The Nikon Z9 is Nikon’s eighth mirrorless camera and by far the most powerful to date. Thanks to Nikon Europe we were able to get our hands on one of these elusive beasts for a week in order to add support to Reikan FoCal.
This post is a few thoughts, photos, and FoCal results that we found interesting related to the Z9. It’s not a super-detailed analysis, although next time we have access to a Z9 there are a whole bunch of tests we’d like to run.
The post is organised into a few pages:
- Nikon Z9 Specs Overview – a look at some of the interesting specifications of the Z9, especially where relevant to FoCal
- Play Time! – we headed out to Durdle Door on the Jurassic Coast World Heritage Site to play around with the Z9 and FoCal
- FoCal Analysis – a look at some interesting results during our testing with the Z9, first Stabilisation Test results showing VR performance comparisons between the Z9 and the Z7, then MultiTest, showing some improvements across the aperture range with various AF Fine-tune values even on mirrorless cameras like the Z9
The latest version of FoCal – release 3.2 – is now available with full support for the Nikon Z9. All test results in this post were captured using FoCal 3.2 Pro.
1.3kg (3lb) of metal, plastic, glass and semiconductor. That’s all the Nikon Z9 is. But it’s all put together rather well, and leads to a pretty impressive set of specifications!
The image capture on the Z9 is handled by the 45.7 megapixel stacked (and therefore implicitly back-side illuminated) sensor. Back-side illumination improves the light capture performance of the sensor but since its introduction in full-frame cameras with the Sony A7rII back in 2015 the technology has become commonplace. Stacking the sensor by effectively placing two integrated circuits on top of each other – one for light capture and the second for logic – allows for faster data readout, reducing rolling shutter artefacts during video and allowing faster image capture rates.
Like quite a few previous Nikon cameras from the D850 through to the Z7ii , the Z9 sensor has an effective pixel count of 45.7 million pixels bringing a full-frame image resolution of 8,256×5,504 pixels. But interestingly the total number of pixels on the sensor is up from 46.89 million on the previous design to a whopping 52.37 million for the Z9 sensor. This is pretty unusual among cameras, and no-one outside of Nikon seems to be quite sure what all the extra pixels are for.
It’s common to have an excess of pixels for various reasons, one of which being that some are deliberately masked (blacked out) and used to measure the intrinsic noise and levels of the sensor in order to pre-process the captured image data and improve quality. But this normally amounts to maybe 1 million pixels at most – not the 6.5 million on the Z9.
Sensor Shift Stabilisation
Introduced with the Z6 and Z7 back in 2018, the Z9 also offers sensor-shift stabilisation. Using gyroscopes and accelerometers, the camera measures the tiny motions introduced by your hand-holding and compensates by fractionally moving the image sensor.
The Z9 sports 5-axis image sensor shift stabilisation, correcting:
- roll – rotation around the optical axis of lens
- yaw – rotation around the vertical axis of the camera
- pitch – rotation around the horizontal axis of the camera
- horizontal – side-to-side movement of the camera body
- vertical – up and down movement of the camera body
We did a fair amount of testing of the system using the FoCal Stabilisation Test so take a look at the FoCal Analysis section for more information.
Still Image Capture
There’s no mechanical shutter in the Z9, the readout from the sensor is purely electronic and the only mechanical parts around the sensor are the VR unit and the protection curtain. What this means is that you get a shutter speed as short as 1/32,000s, and at the long end all the way up to 900s (15 minutes).
The stacked sensor is no slouch at spitting its data out either. At full frame resolution – so 45.7 megapixel images – the camera can capture 20 frames per second of raw data for over 1,000 compressed raw images! If that isn’t enough, switch to JPEG and you’ll get 30 full-frame images per second, or drop down to 11 megapixel and you get… 120fps! One hundred and twenty 11 megapixel still frames every second!
On the assumption that no significant processing is done on the sensor that reduces the data rate, the readout is 45.7 million pixels x 30 per second, which is 1.37 gigapixels per second. Each pixel has 14-bits of information, so the total data rate off the sensor is a little over 19Gb/s.
The Expeed 7 processor quickly (well, very quickly!) compresses this data to either JPEG, resulting in image files up to about 17MB, or the new High Efficiency Raw format resulting in between 22 and 33MB maximum depending on the settings. The result is a need to push data to the card at around 500-600MB/s, very achievable with current fast CFExpress cards which offer write speeds of around 1500MB/s.
Sorry, your stash of CompactFlash or SD cards isn’t going to be much use here. The Z9 has 2 card slots, both for CFExpress Type B or XQD cards. The best SD cards top out at around 280MB/s write speed, and looking at the figures above for the required write speed at those high frame rates, that’s just not fast enough!
