Some trouble with stacking artifacts...

rjlittlefield · Post by **rjlittlefield** » Sun Jan 10, 2016 6:40 pm

etalon wrote:Maybe it is a kind of tilting the sensor while the torque moment of the shutter motors affect the sensor board in the time of shutter action... I don´t know it, but because of the very small DOF, some micron would be enough...

I apologize for not spotting this issue earlier, but your thinking here is very far from correct.

The depth of field at the subject is indeed microns. But at the sensor, depth of focus is larger than that, by a factor of magnification squared.

On the sensor side, the optics are working at NA 0.028 (=0.28/10), about f/18. The computed depth of focus at the sensor, measured as single-sided distance from perfect focus, is 0.00055/(2*0.028^2) = 0.351 mm (for green light, using the 1/4-lambda rule).

--Rik

mawyatt · Post by **mawyatt** » Mon Jan 11, 2016 9:35 am

Saul wrote:If you have Iphone or Android phone , you can use Sensors Kinetics app - you can see all your vibrations easily - after playing 5 mins with this app, I bought Pro version immediately, it helped to solve my remaining problems with vibrations . Even more - I could adjust my delay times in my PLC according real data - it means my stacking process became shorter.

Saul,

Thanks for the App info, I just downloaded it and plan on spending some time trying to drive vibration source down further. I posted this info with your credit for the folks over at Ugly Headgehog.

Cheers,

Mike

etalon · Post by **etalon** » Mon Jan 11, 2016 9:38 am

Saul, thank you for your advise. That sounds interesting. Are you sure, that the sensors in a smart phone are sensitive enough to detect such small vibrations? How made you a stiff enough coupling of the smart phone to avoid a loss in sensitivity?

I apologize for not spotting this issue earlier

No problem, Rik. It´s early enough for me

Thanks for the correction. The truth is always highly appreciated in my threads. 0.35mm seems to be quite a little bit to much for an explanation, as well for me, too.

But this helps me not really to understand, what kind of vibration affects only a part of the sensor...

Cheers Markus

Saul · Post by **Saul** » Mon Jan 11, 2016 1:20 pm

For me it was enough to
put on the top in order to understand what is coming from

JH · Post by JH » Mon Jan 11, 2016 2:48 pm

Hi
I still think that one explanation could be that the sensor is exposed by light sequentially. Even if there is a time when the sensor is totally open the exposure does not start and ends at the same time over the entire sensor.

A theory from my side is that this means that the start of the second curtain will affect the picture of one end of the sensor more severely than the other.

This is a picture to show what I mean at 1/320 sec one part of the picture is exposed by the continuous light and one part is exposed buy the flash and the continuous light.

Regards Jörgen

etalon · Post by **etalon** » Mon Jan 11, 2016 3:22 pm

Hi Jörgen,

thanks for that explanation.

I still Think that one explanation could be that the sensor is exposed by light sequentially.

Are you sure? I think at 1/320sec the curtain should open completely. Sequentially opening of the curtain should be first at less than 1/1000sec? (please correct me if I´m wrong)

Even if there is a time when the sensor is totally open the exposure does not start and ends at the same time over the entire sensor.

I don´t belive that, because those cameras use normally CMOS devices, which have an output amp for each pixel. So readout would happen for the entire sensor at the same time (not like in a CCD, where all electrons moved through one readout amp one after another). Collecting photoelectrons will start when light is falling on the sensor. So your shown pic seems to me like the flash has fired before the curtain was completely open (flash was set to 1. curtain?), or nearly closed again (flash is set to the 2. curtain)...
Maybe is that the reason, why normally the shortest flash sync time is 1/200sec?

Cheers Markus

rjlittlefield · Post by **rjlittlefield** » Mon Jan 11, 2016 4:14 pm

etalon wrote:Are you sure? I think at 1/320sec the curtain should open completely. Sequentially opening of the curtain should be first at less than 1/1000sec? (please correct me if I´m wrong)

Consider yourself corrected.

The time needed for opening the curtain is only a little less than what the manufacturer quotes as the minimum shutter speed suitable for use with electronic flash. For the Canon 5D Mark II, the User's Manual says that "The camera can synchronize with non-Canon compact flash units at 1/200 second and slower speeds."

