--- In DynoMotion@yahoogroups.com, Tom Kerekes <tk@...> wrote:
>
> Hi Grace,
>  
> My understanding of how most cameras work is the sensor's pixels are like tiny capacitors that charge up as light strikes them.  So they "integrate" the light over time.  In normal operation the image intensity is constant so the longer it is allowed to integrate the brighter the image is.  Then the chip reads out the charge for every pixel.  I assume in your case you are using an LED as basically a strobe light to freeze and capture the frame. 
>  
> Here are some thoughts:
>  
> 30 fps corresponds to 33ms per frame.
>  
> Assuming an accuracy of 1/2000th of a frame the trigger time would need an accuracy of ~16us.
>  
> KFLOP can't quite do that in software.  The best a KFLOP User program can do is execute every 90us (using the User callback every servo sample).
>  
> It seems like a hardware timer with microsecond resolution would be required.  You could do something like sample the capstan encoder position every 90us.  The last few 90us samples could be used to determine your average position and velocity.  Using the position and velocity information you could predict very precisely in the future when the capstan should arrive at the target position.  When the time to target becomes less than say 120us you could program the hardware timer to trigger at that time.
>  
> In order to syncronize the film position to the film sprockets, you might feed a optical sensor output into a Kanalog ADC.  The optical sensor should be configured/focused such that the ADC value ramps over a few 90us samples.  KFLOP could then use two or more samples to interpolate in time exactly when the signal crossed a threshold.  For example at one 90us sample the ADC reads -300, at the next -100, then +100.  We could assume it crossed zero 45us ago.
>  
> KFLOP doesn't currently have any programmable timers, but I think it might be possible to use a Step/Dir generator as a Timer. 
>  
> Sounds like fun :}
>  
> TK
>  
>  
>  
>
> From: gracemckay <grace@...>
> To: DynoMotion@yahoogroups.com
> Sent: Sunday, November 20, 2011 12:48 PM
> Subject: [DynoMotion] Re: New Machine Idea- Would KFlop be suitable?
>
>
>  
>
> Hi Quelisto--
>
> the frame rate will be somewhere in the 12 to 30 frame per second range. the camera is capable of 40 fps, but I think for the time being I am better off not approaching the limits. I am not sure what this term means- integration time-- can you clarify?
>
> I do agree that getting the frame registration to work correctly is the key to the whole thing. I have several ideas about how to make that work. The first one is to use an optical encoder in the film path to sense motion and position of the film, and to use that info to trip both the camera shutter and the led strobe. If I can't make that work, second is to use an optical trigger looking at the film sprockets to do the trigger. A mechanical sprocket is also possible to drive the capture but is my last choice, at least right now. The idea is for the film to have a constant motion, with the frame capture to happen as the film passes the gate.
>
> I have hopes that I can use the Kflop to handle the trigger signals, but that is something I will need to figure out. Maybe the Kanalog will have the functionality for that. My understanding of the whole system is at the extreme novice stage. So it will be fun to figure out.
>
> The motor I have selected for the capstan is a brushed servo with an on-board encoder. I think you are right about a stepper being too rough for this application. I may want to use steppers for the reel motors, but probably servos will work better here as well.
>
> Grace McKay
> ScanBox
> High Definition Motion Scanning
> Scanbox.tv
>
> --- In DynoMotion@yahoogroups.com, "quelisto" <quelisto@> wrote:
> >
> > How fast are you going to try and do this (frame rate wise)? I certainly think what you want to do is quite possible so long as you aren't trying to do something at a really high frame rate. What also matters here is the integration time for your image sensor. The faster the integration time 9typically measured in microseconds), the easier it will be to get the mechanical system to accomodate it.
> >
> > The biggest challenge will be doing frame registration. You really need to synchronize digital image capture (frame grabber) to with some sort of fiducial feature on the film itself. As you know, in a mechanical camera, frame synchronization is done with a claw-like mechanism that holds the film in registration for a standard period of time. If you do not use this type of mechanical film registration, you'll just have to do a fast image capture once you trigger on some element of the film geometry.
> >
> > All of the other stuff with the film (constant speed capstan) is fall-off-a-log easy with the KFlop and any sort of motor that has feedback (whether encoder or resolver), though I'd argue in favor of a brushed or brushless DC motor for smoothness (steppers have WAY more torque ripple).
> >
> > Holler back to discuss further.
> >
> > --- In DynoMotion@yahoogroups.com, Grace McKay <grace@> wrote:
> > >
> > >
> > > Hi guys, my first post here. I have been reading the archive and
> > > watching the live discussions for a couple of weeks. So here is(are) my
> > > question(s)...
> > >
> > >
> > > I am trying to build a machine for transferring super8 film to digital
> > > files. I have a basic mechanical design and rough prototype. My choice
> > > for the main drive motor and the reel motors are currently nema 23
> > > servos. I have a test motor from Leadshine, along with the matching
> > > driver and power supply. the driver is expecting a step/dir type of
> > > signal which I believe is the same as typical stepper drivers. From my
> > > research so far, it looks like the kflop board and the associated
> > > software might work for my design.