XQD version 2 cards are capable of up to 8Gb/s (1GB/s), while CFExpress Type B cards are capable of 2 times this speed. Both would be acceptable for the highest possible output from the Z9. However, it’s important to look at the sustained write speed the cards offer – not the headline burst rate shown on the card marketing material. This can differ widely, and it’s worth researching the cards to make sure you pick the most appropriate combination of cost and performance for you.
The Z9 offers 3 compression levels for raw recording: Starting with the uncompressed raw image data straight off the sensor, which amounts to almost exactly 80MB of data, the options area:
- Lossless Compression with a file size up to around 60MB (25% saving)
- High Efficiency*, file size up to around 33MB (58% saving)
- High Efficiency, file size up to around 22MB (73% saving)
The two High Efficiency modes use a new codec based on TicoRAW from intoPIX, which offers computationally efficient compression allowing the huge volume of data from the sensor to be compressed by the Expeed 7 processor fast enough to allow those headline burst rates.
Right now, FoCal does not support the raw format for the Z9. We will be adding support as soon as we can, but it’s worth noting that most people should not be using raw for analysis in FoCal – it’s slower, has limited comparison data and is only really needed for very specific cases (e.g. testing with Canon’s Dual-Pixel Raw, or for analysing separate colour channels).
A bit of background…
Like all mirrorless cameras, the Z9 has no separate autofocus sensor but instead uses the image sensor to determine the focus position. The sensor incorporates special phase-detect autofocus pixels, and uses a mix of phase-detect and contrast-detect under various situations to get the best compromise between focus speed and accuracy.
Phase-detect autofocus works by measuring the difference between two light paths through the lens and using this single measurement to determine the focus vector – the amount of focus required and the direction in which to focus. It’s quick as you get this vector from a single measurement, but the overall focus performance is based on the total autofocus system performance, which includes the optics, the mechanics and the electronics of the camera and lens. Fractionally misaligned lens mounts or internal lens elements, worn motor drive mechanics, optical aberrations and more can all lead to a suboptimal focus result.
Contrast-detect autofocus works in pretty much the same way as you do when you manually focus a lens. Look at the blurred image, pick a direction and move the focus ring a bit and see if the image gets better or worse. Once you get going, you know which direction to move in and you can hone in on the best focus position. However… you can’t know you’ve got the best position until you’ve gone past the best position, at which point you then have to come back again. Because of this, you can’t be 100% sure you’ve got exactly the best focus (as you’d have to move from it to find out), and the movement back and forth relies on the lens moving in very precise and repeatable ways – something that is certainly not the case for older F-mount lenses especially. All these moments of checking and driving the lens back and forth make contrast detect focus quite a bit slower than phase detect, but the end result is usually more accurate.
The Z9, like other Nikon mirrorless cameras, uses hybrid autofocus – so a combination of phase-detect and contrast-detect, although the latter slower but more accurate method is not often in “normal” circumstances.
Autofocus on the Z9
The Z9 had 493 discrete AF points laid out in a grid of 29 x 17 which can be manually selected. The outside row of points is not used for the auto modes, meaning 405 are available for use in the various auto-selection AF modes.
As the image sensor is used for autofocus, the processor can analyse the image content and the opportunity to focus on subjects rather than just points is available for mirrorless cameras. This is used to good effect in the Z9 – not only can you choose to focus on people’s faces or eyes, but it will also detect animals and vehicles for locking and tracking focus.
Like all other Nikon Z series cameras, the Z9 offers AF Fine-tune for all lenses. But why do you need to fine tune autofocus on a mirrorless camera?
As mentioned above, the hybrid autofocus system relies on a perfect autofocus system – not just the sensor. This system includes the optical and mechanical components of the camera and lens, and manufacturing tolerances, physical abuse and even temperature variations can affect the performance, and therefore the focus accuracy.
Many user of FoCal choose to upload their test results to Reikan, where we use them to build profiles of lens behaviour on all cameras and feed the information back in order to allow comparison with real-world users of the same equipment. With all this data, it seemed sensible to take a look at the AF Fine-tune results for Nikon mirrorless cameras, so we could definitively answer the question of whether you should calibrate lenses on a Z-camera.
