CHDKPTP - PC Remote Control Performance Analysis - page 33 - RAW Shooting and Processing - CHDK Forum

CHDKPTP - PC Remote Control Performance Analysis

  • 465 Replies
  • 120360 Views
*

Offline SticK

  • *****
  • 779
Re: CHDKPTP - PC Remote Control Performance Analysis
« Reply #320 on: 30 / September / 2012, 02:02:20 »
Advertisements
PHASE II:  S90+CHDKPTP equipment compatibility - results of the day

I know I've been hammering at you because I want the already very good CHDK to be the best on planet - because - it will drive instrumentation, and allow me to squeeze the maximum possible out this fantastic Canon hardware.  So I have some good news today, and wish to report it too. 

The instrumentation computer is heavily loaded with 10 USB ports and LPT port motion control.  Most of the I/O operates "simultaneously."

1) LibUSB coexists with Canon drivers.  This is good news because I can run the S90s on this machine w/o uninstalling the Canon driver for the S50.  Pass.

2) Resource contentions.  Occasionally a shoot in CHDKPTP briefly interferes with the TV videostream which shows up a benign passing horizontal artifact.   Other than that, since the tests began yesterday with S90+CHDKPTP, I have had no major resource contentions in the machine.  In fact I did a tough stress test where I had two live viewers going (RemoteCapture & TV video stream) and CHDKPTP, with asynchronous shooting from both remote controllers - one shooting the S50 and the other shooting the S90 in various combinations.  All have worked perfectly all day under simulated use, both in the stress test, and as if the S90 had already replaced the S50 (the more relaxed condition).  Pass.

3) CPU.  With worst case usage of all instrumentation components, CPU was under a comfortable 30%.  Pass.

4) Motion.  The visual performance XYZ motion and visual focus tracking are very important in my instrument.  Simulating specimen motion with IU OFF, tearing is not serious and visual action at 13 fps is definitely acceptable.  Pass.

5) Liveview image quality.  With RemoteCapture on the S50 you have poor quality compression compounded unavoidable liveview noise in the camera.  That results in twinkling/shimmering pixels so it is difficult to focus, especially when examining partially translucent specimens.  CHDKPTP liveview with your direct frame transfer technique is nothing short of amazing.  The image (both small and large formats) is crisp and noise-free: that's a beautiful combination of CHDKPTP and the S90.  The ability to go to 640x480 "high-mag liveview" will be very much in demand once this is implemented.  In this respect, RemoteCapture is a dud by comparison.  Pass with flying colors.

All these give me increasingly more energy to continue with the remainder of Phase II and on to Phase III once the lens arrives.  This is looking quite good I'd say.

Quote "You can reset it by toggling the live view on and off"

I didn't know about lifetime avg // this helps // thank you.  I tried it just now and read 15 fps.  If I turn on UI and toggle, I get 11 fps // all makes sense.

Quote "I don't currently have time. (edit: probably through the next week at least)"

Thanks! I am looking forward to it ... and thanks for the USB controller tip too.



« Last Edit: 30 / September / 2012, 02:06:09 by SticK »

*

Offline srsa_4c

  • ******
  • 4451
Re: CHDKPTP - PC Remote Control Performance Analysis
« Reply #321 on: 30 / September / 2012, 09:59:54 »
@SticK
I have (hopefully) implemented live framebuffer support. If I got it right, you may experience either
- almost no tearing (if frame transfer time is way below 100ms)
or
- continuous tearing
depending on how correct this implementation is.

The binaries are here: https://subversion.assembla.com/svn/chdk-s1.bin/files/
I don't know which build options reyalp's builds use, I have only activated OPT_LUA_CALL_NATIVE in mine.

I'll soon have preliminary support for detection of file save completion based on PT_CompleteFileWrite in the remote capture test branch. That branch is based on the development version of CHDK with a changed menu and modules which are incompatible with CHDK 1.1 .

@reyalp
What would be a good way to test the effectiveness of this? Furthermore, I wonder whether it is a good idea to always run those event procedures at startup (which I currently do).
I'm doing what is described here: http://chdk.setepontos.com/index.php/topic,5690.0.html

*

Offline SticK

  • *****
  • 779
Re: CHDKPTP - PC Remote Control Performance Analysis
« Reply #322 on: 30 / September / 2012, 10:31:16 »
Quote "I don't know which build options reyalp's builds use"

Should I wait for reyalp to verify the build options, just to be sure?  I currently have installed S90_100c: 2163 test 1 and S90_101a 2142 test-1.  Functionality-wise, do yours match or supersede the more recent of those?

Quote "I'll soon have preliminary support for detection of file save completion based on PT_CompleteFileWrite in the remote capture test branch. That branch is based on the development version of CHDK with a changed menu and modules which are incompatible with CHDK 1.1 ."