> > >
> > > If you think it might not work, can you help me understand why not? It
> > > looks like I would need to create software to run on the board, starting
> > > from the examples you give on the website. This is a problem for me -- I
> > > am not a coder, but maybe there are experts here that can help, or point
> > > me to freelancers who understand the system. I see the lively exchange
> > > of info on the list, and I am encouraged that there is a lot of
> > > knowledge flowing.
> > >
> > > So let me give you a little background on this project. I have been
> > > involved with film transfer for around 15 years, but the machines I have
> > > are all standard definition and the market has moved to hi-def. If I can
> > > make this machine work, maybe I can compete in the new market. The
> > > machine concept itself is modeled after my current telecine- a BTS
> > > Quadra, http://en.wikipedia.org/wiki/File:BTS_Quadra_444_Telecine.JPG as
> > > well as a machine built in the Netherlands, that I think has some good
> > > ideas. http://www.filmfabriek.nl/home.html..
> > >
> > > So, three motors, one is the capstan, which determines the speed the
> > > film will run through, and two reel motors. The capstan speed is
> > > constant, and the reel motors speed is variable, with feedback from
> > > linear encoders on each side, mounted with an idler roller to sense the
> > > load of the filmpacks. . I have a camera that accepts a remote trigger,
> > > and I am designing an led light that will accept or send a pulse. there
> > > will be a rotary encoder that controls the trigger, maybe.
> > >
> > > So in some ways it is like a CNC machine, in that there are motors (3)
> > > and feedback devices to control them, as well as the camera and light
> > > system.
> > >
> > > So please, any ideas, thoughts, questions?
> > >
> > > Grace McKay
> > >
> >
>

--- In DynoMotion@yahoogroups.com, Tom Kerekes <tk@...> wrote:
>
> Hi Grace,
>  
> My understanding of how most cameras work is the sensor's pixels are like tiny capacitors that charge up as light strikes them.  So they "integrate" the light over time.  In normal operation the image intensity is constant so the longer it is allowed to integrate the brighter the image is.  Then the chip reads out the charge for every pixel.  I assume in your case you are using an LED as basically a strobe light to freeze and capture the frame. 
>  
> Here are some thoughts:
>  
> 30 fps corresponds to 33ms per frame.
>  
> Assuming an accuracy of 1/2000th of a frame the trigger time would need an accuracy of ~16us.
>  
> KFLOP can't quite do that in software.  The best a KFLOP User program can do is execute every 90us (using the User callback every servo sample).
>  
> It seems like a hardware timer with microsecond resolution would be required.  You could do something like sample the capstan encoder position every 90us.  The last few 90us samples could be used to determine your average position and velocity.  Using the position and velocity information you could predict very precisely in the future when the capstan should arrive at the target position.  When the time to target becomes less than say 120us you could program the hardware timer to trigger at that time.
>  
> In order to syncronize the film position to the film sprockets, you might feed a optical sensor output into a Kanalog ADC.  The optical sensor should be configured/focused such that the ADC value ramps over a few 90us samples.  KFLOP could then use two or more samples to interpolate in time exactly when the signal crossed a threshold.  For example at one 90us sample the ADC reads -300, at the next -100, then +100.  We could assume it crossed zero 45us ago.
>  
> KFLOP doesn't currently have any programmable timers, but I think it might be possible to use a Step/Dir generator as a Timer. 
>  
> Sounds like fun :}
>  
> TK
>  
>  
>  
>
> From: gracemckay <grace@...>
> To: DynoMotion@yahoogroups.com
> Sent: Sunday, November 20, 2011 12:48 PM
> Subject: [DynoMotion] Re: New Machine Idea- Would KFlop be suitable?
>
>
>  
>
> Hi Quelisto--
>
> the frame rate will be somewhere in the 12 to 30 frame per second range. the camera is capable of 40 fps, but I think for the time being I am better off not approaching the limits. I am not sure what this term means- integration time-- can you clarify?
>
> I do agree that getting the frame registration to work correctly is the key to the whole thing. I have several ideas about how to make that work. The first one is to use an optical encoder in the film path to sense motion and position of the film, and to use that info to trip both the camera shutter and the led strobe. If I can't make that work, second is to use an optical trigger looking at the film sprockets to do the trigger. A mechanical sprocket is also possible to drive the capture but is my last choice, at least right now. The idea is for the film to have a constant motion, with the frame capture to happen as the film passes the gate.
>
> I have hopes that I can use the Kflop to handle the trigger signals, but that is something I will need to figure out. Maybe the Kanalog will have the functionality for that. My understanding of the whole system is at the extreme novice stage. So it will be fun to figure out.
>
> The motor I have selected for the capstan is a brushed servo with an on-board encoder. I think you are right about a stepper being too rough for this application. I may want to use steppers for the reel motors, but probably servos will work better here as well.
>
> Grace McKay
> ScanBox
> High Definition Motion Scanning
> Scanbox.tv
>
> --- In DynoMotion@yahoogroups.com, "quelisto" <quelisto@> wrote:
> >
> > How fast are you going to try and do this (frame rate wise)? I certainly think what you want to do is quite possible so long as you aren't trying to do something at a really high frame rate. What also matters here is the integration time for your image sensor. The faster the integration time 9typically measured in microseconds), the easier it will be to get the mechanical system to accomodate it.
