Dot-Tune: Autofocus Fine Tuning in under 5 minutes
p.35 #1Dot-Tune: Autofocus Fine Tuning in under 5 minutes
snapsy wrote:
Can you share a reference from 2007 of this method being used? Nikon's implementation of their automated tuning method in the D5/D500 appears to function in the same manner as DotTune (based on the description provided in their tech sheet), so apparently Nikon finds value in it even if some don't share that opinion.I don't have the time to dig around for 8-9 year old forums posts with no easily searchable key word, sorry 
An employee at a local camera store was talking about this general method way back in the day when the D3 was still new and everyone was excited about AF fine tune, that's where I very first heard of it myself.
Nikon's automated implementation on the D5/D500 just compares CDAF to PDAF as far as I can tell, which has been the most basic way to do it since the beginning - there is no mention of viewfinder focus indicator unless perhaps I missed something. With the D5 apparently already out in some parts of the world, we should see the finer details on the process revealed soon enough. Automating the AFFT process and making it quick & easy will primarily be useful for quickly determining if you need to exchange a lens or not, and may be quick enough to be done in a camera store, which I think could be very useful.
I think lens returns & exchanges are going to go way up after the D5/D500 because customers are going to now have a numerical value generated by the camera itself to 'prove' a lens is out of calibration using an easily understandable quantitative scale. How I see this happening is they will run the calibration at a few combinations, and if the values are different across the range, it probably can't be reliably tuned and people will exchange it. As far as I know it still only stores a value for a single combination of focal length and distance which is again at best a band-aid solution for most people, especially if you're picky.Nikon's D5/D500 mechanism uses CDAF to focus and then likely cycles through all the PDAF AF tune values to find the one with the strongest PDAF focus confirmation. It's the same method used by DotTune but automated - the feedback from the focus confirmation is sampled directly by firmware from the same mechanism that drives the VF rangefinder display. It's how the automated version of DotTune in Magic Lantern works on Canon bodies - the ML firmware samples a firmware PDAF confirmation flag in memory, the same flag that is used to drive the VF confirmation dot.If it ends up working exactly like that, it sounds like they just automated the most obvious way to do it using a guess & check algorithm that people have been doing manually for years. Definitely still a time saver, and I am curious to read about it in detail. I wonder if it could also be done by taking pictures and comparing edge contrast, but I imagine that would be far more complex. Regardless, like I was saying, I think they are going to see lens returns & exchanges go way up! I can see it now... "The camera says the lens is out by -15, I need to exchange please."Many times the best inventions are those which are obvious in retrospect. If it was obvious to Nikon they would have implemented it years ago. Btw Canon has a patent on a similar technique. There's no need to take pictures because the PDAF differential can be sampled from the AF sensor once perfect focus has been established (either via CDAF or manual focus).
I understand that, I was just thinking out loud because I was curious if they could also do it by comparing actual photos, necessary or not I'm sure it's faster to just do it with the hardware as it is likely done.
CanadaMark wrote
snapsy wrote
CanadaMark wrote
snapsy wrote
CanadaMark wrote
I would agree and I find it's a band-aid solution at best. People have been using this method since ~2007 albeit not with the same catchy name. If the gear cannot be calibrated perfectly at all combinations of focal length and subject distance (extremely rare for a zoom), it should be exchanged or sent along with the body for factory calibration in my opinion. An employee at a local camera store was talking about this general method way back in the day when the D3 was still new and everyone was excited about AF fine tune, that's where I very first heard of it myself. Nikon's automated implementation on the D5/D500 just compares CDAF to PDAF as far as I can tell, which has been the most basic way to do it since the beginning - there is no mention of viewfinder focus indicator unless perhaps I missed something. With the D5 apparently already out in some parts of the world, we should see the finer details on the process revealed soon enough. Automating the AFFT process and making it quick & easy will primarily be useful for quickly determining if you need to exchange a lens or not, and may be quick enough to be done in a camera store, which I think could be very useful.
I think lens returns & exchanges are going to go way up after the D5/D500 because customers are going to now have a numerical value generated by the camera itself to 'prove' a lens is out of calibration using an easily understandable quantitative scale. How I see this happening is they will run the calibration at a few combinations, and if the values are different across the range, it probably can't be reliably tuned and people will exchange it. As far as I know it still only stores a value for a single combination of focal length and distance which is again at best a band-aid solution for most people, especially if you're picky.
