CHDKPTP - PC Remote Control Performance Analysis - page 29 - RAW Shooting and Processing - CHDK Forum supplierdeeply

CHDKPTP - PC Remote Control Performance Analysis

  • 465 Replies
  • 98120 Views
*

Offline reyalp

  • ******
  • 13746
Re: CHDKPTP - PC Remote Control Performance Analysis
« Reply #280 on: 24 / September / 2012, 16:45:24 »
Advertisements
Camera crashed early this morning at 193136, with (2911 - 1) successful shoots.  See attached ROMLOG, ReadFDir error.

Can you please offer your comments?  You mentioned CHDK should work down to 150K.  Is 193K free within range of crash failure considering that native RAWs are being processed with JPGs?
This is the same malloc failure, so it's safe to say my 150K WAG was too low.
Don't forget what the H stands for.

*

Offline SticK

  • *****
  • 779
Re: CHDKPTP - PC Remote Control Performance Analysis
« Reply #281 on: 24 / September / 2012, 23:31:54 »
S90_100c JPG-only memory leak operational cutoff test

Result: (888808-192032)/4017 shoots = 173 bytes/shoot.  Test ran all the way through clean, pending your confirmation of malloc error in the ROMLOG attached.

Preliminary conclusions 

The last JPG+CR2 test showed 239 bytes/shoot.  Therefore at this high volume bytes/shoot are close despite JPG only.  Here camera crash occurred at 192k vs 193k for both file types.  My prediction for CR2 the performance will likely be nearby to what we have thus far.

Next ........

Because of the success of this test and encouraging results (very fast write to SD with small file -- less probability of garbage collection on SD), for the native CR2 test to begin this evening, I will increase the post-shoot delay by 0.8s.  This won't guarantee the test to be crash-proof against SD write stall, but, we might get lucky.  Because these results will be directly applicable to srsa's solution, the objective now is more to solidify the operational cutoff memory point for this camera rather than worry about write performance.

edit: The value for JPG-only calculated was incorrect.  The correct value is 173 bytes/shoot.
« Last Edit: 27 / September / 2012, 08:21:55 by SticK »

*

Offline reyalp

  • ******
  • 13746
Re: CHDKPTP - PC Remote Control Performance Analysis
« Reply #282 on: 25 / September / 2012, 01:50:59 »
the objective now is more to solidify the operational cutoff memory.
This seems like a waste of time. If you reboot when you get down to 400k you'll get over 1000 shots per session and have plenty of margin.
Quote
pending your confirmation of malloc error in the ROMLOG attached.
You can do this yourself. If it says "ReadFDir.c Line 335" then it's the same assert. If it's a different file name or different number, then it's not.
Don't forget what the H stands for.

*

Offline SticK

  • *****
  • 779
Re: CHDKPTP - PC Remote Control Performance Analysis
« Reply #283 on: 25 / September / 2012, 08:58:07 »
Quote "This seems like a waste of time. If you reboot when you get down to 400k you'll get over 1000 shots per session and have plenty of margin."

Given the memory limitation, that's the final intention once the camera is accepted as a CCD donor, but only after Phase IV.  To pass Phase II compatibility testing, automation will simulate 1000's images in a row, in fact, as many as the camera can handle.  These are very long equipment tests where we need predictable performance from the camera: I have to control spurious errors wherever possible and if there is an equipment contention, I want be sure the camera is fine on its own, and not be led down a garden path due to SD card stall for example (while waiting for the srsa solution - which will also have to be thoroughly tested).  Please remember all this, as I have mentioned it in another way before.  Fundamentally, if it were a waste of time, believe me, I would not be doing it.  Recall the +9 f-stop performance the Canon hardware will give my instrumentation.

In addition the CHDK community gets accurate bytes/shoot that you can add to your memory page once all these Phase I tests are done.  Already you can count on S90JPG and S90JPG+CR2 as being accurate and you can publish those very useful results on your site for folks intending to use CHDKPTP.


*

Offline SticK

  • *****
  • 779
Re: CHDKPTP - PC Remote Control Performance Analysis
« Reply #284 on: 26 / September / 2012, 17:06:01 »
S90_100c shoot+dcimdl CHDKPTP RESULTS SUMMARY

#89 CanonCR2 only

The final CanonCR2-only test was set to 2.4 seconds delay from shoot to dcimdl, using a C10 Duracell 4GB SD card (markings 200X) formatted as FAT16.  In the S90 this SD card has a typical write performance estimated at very approximately 15 MB/second.  Average file size was 10MB.  Also this time the dedicated computer was lightly used with no interference.  malloc failed (ReadFDir) at 194k.  4010 files were transferred successfully before malloc fail.  This was the second test not to experience a spurious error before the expected malloc fail. 

Code is here:
http://chdk.setepontos.com/index.php?topic=8613.msg91217#msg91217
*** Note that the delay in the is set to sys.sleep(1600), 1.6s.  This value has to be optimized for specific camera and SD card.