> >
> > The biggest challenge will be doing frame registration. You really need to synchronize digital image capture (frame grabber) to with some sort of fiducial feature on the film itself. As you know, in a mechanical camera, frame synchronization is done with a claw-like mechanism that holds the film in registration for a standard period of time. If you do not use this type of mechanical film registration, you'll just have to do a fast image capture once you trigger on some element of the film geometry.
> >
> > All of the other stuff with the film (constant speed capstan) is fall-off-a-log easy with the KFlop and any sort of motor that has feedback (whether encoder or resolver), though I'd argue in favor of a brushed or brushless DC motor for smoothness (steppers have WAY more torque ripple).
> >
> > Holler back to discuss further.
> >
> > --- In DynoMotion@yahoogroups.com, Grace McKay <grace@> wrote:
> > >
> > >
> > > Hi guys, my first post here. I have been reading the archive and
> > > watching the live discussions for a couple of weeks. So here is(are) my
> > > question(s)...
> > >
> > >
> > > I am trying to build a machine for transferring super8 film to digital
> > > files. I have a basic mechanical design and rough prototype. My choice
> > > for the main drive motor and the reel motors are currently nema 23
> > > servos. I have a test motor from Leadshine, along with the matching
> > > driver and power supply. the driver is expecting a step/dir type of
> > > signal which I believe is the same as typical stepper drivers. From my
> > > research so far, it looks like the kflop board and the associated
> > > software might work for my design.
> > >
> > > If you think it might not work, can you help me understand why not? It
> > > looks like I would need to create software to run on the board, starting
> > > from the examples you give on the website. This is a problem for me -- I
> > > am not a coder, but maybe there are experts here that can help, or point
> > > me to freelancers who understand the system. I see the lively exchange
> > > of info on the list, and I am encouraged that there is a lot of
> > > knowledge flowing.
> > >
> > > So let me give you a little background on this project. I have been
> > > involved with film transfer for around 15 years, but the machines I have
> > > are all standard definition and the market has moved to hi-def. If I can
> > > make this machine work, maybe I can compete in the new market. The
> > > machine concept itself is modeled after my current telecine- a BTS
> > > Quadra, http://en.wikipedia.org/wiki/File:BTS_Quadra_444_Telecine.JPG as
> > > well as a machine built in the Netherlands, that I think has some good
> > > ideas. http://www.filmfabriek.nl/home.html..
> > >
> > > So, three motors, one is the capstan, which determines the speed the
> > > film will run through, and two reel motors. The capstan speed is
> > > constant, and the reel motors speed is variable, with feedback from
> > > linear encoders on each side, mounted with an idler roller to sense the
> > > load of the filmpacks. . I have a camera that accepts a remote trigger,
> > > and I am designing an led light that will accept or send a pulse. there
> > > will be a rotary encoder that controls the trigger, maybe.
> > >
> > > So in some ways it is like a CNC machine, in that there are motors (3)
> > > and feedback devices to control them, as well as the camera and light
> > > system.
> > >
> > > So please, any ideas, thoughts, questions?
> > >
> > > Grace McKay
> > >
> >
>

--- In DynoMotion@yahoogroups.com, "himykabibble" <jagboy@...> wrote:
>
> Having been VP of Engineering for two different digital camera companies, this is something I know a little bit about. I think the bigger problem will be that, depending on exactly what device you are using to do the capture, the latency from the time you signal you want to do a capture, and when that capture actually occurs is likely to be quite significant, and quite variable from frame to frame. So, before you go any further, you need to understand exactly how the capture device you intend to use actually operates. In video mode, their frame rate is typically a function of their internal clocks, and you will have no control over that. Trying to do it as a rapid-fire sequence of still frames will not work, due to the time required to "setup" each frame - (AE, AF, color balance, etc). Nor will you be able to control the phase of the capture to your trigger signal in video mode. - the first frame will generally be captured some indeterminate period of time after to tell it to begin video capture. Doing what you're suggesting will likely require building your own capture device, and controlling the sensor and analog front end directly, which is non-trivial, especially if you're using a CCD, rather than a CMOS sensor.
>
> Regards,
> Ray L.
>
>
> --- In DynoMotion@yahoogroups.com, Tom Kerekes tk@ wrote:
> >
> > Hi Grace,
> >  
> > My understanding of how most cameras work is the sensor's pixels are like tiny capacitors that charge up as light strikes them.  So they "integrate" the light over time.  In normal operation the image intensity is constant so the longer it is allowed to integrate the brighter the image is.  Then the chip reads out the charge for every pixel.  I assume in your case you are using an LED as basically a strobe light to freeze and capture the frame. 
> >  
> > Here are some thoughts:
> >  
> > 30 fps corresponds to 33ms per frame.
> >  
> > Assuming an accuracy of 1/2000th of a frame the trigger time would need an accuracy of ~16us.
> >  
> > KFLOP can't quite do that in software.  The best a KFLOP User program can do is execute every 90us (using the User callback every servo sample).
> >  
> > It seems like a hardware timer with microsecond resolution would be required.  You could do something like sample the capstan encoder position every 90us.  The last few 90us samples could be used to determine your average position and velocity.  Using the position and velocity information you could predict very precisely in the future when the capstan should arrive at the target position.  When the time to target becomes less than say 120us you could program the hardware timer to trigger at that time.
