ever produced.
So in my eyes A7IV is just a 33MP version A7C that has better ergonomics. Both cameras don’t have high res or high speed, they’re just temporary solutions for me. Am I paying 1000+ euro just to get better ergonomics on a camera that probably won’t accompany me for a long time? Fucking hell no.
So let’s talk about A7C…
When A7C was announced I already wrote some posts about how underrated it is so let’s keep this short. The AF system is very good and I’m confident to say that you won’t find A7IV noticeably better (except for the bird eye AF). It’s definitely not a smaller A7III, don’t make that stupid mistake. A7III is not bad but it can’t AF like A7C.
So it’s a very powerful machine. But I guess most people are more concerned about its crippled ergonomics. This is the part where I’m actually surprised when I got the camera.
First of all I don’t have the slightest bit of complain about the grip. It’s not like “ah it’s bad but I can survive it”. It’s “I don’t notice it at all.”
The lack of buttons is a bit inconvenient at the beginning but it’s totally manageable. I have an RX100M7 that has less buttons and dials yet I can still find a way to configure it to suit my needs. And if you have a lens with AF/MF switch and a focus hold button (like my 35GM) then I’d say it’s another “I don’t notice it at all” situation. But of course it depends on the user case. Anyway, the big plus of A7C is that it has three dials instead of two. I always find two-dial cameras like A6X00 or RX100 a bit restrictive, given that three dials are enough to control every aspect of the exposure.
The EVF is indeed problematic. It’s just too small and not enjoyable to use. More like a last resort when you can’t see shit on the LCD under strong light. Even the A6X00 series has a much larger EVF than it. This is something that I actually have to put up with.
But overall I feel like ergonomic is really not that important for me. Even back in the A6000 days I was complaining more about the software instead of the hardware. Maybe it’s just me but I always feel like it’s just an excuse for fanboys to attack each other. My camera sucks but it has better ergonomic so it’s a real camera blablabla. Turned out his F brand real camera doesn’t even have a grip and the EVF is barely larger than the A7C. At the end of the day it’s a camera, not a massage chair.
But everything has a bottom line, and this is off limit for me:
The flip screen is a fucking life saver. Whatever “advantages” a tilt screen has is just insignificant compared to what a flip screen can offer. It’s like saying a $1 coin is better than a $20 note because it’s made of metal. Completely ridiculous.
Overall, I feel like if you’re not that into high-end video spec, then A7C is really the best mainstream hobbyist FF camera on the market now. It’s like the reincarnation of A6000, strong performance in a compact chassis with poor ergonomics, and it’s cheap for what it can do. Peak old Sony philosophy of making cameras. Now they make their products more and more user friendly to the extent that there is actually a bokeh button on the ZV series. Yet you don’t see performance increment like in the old days.
It hits Canon’s weakness surgically (33MP vs 20MP) but other than that… it’s a pretty mediocre camera and leaves a lot to be desired.
Sensor Analysis
I’ve seen some contradicting numbers so what’s reported here might not be accurate, but I believe it won’t be too far away either. DPR has reported that the full sensor readout speed is only 15FPS. So basically the total speed is very similar to A7III or A7RIII or A7RIV, no fucking improvement at all, damn… It’s also a bit slower than the R6 sensor (20MP*30FPS in normal operation). For a company that’s famous for their sensors they’ve been trapped in the 500MPPS zone for a very long time now.
However this doesn’t mean that the burst speed should drop to 5-6 FPS when you’re not shooting compressed RAW, it just doesn’t make sense. A7RIV does this because the sensor is even slower. 15FPS on the other hand should be faster enough for 10FPS + lossy compression. I hope this is just some pre-production firmware issue but maybe it’s just my pipe dream.
