On the web there is quite some confusion now about the kind of repairs that are offered to repair the D800 auto focus issues by the different Nikon Service Centers.
My previous repair resulted in my 14-24/28 lens to show some unsharpness at 14mm, so I made an appointment last Thursday to have that fixed. After the check up of my 24-70 (extremely good) and my 14-24 (some adjustment), I was offered a tour through the Dutch Service Center with some explanation of the way autofocus is handled by them. Please be aware I’m not an expert in optical physics, so what follows is a summary of what I understood.
2. Repair overview
The repair procedure has been developed by Evert van Stijn, a technician at the Nikon Service Points in the Netherlands. He has invented this procedure in the past, for a Nikon D2(x or h), that did show AF problems as well. The repair consist of several stages:
- Align the sensor and bayonet
- Determine and set the distance of the sensor to the bayonet
- Adjust the angle and of the first and second mirror
- Straighten the AF CAM module
- Determine the relative displacement of the AF points by software
- AF-fine tune lenses
- Final checks
2.1 Align sensor and bayonet
In the days of 35mm film, the bayonet and the film could easily be aligned by using a dial indicator. The camera is laid down on its bayonet, the backdoor is opened, and the film holder is measured at several places to see if it has the same distance. To measure the sensor tilt in a digital camera, it needs to be dismantled. This gives access to some reference points on the sensor PCB, so that the relative sensor distance can be measured.
To be able to measure the sensor alignment to the bayonet without opening the camera, or to determine whether an AF focus module is tilted, a big setup has been made with a tightly fixed and calibrated lens that points towards a scene with 4 bended scales in the corners. By attaching a camera to this lens, one can see from the 4 bended scales whether the bayonet and sensor/AF module are misaligned.
2.2 Determine and set the distance of the sensor to the bayonet
In case the sensor and bayonet are straight, the exact distance may still be a bit off. To be able to determine the exact distance, a manual focus lens, calibrated to the distance is being used. The lens was attached to a large tube with a kind of microscopic viewfinder (I think it is a collimator), where you could see green chroma, magenta chroma or no chroma at all when turning the focus ring. This principle can be used to determine the pixel sharpness on the sensor.
2.3 Adjust the angle of the first and second mirror
By using a laser-based mirror angle-inspection tool (developed by a service centre in Germany, and refined and issued to all service centers by Nikon), the primary mirror can be checked for right-left and up-down deviations and put under and angle of exactly 45 degrees. Secondly, the secondary mirror can be adjusted to 54 degrees, by checking for up-down deviations. (picture from the web, click on picture to see the source)
2.4 Straighten the AF CAM module
The AF CAM module can also be aligned, with screws and washers. I do not have much details on this specific phase.
2.5 Determine the relative displacement of the AF points by software
The camera is attached to a computer, and directed towards an illumination box at 1m distance holding a card with vertical lines. A computer program determines the autofocus offset per AF field, and puts these offsets in a flash/EEPROM memory as compensation values.
UPDATE: The software consists of 3 stages: (1) it checks for dust on the AF sensor, (2) it determines the offset per AF-field and stores this in the memory, (3) it re-checks the new values by doing a measurements again. I’ve learned that the software only checks the AF module. During this test NO photos are being taken, so only the AF sensor is checked for asymmetric behavior, and no check is done with its relative behavior towards the sensor. This explains why next to this software step AF-fine tuning may still be required after the repair (unless this is compensated by the repair as well)!
UPDATE 2: A reference D800 camera (sent by Nikon) is used to analyse the software test setup that is used to re-program the AF field offsets. This reference camera is used to determine the measured values for all the AF points for the specific test setup of the service centre, and these values are stored in the software. So it essentially calibrates the test-setup. Subsequently, a victim camera goes through the test, the measured values are compared to the values measured by the reference camera, and adapted accordingly.
UPDATE 3: In some cases, additional offsets will be applied to the software calibration setup, effectively pushing the outer AF fields even more towards a certain direction. The test setup to perform the software calibration is carried out with a 50mm/1.4D lens (@f2.2) at a distance of 1m to the test card, a situation where not all deltas may be discovered at a sufficient resolution to determine the right compensation values. This means that with certain lenses, and especially at larger distances, focus issues may still be present inside the camera. The additional offset of focus points is performed by means of editing the offsets in a reference data file created by the reference camera. NSP has created multiple versions, from moderate to aggressive. Depending on the calibration results (that need to be checked by hand), they choose a certain hand-made setting and re-calibrate again, until satisfactory results have been accomplished.
2.6 AF fine-tune setting (optional)
A slant-chart is used to determine the AF-fine tune setting per lens (UPDATE: this is not a standard part of the repair procedure; only when a camera-lens combination needs to be checked, normally done by the consumer himself). The default lens for Nikon is the 50mm/1.4D, at a f2.2 setting. (picture from the web, click on picture to see the source)
2.7 Final Checks
An autofocus reference card is used to check and judge the overall result.
3. Other type of repairs
On the forums, and on Ming’s blog or Mansurovs’s blog repairs are mentioned without mechanical adjustments. These repairs do mention the use of a laser and jig, so we suspect that only stage 3 (mirror alignment) and stage 5 (software calibration) are carried out in those cases.
As can be seen, the repair consists of a mechanical part, and a software part. The mechanical part first makes sure the sensor and bayonet are well aligned, the mirror is well aligned, and then the software kicks in to register the relative differences that are still left.