> >  
> > In order to syncronize the film position to the film sprockets, you might feed a optical sensor output into a Kanalog ADC.  The optical sensor should be configured/focused such that the ADC value ramps over a few 90us samples.  KFLOP could then use two or more samples to interpolate in time exactly when the signal crossed a threshold.  For example at one 90us sample the ADC reads -300, at the next -100, then +100.  We could assume it crossed zero 45us ago.
> >  
> > KFLOP doesn't currently have any programmable timers, but I think it might be possible to use a Step/Dir generator as a Timer. 
> >  
> > Sounds like fun :}
> >  
> > TK
> >  
> >  
> >  
> >
> > From: gracemckay grace@
> > To: DynoMotion@yahoogroups.com
> > Sent: Sunday, November 20, 2011 12:48 PM
> > Subject: [DynoMotion] Re: New Machine Idea- Would KFlop be suitable?
> >
> >
> >  
> >
> > Hi Quelisto--
> >
> > the frame rate will be somewhere in the 12 to 30 frame per second range. the camera is capable of 40 fps, but I think for the time being I am better off not approaching the limits. I am not sure what this term means- integration time-- can you clarify?
> >
> > I do agree that getting the frame registration to work correctly is the key to the whole thing. I have several ideas about how to make that work. The first one is to use an optical encoder in the film path to sense motion and position of the film, and to use that info to trip both the camera shutter and the led strobe. If I can't make that work, second is to use an optical trigger looking at the film sprockets to do the trigger. A mechanical sprocket is also possible to drive the capture but is my last choice, at least right now. The idea is for the film to have a constant motion, with the frame capture to happen as the film passes the gate.
> >
> > I have hopes that I can use the Kflop to handle the trigger signals, but that is something I will need to figure out. Maybe the Kanalog will have the functionality for that. My understanding of the whole system is at the extreme novice stage. So it will be fun to figure out.
> >
> > The motor I have selected for the capstan is a brushed servo with an on-board encoder. I think you are right about a stepper being too rough for this application. I may want to use steppers for the reel motors, but probably servos will work better here as well.
> >
> > Grace McKay
> > ScanBox
> > High Definition Motion Scanning
> > Scanbox.tv
> >
> > --- In DynoMotion@yahoogroups.com, "quelisto" <quelisto@> wrote:
> > >
> > > How fast are you going to try and do this (frame rate wise)? I certainly think what you want to do is quite possible so long as you aren't trying to do something at a really high frame rate. What also matters here is the integration time for your image sensor. The faster the integration time 9typically measured in microseconds), the easier it will be to get the mechanical system to accomodate it.
> > >
> > > The biggest challenge will be doing frame registration. You really need to synchronize digital image capture (frame grabber) to with some sort of fiducial feature on the film itself. As you know, in a mechanical camera, frame synchronization is done with a claw-like mechanism that holds the film in registration for a standard period of time. If you do not use this type of mechanical film registration, you'll just have to do a fast image capture once you trigger on some element of the film geometry.
> > >
> > > All of the other stuff with the film (constant speed capstan) is fall-off-a-log easy with the KFlop and any sort of motor that has feedback (whether encoder or resolver), though I'd argue in favor of a brushed or brushless DC motor for smoothness (steppers have WAY more torque ripple).
> > >
> > > Holler back to discuss further.
> > >
> > > --- In DynoMotion@yahoogroups.com, Grace McKay <grace@> wrote:
> > > >
> > > >
> > > > Hi guys, my first post here. I have been reading the archive and
> > > > watching the live discussions for a couple of weeks. So here is(are) my
> > > > question(s)...
> > > >
> > > >
> > > > I am trying to build a machine for transferring super8 film to digital
> > > > files. I have a basic mechanical design and rough prototype. My choice
> > > > for the main drive motor and the reel motors are currently nema 23
> > > > servos. I have a test motor from Leadshine, along with the matching
> > > > driver and power supply. the driver is expecting a step/dir type of
> > > > signal which I believe is the same as typical stepper drivers. From my
> > > > research so far, it looks like the kflop board and the associated
> > > > software might work for my design.
> > > >
> > > > If you think it might not work, can you help me understand why not? It
> > > > looks like I would need to create software to run on the board, starting
> > > > from the examples you give on the website. This is a problem for me -- I
> > > > am not a coder, but maybe there are experts here that can help, or point
> > > > me to freelancers who understand the system. I see the lively exchange
> > > > of info on the list, and I am encouraged that there is a lot of
> > > > knowledge flowing.
> > > >
> > > > So let me give you a little background on this project. I have been
> > > > involved with film transfer for around 15 years, but the machines I have
> > > > are all standard definition and the market has moved to hi-def. If I can
> > > > make this machine work, maybe I can compete in the new market. The
> > > > machine concept itself is modeled after my current telecine- a BTS
> > > > Quadra, http://en.wikipedia.org/wiki/File:BTS_Quadra_444_Telecine.JPG as
> > > > well as a machine built in the Netherlands, that I think has some good
> > > > ideas. http://www.filmfabriek.nl/home.html..
> > > >
> > > > So, three motors, one is the capstan, which determines the speed the
> > > > film will run through, and two reel motors. The capstan speed is
> > > > constant, and the reel motors speed is variable, with feedback from
> > > > linear encoders on each side, mounted with an idler roller to sense the
> > > > load of the filmpacks. . I have a camera that accepts a remote trigger,
> > > > and I am designing an led light that will accept or send a pulse. there
> > > > will be a rotary encoder that controls the trigger, maybe.
