Man, if anyone has any ideas on how to test for the number of AF steps please pm them to me. I've tried everything (including taking out sensors and looking them up in catalogue).
RCicala wrote:
Man, if anyone has any ideas on how to test for the number of AF steps please pm them to me. I've tried everything (including taking out sensors and looking them up in catalogue).
I can only think that a single MA step equates to a lens step, so it would be down to measuring focus shift as MA is adjusted. Not exactly something that is repeatable. (and likely not representative of the number of af steps)
I cannot wait till all lenses work like stepper motors.
SloPhoto wrote:
I can only think that a single MA step equates to a lens step, so it would be down to measuring focus shift as MA is adjusted. Not exactly something that is repeatable. (and likely not representative of the number of af steps)
I cannot wait till all lenses work like stepper motors.
Yeah, there is only an indirect correlation between AF tune steps and lens stepper motor steps. The AF tune value is actually a bias that feeds into the PDAF detection logic on the input side; it's not an output bias that gets added to the "focus to xxx" command sent by the body.
snapsy wrote:
Yeah, there is only an indirect correlation between AF tune steps and lens stepper motor steps. The AF tune value is actually a bias that feeds into the PDAF detection logic on the input side; it's not an output bias that gets added to the "focus to xxx" command sent by the body.
Thanks, I was hoping you would chime in. I presumed that was it (unfortunately). Can you tell how much I wished it was an offset
But then I cannot figure out why the focus sweep between -20 and 20 on a 70-200 is small, and on the sigma85 you could just about hide a house in the focus sweep.
Do usm/hsm systems at least use an optical encoder to determine distance?
RCicala wrote:
Man, if anyone has any ideas on how to test for the number of AF steps please pm them to me. I've tried everything (including taking out sensors and looking them up in catalogue).
Would something less scientific work, like strapping a sensitive microphone to the housing and moving through the range at a constant rate?
carstenw wrote:
Would something less scientific work, like strapping a sensitive microphone to the housing and moving through the range at a constant rate?
Tried that, and tried an oscilloscope. RED released the number of steps for certain lenses, but I'm trying (with no success) to find similar information for cameras. Lenses can vary from 550 to 2700 steps that I know of, but I can't find any way to tell how many various cameras can process.
RCicala wrote:
Man, if anyone has any ideas on how to test for the number of AF steps please pm them to me. I've tried everything (including taking out sensors and looking them up in catalogue).
Not sure if there's a way to mechanize/automate the process of triggering a single step of focus to the body. How about a rig like slrgear's that moves the target away in precise increments on a dolly and manually focusing for each step? You can distinguish the steps by correlating the measured MTFs. Sounds like a lota work ;(
RCicala wrote:
Tried that, and tried an oscilloscope. RED released the number of steps for certain lenses, but I'm trying (with no success) to find similar information for cameras. Lenses can vary from 550 to 2700 steps that I know of, but I can't find any way to tell how many various cameras can process.
Usm rings have holes right? Could those be used for encoder ticks? If so that may be more relevant with a usm system.
Robert, thanks for your report. Although I'm a Nikon fanboy, my gut told me to preorder this lens and I'm glad I did. So far this has convinced me I made the right decision not to order Nikon as usual.
The real question is the bokeh. Even if it is sharp as a tack if the bokeh is busy, it may not be a good choice. I think people who prefer f1.4 over f2 are the ones who prefer shallow DOF. My unscientific assessment of Nikon 35g was that its bokeh was smoother than Canon's 35L.
Sigma should enjoy why they can. Nikon will probably try to release a 35mm f2 or or 1.8 at a much lower price to try and take some wind out of sigma's sale. i'm sure at this point nikon has sold the lions share of 35mm 1.4's and will be looking to maintain market share for the 35mm focal length. Because at $899 a 1.4 sounds very nice, but a 35mm 1.8 or f2 at $399 would sound much better :-)
I really doubt nikon care much about sigma, they just keep releasing lenses at ever higher prices and people keep buying them because of the nikon badge.
Desmond79 wrote:
Sigma should enjoy why they can. Nikon will probably try to release a 35mm f2 or or 1.8 at a much lower price to try and take some wind out of sigma's sale. i'm sure at this point nikon has sold the lions share of 35mm 1.4's and will be looking to maintain market share for the 35mm focal length. Because at $899 a 1.4 sounds very nice, but a 35mm 1.8 or f2 at $399 would sound much better :-)
If I had a crystal ball and knew Nikon's plan, I would wait for their FF version of the 35/1.8 or 2 if it were really priced at $399.
carstenw wrote:
Would something less scientific work, like strapping a sensitive microphone to the housing and moving through the range at a constant rate?
