Forum

Author Topic: Samsung NX1000  (Read 102983 times)

andy_s

  • Full Member
  • ***
  • Posts: 147
    • View Profile
Re: Samsung NX1000
« Reply #30 on: February 04, 2013, 01:14:06 AM »
Great stuff Merry - i'll maybe stick those [recent purchase] 4 x 550D's back on gumtree  :)

Silly question time [from the inexperienced] - maybe you / someone can comment about:

"The continuous shooting mode of 8 fps is very respectable."

Does that mean I should [if wanted] be able to re-trigger shutter in 125ms [plus a little padding e.g in arduino - delay(150);]   

Magnus

  • Jr. Member
  • **
  • Posts: 83
    • View Profile
Re: Samsung NX1000
« Reply #31 on: February 04, 2013, 08:54:02 AM »
Hello!

I just thought I'd add this http://www.photozone.de/samsungnx/703_samsungnx2050f3556?start=1 The kit lens does look really, really good!

Best, Magnus.

RalfH

  • Sr. Member
  • ****
  • Posts: 344
    • View Profile
Re: Samsung NX1000
« Reply #32 on: February 04, 2013, 10:45:14 AM »
andy_s,

continuous shooting mode refers to keeping the shutter pressed. This is much faster than pressing-releasing the shutter (because the camera does not change focus etc.), but you loose the ability to exactly control when the picture is taken. With rapid press-and-release through remote control, you will not reach the same speed but be able to control when the image is taken. You'll have to do some testing to see how short your intervals can be.

Also, depending on how fast the camera can write the pictures into memory, 8 pcitures per second in continuous mode will only be achievable in the first second or so and then drop to less.

Regarding a NX1000 test I've read, autofocus of this camera is quite slow (more than half a second). For synchronous release, this is problematic, so you'd have to pre-focus.
« Last Edit: February 04, 2013, 11:00:22 AM by RalfH »

andy_s

  • Full Member
  • ***
  • Posts: 147
    • View Profile
Re: Samsung NX1000
« Reply #33 on: February 04, 2013, 12:15:27 PM »
Appreciate that RalfH.

The idea is to to [possibly] use the [prior described] remote shutter release [usb] under arduino control and take a series of frames [like quick timelapse] as a turntable/external carousel rotates - "normal" shooting mode / pre-focus [hope these are the correct terms  :-[].   

I'm really trying to understand what "bottleneck(s)" might exist in that flow as referred to by your comment:

"Also, depending on how fast the camera can write the pictures into memory, 8 pcitures per second in continuous mode will only be achievable in the first second or so and then drop to less."

I found a good article [google "Biggest cam secret is Cache memory ?"] that describes the theory but I still need to be able to apply the theory to this particular camera.

While I understand the spec may not exist it'd be nice if it did so that "You'll have to do some testing to see how short your intervals can be" could mostly be circumvented.

Meanwhile - back on planet earth  :)

Thanks again for your comments.

RalfH

  • Sr. Member
  • ****
  • Posts: 344
    • View Profile
Re: Samsung NX1000
« Reply #34 on: February 04, 2013, 12:52:11 PM »
From my experience with the continuous shooting mode (which has to be activated in the camera settings) with other cameras (e.g. most recently the Canon G12), keeping the shutter pressed will result in a rapid series of photos taken which (judging from the shutter clicks) in the beginning appears to be regularly spaced in time but then becomes a bit "bumpy" as the camera has to move data to the SD card to free internal memory while still trying to take pictures. Something like x-x-x-x-x-x-x-x-x-x-x---x---x-x-x---x---x--x-x-x-x---x-x---x-x-x where "x" is shutter and "-" is the pause in between. In fact, "continuous" mode is not really great for really continuous shooting if you want equal spacing in time. I do use it for "document-as-I-walk-along" approaches and feed this into PhotoScan, but in that case my object is large and not moving, and there is no need to synchonize.

Not so very much you can do about this. Using the fastest SD card you can get will definitely be better than using a slow one (i.e., the truly continuous series will be longer until intervals become irregular), but part of the problem is somewhere in the camera where you can't do anything about it.