> > > >
> > > > So in some ways it is like a CNC machine, in that there are motors (3)
> > > > and feedback devices to control them, as well as the camera and light
> > > > system.
> > > >
> > > > So please, any ideas, thoughts, questions?
> > > >
> > > > Grace McKay
> > > >
> > >
> >
>

--- In DynoMotion@yahoogroups.com, "gracemckay" <grace@...> wrote:
>
> Hi Ray,
>
> The camera is a Prosilica GX1920. I hope to be using the still
> capture mode, which can go as fast as 40 frames per second, if I
> read the datasheet correctly. But you are right, I don't really
> know. The whole system is manual- manual focus, exposure, white
> balance, so I presume we will lose no time with those functions. As
> near as I can tell, the speed of the hard drive system is the
> constraint that will matter the most on how fast we can actually
> capture. The manufacturer is recommending a 4 drive hardware raid to
> go at normal speeds. Right now, I am still trying to figure out the
> transport stuff. If I can get 1 frame per second to start, I will
> consider that a victory. Ray, I am glad you are here, your
> background and knowledge may keep me from stepping in a mud puddle
> or two. Thanks!
>
>
> Grace McKay
> ScanBox
> High Definition Motion Scanning
> Scanbox.tv
> --- In DynoMotion@yahoogroups.com, "himykabibble" <jagboy@> wrote:
> >
> > Having been VP of Engineering for two different digital camera
> companies, this is something I know a little bit about. I think the
> bigger problem will be that, depending on exactly what device you are
> using to do the capture, the latency from the time you signal you want
> to do a capture, and when that capture actually occurs is likely to be
> quite significant, and quite variable from frame to frame. So, before
> you go any further, you need to understand exactly how the capture
> device you intend to use actually operates. In video mode, their frame
> rate is typically a function of their internal clocks, and you will have
> no control over that. Trying to do it as a rapid-fire sequence of still
> frames will not work, due to the time required to "setup" each frame -
> (AE, AF, color balance, etc). Nor will you be able to control the phase
> of the capture to your trigger signal in video mode. - the first frame
> will generally be captured some indeterminate period of time after to
> tell it to begin video capture. Doing what you're suggesting will
> likely require building your own capture device, and controlling the
> sensor and analog front end directly, which is non-trivial, especially
> if you're using a CCD, rather than a CMOS sensor.
> >
> > Regards,
> > Ray L.
> >
> >
> > --- In DynoMotion@yahoogroups.com, Tom Kerekes tk@ wrote:
> > >
> > > Hi Grace,
> > > Â
> > > My understanding of how most cameras work is the sensor's pixels are
> like tiny capacitors that charge up as light strikes them. So they
> "integrate" the light over time. In normal operation the image
> intensity is constant so the longer it is allowed to integrate the
> brighter the image is. Then the chip reads out the charge for every
> pixel. I assume in your case you are using an LED as basically a
> strobe light to freeze and capture the frame.Â
> > > Â
> > > Here are some thoughts:
> > > Â
> > > 30 fps corresponds to 33ms per frame.
> > > Â
> > > Assuming an accuracy of 1/2000th of a frame the trigger time would
> need an accuracy of ~16us.
> > > Â
> > > KFLOP can't quite do that in software. The best a KFLOP User
> program can do is execute every 90us (using the User callback every
> servo sample).
> > > Â
> > > It seems like a hardware timer with microsecond resolution would
> be required. You could do something like sample the capstan encoder
> position every 90us. The last few 90us samples could be used to
> determine your average position and velocity. Using the position
> and velocity information you could predict very precisely in the future
> when the capstan should arrive at the target position. When the
> time to target becomes less than say 120us you could program the
> hardware timer to trigger at that time.
> > > Â
> > > In order to syncronize the film position to the film sprockets, you
> might feed a optical sensor output into a Kanalog ADC. The optical
> sensor should be configured/focused such that the ADC value ramps over a
> few 90us samples. KFLOP could then use two or more samples to
> interpolate in time exactly when the signal crossed a threshold.Â
> For example at one 90us sample the ADC reads -300, at the next -100,
> then +100. We could assume it crossed zero 45us ago.
> > > Â
> > > KFLOP doesn't currently have any programmable timers, but I think it
> might be possible to use a Step/Dir generator as a Timer.Â
> > > Â
> > > Sounds like fun :}
> > > Â
> > > TK
> > > Â
> > > Â
> > > Â
> > >
> > > From: gracemckay grace@
> > > To: DynoMotion@yahoogroups.com
> > > Sent: Sunday, November 20, 2011 12:48 PM
> > > Subject: [DynoMotion] Re: New Machine Idea- Would KFlop be suitable?
> > >
> > >
> > > Â
> > >
> > > Hi Quelisto--
> > >
> > > the frame rate will be somewhere in the 12 to 30 frame per second
> range. the camera is capable of 40 fps, but I think for the time being I
> am better off not approaching the limits. I am not sure what this term
> means- integration time-- can you clarify?