If the shutter could open significantly faster than that, then the manufacturer would quote a faster sync speed.

Even if there is a time when the sensor is totally open the exposure does not start and ends at the same time over the entire sensor.
I don´t belive that,

JH is correct, though I'm not sure how his explanation relates to his picture. [Edit: Ah, I see. On the monitor I was using, it appeared that the bottom part of the image was completely black. On another monitor, or with a levels adjust, I see that the bottom part of the image is indeed dimly exposed, and with a different color balance.]

Exposure begins when light begins to hit the sensor, after the sensor has been electronically cleared.

With continuous illumination and mechanical shutter opening, the electronic clearing happens while the shutter is still closed. In this case exposure will begin as soon as the shutter curtain moves from in front of each portion of the sensor. The side where opening occurs first will begin its exposure about 1/200 second before the side where opening occurs last.

With continuous illumination and EFSC, the electronic clearing is done incrementally while the mechanical shutter is already open. It proceeds sequentially across the sensor mimicking the motion of the mechanical shutter. Again, the side where clearing occurs first will begin its exposure about 1/200 second before the side where clearing occurs last.

In both cases, exposure ends when the light is cut off by the mechanical shutter moving in front of the sensor.

--Rik

JH · Post by JH » Mon Jan 11, 2016 4:21 pm

The picture I posted is not with EFCS and with a slow and cheap studio flash so I am just happy that it syncs at all (at 1/100).

My point with the picture is to show that all of the exposure of the sensor does not necessarily happens at exactly the same time. Wich lead me to my theory that in EFCS the start of the curtain can have a different effect on different areas of the sensor/picture.

Regards Jörgen

rjlittlefield · Post by **rjlittlefield** » Mon Jan 11, 2016 9:11 pm

etalon wrote:...because those cameras use normally CMOS devices, which have an output amp for each pixel. So readout would happen for the entire sensor at the same time (not like in a CCD, where all electrons moved through one readout amp one after another).

By the way, I am puzzled by these sentences. It's true that CMOS sensors have an output amp for each pixel. But they certainly do not have an analog-to-digital converter for each pixel, or likely even a separate A/D converter for each row or column. Instead there are multiplexors that sequentially route the pixel-amp outputs through a much smaller number of A/D converters, in principle as few as a single converter. See for example the slides labeled "Basic CCD Camera" and "Basic CMOS camera" at https://courses.cs.washington.edu/cours ... magers.pdf.

So, by my understanding the readout does not happen for a CMOS entire sensor at the same time. The reading process may be faster with CMOS than with CCD, but in both cases the pixel data ends up getting clocked sequentially through a narrow bottleneck of A/D conversion.

It doesn't really matter, of course, since in either case no readout will be done until the exposure has ended. But I am curious...

Can you explain in more detail your thinking on this?

--Rik

etalon · Post by **etalon** » Tue Jan 12, 2016 12:26 am

If the shutter could open significantly faster than that, then the manufacturer would quote a faster sync speed.

This sounds logical to me

My point with the picture is to show that all of the exposure of the sensor does not necessarily happens at exactly the same time. Wich lead me to my theory that in EFCS the start of the curtain can have a different effect on different areas of the sensor/picture.

Jörgen, now I begin slowly to understand, how vibrations can affect only a part of the sensor, because all DSLR camera sensors work till now with a rolling (electronical) shutter. But, as I read, Canon is working on CMOS sensors with global shutter for the new DSLR models... Please excuse my slow understanding. My opinion was, that are global shutter CMOS are standard in these days...

Can you explain in more detail your thinking on this?

Well, as I wrote to Jörgen, I assumed, that the common CMOS tecnique in DSLR today is a global shutter CMOS. It works like an interline transfere CCD, which shifted the loads to a light dense register before they are read out. If no further electrons are added through light hits, the D/A converting process, which in fact is the bottleneck on a CMOS, would not be time critical anymore. Since that is a common technique in video cameras today, I assumed wrong, that these techniques are implemented in DSLR, too.

Cheers Markus

Pau · Post by **Pau** » Tue Jan 12, 2016 1:44 am

JH wrote:The picture I posted is not with EFCS and with a slow and cheap studio flash so I am just happy that it syncs at all (at 1/100).