For 4K FF video the speed is boosted to ~40FPS, which is faster than R6 (still 30FPS for 4K24P video, it’s only boosted to 50FPS for 4K60P or 20FPS burst shooting) and on par with A7III. So the rolling shutter is actually quite decent, I don’t know why people are complaining, maybe a bit spoiled by A1/A7SIII? If the 5FPS burst rate drop issue is not caused by firmware bugs then I believe that when you use continuous shooting + compressed RAW this 40FPS boost mode will be automatically engaged, and you’ll see some dynamic range drops.
Anyway, it should be very easy for Sony to achieve 20FPS burst rate with such a 40FPS sensor. It’s going to have some rolling shutter but it’s still better than 10FPS. Big fail.
The 40FPS speed should also be enough for S35 4K60P but interestingly they further boost it to ~80FPS. It means theoretically it can even output FF 4K60P oversampled video (or with a tiny amount of crop just like R6) if the I/O speed is faster enough. Unfortunately a mediocre sensor like this is most likely still using the 1GPPS I/O, whereas the oversampled FF 4K60P requires 1.68GPPS. But of course you can also use pixel binning to walk around the I/O limitation.
The 80FPS faster readout means Sony still has a lot of room to decrease the crop factor. Ideally they should spend the extra bandwidth on active stabilization/focus breathing compensation so you won’t see any extra crop in the S354K60P mode. I don’t know if that’s actually the case though.
All in all the sensor is Sony’s old shit but still capable enough to make a better camera, with 20FPS burst rate and FF pixel-binning 4K60P video. But it’s Sony, you know… as long as the business is good there is no need to exploit their poor little sensors.
Why Is 20FPS Important
As I’ve said many times, mirrorless cameras are different from DSLR. Burst rate doesn’t only mean how many photos you can get per second, but also how many frames you can see when shooting (or at least in most cases, A1&A9 are not bound by the rule). 10FPS vs 20FPS makes a huge difference for keeping track of fast objects. Everybody knows that a smooth motion requires >24FPS (and I won’t really call 24FPS smooth, nowadays I rely on my TV’s motion interpolation to boost everything to 120FPS, otherwise it looks very jumpy) so 20FPS is definitely not too much to ask. If Sony can put some frames only for view-finding in-between like A9/A1, i.e, you’re shooting 10FPS but get a 20FPS viewfinder refresh rate, then I’m fine with it. Once you tried a 20FPS camera, it’s extremely difficult to go back.
Focus Breathing Compensation
Besides the all-time bluetooth connection and the focus map, this is one of the few things that’s actually worth mentioning, although I guess most people outside the filmmaking community don’t even know what focus breathing is and they won’t notice it at all.
The closer is you focusing distance, the narrower FOV you get. And the nominal FOV/focal length is determined by when the lens is focusing at infinity. This means that when you engage this function you’re definitely going to lose some FOV. It’s not like your 15 degrees FOV lens becomes a 25 degrees one when focusing at a closer distance so even with the compensation it’s still a 15 degrees lens in any case. It’s the other way around. The worse the focus breathing is, the more FOV you lose.
VS EOS R6
For still shooters the 33MP resolution is really hard to resist. For video shooters though, R6 only has some minimal cropping in the 4K60P mode, and the IBIS is 100 times better than Sony’s poor shit. Sony does have focus breathing compensation but I don’t think it’s really that important. Canon also supports All-I codec but only for time-lapse, while Sony doesn’t have the same limitation. In general, if you’re fine with 20MP then I think R6 is a better tool, even though the card slots implementation is really stupid (it’s going to haunt Canon in the coming years). However, personally I’m not fine with 20MP, and I can’t ignore the fact that Sony has a much better lens line-up at the moment. Personally I can imagine myself switching to EOS R5 (which is the closest thing to “the last camera you’ll ever buy”) if they have something like 24GM or 35GM.
VS A7III/A7RIII
The video performance is greatly boosted for sure. For still shooters the real-time tracking is also a killer feature that I don’t want to miss. Plus A7IV also has a flip screen. Yeah I know some people claim that they hate the flip screen (the keyword here is “claim”). So I have to admit reluctantly that it’s still a worthy upgrade no matter what your user case is. However if the 3rd gen bodies also have real-time tracking then I don’t think I’m gonna be remotely interested in the A7IV. And that’s probably why Sony won’t give you that upgrade.