I understand that, I was just thinking out loud because I was curious if they could also do it by comparing actual photos, necessary or not
I'm sure it's faster to just do it with the hardware as it is likely done.Mar 17, 2016 at 02:12 PM
p.35 #2Dot-Tune: Autofocus Fine Tuning in under 5 minutes
woutgeo wrote:
When I did dot-tune I noticed that I had lenses with quite small ranges (6 steps) of acceptable focus and others with much larger ranges (14 steps). Does this have any meaning?
Lenses with greater DOF will be less sensitive to fine tune values, but without more info I'm not sure exactly if that applies to your situation.
When I did dot-tune I noticed that I had lenses with quite small ranges (6 steps) of acceptable focus and others with much larger ranges (14 steps). Does this have any meaning?
Lenses with greater DOF will be less sensitive to fine tune values, but without more info I'm not sure exactly if that applies to your situation.
Mar 17, 2016 at 02:13 PM
p.35 #3Dot-Tune: Autofocus Fine Tuning in under 5 minutes
CanadaMark wrote:
Lenses with greater DOF will be less sensitive to fine tune values, but without more info I'm not sure exactly if that applies to your situation.
I should have clarified that this was happening at the same f-stop. Specifically, I noticed a big difference between my 100mm f/2.8L Macro and my 14mm f/2.8L (both set up at 50x). The only thing I could chalk this up to was user error.
woutgeo wrote
When I did dot-tune I noticed that I had lenses with quite small ranges (6 steps) of acceptable focus and others with much larger ranges (14 steps). Does this have any meaning?I should have clarified that this was happening at the same f-stop. Specifically, I noticed a big difference between my 100mm f/2.8L Macro and my 14mm f/2.8L (both set up at 50x). The only thing I could chalk this up to was user error.
Mar 17, 2016 at 02:17 PM
p.35 #4Dot-Tune: Autofocus Fine Tuning in under 5 minutes
woutgeo wrote:
When I did dot-tune I noticed that I had lenses with quite small ranges (6 steps) of acceptable focus and others with much larger ranges (14 steps). Does this have any meaning?
It has meaning and relates to how the phase-differential sampling can be affected by different lenses and optics (focal length, contrast differential between focused and slightly OOF, etc...). Fortunately it doesn't have significance for DotTune because as long as the confirmed ranged fits within the full +/-20 range the midpoint of the range will still represent the optimal AF tune value irrespective of how large the range is.
When I did dot-tune I noticed that I had lenses with quite small ranges (6 steps) of acceptable focus and others with much larger ranges (14 steps). Does this have any meaning?
It has meaning and relates to how the phase-differential sampling can be affected by different lenses and optics (focal length, contrast differential between focused and slightly OOF, etc...). Fortunately it doesn't have significance for DotTune because as long as the confirmed ranged fits within the full +/-20 range the midpoint of the range will still represent the optimal AF tune value irrespective of how large the range is.
Mar 17, 2016 at 02:18 PM
p.35 #5Dot-Tune: Autofocus Fine Tuning in under 5 minutes
snapsy wrote:
It has meaning and relates to how the phase-differential sampling can be affected by different lenses and optics (focal length, contrast differential between focused and slightly OOF, etc...). Fortunately it doesn't have significance for DotTune because as long as the confirmed ranged fits within the full +/-20 range the midpoint of the range will still represent the optimal AF tune value irrespective of how large the range is.
Does this mean there *should* be a standard range for a specific lens/body combination? In other words, should the range of my 100mm f/2.8 on a 5D MK2 be the same as someone else's? If so, does this simplify the process even more, allowing you to just find one end of the range?
woutgeo wrote
When I did dot-tune I noticed that I had lenses with quite small ranges (6 steps) of acceptable focus and others with much larger ranges (14 steps). Does this have any meaning?Does this mean there *should* be a standard range for a specific lens/body combination? In other words, should the range of my 100mm f/2.8 on a 5D MK2 be the same as someone else's? If so, does this simplify the process even more, allowing you to just find one end of the range?