My point with the picture is to show that all of the exposure of the sensor does not necessarily happens at exactly the same time. Wich lead me to my theory that in EFCS the start of the curtain can have a different effect on different areas of the sensor/picture.

IMO the only effect evident with your test picture is that 1/320s is too fast for flash sync and when the flash fires the mechanical shutter is not full open, this only happens at 1/200 or slower, a well stablished point affecting all focal plane shutters
Here you have a link to a most interesting video illustrating how the shutter works:
http://www.photomacrography.net/forum/v ... hp?t=27819

JH · Post by JH » Tue Jan 12, 2016 12:58 pm

Hi Pau

With his very good testing skills Markus have taken pictures that with my eyes seems to to show that there with EFCS can be differences between the top and the bottom of a picture. This differences (again with my eyes) doesn't seem to bee there with flash, indicating that the differences could be caused by vibrations.

What causes this? My theory (for the moment) is that this i caused by second curtain vibrations. For this to be plausible I either needed more information which Rik with his vast knowledge gave us above or take a picture that show that all of the exposure of the sensor does not necessarily happens at exactly the same time.

There is probably something in this technical discussion that I misinterpret but I thought that a picture at 1/320 would be enough to show what I wanted.

Is it better to post a picture taken at 1/200 sec?

Regards Jörgen

rjlittlefield · Post by **rjlittlefield** » Tue Jan 12, 2016 1:46 pm

JH wrote:With his very good testing skills Markus have taken pictures that with my eyes seems to to show that there with EFCS can be differences between the top and the bottom of a picture. This differences (again with my eyes) doesn't seem to bee there with flash, indicating that the differences could be caused by vibrations.

What causes this? My theory (for the moment) is that this i caused by second curtain vibrations..

I think that's a viable theory.

We usually say that with EFSC you don't have to worry about vibration because there's no mechanical movement at the start, and at the end it's too late to matter.

But of course that's only an approximation. It's more true on the edge that closes first than on the edge that closes last, and it's more true for long exposures than for short ones.

With something like a 1/250 second exposure time -- the value shown in the EXIF data of Markus's images -- the last edge of the sensor will see second curtain movement during the entire exposure, while the first edge of the sensor will see it for a much briefer period, only between second curtain activation and whenever the shutter actually starts to cross the sensor.

This story could work to explain the "with liveview" results at http://www.photomacrography.net/forum/v ... 768#182768, where the image from top of sensor (bottom of picture) is sharp but from bottom of sensor (top of picture) is blurred.

However, it is more challenging to explain other results in that same post, in which the bottom of sensor (top of picture) is more blurred with liveview than it is with mirror lockup. To explain that pair by shutter movement, it seems like we need some sort of cancellation between first curtain movement and second curtain movement. That's not impossible, but it seems unlikely enough that I would work hard to rule out other causes.

--Rik

etalon · Post by **etalon** » Tue Jan 12, 2016 2:35 pm

I think that's a viable theory.

I agree completely. For me, the key fact was that DSLRs CMOS use rolling shutters for the readout. When the second curtain starts, it introduce vibrations. But because of the sensors mass it need time to accelerate that mass. In this time, the first part of the sensor is already be read out. When the acceleration of the sensor (caused by the 2. curtain) became an amount which make it visible, only the remaining sensor part, which is still not read out by the rolling shutter process in this moment, would be affected. This is because of the light is still falling on parts of the sensors surface while the mechanical shutter is not completely closed yet. On that part (upper side of the sensor) vibrations became visible, because geometrical informations in the image would be placed on different pixels (caused by laterally shifting occurred by vibration) and smear that part until the shutter is completely closed or/and the readout process is finished over the whole sensor.

Just my two cents, but I see much more clearer now on what maybe happens.

Cheers Markus

EDIT: When I talk about time in that case, I talk about milli seconds...

Pau · Post by **Pau** » Tue Jan 12, 2016 3:21 pm

Jörgen, I was only referring to your partially flash exposed frame image. Undoubtly to do flash tests the speed must be 1/200 or slower with this camera. At speeds significatively slower you could also play with 1st or 2nd curtain sync option effect (at the max sync speed this doesn't apply)

I haven't yet a clear enough opinion about Markus' unsharpness issues to seriously discuss it.