Fingers crossed ...  Does this mean I can only install the builds only you post from now on (until reyalp integrates your changes in the full release set)?

edit:
Quote "I have (hopefully) implemented live framebuffer support."
Worst case on the 2.0 machine I have seen is 60 ms.  Right now with UI OFF I see very fast 20 fps.
« Last Edit: 30 / September / 2012, 10:41:28 by SticK »

*

Offline srsa_4c

  • ******
  • 4451
Re: CHDKPTP - PC Remote Control Performance Analysis
« Reply #323 on: 30 / September / 2012, 10:47:13 »
Functionality-wise, do yours match or supersede the more recent of those?
Good question. There are some debug-related options which I didn't activate.
These builds are stock 1.1 code base plus the live buffer code. I suggest you to try whether live view is better, using these. Your test scripts may not produce some of the debug text or debug dumps (haven't checked). If the live view becomes better, the patch can be integrated into the official code.
Quote
Quote "I'll soon have preliminary support for detection of file save completion based on PT_CompleteFileWrite in the remote capture test branch. That branch is based on the development version of CHDK with a changed menu and modules which are incompatible with CHDK 1.1 ."

Fingers crossed ...  Does this mean I can only install the builds only you post from now on (until reyalp integrates your changes in the full release set)?
This is a sort of warning reminder that you'll have to update the modules too if you'll decide to try builds from that (test) branch. CHDK 1.1 and 1.2 modules are not 100% compatible.
« Last Edit: 30 / September / 2012, 11:25:25 by srsa_4c »


*

Offline SticK

  • *****
  • 779
Re: CHDKPTP - PC Remote Control Performance Analysis
« Reply #324 on: 30 / September / 2012, 11:08:13 »
Quote "There are some debug-related options which I didn't activate."

Right now the 101a is working undergoing acceptance with the instrumentation computer and I'd rather not touch it until there is certainty in the build version & options.  For verifying your implementation we have the 100c which can we consider the "development (sacrificial) camera" connected to a standalone USB 2.0 testbed.  For example, once 100c has passed your expectations and my testing, then you can emit 101a.  Is that reasonable?  The 100c also contains the fast C10 SD card (101a is a C4), but I think you are no longer relying on that anyway.

Quote "I suggest you to try whether live view is better, using these."

Thus I will try the liveview on the 100c, although it's a bit of a technical kludge because the 2nd AC adapter hasn't arrived yet // but I expect it will be more doable now that I have two batteries to work with.  Let's see...

*

Offline srsa_4c

  • ******
  • 4451
Re: CHDKPTP - PC Remote Control Performance Analysis
« Reply #325 on: 30 / September / 2012, 11:29:40 »
Thus I will try the liveview on the 100c, although it's a bit of a technical kludge because the 2nd AC adapter hasn't arrived yet // but I expect it will be more doable now that I have two batteries to work with.  Let's see...
Yes, I forgot to mention that the live buffer support code is the same for every fw revision of the camera, so it's OK to only test on one of your cams. The battery will surely last for those few minutes (you don't need to test anything else with this build).

*

Offline SticK

  • *****
  • 779
Re: CHDKPTP - PC Remote Control Performance Analysis
« Reply #326 on: 30 / September / 2012, 11:46:39 »
Yes this helps because on top of it all, the target instrumentation computer has the stripped-down S90_101a.  So I want to avoid handling it at the code level for acceptance testing, and for AC power, if possible.  Also, if you can coordinate 100c with reyalp even for the testbed, that would be useful because there I have memory probing active in the shoot+dcimdl and will be able to test your file transfer implementation and compare memory behavior.

*

Offline SticK

  • *****
  • 779
Re: CHDKPTP - PC Remote Control Performance Analysis
« Reply #327 on: 30 / September / 2012, 12:41:17 »
Quote "If I got it right, you may experience either - almost no tearing (if frame transfer time is way below 100ms)"

This testbed (with 100c) is slower than the target machine (101a) where I get up to 20 fps with UI OFF.  On this testbed I get max 13 fps.   I swung my pencil around in front of the lens for good five minutes and all's fine subjectively.  Then I tried fast-panning camera and even at 10 fps with UI ON -- smooth-n-delicious.  What a difference! .. CHDKPTP is like the camera LCD // you got it right.  Nice job. 

Shoot+dcimdl also work OK on this build.  That is, I see my memory sniff (912k free instead of the usual 888k because of lighter binary I guess), and my file transfer report and freemem in the console that I am used to seeing.  So this build must be the current version, but w/o debug as you suggested.

Anything else you'd like me to check with this version ?


*

Offline srsa_4c

  • ******
  • 4451
Re: CHDKPTP - PC Remote Control Performance Analysis
« Reply #328 on: 30 / September / 2012, 12:59:22 »
This testbed (with 100c) is slower than the target machine (101a) where I get up to 20 fps with UI OFF.  On this testbed I get max 13 fps.   I swung my pencil around in front of the lens for good five minutes and all's fine subjectively.  Then I tried fast-panning camera and even at 10 fps with UI ON -- smooth-n-delicious.  What a difference! .. CHDKPTP is like the camera LCD // you got it right.  Nice job.
Thanks, I'll commit this change then.
Quote
Anything else you'd like me to check with this version ?
No, there are no other changes.

Oh, I forgot:
Can you check whether live view still works in movie mode?
« Last Edit: 30 / September / 2012, 13:04:06 by srsa_4c »

*

Offline SticK

  • *****
  • 779
Re: CHDKPTP - PC Remote Control Performance Analysis
« Reply #329 on: 30 / September / 2012, 13:17:29 »
Quote "Can you check whether live view still works in movie mode?"

I set it to VIDEO_STD (Canon movie icon upper right) and all's fine // went back to M and all's fine.  BTW, I installed your version on the target 101a machine, and that works OK there // in the movie mode too.  That's what I like to see .. two cameras on two machines doing well.

 

Related Topics