VS A7C
Funny thing.
It has both the real time tracking and a flip screen. If you don’t shoot video and can live with A7C’s obvious ergonomic issues then A7IV is basically just an A7C with more pixels.
That’s why I immediately bought one after the disappointing announcement. As a long time A6000 user I don’t think the ergonomic issue is going to bother me. A7C is actually a force that should be recognized with, it looks like the best entry-level FF camera that you can buy on the market right now.
I Don’t Care Anymore
We have to accept the fact that Sony is no longer the company that keeps pushing the boundary. Every camera they make now looks like a well-calculated move by those marketing guys instead of a passionate creation of brilliant engineers. You can’t say their cameras are bad. It’s going to sell well. It’s just boring.
My previous combo is Tamron 17-28F2.8 + Sigma 56mmF1.4 which I thought I could use till the end of the Universe. But after my son was born I was starting to realize that a 85mm equivalent lens is just not the best choice for babies, even though it has a fairly large image circle in the full frame mode for some extra flexibility. Meanwhile I feel that T17-28 is a bit on the slow side (both the aperture and the AF tracking speed) but as a zoom lens it still offers great versatility, so now I’m switching to the T17-28 + 35GM combo and it’s indeed much much better. I believe a 35mmF1.8 would work fine as well.
I guess most people would think that T17-28 + 50GM is a more sensible choice since 28mm is just too close to 35mm. But a 50mm is still too long for babies (I often have to lean back a little bit even with a 35mm). So let me explain why you need a shorter focal length:
1: When you’re looking after the baby by yourself you need to stay close to him.
2: The baby usually pays attention to people/subjects that’s closer to him. You want him to look at you then you better stay close.
3: When the baby was just born he’s lying in the crib for the most of the time. This means you often have to shoot from above. A shorter focal length is certainly more handy:
4: The baby stays mostly indoor for the first few months. The space could be too tight for a 50mm lens even though the baby is very small. Remember you need to buy a shit tons of baby stuffs that occupy a lot of room. Your home is probably going to be very messy.
5: As the baby’s mobility grows you need some fences to restrict him from those dangerous places. The fences also restrict you:
6: Now your baby can walk. He walks fast and he doesn’t need too much protection anymore. You’d imagine now it’s a good time to switch to a longer focal length. Well it’s not. Your baby doesn’t cooperate. Sometimes he’s looking at the north and sometimes the south. If you want the best angle then you need to move your own fat ass. And the longer focal length you have the longer distance you travel.
7: When you take your kids to a crowded playground and shoot from a distance the other parents could think that you’re violating their kids’ privacy, even though your fancy lens is going to blur everything. But if you just abuse the shutter next to your kid then nobody is going to say shit.
There are probably some other reasons that I couldn’t remember right now but here is my general feeling:
When your baby is still young there isn’t really a lot of occasions that only a 50/85mm lens would work, which implicates that you’re somehow separated from your baby. On the other hand there are a lot of occasions that they just don’t work.
Of course you can prepare a full set of lenses, but if you’re going to carry as few lenses as possible since now you have to carry a lot of other baby stuffs, then a shorter focal length is definitely the way to go. Sometimes even a 35mm lens would be too long. You need 24mm, 18mm or even a fisheye. So what’s wrong with the Tamron 17-28 or a Sony 16-35GM that cover more focal range? Well, the aperture.
As I said you’re often shooting indoor where you need to collect more light. Also the bokeh of F2.8 is just a bit lacking. Don’t get me wrong I’m definitely not a bokeh maniac. I often stop down my Sigma 56 to make the background more comprehensible and I also stop down the 35GM when shooting upper-body portraits. And 28mmF2.8 can produce some bokeh, otherwise I’d had replaced it a long time ago. But now that my son’s:
1: growing in size (55cm -> almost 90cm)
2: fully erected instead of sitting/crawling/curl up in bed
3: can run away pretty fast so I often need to crop the image a little bit
Getting a decent amount of bokeh just seems more and more difficult for the T17-28, so I do feel the desire to upgrade my kits a little bit.
