CHDKPTP: S90 Primary Focal Plane Configuration - hacking out the CCD - page 2 - Creative Uses of CHDK - CHDK Forum supplierdeeply

CHDKPTP: S90 Primary Focal Plane Configuration - hacking out the CCD

  • 704 Replies
  • 73045 Views
Re: CHDKPTP: S90 Primary Focal Plane Configuration - hacking out the CCD
« Reply #10 on: 14 / October / 2012, 09:05:52 »
Advertisements
The accuracy and detail of your disassembly documentation will be proved if you can reassemble the lens\CCD module.

*

Offline SticK

  • *****
  • 779
Re: CHDKPTP: S90 Primary Focal Plane Configuration - hacking out the CCD
« Reply #11 on: 14 / October / 2012, 09:48:58 »
My intent as I had described in the INTRODUCTION page is not to reassemble the lens ... the whole objective of this thread // lest we not lose sight of it.  I did however add some help and suggestions for folks who do want only to clean their lenses, including the reassembly map, to do just that.  For reassembly, it is very important that you document (index mark) the angular component positions at the step where you disengage the internal components by rotating the entrance lens.  As I mentioned also, print out the procedure separating out the steps if you want to go in reverse.  Fortunately Canon improved the mechanics over the years and there are no tiny loose parts that are difficult reassemble.  Hence reassembly should be a piece-o-cake and is left as an exercise for the reader ;).  Good luck!

edit:
In fact I would add this looks like an optimal mechanical design all aspects considered, and in my opinion, is very elegant ... just those very-low-friction dust-proofing felt rings would be great if Canon Engineering is listening.
« Last Edit: 14 / October / 2012, 10:23:28 by SticK »

*

Offline SticK

  • *****
  • 779
Re: CHDKPTP: S90 Primary Focal Plane Configuration - hacking out the CCD
« Reply #12 on: 17 / October / 2012, 12:17:48 »
LENS COMPONENT FUNCTIONAL DESCRIPTIONS

OPTICS

The lens assembly is constructed from static (bound-to-camera, directly or indirectly) and moving subassemblies.  There are three principal structures that contain simple or compound lenses, in entrance-to-exit direction:
  a) Entrance compound
  b) Exit compound (contains IS and shutter)
  c) focus simple
All three move independently of each other and the camera.  The shutter is mounted on the CCD-side of structure (b).  The iris moves axially between (a) and (b) also with independent motion.  The complex relative motion of structures (a), (b) and iris is mechanically encoded into ring 3B.  Structure (c) moves relative to backshell 1C.

MOTION CONTROL

Camera-relative axial motion control for structures (a), (b) and iris is exclusively provided by the zoom motor (1A, 5 o'clock) via 1B->3B.  The shutter solenoid is a bistable type that drives a magnetic armature: it changes state by a short strong electrical impulse and remains in the state (open or closed) un-powered.  Reversing current direction of the impulse changes state.  The iris vanes are driven by a pinion-geared 2-coil stepper motor, and at 5(OD)x7mm is smallest I've seen.  Focus structure (c) is moved by a larger stepper motor along a 1.5x11 mm leadscrew with a pitch of 4 tpmm.  At the default PUP position in manual focus mode, the motor is stepped by 1/12th revolution per WHEEL L or R index.  That is, the focus lens is stepped axially in increments of approximately 80 microns (width of an eyelash) inside the lens cavity.  The focus motor is very powerful and can transition the lens over the full length of the leadscrew in about a half-second.  With iris, focus and zoom, Canon provides precision intelligent motion control by slightly over-throwing the movement and then backing off to the desired position.

DETECTORS, FEEDBACK and POSITION CONTROL

There are five microminiature optoswitches in the lens assembly.  Three detect stowed positions of the iris, focus lens and zoom lens, and two detect revolutions and can detect motor direction in the zoom motor, where the zoom motor is powered only when in motion.  The zoom motor is in the servo configuration in that firmware always knows the current position of the zoom lens relative to its first PUP un-stowed signal by counting zoom motor optoswitch revolution pulses.  The IS contains two differential Hall-effect sensors (4-wire each) to detect the XY position of the floating lens in structure (b).  The IS lens is suspended on springs and positioned by two linear motors in the XY plane perpendicular to the optical axis.  The maximum throw of the IS lens is ~0.4 mm radius away from center in all radial directions.  Although both iris and focus subassemblies are theoretically open-loop, both have optoswitch stowed detection.  Hence, similar to zoom lens PUP reference calibration, firmware calibrates iris and focus position relative to their first un-stowed PUP signals.  After PUP reference position calibration, because steppers differ from servos in that they are constantly being AC-driven (electrically-held even when motionless), as long as motion is not obstructed (by dirt for example), firmware always knows their current position.  The shutter has no sensors and thus is the only component that operates fully open loop. 