In my eyes, testing is always better than trusting the specs. And with Ardiuno, you can easily change your programm and try different things. Try a series of 50 or so images at 0.5 second intervals and listen carefully if you camera shoots every time, if yes, make the intervals shorter, if no, make them longer. Test again until you find the limit until which your camera reliable shoots at every pulse sent from the Arduino. For real-life applications, make the interval a bit (perhaps 10-20%) longer than the limit you found, just to be on the safe side.
« Last Edit: February 04, 2013, 12:54:06 PM by RalfH »

mala

  • Full Member
  • ***
  • Posts: 109
    • View Profile
Re: Samsung NX1000
« Reply #35 on: February 04, 2013, 01:12:37 PM »
Ralf has summed up the continuous shooting mode,what matters to me is if the shots are evenly spaced in time and always the same.
With the NX1000 I think I can get 10 shots before the buffer is full and it starts to write to card.
As for shooting in normal drive mode at high speed,no way of knowing yet what is possible without some testing,but I doubt it will get anywhere near as fast because I suspect every image is written direct to card and not buffered.
Regarding auto focus I do not use it normally and just stick to manual focus because neither my subject or camera is moving.

@andy...sorry if I've hijacked your thread a bit!

Infinite

  • Sr. Member
  • ****
  • Posts: 366
    • View Profile
Re: Samsung NX1000
« Reply #36 on: February 04, 2013, 01:39:49 PM »
Ralf has summed up the continuous shooting mode,what matters to me is if the shots are evenly spaced in time and always the same.
With the NX1000 I think I can get 10 shots before the buffer is full and it starts to write to card.
As for shooting in normal drive mode at high speed,no way of knowing yet what is possible without some testing,but I doubt it will get anywhere near as fast because I suspect every image is written direct to card and not buffered.
Regarding auto focus I do not use it normally and just stick to manual focus because neither my subject or camera is moving.

@andy...sorry if I've hijacked your thread a bit!

Syncing multiple cameras/sensors is the biggest issue. I can confirm syncing 6x D800's is a dream. They are built like Ferrari's and are consistently reliably up to 1/250th. From reading online you can't sync flash light faster than 1/250th on these types of cameras. So I would imagine that exposure speed has an impact on the overall burst speed? although that gets a bit to complicated for me.

550D's are terrible at syncing, 600D's OK, although I do get the odd black frame now and then. 550D's you can't sync faster than 1/8th reliably (1 image every 0.125 seconds). 600D's 1/50th (1 image every 0.02 seconds) for one shot, multiple fires NO way. You also have to take into account slight delay for syncing the flash.

So far the only cameras I had experience testing quick fire for multiple frames (8 frames in quick succession, 48 images total) are the D800's and that's under 2 seconds. Any faster is impossible because of the flashes I use aren't fast enough and they have super fast refresh times! plus you are stuck on 1/250th speed anyway.
« Last Edit: February 04, 2013, 02:25:11 PM by Infinite »
_______________________________________________
I N F I N I T E
www.ir-ltd.net

andy_s

  • Full Member
  • ***
  • Posts: 147
    • View Profile
Re: Samsung NX1000
« Reply #37 on: February 04, 2013, 01:55:30 PM »
Appreciate all the comment Lee, Merry, Ralf - precisely the conversation I was hoping for. Thanks...

You may also find the following interesting - google "NX1000 Slooooooow write speeds"

hsmith

  • Newbie
  • *
  • Posts: 22
    • View Profile
Re: Samsung NX1000
« Reply #38 on: February 04, 2013, 05:39:45 PM »
Everybody's comments are right on. 

From what I've learned from multiple sync for Canon cameras, the shortest lag time between single shots is never shorter than what is acheived with continuous ("burst") mode; i.e., 8 fps your NX1000.  The real limitations, as Lee pointed out, is the time it takes to write to the disk (perhaps related to maximum number of shots that at a rate that you can get consistent results with continuous mode), as well as the recycle time for flash.