Summary of all tests

  JPG+CR2 (888800-193136)/2910 = 239 bytes/{shoot+dcimdl} 12MB/shoot - 1.6s delay - one spurious failure
  JPG-only (888808-192032)/4017 = 173 bytes/{shoot+dcimdl}   2MB/shoot - 1.6s delay - no spurious failures
  CR2-only (888864-193832)/4010 = 173 bytes/{shoot+dcimdl} 10MB/shoot - 2.4s delay - no spurious failures

Some preliminary conclusions

These results reveal that with an appropriately long delay, continuous shooting in a remote control environment using CHDKPTP may be possible without spurious failures.  All three tests fail malloc at approximately 192k-194k ranging from 2900 shoots for 2 files to 4000 shoots for 1 file, sufficient to claim that these are the cutoff points for the S90. These are the best-case values and assume no other CHDK features are used except shoot & dcimdl. 

The case of the CR2 is interesting because 4000 files implies 40GB data was transferred to PC over a 4 GB SD card.  That means the card was repeatedly filled at least 10 times, and either all garbage collection is very fast and was covered by the 2.4s delay, or, slow garbage collection did not yet have a chance to occur.  Overall, the results indicate that spurious errors are most likely due to card data reorganization and/or garbage collection initiated with a write to the card by the camera.

Canon guarantees a maximum of 1000 continuous shots.  Because the S90 has perhaps one of the largest freemem pools of the PowerShots, this may be close to the real limit of the PowerShot line under CHDKPTP continuous shooting.  Thus if one is considering CHDKPTP and a camera for long sequence-critical remote control without camera reboot, free memory should be examined.

I attribute the success of this particular large-file test to the very long delay time.  Phase II equipment compatibility testing will begin soon // good news obviously.
« Last Edit: 26 / September / 2012, 17:13:19 by SticK »

*

Offline SticK

  • *****
  • 779
Re: CHDKPTP - PC Remote Control Performance Analysis
« Reply #285 on: 27 / September / 2012, 09:34:51 »
REBOOT obstacle

Rebooting the SX110 while in REC mode does what is expected: camera lens retracts, shuts down and restarts into PLAY mode.  I press Connect and REC.  The Viewfinder image flickers rapidly for about 10 seconds (probably not normal) but then settles and all's fine.

The S90 is different.  In camera in REC mode, I press REBOOT the sometimes camera simply turns OFF and the lens remains extended (the only way to get control is reinsert the battery).  Sometimes the lens retracts and powers OFF, and I can PUP with camera switch to get control.  Sometimes the lens will retract and "E10" appears in the lower left corner before powering OFF, but camera does not PUP on its own.  Occasionally it does work like the SX110 but that's rare.

What could be the problem?
« Last Edit: 27 / September / 2012, 09:53:31 by SticK »

*

Offline SticK

  • *****
  • 779
Re: CHDKPTP - PC Remote Control Performance Analysis
« Reply #286 on: 27 / September / 2012, 22:01:50 »
I need some help on invoking and running CHDK menus with CHDKPTP on S90.

The GUI MENU button invokes Canon menus OK.  However, on the GUI there is no button that sets <alt> on the S90.  So I tried =enter_alt() // that works, but then pressing GUI MENU does not invoke CHDK menus, instead it invokes Canon menus as if <alt> were not set.

How can I get CHDK menus to work and manipulate them from CHDKPTP?
« Last Edit: 27 / September / 2012, 22:12:15 by SticK »

*

Offline srsa_4c

  • ******
  • 4450
Re: CHDKPTP - PC Remote Control Performance Analysis
« Reply #287 on: 27 / September / 2012, 22:24:48 »
How can I get CHDK menus to work and manipulate them from CHDKPTP?
You won't be able to do this, it doesn't work. What exactly would you like to do in that menu?
REBOOT obstacle
...
What could be the problem?
I don't know what the problem is. I can only report that one of my cameras (A470), which I use for testing things, behaves similarly unstable when trying to reboot it while in rec mode.
Try to switch the camera to play mode before you attempt to reboot. Set lens retraction time to zero in the Canon menu before.


*

Offline SticK

  • *****
  • 779
Re: CHDKPTP - PC Remote Control Performance Analysis
« Reply #288 on: 27 / September / 2012, 22:45:44 »
Quote "What exactly would you like to do in that menu?"

Hmmm .. this is very important because I will have no camera button access once installed in the instrumentation.  For now I just need to set "Disable LCD Off" to Always to continue electrical exploration.  Later I may need to set/reset other parameters.

Quote "Try to switch the camera to play mode before you attempt to reboot. Set lens retraction time to zero in the Canon menu before."

I tried it 5 times and it works, thanks // hope all that can done programmatically.
« Last Edit: 27 / September / 2012, 22:54:25 by SticK »

*

Offline SticK

  • *****
  • 779
Re: CHDKPTP - PC Remote Control Performance Analysis
« Reply #289 on: 28 / September / 2012, 00:39:56 »
NOTE #1:  CCD temp is high by about 12degC 

CHDK    Actual S90 (back of chip measured with IR thermometer)
42C      33.2C
44C      33.8C
45C      34.2C
47C      34.9C

NOTE #2:  So that camera doesn't turn OFF, I set Canon Auto Power Save OFF.  Because I can't access CHDK menus via CHDKPTP, in addition I programmatically set "Disable LCD Off Always" by sending =set_config_value(67,1) which seems to work.  Is this correct?
ref:
http://chdk.wikia.com/wiki/Script_commands#set_config_value

 

Related Topics