Those 100ms lightning motion detection models, which are they? - page 4 - General Help and Assistance on using CHDK stable releases - CHDK Forum supplierdeeply

Those 100ms lightning motion detection models, which are they?

  • 40 Replies
  • 21558 Views
Re: Those 100ms lightning motion detection models, which are they?
« Reply #30 on: 11 / April / 2008, 06:30:57 »
Advertisements
Deleted
« Last Edit: 22 / April / 2008, 16:20:43 by Barney Fife »
[acseven/admin commented out: please refrain from more direct offensive language to any user. FW complaints to me] I felt it imperative to withdraw my TOTAL participation. Nobody has my permission, nor the right, to reinstate MY posts. Make-do with my quoted text in others' replies only. Bye

Re: Those 100ms lightning motion detection models, which are they?
« Reply #31 on: 11 / April / 2008, 08:05:48 »
I was also interested in what DataGhost mentioned about similarities between the MD code and the histogram, and why the histogram seems to update so fast. I wonder if, for the sake of lightning and such where you only want to detect overall intensity change and not worry about cells etc, doing a histogram comparison and checking for level changes would be faster. Considering it's already doing most of the calculations needed already, isn't it?

*

Offline DataGhost

  • ****
  • 314
  • EOS 40D, S5IS
    • DataGhost.com
Re: Those 100ms lightning motion detection models, which are they?
« Reply #32 on: 11 / April / 2008, 08:09:00 »
Histogram also has to process the pixels returned by the sensor the same way motion detection needs to calculate change. The Canon one just seems much more efficiently coded, but I haven't really looked into that yet (busy with other stuff). Maybe it's done in hardware (by the sensor?) which would be a great help for intensity-change detection indeed.

Re: Those 100ms lightning motion detection models, which are they?
« Reply #33 on: 11 / April / 2008, 08:19:35 »
Well I wasn't sure if the histogram code was disabled totally when it's not being shown. I guess it is. Another idea would be to use the built-in histogram data. Is that possible? Either way, maybe it's correct the slow response is caused more from shutter/button lag than anything else.

Edit: Oh I see. I thought you were talking about CHDK's histogram.
« Last Edit: 11 / April / 2008, 08:37:13 by jonnythe »


*

Offline ISO64

  • *
  • 20
  • S3 IS + A550
Re: Those 100ms lightning motion detection models, which are they?
« Reply #34 on: 13 / April / 2008, 20:01:34 »
I did some testing using script MD LIghtningB and dataghost md_meter web page. S3 IS set to manual (1/100 s, f=4), manual focus on screen, only variable ISO sensitivity, per Canon options. 10 shots on each ISO setting and the results for average motion detection times are:
ISO  80  205 ms,
ISO 100 153 ms,
ISO 200 134 ms,
ISO 400 119 ms,
ISO 800 137 ms.

Not bad at all, and gettting it close to 120 ms is pretty amazing. A body in free fall gets only 70 mm far from stationary point.

If I try to shoot without CHDK, just looking at the screen and keeping my finger on the button in half shoot position, times are on average, 252 ms. You can not beat the machine! :'(
« Last Edit: 13 / April / 2008, 20:03:33 by ISO64 »

*

Offline Bg~

  • *
  • 27
Re: Those 100ms lightning motion detection models, which are they?
« Reply #35 on: 20 / April / 2008, 15:13:37 »
I've made a little setup with a ruler, an upright rod, and a weighted piece of cardboard. The cardboard slides down the rod while the camera takes a picture of its position. I've gotten ~170 ms (comparing the release point and the point captured by the camera) reaction time, which agrees with what the web tester gave me (A720). I used the MD LightningB script.

Re: Those 100ms lightning motion detection models, which are they?
« Reply #36 on: 22 / April / 2008, 00:18:16 »
I rewrote my test program to display the actual shot delay in ms. Here's some typical results on the S5 at ISO 800, 200 and 80 respectively. I'm not really seeing a difference in speed with different ISO settings.

Re: Those 100ms lightning motion detection models, which are they?
« Reply #37 on: 20 / January / 2009, 13:55:19 »
Hi guys, this is my first post after a bit of lurking.

AFAI have seen, more or less everybody obtains a response time of 100-200 ms.

I've got an ixus 80is, and with the md_testv3 I get between 40 and 80ms. Given that it seems to me unbelievable, let's see how I do it: given that the shot takes 1/5 seconds there are many vertical lines...and I take the one to the left side. The total span in ms (from the first to the last line) is exactly 1/5 second. I'm running it on a LCD screen, so I was wandering if the latency time of the LCD may affect. In other words when I take the shot, the sensor registers not only the line that is being drawn (and the followings) but also the lines that have already been drawn, but that are still on the screen due to latency time.

What do you think?
Ixus 80IS fw ver. 101a - CHDK ver. 0.9.1-678 (13 jan 2009) - GCC 4.3.2


Re: Those 100ms lightning motion detection models, which are they?
« Reply #38 on: 21 / January / 2009, 08:26:56 »
Ehi, I reply myself, after having thought a little...

Well, given that more or less the vertical lines cover a timespan equal to the exposure time, this means that "ghosts" due to latency time are already gone...

So it's official: I've got a 40-80ms response machine;)

Now I only wait the first lightning storm to test the stuff;)
Ixus 80IS fw ver. 101a - CHDK ver. 0.9.1-678 (13 jan 2009) - GCC 4.3.2

Re: Those 100ms lightning motion detection models, which are they?
« Reply #39 on: 30 / April / 2009, 01:34:41 »
Hi,

I loaded the MDFB-080914 script with the latest CHDK autobuild onto my A650, and took photos of the DatagHost timer web site.  Can someone please tell me how to read the photo, because it isn't clear to me how fast my camera is.  For example, in the attached photo, is it 200ms (the first square that isn't too dark), or is it maybe 250ms (the first square that is all white)?

 

Related Topics