Andy, I wonder if a rate of 2 or 3fps would be sufficient for a lazy-susan/camera dolly approach, (assuming we get the non-rigid registration working).  Even if you could get sufficiently fast flash recycle times, at one "chunk" taken every 30degrees, for example, that would be 15rpm.  (30degrees/360degrees per revoution x  3 frames per second= .25 revolutions/second, which multiplied by 60 seconds/minute gives 15rpm).

At 8 frames/second, that would be 40rpm, which seems pretty fast.

Just my 2 cents worth...

Harold

andy_s

  • Full Member
  • ***
  • Posts: 147
    • View Profile
Re: Samsung NX1000
« Reply #39 on: February 04, 2013, 10:11:24 PM »
Hi Harold, thx for your comment - please keep them coming !

Lee's an old hand at this stuff [i'm sure] but here's a 3FPS [equivalent] then 1.5FPS [equivalent] Excel copy/paste [below] in which I see the main tradeoffs - bit long winded i'm afraid.

I haven't got photoscan yet [and those 4 cameras are still in their boxes] so I think we'll just need to discuss and experiment a way to solution. I've never implemented a stepper or servo or dc/encoder solution before so just trying to understand the tradeoffs there at the moment [a nice little tool looks to be virtualbreadboard for a clean finish if interested].

Magnus is doing some fantastic work that will impact everything that happens. There's tradeoffs there as well - a capture technique that may substantially increase quality. It could decrease the number of cams required but increase capture time...all good fun !

I've long ago forgotten the maths of "mechanics of machines" but I do have a picture of the "light tent" taking off etc if we're not careful  :). Some is just a repeat of what you said [for the record] with a few extra notes - but tread with care - I have [on occasion] measured wallpaper 3 times [before hanging] and still cut short  ;)

So [say] 3FPS

Frames   Seconds   
3          1   
1          0.33   
         
Scenario 1: [say] 4 Groups of 4 Vertical Cams @ 90 degree offset [to each other] and 30 degree movement chunks [3FPS easily achieveable => 0.33 sec inter frame capture]. 48 Frame total capture in 660ms:         
e.g. timing for 1st vertical group of 4 cams:         
Frame No   Frame capture Start Time (secs)   Degrees (from starting position)   
1           0                                                    0   
2           0.33                                                   30   
3           0.66                                                   60   
RPM Calculation:   Time (secs)   Degrees   
                           3.96                   360   
                           60                   5455   => 15.15 rpm, 360 degree full body capture in 660ms  - [shake, rattle, roll] - but maybe forget about 'non-rigid registration' solution ? [wish I'd listened to the mechanics lecturer more closely :)]
         
Scenario 1[a]: Slow it down with same camera positional setup ?? - double interframe capture time to 666ms [1.5FPS]=>   7.58 rpm. Full 360 body capture in 1.32secs. Manageable mechanical stress ? But 'Hao-Li 'non-rigid registration' solution now required [or is it] ?      
                  
Scenario 2: [say] 1 Groups ONLY of 4 Vertical Cams and 30 degree movement chunks [3FPS easily achieveable => 0.33 sec inter frame capture]. 48 Frame total capture in 3.63 secs         
eg timing for:         
Frame No   Frame capture Start Time (secs)   Degrees (from starting position)   
1          0                                                    0   
2           0.33                                                   30   
3           0.66                                                   60   
4           0.99                                                   90   
5           1.32                                                   120   
6           1.65                                                   150   
7           1.98                                                   180   
8           2.31                                                   210   
9           2.64                                                   240   
10           2.97                                                   270   
11           3.3                                                   300   
12           3.63                                                   330   
         
RPM Calculation:   Time (secs)   Degrees   
                            3.96           360   
                            60                   5455   => 15.15 rpm, 360 degree full body capture in 3.63 sec  - [shake, rattle, roll again] - 'non-rigid registration' already required [or is it] ? [still wish I'd listened to the mechanics lecturer more closely :)]

hsmith

  • Newbie
  • *
  • Posts: 22
    • View Profile
Re: Samsung NX1000
« Reply #40 on: February 04, 2013, 10:38:32 PM »
Very interesting discussion, Andy.