Nope. What you'd hear then is just motor drive noise. The actual encoder resolution is (nowadays):
MR tape pulse resolution x encoder length.
or
optical guide resolution x encoder length
Very few new systems use mechanical or electrical resistance slider systems. You use either a MR head and a premagnetized substrate, or an optical system (but that's a lot more sensitive to dirt and mechanical misalignments...). Older systems like for instance the EF50/1.4 used mechanical systems, like a Hall-effect rotor or similar systems.
Large diameter systems get higher resolutions from natural causes, often you use similar turning angles - but the linear travel is angular travel x diameter.
A system like the Nikon 70-200 has about 6100 pulses per full travel. But this is an internal parameter, on the lens mainboard you have a logic circuit (part of the motor control circuit) that is responsible for translating between command sequences and actual lens movement. But since the (Nikon) communication protocol uses half-word (two bytes) to send this command, you could guess that with data padding you have 12-bit (1/4096) range resolution.
I know the data sent varies with lens - so maybe the camera knows what resolution the command sets should use from the firmware lens data in the camera. This might explain why some third party lenses react differently to MA than what you would expect. This goes for both Nikon and Canon.
Some Canon lenses I know of have about 2700 pulses as Roger said, and that seems to be more than enough. That's an external number, as communicated by the logic circuit - not the actual encoder resolution (I think). Higher numbers would just minimize integer round-off errors when translating input>output comparisons.
To get lens internal position measurement resolution you just need a pulse counter - which you have in most any modern digital oscilloscope that has a software connection - and the correct voltage feed to the lens control cirquit (or just the MR head..) to get it working. The resolution will then be the MR head pulse-count through a full focusing rack (-5% to account for focusing past infinity)
theSuede wrote:
Nope. What you'd hear then is just motor drive noise. The actual encoder resolution is (nowadays):
MR tape pulse resolution x encoder length.
or
optical guide resolution x encoder length
Very few new systems use mechanical or electrical resistance slider systems. You use either a MR head and a premagnetized substrate, or an optical system (but that's a lot more sensitive to dirt and mechanical misalignments...). Older systems like for instance the EF50/1.4 used mechanical systems, like a Hall-effect rotor or similar systems.
Large diameter systems get higher resolutions from natural causes, often you use similar turning angles - but the linear travel is angular travel x diameter.
A system like the Nikon 70-200 has about 6100 pulses per full travel. But this is an internal parameter, on the lens mainboard you have a logic circuit (part of the motor control circuit) that is responsible for translating between command sequences and actual lens movement. But since the (Nikon) communication protocol uses half-word (two bytes) to send this command, you could guess that with data padding you have 12-bit (1/4096) range resolution.
I know the data sent varies with lens - so maybe the camera knows what resolution the command sets should use from the firmware lens data in the camera. This might explain why some third party lenses react differently to MA than what you would expect. This goes for both Nikon and Canon.
Some Canon lenses I know of have about 2700 pulses as Roger said, and that seems to be more than enough. That's an external number, as communicated by the logic circuit - not the actual encoder resolution (I think). Higher numbers would just minimize integer round-off errors when translating input>output comparisons.
To get lens internal position measurement resolution you just need a pulse counter - which you have in most any modern digital oscilloscope that has a software connection - and the correct voltage feed to the lens control cirquit (or just the MR head..) to get it working. The resolution will then be the MR head pulse-count through a full focusing rack (-5% to account for focusing past infinity)...Show more →
Is the nikon communication protocol published? If it is I wonder if it would be possible to create a test rig that counts up from zero on that command. Combine that with a microphone and you should be able to tell the minimum spread between motion. After you find the upper and lower bounds, you may be able to determine total steps.
That would be an immense amount of work, and would only tell you external steps. I am actually more interested in internal steps as that has (in my very very limited experience with motor control) had a larger impact on smoothness. (I do UI layer stuff most of the time)
How did you find out the number of steps in the 70-200?
Pulse counter in parallel over the input from the encoder, right at the logic chip on the lens mainboard. One full rack > 6120 pulses. This was the 70-200/2.8VRmk1, not the newest version. Since you can often get at the mainboard just by disassembling the rear outer barrel and mount of the lens, this is the easiest point of access. But some lenses have "access windows" over the MR head in the lens barrel.
Actually I'm a bit confused about how to interpret the result - since the polarity (direction of the motion) is derived from having two heads with a half-pulse offset the actual resolution is half of 1/6120, or 1/12200. Close to 16-bit resolution.
Nikon has never published their actual specification, so most of it is stuff I can't write about here. Reverse engineering isn't illegal per se, but you can get in trouble for publishing your results.