The 35GM certainly doesn’t disappoint. Not gonna write a review because there are already tons of reviews out there, plus it’s almost perfect. Nothing to write about. The focus breathing is quite scary though, I almost thought that I’ve bought a zoom lens…
Anyway another advantage of a Sony lens is that the AF tracking is much better. Previously I was only shooting landscapes or my wife so I never feel that the AF of either Tamron 17-28 or Sigma 56 is incapable. They get the job done. Now with the baby the problem starts to show. The hit rate is still usable but I definitely want it to be better. There are also some rare occasions that I missed the moment completely and regretted not having a Sony lens.
BTW… previously I sold my 16-35GM and got the T17-28 instead and was super happy. But now my perspective has changed a little bit because of the baby. As I pointed out in my comparison review the biggest difference between the two lenses is not 16mm vs 17mm but rather 28mm vs 35mm. The GM gives you longer reach and noticeably more bokeh. And of course the AF tracking is better. If I still had the 16-35GM I guess I’ll be less motivated to buy the 35GM. The biggest problem for me personally is that 16-35GM is too long to be placed vertically in my Cosyspeed bag, so it’s difficult to carry two lenses together. T17-28 doesn’t have this issue. It fits nicely with the camera body plus the 35GM in the Cosyspeed.
Now I’ve improved my lenses, I do have some complaints about the camera body:
1: A7RIII’s eye-AF is outdated. You can easily feel it when you also have an RX100M7 with better AF algorithm. Your kid wears a hat or turn his head a little bit and suddenly your F1.4 lens becomes a focus-on-the-nose joke.
2: Not having a 2-axis screen is purely fucking abysmal. When you have a kid you’re going to shoot a crazy amount of low-angle vertical shots and it fucking kills me on a Sony body! Also now I’m more willing to take selfies with my kid. If Sony keeps doing this stupid shit like what they did on the A1 I’m going to switch to Canon at the right time. The IBIS is also much better!
So if you don’t care about resolution then the best Sony full frame camera for baby photography is actually… A7C?
It seems like Sony has significantly improved their lens design since the debut of 24GM. I don’t know what happened but Sony did bring us a winning streak with 24GM, 20G, 135GM, 12-24GM, 35GM etc. And the new coating used on the 12-24GM is probably one of the best on the market right now. However, neither Sony nor Canon, the old king of lens design, have announced anything in this time period that share similar specs for us to do a direct comparison, so it’s difficult to tell how good the “New Sony” actually is. Now we finally have the first head to head clash and I have to say… I do think Sony is the best lens maker right now.
The major distinguishing feature between Sony and Canon is how they autofocus. RF50 is basically a whole-group-focusing lens instead of front-group-focusing even though the rear group doesn’t move. I’m saying this because the AF motor even has to drag the aperture unit with the glasses together, which is exactly what the front-group-focusing lenses try to avoid.
Sony didn’t reveal the details of 50GM except that it has two internal focusing groups. But if you ask me to guess then it’s most probably like this:
As you can see Canon’s AF motor really has to do a lot more works than the Sony. The end results is as what Jared Polin said: you can hear and feel the autofocus of the RF50, but you can’t hear or feel the GM at all. Not only does RF50 makes sounds that can be picked up in your video and vibrates in you hand, it’s also a bit slower. It’s probably good enough for most everyday scenarios but when things get really serious just don’t expect it to track your subject as good as the GM. It’s not really a deal breaker for me (nobody is birding with a 50mm lens) but I never really appreciate these kind of lenses either.