IMAGE STABILIZER OPERATION DETAIL

When the camera is PUPed, firmware establishes the IS lens bounds by throwing the lens to its XY extremes, calibrating the reference positions and then centering the lens.  Hence the linear motors are always powered and even if you set IS to OFF in the Canon menu, there is very little, if any, operating time to be gained.  Hand vibration is 10-20 Hz, and the correction is applied to the XY plane of the IS subassembly.  You can think of anchoring an optical viewfinder camera with a telephoto lens so that its focal plane on the CCD is anchored against your nose.  Random hand vibration squiggles will then trace out a patch on a tiny area of a spherical surface (spherical sector) at the entrance pupil of the lens, called pitch and yaw, as in aircraft motion.  The farther you hold the camera away from your face, the flatter the spherical patch becomes.  Hence, vibration detection is done by two mutually-perpendicular electronic MEMS gyros that detect pitch and yaw, and then movement correction gets mapped into the flat XY plane of the lens subassembly.  In general, because the actual vibration movement is very small relative to the spherical radius at pitch and yaw origin, the error caused by flattening pitch and yaw into an XY plane is negligible.  I estimate the anti-vibration correction movement of the lens to be very small, in the low units of microns range.  The gyro chip is situated on the main camera board.

Next post: signal descriptions.

Re: CHDKPTP: S90 Primary Focal Plane Configuration - hacking out the CCD
« Reply #13 on: 17 / October / 2012, 14:10:58 »
Do you know the resistance of the shutter solenoid, it will be very low ?

Have you been able to drive the tiny stepper motor ?


*

Offline SticK

  • *****
  • 779
Re: CHDKPTP: S90 Primary Focal Plane Configuration - hacking out the CCD
« Reply #14 on: 17 / October / 2012, 14:46:01 »
The shutter solenoid is 10 ohms, 2 mH, if I recall.  I will be revisiting these parameters in the next post anyway, and more in detail when I present a shutter driver interface for your own shutter, if I ever get that far.  We'll see.

Although I have the hardware in my lab to drive stepper motors, in the context of the objective here of ideally eliminating electromechanics there is no gain.  One wants to characterize operation and then either remove unnecessary Canon hardware that doesn't generate shutdown errors, or replace as much as is reasonably feasible by mimicking the feedback mechanisms in external solid-state hardware.  Ideally, the entire lens assembly and contents should disappear leaving only the CCD, but I feel at this point although definitely not impossible, is unlikely.  In fact, if the iris and IS stay, the solution is not bad at all: because of their long ribbon cable, you can either locate them in the front (beside the main lens assembly) or back of the camera, leaving all flexible options to face the CCD in any direction relative to the camera body that is suitable for your application.

Re: CHDKPTP: S90 Primary Focal Plane Configuration - hacking out the CCD
« Reply #15 on: 17 / October / 2012, 15:01:12 »
I mentioned the stepper motor because you said "The focus motor is very powerful and can transition the lens over the full length of the leadscrew in about a half-second."

*

Offline SticK

  • *****
  • 779
Re: CHDKPTP: S90 Primary Focal Plane Configuration - hacking out the CCD
« Reply #16 on: 17 / October / 2012, 16:45:15 »
When you go full zoom the focus assembly follows outwards towards the end of leadscrew travel.  When you PDN from there, the assembly flies back to its stowed position (~1 cm) in about a half second ... it's very fast and must in good part contribute to the high speed autofocus in this camera.

*

Offline SticK

  • *****
  • 779
Re: CHDKPTP: S90 Primary Focal Plane Configuration - hacking out the CCD
« Reply #17 on: 19 / October / 2012, 13:15:45 »
LENS ELECTRICAL COMPONENTS and ELIMINATION POSSIBILITIES

 1) Shutter solenoid coil:  9 ohms, 1.9 mH
 2) Iris 2-winding stepper motor:  20 ohms, 1.2 mH,  per coil
 3) Focus 2-winding stepper motor: 20 ohms, 3.1 mH, per coil
 4) IS: 30 ohms, 0.5 mH, per linear motor
 5) Zoom lens motor:  6 ohms, 0.7 mH

These are the five independent electrical subsystems that can generate a shutdown error if interfered with.  Camera firmware is quite astute in reporting lens problems.  The Canon error is E32 for all that I have tested so far.  Typically, after an E32 occurs the camera shuts down within a minute.  So far, I found this error to be fully recoverable by reconnecting the component.  Note that you must power cycle to cold by disconnecting the AC adapter or removing battery to clear the error.