Mar 17, 2016 at 02:24 PM
p.35 #6Dot-Tune: Autofocus Fine Tuning in under 5 minutes
woutgeo wrote:
It has meaning and relates to how the phase-differential sampling can be affected by different lenses and optics (focal length, contrast differential between focused and slightly OOF, etc...). Fortunately it doesn't have significance for DotTune because as long as the confirmed ranged fits within the full +/-20 range the midpoint of the range will still represent the optimal AF tune value irrespective of how large the range is.Does this mean there *should* be a standard range for a specific lens/body combination? In other words, should the range of my 100mm f/2.8 on a 5D MK2 be the same as someone else's? If so, does this simplify the process even more, allowing you to just find one end of the range?
I suspect the range would be similar for a given body+lens combination but only if the focus distance and target was precisely the same, which would be hard to achieve in practice. I don't think it would help the process much since the end of the range for a specific body+lens copy would still need to be arrived at empirically.
snapsy wrote
woutgeo wrote
When I did dot-tune I noticed that I had lenses with quite small ranges (6 steps) of acceptable focus and others with much larger ranges (14 steps). Does this have any meaning?I suspect the range would be similar for a given body+lens combination but only if the focus distance and target was precisely the same, which would be hard to achieve in practice. I don't think it would help the process much since the end of the range for a specific body+lens copy would still need to be arrived at empirically.
Mar 17, 2016 at 02:26 PM
p.35 #7Dot-Tune: Autofocus Fine Tuning in under 5 minutes
snapsy wrote:
It has meaning and relates to how the phase-differential sampling can be affected by different lenses and optics (focal length, contrast differential between focused and slightly OOF, etc...). Fortunately it doesn't have significance for DotTune because as long as the confirmed ranged fits within the full +/-20 range the midpoint of the range will still represent the optimal AF tune value irrespective of how large the range is.Does this mean there *should* be a standard range for a specific lens/body combination? In other words, should the range of my 100mm f/2.8 on a 5D MK2 be the same as someone else's? If so, does this simplify the process even more, allowing you to just find one end of the range?I suspect the range would be similar for a given body+lens combination but only if the focus distance and target was precisely the same, which would be hard to achieve in practice. I don't think it would help the process much since the end of the range for a specific body+lens copy would still need to be arrived at empirically.
And lighting, I suspect.
woutgeo wrote
snapsy wrote
woutgeo wrote
When I did dot-tune I noticed that I had lenses with quite small ranges (6 steps) of acceptable focus and others with much larger ranges (14 steps). Does this have any meaning?And lighting, I suspect.
Mar 17, 2016 at 02:28 PM
p.35 #8Dot-Tune: Autofocus Fine Tuning in under 5 minutes
woutgeo wrote:
It has meaning and relates to how the phase-differential sampling can be affected by different lenses and optics (focal length, contrast differential between focused and slightly OOF, etc...). Fortunately it doesn't have significance for DotTune because as long as the confirmed ranged fits within the full +/-20 range the midpoint of the range will still represent the optimal AF tune value irrespective of how large the range is.Does this mean there *should* be a standard range for a specific lens/body combination? In other words, should the range of my 100mm f/2.8 on a 5D MK2 be the same as someone else's? If so, does this simplify the process even more, allowing you to just find one end of the range?I suspect the range would be similar for a given body+lens combination but only if the focus distance and target was precisely the same, which would be hard to achieve in practice. I don't think it would help the process much since the end of the range for a specific body+lens copy would still need to be arrived at empirically.And lighting, I suspect.
Definitely - lighting has a significant effect on the result, particularly the lighting temperature but also the diffuseness of the lighting as well. High-contrast lighting produces the worst result because it creates aberrations off the target.
snapsy wrote
woutgeo wrote
snapsy wrote
woutgeo wrote
When I did dot-tune I noticed that I had lenses with quite small ranges (6 steps) of acceptable focus and others with much larger ranges (14 steps). Does this have any meaning?Definitely - lighting has a significant effect on the result, particularly the lighting temperature but also the diffuseness of the lighting as well. High-contrast lighting produces the worst result because it creates aberrations off the target.
Mar 17, 2016 at 02:58 PM
p.35 #9Dot-Tune: Autofocus Fine Tuning in under 5 minutes
I'm using studio CFL's that have a temp of 5,400.