> > >
> > > I do agree that getting the frame registration to work correctly is
> the key to the whole thing. I have several ideas about how to make that
> work. The first one is to use an optical encoder in the film path to
> sense motion and position of the film, and to use that info to trip both
> the camera shutter and the led strobe. If I can't make that work, second
> is to use an optical trigger looking at the film sprockets to do the
> trigger. A mechanical sprocket is also possible to drive the capture but
> is my last choice, at least right now. The idea is for the film to have
> a constant motion, with the frame capture to happen as the film passes
> the gate.
> > >
> > > I have hopes that I can use the Kflop to handle the trigger signals,
> but that is something I will need to figure out. Maybe the Kanalog will
> have the functionality for that. My understanding of the whole system is
> at the extreme novice stage. So it will be fun to figure out.
> > >
> > > The motor I have selected for the capstan is a brushed servo with an
> on-board encoder. I think you are right about a stepper being too rough
> for this application. I may want to use steppers for the reel motors,
> but probably servos will work better here as well.
> > >
> > > Grace McKay
> > > ScanBox
> > > High Definition Motion Scanning
> > > Scanbox.tv
> > >
> > > --- In DynoMotion@yahoogroups.com, "quelisto" <quelisto@> wrote:
> > > >
> > > > How fast are you going to try and do this (frame rate wise)? I
> certainly think what you want to do is quite possible so long as you
> aren't trying to do something at a really high frame rate. What also
> matters here is the integration time for your image sensor. The faster
> the integration time 9typically measured in microseconds), the easier it
> will be to get the mechanical system to accomodate it.
> > > >
> > > > The biggest challenge will be doing frame registration. You really
> need to synchronize digital image capture (frame grabber) to with some
> sort of fiducial feature on the film itself. As you know, in a
> mechanical camera, frame synchronization is done with a claw-like
> mechanism that holds the film in registration for a standard period of
> time. If you do not use this type of mechanical film registration,
> you'll just have to do a fast image capture once you trigger on some
> element of the film geometry.
> > > >
> > > > All of the other stuff with the film (constant speed capstan) is
> fall-off-a-log easy with the KFlop and any sort of motor that has
> feedback (whether encoder or resolver), though I'd argue in favor of a
> brushed or brushless DC motor for smoothness (steppers have WAY more
> torque ripple).
> > > >
> > > > Holler back to discuss further.
> > > >
> > > > --- In DynoMotion@yahoogroups.com, Grace McKay <grace@> wrote:
> > > > >
> > > > >
> > > > > Hi guys, my first post here. I have been reading the archive and
> > > > > watching the live discussions for a couple of weeks. So here
> is(are) my
> > > > > question(s)...
> > > > >
> > > > >
> > > > > I am trying to build a machine for transferring super8 film to
> digital
> > > > > files. I have a basic mechanical design and rough prototype. My
> choice
> > > > > for the main drive motor and the reel motors are currently nema
> 23
> > > > > servos. I have a test motor from Leadshine, along with the
> matching
> > > > > driver and power supply. the driver is expecting a step/dir type
> of
> > > > > signal which I believe is the same as typical stepper drivers.
> From my
> > > > > research so far, it looks like the kflop board and the
> associated
> > > > > software might work for my design.
> > > > >
> > > > > If you think it might not work, can you help me understand why
> not? It
> > > > > looks like I would need to create software to run on the board,
> starting
> > > > > from the examples you give on the website. This is a problem for
> me -- I
> > > > > am not a coder, but maybe there are experts here that can help,
> or point
> > > > > me to freelancers who understand the system. I see the lively
> exchange
> > > > > of info on the list, and I am encouraged that there is a lot of
> > > > > knowledge flowing.
> > > > >
> > > > > So let me give you a little background on this project. I have
> been
> > > > > involved with film transfer for around 15 years, but the
> machines I have
> > > > > are all standard definition and the market has moved to hi-def.
> If I can
> > > > > make this machine work, maybe I can compete in the new market.
> The
> > > > > machine concept itself is modeled after my current telecine- a
> BTS
> > > > > Quadra,
> http://en.wikipedia.org/wiki/File:BTS_Quadra_444_Telecine.JPG as
> > > > > well as a machine built in the Netherlands, that I think has
> some good
> > > > > ideas. http://www.filmfabriek.nl/home.html..
> > > > >
> > > > > So, three motors, one is the capstan, which determines the speed
> the
> > > > > film will run through, and two reel motors. The capstan speed is
> > > > > constant, and the reel motors speed is variable, with feedback
> from
> > > > > linear encoders on each side, mounted with an idler roller to
> sense the
> > > > > load of the filmpacks. . I have a camera that accepts a remote
> trigger,
> > > > > and I am designing an led light that will accept or send a
> pulse. there
> > > > > will be a rotary encoder that controls the trigger, maybe.
> > > > >
> > > > > So in some ways it is like a CNC machine, in that there are
> motors (3)
> > > > > and feedback devices to control them, as well as the camera and
> light
> > > > > system.
> > > > >
> > > > > So please, any ideas, thoughts, questions?
> > > > >
> > > > > Grace McKay
> > > > >
> > > >
> > >
> >
>

--- In DynoMotion@yahoogroups.com, "himykabibble" <jagboy@...> wrote:
>
> Take a good look at pages 42-44 of the Prosilica Technical Manual, which spells out the latency, and latency jitter, from the time you request a capture, until it actually takes place. They can be quite short (1.5uSec, +/-0.5 uSec), under the right conditions, but you may not be able to provide those conditions at higher frame rates.