So why did Canon use such kind of design? Well, a bigger focusing group usually gives you some extra advantages: smaller size, better optical quality, or larger magnification ratio. There are two companies that are particularly fond of it: Fujifilm and Canon. Most of Fuji’s fast primes in the early phase of the X-mount use front/rear-group-focusing and the XF35F1.4 is a whole-group-focusing lens. This helped them gaining a lot of fame in the beginning days of mirrorless since the camera itself was so slow with the contrast-based autofocus that nobody cared about how fast the lenses are. Sony on the other hand is an extreme advocate of lightweight focusing groups. They try to avoid front/rear/whole group design as much as possible and most of their lenses use only one element to change the focusing distance. They chose this strategy probably to optimize the video AF performance and also they knew that on-chip PDAF is going to be a thing in the near future. So later when A6000 became the hottest APS-C camera on the market Sony’s strategy started to pay off, and Fuji had to make a new F2 line to embrace the modern internal-focusing design due to all the criticisms.
Canon is still a big fan of big focusing group till this day. They did make some nice/unique lenses such as the RF35F1.8 or the RF85F2, both having a very impressive 0.5X magnification ratio. And it’s commonly used on their big guns to achieve the ultimate optical results. So before the announcement I didn’t really expect 50GM to deliver better results than the RF. I thought maybe it’s a fast tracker with some minor optical compromises. Well well well… I was wrong. It turned out that the 50GM actually has better loCA control and slightly less cat-eye effect than the RF. Maybe Canon also has some advantages here or there for example perhaps it’s sharper on a 150MP camera or has less focus breathing. But Sony does get the more important things right. Considering that Sony is: 1) using an internal focusing design 2) does have a smaller mount that makes it more difficult to design lenses in theory (although I can’t quantize how much impact it is, maybe it’s nothing at all), yet still achieving a better optical results together with the 170g lighter weight and $300 cheaper price tag, well, what can I say?
Oh yeah there is another 50F1.2 from Nikon… which is a joke so let’s not talk about it. You gotta wonder what Nikon is thinking these days.
If I have to say something negative about the GM then this lens is not what I really wished for. I’ve seen quite some 50ZA vs RF50 or 85GM vs RF85 comparisons to see if F1.2 really pays off. Unfortunately the difference is so subtle that you really have to stare at the photos for a while to see it. So I wouldn’t be too surprised if someone compares the 50ZA and the 50GM later and say: “huh… that’s it??” There might be some difference when pixel peeping at 100% but for most users the images they produce just look almost the same. In this case 50GM is just another 50ZA with similar weight/size and a much higher price tag. I would rather see Sony making a new 50mmF1.4 that improves the compactness/image quality and also lowers the price (like 35GM vs 35ZA) instead of pursuing some diminishing returns. But I guess Sony has an ego war to win here since there are lot of people saying that E-mount is too small. Some brain damaged Fuji fanboy even wrote a super long article predicting the death of E-mount. And Sony has won it beautifully. This is basically Sony’s “shut the fuck up” lens. lol
Hopefully Sony will also update some of their “tomorrow’s lens today” in the near future because they are kind of embarrassing with all the QA issues (70-200GM for starter…). And I’m really curious about what Sony’s lens designers can do these days!
IBIS is becoming Sony’s Achilles heel and we really should have a discussion about it. Discussion, not conclusion. I don’t have any firm conclusions but only observations and speculations.
It’s by no means the most scientific calculation but you can get some rough ideas. Below 200mm you only need to move the sensor by 2mm. Sony’s IBIS can easily do that. The maximum displacement in the horizontal/vertical direction is approximately 3mm just by eyeballing.
The graph also tells us that IBIS is quite useless for telephotos. It is good for lateral camera movements (horizontal, vertical, roll) but no so effective for angular movements (pitch and yaw). In contrary, OSS does nothing to lateral movements but is very good at correcting angular movements. You need both mechanisms to work together to achieve the optimal results. A good OSS also makes sure that the sensor doesn’t have to move too far. Win win. Sony also claims to have Dual IS but it seems to me that it works like shit. It’s not clear if there’s any body-lens communications regarding the stabilization. Their graph actually shows that the sensor is not participating the pitch/yaw compensation at all.