Mar 17, 2016 at 03:05 PM
p.35 #10Dot-Tune: Autofocus Fine Tuning in under 5 minutes
Can I ask a quick question about one of the steps in the dot tune method? In the step where you switch to live view, use autofocus on a target and then adjust the focus manually in live view at 10x, Would we not want to leave the autofocus as the camera determines it and then move to the step where you adjust the afma to find each end of the spectrum? It seems like if I manually adjust the focus point I am defeating the purpose of the process, though I could be wrong here just hoping for some clarification.
Thanks!
Thanks!
Mar 18, 2016 at 11:08 AM
|
p.35 #11Dot-Tune: Autofocus Fine Tuning in under 5 minutes
Noel West wrote:
Can I ask a quick question about one of the steps in the dot tune method? In the step where you switch to live view, use autofocus on a target and then adjust the focus manually in live view at 10x, Would we not want to leave the autofocus as the camera determines it and then move to the step where you adjust the afma to find each end of the spectrum? It seems like if I manually adjust the focus point I am defeating the purpose of the process, though I could be wrong here just hoping for...Show more →
You can use LV AF or LV manual focus or a combination of the two. Whatever you feel is the best method to get precise focus before starting the AF tune scaling steps.
Can I ask a quick question about one of the steps in the dot tune method? In the step where you switch to live view, use autofocus on a target and then adjust the focus manually in live view at 10x, Would we not want to leave the autofocus as the camera determines it and then move to the step where you adjust the afma to find each end of the spectrum? It seems like if I manually adjust the focus point I am defeating the purpose of the process, though I could be wrong here just hoping for...Show more →
You can use LV AF or LV manual focus or a combination of the two. Whatever you feel is the best method to get precise focus before starting the AF tune scaling steps.
Mar 18, 2016 at 11:13 AM
p.35 #12Dot-Tune: Autofocus Fine Tuning in under 5 minutes
woutgeo wrote:
Lenses with greater DOF will be less sensitive to fine tune values, but without more info I'm not sure exactly if that applies to your situation.I should have clarified that this was happening at the same f-stop. Specifically, I noticed a big difference between my 100mm f/2.8L Macro and my 14mm f/2.8L (both set up at 50x). The only thing I could chalk this up to was user error.
At 14mm ( 50x ) you need to be at 0.7m or 2.29 feet from your target.
At 100mm ( 40x ) you need to be at 4m or 13.12 feet from your target.
That should get you better result for M.A.
Francois
CanadaMark wrote
woutgeo wrote
When I did dot-tune I noticed that I had lenses with quite small ranges (6 steps) of acceptable focus and others with much larger ranges (14 steps). Does this have any meaning?At 14mm ( 50x ) you need to be at 0.7m or 2.29 feet from your target.
At 100mm ( 40x ) you need to be at 4m or 13.12 feet from your target.
That should get you better result for M.A.
Francois
Mar 18, 2016 at 12:59 PM
p.35 #13Dot-Tune: Autofocus Fine Tuning in under 5 minutes
I just installed magic lantern on my 6d but dot tune is not in the menu. Do I maybe have the wrong version of ML? Can anyone point me in the right direction?
Mar 18, 2016 at 01:34 PM
p.35 #14Dot-Tune: Autofocus Fine Tuning in under 5 minutes
At 100mm ( 40x ) you need to be at 4m or 13.12 feet from your target.
Huh. Can you explain why it's better to be at 40x for the 100mm lens? Thanks!
Mar 18, 2016 at 02:28 PM
p.35 #15Dot-Tune: Autofocus Fine Tuning in under 5 minutes
woutgeo wrote:
Huh. Can you explain why it's better to be at 40x for the 100mm lens? Thanks!
Here a simple explication from that link using those scale give's me perfect focus . M.A. many lens and it work's great
http://s449182328.websitehome.co.uk/focal/dl//Docs/FoCal%20Test%20Distance_1.1.pdf
Francois
Earlier quote
At 100mm ( 40x ) you need to be at 4m or 13.12 feet from your target.Here a simple explication from that link using those scale give's me perfect focus . M.A. many lens and it work's great
http://s449182328.websitehome.co.uk/focal/dl//Docs/FoCal%20Test%20Distance_1.1.pdf
Francois
Mar 18, 2016 at 03:08 PM
p.35 #16Dot-Tune: Autofocus Fine Tuning in under 5 minutes
qc_mountain wrote:
Huh. Can you explain why it's better to be at 40x for the 100mm lens? Thanks!Here a simple explication from that link using those scale give's me perfect focus . M.A. many lens and it work's great
http://s449182328.websitehome.co.uk/focal/dl//Docs/FoCal%20Test%20Distance_1.1.pdf
Francois
I use that link as well.