>
> Regards,
> Ray L.
>
> --- In DynoMotion@yahoogroups.com, "gracemckay" <grace@> wrote:
> >
> > Hi Ray,
> >
> > The camera is a Prosilica GX1920. I hope to be using the still
> > capture mode, which can go as fast as 40 frames per second, if I
> > read the datasheet correctly. But you are right, I don't really
> > know. The whole system is manual- manual focus, exposure, white
> > balance, so I presume we will lose no time with those functions. As
> > near as I can tell, the speed of the hard drive system is the
> > constraint that will matter the most on how fast we can actually
> > capture. The manufacturer is recommending a 4 drive hardware raid to
> > go at normal speeds. Right now, I am still trying to figure out the
> > transport stuff. If I can get 1 frame per second to start, I will
> > consider that a victory. Ray, I am glad you are here, your
> > background and knowledge may keep me from stepping in a mud puddle
> > or two. Thanks!
> >
> >
> > Grace McKay
> > ScanBox
> > High Definition Motion Scanning
> > Scanbox.tv
> > --- In DynoMotion@yahoogroups.com, "himykabibble" <jagboy@> wrote:
> > >
> > > Having been VP of Engineering for two different digital camera
> > companies, this is something I know a little bit about. I think the
> > bigger problem will be that, depending on exactly what device you are
> > using to do the capture, the latency from the time you signal you want
> > to do a capture, and when that capture actually occurs is likely to be
> > quite significant, and quite variable from frame to frame. So, before
> > you go any further, you need to understand exactly how the capture
> > device you intend to use actually operates. In video mode, their frame
> > rate is typically a function of their internal clocks, and you will have
> > no control over that. Trying to do it as a rapid-fire sequence of still
> > frames will not work, due to the time required to "setup" each frame -
> > (AE, AF, color balance, etc). Nor will you be able to control the phase
> > of the capture to your trigger signal in video mode. - the first frame
> > will generally be captured some indeterminate period of time after to
> > tell it to begin video capture. Doing what you're suggesting will
> > likely require building your own capture device, and controlling the
> > sensor and analog front end directly, which is non-trivial, especially
> > if you're using a CCD, rather than a CMOS sensor.
> > >
> > > Regards,
> > > Ray L.
> > >
> > >
> > > --- In DynoMotion@yahoogroups.com, Tom Kerekes tk@ wrote:
> > > >
> > > > Hi Grace,
> > > > Â
> > > > My understanding of how most cameras work is the sensor's pixels are
> > like tiny capacitors that charge up as light strikes them. So they
> > "integrate" the light over time. In normal operation the image
> > intensity is constant so the longer it is allowed to integrate the
> > brighter the image is. Then the chip reads out the charge for every
> > pixel. I assume in your case you are using an LED as basically a
> > strobe light to freeze and capture the frame.Â
> > > > Â
> > > > Here are some thoughts:
> > > > Â
> > > > 30 fps corresponds to 33ms per frame.
> > > > Â
> > > > Assuming an accuracy of 1/2000th of a frame the trigger time would
> > need an accuracy of ~16us.
> > > > Â
> > > > KFLOP can't quite do that in software. The best a KFLOP User
> > program can do is execute every 90us (using the User callback every
> > servo sample).
> > > > Â
> > > > It seems like a hardware timer with microsecond resolution would
> > be required. You could do something like sample the capstan encoder
> > position every 90us. The last few 90us samples could be used to
> > determine your average position and velocity. Using the position
> > and velocity information you could predict very precisely in the future
> > when the capstan should arrive at the target position. When the
> > time to target becomes less than say 120us you could program the
> > hardware timer to trigger at that time.
> > > > Â
> > > > In order to syncronize the film position to the film sprockets, you
> > might feed a optical sensor output into a Kanalog ADC. The optical
> > sensor should be configured/focused such that the ADC value ramps over a
> > few 90us samples. KFLOP could then use two or more samples to
> > interpolate in time exactly when the signal crossed a threshold.Â
> > For example at one 90us sample the ADC reads -300, at the next -100,
> > then +100. We could assume it crossed zero 45us ago.
> > > > Â
> > > > KFLOP doesn't currently have any programmable timers, but I think it
> > might be possible to use a Step/Dir generator as a Timer.Â
> > > > Â
> > > > Sounds like fun :}
> > > > Â
> > > > TK
> > > > Â
> > > > Â
> > > > Â
> > > >
> > > > From: gracemckay grace@
> > > > To: DynoMotion@yahoogroups.com
> > > > Sent: Sunday, November 20, 2011 12:48 PM
> > > > Subject: [DynoMotion] Re: New Machine Idea- Would KFlop be suitable?
> > > >
> > > >
> > > > Â
> > > >
> > > > Hi Quelisto--
> > > >
> > > > the frame rate will be somewhere in the 12 to 30 frame per second
> > range. the camera is capable of 40 fps, but I think for the time being I
> > am better off not approaching the limits. I am not sure what this term
> > means- integration time-- can you clarify?