SHUTTER.  I have not disconnected the shutter solenoid for error tests because that circuit has to be left in-place anyway, or it can be replaced by a passive circuit (resistor at least, or at worst resistor series inductor to simulate the net reactance above), so that its signal can be used to drive your own shutter.  If the camera shutter driver has a fault detector, it would be on the main board and can be fooled by the passive network.  Hence replacing the shutter solenoid with a passive network will work perfectly.

IRIS and FOCUS.  Because motor drive is open loop, stepper motor coils can be replaced by passive reactive circuits.  However, both these devices have stowed switches whose first ON pulse after a PUP calibrates the reference position of these two components.  Because of the stowed switch, an active circuit would have to be added to mimic switch operation.  If firmware sets up expected error time windows for the unstowed/stowed signals, then the fooling circuit can be made of 555 timers and some glue logic.  If firmware tracks rotations wrt the unstowed/stowed signal, then a solid-state replacement solution becomes very complicated.  Either way, because steppers are constantly AC-driven even when still, just detecting motor motion is non-trivial.  I have not explored any of these yet, but these are certainly good possibilities.  So my guess is this: powering the motor can easily be detected.  So for PUP we detect the first motion signal into a coil, timeout the unstowed signal and deliver that to firmware by injecting it into the switch photodetector circuit.  That should turn the camera ON w/o error.  However, PDN is different.  Because the motor is powered all the time, we have to detect motion and then timeout the stowed signal so firmware is happy.  Not easy.

IS.  After having understood well the drive and feedback architecture I did a simple error test: before PUP, I wedged a toothpick between the lens and its cavity fixing the lens skewed over to one side.  After PUP I got the E32 error and a shutdown within a minute.  So this simple test explains fully the intricacy of this component.  The IS is a tightly firmware-controlled system that has zero tolerance to any perturbations of the lens.  It is unlikely that error expectancy override can be simulated within reason in external hardware.  The signalling is very complex.  To learn more about how the IS operates refer to this excellent article:
http://www.invensense.com/jp/mems/gyro/documents/whitepapers/ImageStabilizationWhitepaper_051606.pdf
Figure 11 shows the analogue interface circuit (located on the main board) to the 4-wire Hall effect sensors located beside the assembly's linear motors.

ZOOM.  Eliminating this one (1A+1B+1C+3A) on its own is quite important when you have a space-critical application because you can shrink your entire camera volume by almost two-thirds.  Because the zoom motor is a servo type, it can be replaced more easily than the steppers.  Armature rotation optoswitches are aligned to a triple blade fan shaft encoder, where one is ON (no blade) while the other is OFF (blade).  A trivial up/down counter that simulates rotation counting can be excited by the polarity of the motor terminals, with 555s and some glue logic serving to simulate rotation timing signals, and delivering them to the switches.  Here the unstowed/stowed (the assembly's 3rd switch, 1C, 11 o'clock) handling is again trivial because motor motion (and direction) can be easily detected by external hardware.  Hence the ZOOM external assembly has the highest chance of being eliminated, and it is by far the bulkiest component.  Naturally, because the focus mechanism is part of the backshell 1C, it would also have to be handled, or cut out leaving just the motor, leadscrew, and optoswitch.

Next: an attempt at an external shutter controller.
« Last Edit: 19 / October / 2012, 20:51:25 by SticK »


Re: CHDKPTP: S90 Primary Focal Plane Configuration - hacking out the CCD
« Reply #18 on: 19 / October / 2012, 19:25:30 »
Very interesting.

I am confident that information and detail is unique.

You can obtain compatible shutters, I have one.

*

Offline SticK

  • *****
  • 779
Re: CHDKPTP: S90 Primary Focal Plane Configuration - hacking out the CCD
« Reply #19 on: 19 / October / 2012, 20:40:39 »
I am glad you're finding this useful.  A circular shutter for the S90 CCD has to have a clear aperture of at least 10.0 mm.  The one I'm looking at is the Uniblitz DSS10.  It has an OD of 27 mm that will fit into my 30 mm space, but it's $420.  Reaction speed is within the range of the Canon signals, which I plan to discuss later.  However I'm sure there must be some cheaper ones available, and if you are aware of other sources, it would be nice to know and a fitting addition to this thread.

 

Related Topics