woutgeo wrote
Earlier quote
At 100mm ( 40x ) you need to be at 4m or 13.12 feet from your target.http://s449182328.websitehome.co.uk/focal/dl//Docs/FoCal%20Test%20Distance_1.1.pdf
Francois
I use that link as well.
Mar 18, 2016 at 03:10 PM
p.35 #17Dot-Tune: Autofocus Fine Tuning in under 5 minutes
It's there, I just used it on my 6D. You have to enable it in the setup menu, then go to the control menu to use it. It works quite well.
kkroeker wrote:
I just installed magic lantern on my 6d but dot tune is not in the menu. Do I maybe have the wrong version of ML? Can anyone point me in the right direction?
kkroeker wrote:
I just installed magic lantern on my 6d but dot tune is not in the menu. Do I maybe have the wrong version of ML? Can anyone point me in the right direction?
Mar 23, 2016 at 08:54 AM
p.35 #18Dot-Tune: Autofocus Fine Tuning in under 5 minutes
Snapsy,
IMHO, one of the novel, yet obvious after the fact parts of your dottune method, was the concept of identifying the Range for PD AF confirmation. I would guess that AF confirmation is a threshold, hence no true peak created by pd AF screen signal(imho). If true, then any automation using PD AF signal with 'walking' mfa values would need to also ideal tiny the range.
Now, a more perfect way IS possible in current processor right cameras. Two digital images can be digitally compared through a correlation algorithm of the image by taking the known shape image and comparing it to the walking 1s images. Through correlation, a true peak MFA value can be identified for that FL and distance.
Now that I mention this correlation method, It would be interesting to do this off line (using a PC) to see if the mathematical correlation Peak really does correspond to the center of the range. (it might be more related to the true ratio of front and back dof from Live View optics dof.
snapsy wrote:
I would agree and I find it's a band-aid solution at best. People have been using this method since ~2007 albeit not with the same catchy name. If the gear cannot be calibrated perfectly at all combinations of focal length and subject distance (extremely rare for a zoom), it should be exchanged or sent along with the body for factory calibration in my opinion.Can you share a reference from 2007 of this method being used? Nikon's implementation of their automated tuning method in the D5/D500 appears to function in the same manner as DotTune (based on the description provided in their tech sheet), so apparently Nikon finds value in it even if some don't share that opinion.I don't have the time to dig around for 8-9 year old forums posts with no easily searchable key word, sorry An employee at a local camera store was talking about this general method way back in the day when the D3 was still new and everyone was excited about AF fine tune, that's where I very first heard of it myself.
Nikon's automated implementation on the D5/D500 just compares CDAF to PDAF as far as I can tell, which has been the most basic way to do it since the beginning - there is no mention of viewfinder focus indicator unless perhaps I missed something. With the D5 apparently already out in some parts of the world, we should see the finer details on the process revealed soon enough. Automating the AFFT process and making it quick & easy will primarily be useful for quickly determining if you need to exchange a lens or not, and may be quick enough to be done in a camera store, which I think could be very useful.
I think lens returns & exchanges are going to go way up after the D5/D500 because customers are going to now have a numerical value generated by the camera itself to 'prove' a lens is out of calibration using an easily understandable quantitative scale. How I see this happening is they will run the calibration at a few combinations, and if the values are different across the range, it probably can't be reliably tuned and people will exchange it. As far as I know it still only stores a value for a single combination of focal length and distance which is again at best a band-aid solution for most people, especially if you're picky.Nikon's D5/D500 mechanism uses CDAF to focus and then likely cycles through all the PDAF AF tune values to find the one with the strongest PDAF focus confirmation. It's the same method used by DotTune but automated - the feedback from the focus confirmation is sampled directly by firmware from the same mechanism that drives the VF rangefinder display. It's how the automated version of DotTune in Magic Lantern works on Canon bodies - the ML firmware samples a firmware PDAF confirmation flag in memory, the same flag that is used to drive the VF confirmation dot.