> > > >
> > > > I do agree that getting the frame registration to work correctly is
> > the key to the whole thing. I have several ideas about how to make that
> > work. The first one is to use an optical encoder in the film path to
> > sense motion and position of the film, and to use that info to trip both
> > the camera shutter and the led strobe. If I can't make that work, second
> > is to use an optical trigger looking at the film sprockets to do the
> > trigger. A mechanical sprocket is also possible to drive the capture but
> > is my last choice, at least right now. The idea is for the film to have
> > a constant motion, with the frame capture to happen as the film passes
> > the gate.
> > > >
> > > > I have hopes that I can use the Kflop to handle the trigger signals,
> > but that is something I will need to figure out. Maybe the Kanalog will
> > have the functionality for that. My understanding of the whole system is
> > at the extreme novice stage. So it will be fun to figure out.
> > > >
> > > > The motor I have selected for the capstan is a brushed servo with an
> > on-board encoder. I think you are right about a stepper being too rough
> > for this application. I may want to use steppers for the reel motors,
> > but probably servos will work better here as well.
> > > >
> > > > Grace McKay
> > > > ScanBox
> > > > High Definition Motion Scanning
> > > > Scanbox.tv
> > > >
> > > > --- In DynoMotion@yahoogroups.com, "quelisto" <quelisto@> wrote:
> > > > >
> > > > > How fast are you going to try and do this (frame rate wise)? I
> > certainly think what you want to do is quite possible so long as you
> > aren't trying to do something at a really high frame rate. What also
> > matters here is the integration time for your image sensor. The faster
> > the integration time 9typically measured in microseconds), the easier it
> > will be to get the mechanical system to accomodate it.
> > > > >
> > > > > The biggest challenge will be doing frame registration. You really
> > need to synchronize digital image capture (frame grabber) to with some
> > sort of fiducial feature on the film itself. As you know, in a
> > mechanical camera, frame synchronization is done with a claw-like
> > mechanism that holds the film in registration for a standard period of
> > time. If you do not use this type of mechanical film registration,
> > you'll just have to do a fast image capture once you trigger on some
> > element of the film geometry.
> > > > >
> > > > > All of the other stuff with the film (constant speed capstan) is
> > fall-off-a-log easy with the KFlop and any sort of motor that has
> > feedback (whether encoder or resolver), though I'd argue in favor of a
> > brushed or brushless DC motor for smoothness (steppers have WAY more
> > torque ripple).
> > > > >
> > > > > Holler back to discuss further.
> > > > >
> > > > > --- In DynoMotion@yahoogroups.com, Grace McKay <grace@> wrote:
> > > > > >
> > > > > >
> > > > > > Hi guys, my first post here. I have been reading the archive and
> > > > > > watching the live discussions for a couple of weeks. So here
> > is(are) my
> > > > > > question(s)...
> > > > > >
> > > > > >
> > > > > > I am trying to build a machine for transferring super8 film to
> > digital
> > > > > > files. I have a basic mechanical design and rough prototype. My
> > choice
> > > > > > for the main drive motor and the reel motors are currently nema
> > 23
> > > > > > servos. I have a test motor from Leadshine, along with the
> > matching
> > > > > > driver and power supply. the driver is expecting a step/dir type
> > of
> > > > > > signal which I believe is the same as typical stepper drivers.
> > From my
> > > > > > research so far, it looks like the kflop board and the
> > associated
> > > > > > software might work for my design.
> > > > > >
> > > > > > If you think it might not work, can you help me understand why
> > not? It
> > > > > > looks like I would need to create software to run on the board,
> > starting
> > > > > > from the examples you give on the website. This is a problem for
> > me -- I
> > > > > > am not a coder, but maybe there are experts here that can help,
> > or point
> > > > > > me to freelancers who understand the system. I see the lively
> > exchange
> > > > > > of info on the list, and I am encouraged that there is a lot of
> > > > > > knowledge flowing.
> > > > > >
> > > > > > So let me give you a little background on this project. I have
> > been
> > > > > > involved with film transfer for around 15 years, but the
> > machines I have
> > > > > > are all standard definition and the market has moved to hi-def.
> > If I can
> > > > > > make this machine work, maybe I can compete in the new market.
> > The
> > > > > > machine concept itself is modeled after my current telecine- a
> > BTS
> > > > > > Quadra,
> > http://en.wikipedia.org/wiki/File:BTS_Quadra_444_Telecine.JPG as
> > > > > > well as a machine built in the Netherlands, that I think has
> > some good
> > > > > > ideas. http://www.filmfabriek.nl/home.html..
> > > > > >
> > > > > > So, three motors, one is the capstan, which determines the speed
> > the
> > > > > > film will run through, and two reel motors. The capstan speed is
> > > > > > constant, and the reel motors speed is variable, with feedback
> > from
> > > > > > linear encoders on each side, mounted with an idler roller to
> > sense the
> > > > > > load of the filmpacks. . I have a camera that accepts a remote
> > trigger,
> > > > > > and I am designing an led light that will accept or send a
> > pulse. there
> > > > > > will be a rotary encoder that controls the trigger, maybe.
> > > > > >
> > > > > > So in some ways it is like a CNC machine, in that there are
> > motors (3)
> > > > > > and feedback devices to control them, as well as the camera and
> > light
> > > > > > system.
> > > > > >
> > > > > > So please, any ideas, thoughts, questions?
> > > > > >
> > > > > > Grace McKay
> > > > > >
> > > > >
> > > >
> > >
> >
>