We took all of the data uploaded from Nikon Z cameras (FoCal now works with all of them from the original Z6 and Z7 through the Z50, Z5, Z6ii, Z7ii, Zfc and now the Z9), filtered out any low quality results and then split between Z-mount lenses and adapted F-mount lenses. Each data set is a few thousand results, so has some decent statistical weight. Here’s the results for the Z-mount lenses:
Over 30% of Z-mount lenses need no calibration at all, with around 80% of all Z-mount lenses only needing 1 AF Fine-tune point of adjustment. Realistically, you probably don’t need to tune the autofocus (although take a look at our MultiTest Analysis section of this post to see why it could be useful). These lenses behave very well optically and mechanically, have just a single physical connection point (the Z mount) with a short distance between the optics and the sensor. The motors and mechanics used in the native Z lenses are new, high quality designs, and there’s generally no need to adjust by more than 1 or 2 AF Fine-tune points. Having said that, although a small adjustment of 1 or 2 AF Fine-tune points won’t make a significant difference to any single shot, statistically it could slightly shift the keeper-rate of a large batch of images.
For adapted F-mount lenses, the story is a little different. Add an FTZ adapter and you’ve introduced another physical connection and brought the lens optics away from the sensor. Both of these can introduce more significant deviations in the light path to the sensor. On top of this, the lens you attach may be 30 years old (the first lenses with integrated autofocus motors – required for autofocus using the FTZ adapter – were introduced in 1992). Up to 30 years of knocks and bangs, motor and mechanical wear, and old optical and mechanical design mean that not only may the phase-detect measurement be a little erroneous, but the instructions to drive the lens to a specific focus position could well end in a pretty poor focus result.
The numbers from FoCal users show things are a little different with adapted lenses. 60% of lenses require no calibration or a single AF Fine-tune point of adjustment, but that leaves 40% of all F-mount lenses requiring an adjustment which starts to immediately and obviously improve the image quality.
[Note: since writing this article, we’ve release the FoCal IQ Camera Database which lets you dig around in this sort of information for various cameras.]
With mirrorless cameras, you do remove the need to calibrate out any mechanical and optical variation between the image sensor and autofocus sensor that would be present on a DSLR using when viewfinder focusing. This leads to generally smaller AF Fine-tune requirements for adapted F-mount lenses on Nikon Z cameras, and this can be seen by comparing the mirrorless lens calibration requirements above with those of all Nikon DSLRs from our users (hundreds of thousands of results) below. It’s quite a different picture from the mirrorless cameras!
But mirrorless calibration requirement is not negligible. An AF Fine-tune offset of 5 units can make a significant difference to image quality of a single shot, and we’ve seen quite a few F-mount lenses which have needed this level of adjustment – the data in the charts above is showing hundreds of FoCal users immediately benefiting from calibration.
Fine Tune on the Z9
Operation of AF Fine-tune on the Z9 is almost the same as all other Nikon cameras that support tuning at both ends of the zoom range: from the SETUP menu, you choose AF Fine-tune then simply adjust the appropriate value. However, on the Z9, there is an extra confirmation step when you change a value that is already set for a lens:
The camera has 3 different methods by which is can be connected to the outside world for transferring data: USB, WiFi and Ethernet.
The first is a USB 3.2 Gen 1 USB-C interface capable of 5Gb/s transfer rate. This is the one we use for connecting to FoCal (at the moment… let us know if you’d like to see WiFi or Ethernet connection options in the comments section).
The nice thing about using USB-C is that it’s capable of power delivery as well – enable this in the menu, and you can power peripheral devices from the camera.
WiFi supports brings b, g, n and ac variants. 802.11ac has a typical maximum throughput of around 1600Mbps, but it’s unclear just how fast the transfer would be from the Z9.
At the bottom of the camera is an RJ45 Ethernet connector capable of Gigabit Ethernet speeds.
The supported Global Navigation Systems on the Z9 are GPS, GLONASS and QZSS. GPS is the one everyone knows about, run by the US government and available worldwide. GLONASS is a Russian system which can work better at higher latitudes than GPS. And QZSS is a Japan-specific augmentation system which, when combined with GPS, offers a higher accuracy level and the ability to use the system within typical Japanese cityscapes.
As well as logging a location for each photo taken, you can also sync the camera clock from the satellite system and also create a track log for any journey you do with your camera. One limitation, however, is that the system does not contain an electronic compass so there is no information stored about which direction a photo was taken.
The Z9 can use any of the EN-EL18 batteries (of which there are now 5 generations, the original being supplied back with the D4). The latest version – the EN-EL18d – supplied with the Z9 has a higher capacity of 35.6Wh compared with the 27Wh of previous generations. The supplied charger will only work with EN-EL18b or later batteries.
The supplied EN-EL18d version is rated for around 740 shots or 3 hours of movie capture, but is capable of capturing over 5,000 shots in burst mode where the LCD/EVF use makes up a far smaller chunk of the overall time.