IMHO, one of the novel, yet obvious after the fact parts of your dottune method, was the concept of identifying the Range for PD AF confirmation. I would guess that AF confirmation is a threshold, hence no true peak created by pd AF screen signal(imho). If true, then any automation using PD AF signal with 'walking' mfa values would need to also ideal tiny the range.
Now, a more perfect way IS possible in current processor right cameras. Two digital images can be digitally compared through a correlation algorithm of the image by taking the known shape image and comparing it to the walking 1s images. Through correlation, a true peak MFA value can be identified for that FL and distance.
Now that I mention this correlation method, It would be interesting to do this off line (using a PC) to see if the mathematical correlation Peak really does correspond to the center of the range. (it might be more related to the true ratio of front and back dof from Live View optics dof.
snapsy wrote:
CanadaMark wrote
snapsy wrote
CanadaMark wrote
thepiecesfit wrote
Ive found dot tune to be a bit inconsistent, even the automated version in Magic Lantern. The easiest way is to send your gear to the manufacturer sadly. My 5d3 85 1.2 and 135 brought into Canon and all came back perfect at 0 AMFA. An employee at a local camera store was talking about this general method way back in the day when the D3 was still new and everyone was excited about AF fine tune, that's where I very first heard of it myself. Nikon's automated implementation on the D5/D500 just compares CDAF to PDAF as far as I can tell, which has been the most basic way to do it since the beginning - there is no mention of viewfinder focus indicator unless perhaps I missed something. With the D5 apparently already out in some parts of the world, we should see the finer details on the process revealed soon enough. Automating the AFFT process and making it quick & easy will primarily be useful for quickly determining if you need to exchange a lens or not, and may be quick enough to be done in a camera store, which I think could be very useful.
I think lens returns & exchanges are going to go way up after the D5/D500 because customers are going to now have a numerical value generated by the camera itself to 'prove' a lens is out of calibration using an easily understandable quantitative scale. How I see this happening is they will run the calibration at a few combinations, and if the values are different across the range, it probably can't be reliably tuned and people will exchange it. As far as I know it still only stores a value for a single combination of focal length and distance which is again at best a band-aid solution for most people, especially if you're picky.
Mar 23, 2016 at 10:06 AM
p.35 #19Dot-Tune: Autofocus Fine Tuning in under 5 minutes
Noel West wrote:
Thanks for the reply Schlotz, I have been following those instructions. It is interesting that on my 1D Mk IV body where I am getting images quite out of focus at 400-600mm and at great distances that the camera focuses incorrectly in quick mode but gets much closer (if not perfect) using the live view focus mode. Must be as it is using contrast detection for focus as opposed to the af chip in quick mode.
It is difficult to fine tune focus using live view outdoors and at long distances as you get a lot of camera movement, I...Show more →
Noel,
(I haven't finished reading the thread but thought it would be worth mentioning in case it hasn't.)
Live view, when zoomed in to 10x, can still do Contrast AF so you know exactly where it is AF'ing. You don't want to do Quick Mode in live view as it does phase detect which you are adjusting. Mfa has no affect on Live View Contrast Detect.
For your specific focal lengths, why are you using multiple distances? Distance is known to affect focus, most people choose their most common distance for each FL.
Definitely get the new firmware for your Sigma. The amount of time you have spent is mind boggling. This Dottune method can usually be done in about 5 to 10 minutes using a stationary object with good contrast. You don't need fancy targets, just am object with reliable edges for the PD to work and contrast for the CDAF to work. Snaps's alias is very apprapo!
Mar 23, 2016 at 10:19 AM
p.35 #20Dot-Tune: Autofocus Fine Tuning in under 5 minutes
snapsy wrote:
CanadaMark wrote
thepiecesfit wrote
Ive found dot tune to be a bit inconsistent, even the automated version in Magic Lantern. The easiest way is to send your gear to the manufacturer sadly. My 5d3 85 1.2 and 135 brought into Canon and all came back perfect at 0 AMFA.It sounds like the two of you might bereferring to two different things back in 2007. Sands, I Think CanadaMark was referring to just MFA capabilities on cameras back in 2007, not the dottune method.
Mar 23, 2016 at 10:29 AM