Do You Need A Larger Image Circle?
You’d imagine that the lenses must have a larger image circle to let the sensor move around. Actually… you remember the black corners of Sony FE24-105@24mm? Yep that thing doesn’t even cover the full frame image circle before the correction. It all depends on how long can the sensor’s corner stay in the image circle during the exposure. In most cases you’d expect it to stay there for at least a while to collect some light, so at the end you’ll probably just see more vignetting. In extreme cases though, the sensor corner could be hit by no or very little light, causing a bigger black corner:
Is this a problem? I would say no. Cropping a corner off is still far better than a bad stabilizer. Larger image circle is certainly nice to havebut it’s not essential to achieve a good IBIS. This also means that you don’t need a big mount. It should also be noted that since Sony’s sensor can move approximately 3mm toward each direction, you need a 42mm*32mm image circle to fully cover it. Designing medium format lenses for full frame bodies just doesn’t make too much senses.
So there really isn’t any show stopper. The body is big enough and the mount is big enough. Now the question is, what’s exactly wrong?
The Olympus Recipe
To answer this question I think we have to take a look at what other companies have done. In this video Olympus has shown two weapons:
They’ve developed their own motion sensors (gyroscopes and accelerometers) that are said to be 10 times more precise than the their old sensors. Sony’s gyro-based post-stabilization also yields good results so it looks like that their sensors are very capable at first glance. But if you think about it, videos are actually much less demanding than still photos. Your resolution drops to 4K or 1080P and your shutter speed never really exceeds 1/25s. Even if one or two frames are a little bit blurry on the pixel level most people won’t really notice it since it’s only shown for a fraction of a second. Now here is a graph about how precise the gyro sensors have to be to compensate for a 1-pixel shift on a 24MP sensor:
A 50mm lens already requires a 0.006 degrees resolution. To give you some idea, the gyro sensors in my smartphone only has a resolution of 0.06 degrees. So yeah… it looks like Olympus didn’t develop their own gyroscopes for nothing. I guess nothing on the market can meet their requirements within a reasonable price.
Another interesting thing is that Olympus has a dedicated processor just for the IBIS:
I don’t think the algorithm is really that complicated. This is probably a very simple chip. However using a dedicated chip (or a dedicated core in a chip) can be very beneficial to the speed. If your IBIS module has to share a processor with some other task modules then the processor must decide when to process which task. This procedure along makes things more complicated and probably adds a lot of delays into the system. If your IBIS module has a low priority in the system then it’s gonna be even worse. Maybe the processor has to handle the autofocus first. This is kind of like in the old days when iPhone looks significantly faster than Android because Apple prioritizes the UI rendering more than anything else. For example when you touch an APP on the screen, Apple first render an animation to let you know that you’ve indeed touched the APP icon, then try to load the APP. Android probably try to load the APP first and then show the animation. At the end both phones use the same amount of time to accomplish the task but iPhone just looks way faster.
I should emphasize that I’m not saying Sony’s processor is slow, but rather a universal processor always have a lot of other things to do (run an operating system for example) that most of the time your instructions are just waiting in the queue. It’s not being calculated at all.
A good circumstantial evidence is that Sony can achieve much better results in the gyro-based post-stabilization mode than the active mode. They are basically the same things, except that one is performed by the camera and the other by your computer. This indicates that there is indeed some delays in the system so that the in-body cropping can’t react to the camera motions in a timely fashion.
Anyway, these are only my educated guess. The important take here is that the mount diameter/body size has nothing or very little to do with the IBIS. It’s probably more about resolution and speed. So when will Sony fix this problem? Unfortunately Sony seems to be the only company that can actually increase their revenue in this pandemic. I guess we really need Canon to do a better job to bring out the best of Sony…
Photo Engineering 