I had only briefly considered more than one pole with cameras attached, but the scenario with multiple poles, with each pole only having to travel a fraction of the 360degree path, has a lot more pros than cons, it seems.  I might have to buy a few more cameras!

I think that only with experimentation can we determine if a living subject can hold still long enough (2-4 seconds?) that a non-rigid registration can be avoided.
Can't wait to try!

Harold

andy_s

  • Full Member
  • ***
  • Posts: 147
    • View Profile
Re: Samsung NX1000
« Reply #41 on: February 05, 2013, 04:34:28 PM »
Some further discussion -

perhaps a little more mechanically realistic [I think]. I also include CFL in the example as a further little addition. I haven't read all the super comments about light in detail yet and what impact long exposure [40ms] may have on image quality /  pscan chunk quality - other than to suppose it's not good [Magnus, i'll read that previous link about "flash should be the primary thing one should look into" really soon :-)]

Here's another [Excel copy/paste] re-jig [pardon the pun] of the 16 cam scenario.  The extra exposure times may be a red herring.

At a total 2.18 sec capture we [probably ?] need a help from Hao-Li [read: how much will it cost and will it work with 16 chunks to provide "quality" output anyway].

With the increased software / motion control / lower overall quality an extra 32 cams etc etc [for instant capture] may be the better option.

Here's the current drivel:

Scenario 1 (revisited): [say] 4 Groups of 4 Vertical Cams @ 90 degree offset [to each other] and 30 degree movement chunks. Total capture time 2.18secs:
         
e.g. timing for 1st vertical group of 4 cams:         
         
Call individual [variable] shutter lag after short-circuit "Td". Assume Td(max) = 20ms ?         
Consider multiple 800Watt CFL ? Exposure 40ms [get the "shades" ready !]         
Start Time = 0      
Times stated below are absolute dfifferences from the start time:

Frame-No   Start shutter short (sec)   1st Curtain Open (sec)   2nd Curtain Close (sec)
1           0                                   0.02                                   0.06
         
Start motion motor (sec) =>   0.06      

Stop motion motor 1 second later [to include 30degrees (Accelerate / Decelerate/Stop/Stop vibration)] (sec)              =>   1.06      
         
Frame-No   Start shutter short (sec)   1st Curtain Open (sec)   2nd Curtain Close (sec)
2           1.06                                   1.08                                   1.12
         
Start motion motor (sec) =>   1.12      

Stop motion motor 1 second later [to include 30degrees (Accelerate / Decelerate/Stop/Stop vibration)] (sec)              =>     2.12      
         
Frame-No   Start shutter short (sec)   1st Curtain Open (sec)   2nd Curtain Close (sec)
3           2.12                                   2.14                                   2.18

andy_s

  • Full Member
  • ***
  • Posts: 147
    • View Profile
Re: Samsung NX1000
« Reply #42 on: February 05, 2013, 11:10:37 PM »
I guess we should supply Hao-Li 16 point clouds taken from a manual turntable experiment with single 4 cam vertical pole  and see what he comes up with. I don't mean going for a smoke between each chunk or anything - but input some significant change between each angle and see what he comes back with. If it's as good as his video we may be laughing - before we get a price anyway  :)



hsmith

  • Newbie
  • *
  • Posts: 22
    • View Profile
Re: Samsung NX1000
« Reply #43 on: February 06, 2013, 05:31:56 AM »
That is a great plan.  I wouldn't be able to produce anything for a couple of weeks, at the earliest.
In the meantime, I'll try to figure out how to use 3d slicer as a possible back-up.

Harold

andy_s

  • Full Member
  • ***
  • Posts: 147
    • View Profile
Re: Samsung NX1000
« Reply #44 on: February 07, 2013, 11:59:33 AM »
Appreciate that Harold.

[If it works at all] In a manual turntable situation [even accounting for added distance/ time] my suspicion is the sequence:
0°, 330°, 30°, 300°, 60°, 270°, 90°, 240°, 120° 210°, 150°, 180° might provide better results than a simple linear progression.