SG9665GC High Contrast/Dynamic Range Flaw and other Discouragements

Wow, marz, that was a powerful post!

From what I can see, everyone who reports shortcomings with the camera is respectful to the developers etc, does appreciate the support and effort that is being put in to make it an even better product.

I don't know what "chill pills" contain in your part of the world, but if someone perceives there to be an issue, and in this case a demonstrable issue, what is wrong with them highlighting it - that is how things get resolved. Chilling out over the issue doesn't help get it resolved. Do you think people are being unreasonable over this?
 
Wow, marz, that was a powerful post!

From what I can see, everyone who reports shortcomings with the camera is respectful to the developers etc, does appreciate the support and effort that is being put in to make it an even better product.

I don't know what "chill pills" contain in your part of the world, but if someone perceives there to be an issue, and in this case a demonstrable issue, what is wrong with them highlighting it - that is how things get resolved. Chilling out over the issue doesn't help get it resolved. Do you think people are being unreasonable over this?

Well said, and thank you.

I've been trying to understanding the thinking of a new member who has been here for all of two weeks, posted only 12 times to date and who may not even yet own a dash cam, much less an SG9665GC suggesting that other longtime members discussing an ongoing issue should "****en leave".

After all, DashCamTalk is a site dedicated entirely to the discussion, review, analysis, troubleshooting and marketing of dash cams. Talking about and bringing attention to a significant exposure flaw in an otherwise excellent product can only make it better, particularly with a developer like @jokiin.
 
Last edited:
gPcqWOU.jpg


As dashmellow already mentioned, the dynamic range is not optimal yet. This is on CENTER with WDR ON.
As you can see the weather conditions are excellent.
Guess the AE emphasis is still too much on the top part of the image?




Thanks for posting this. It is another confirmation of my identical experience where the new firmware seems to be overcompensating for bright lighting conditions making the footage way too dark on a sunny day. As in the screen shot you've posted here I find that I am sometimes ending up with footage on a bright sunny day that is so dark that I would not capture vital key details of evidence if a traffic incident were to occur. Of course, THAT is the whole point of this concern rather than "refinement" or nitpicking as some have suggested.

yeah we haven't got into the contrast and brightness tuning as yet, getting the tables right is first, will have another update soon, still doing some work on that, the file you sent will help, thanks for that

Glad to hear this is the next step and I'm looking forward to it. The current beta with the optimized Center Weighted AE tables is a significant improvement. Somehow though, one way or another the problem with this camera always comes down to high contrast and poor dynamic range and the current firmware with all of its apparent improvement is still a reminder of why I originally gave this thread the title that it has.
 
getting the contrast right I think is going to take a bit of work, my preference would be to have some user adjustable brightness and contrast controls, we tried that previously but there were some other issues from that which caused problems but it was in the old SDK so not sure if that limitation still exists or not, that's one of those situations where we'll just have to try it and see, if not it will be a case of trying it in steps, small change, analyse, change again, analyse etc and looking for a balance, I did notice something odd with the screenshot that @Feitelijk posted when I viewed the movie file it came from but don't want to jump to any conclusions on that, needs more research, as always though if there's a screenshot we need the movie file it came from for it to be truly useful
 
user control for brigntness? That would be the EV right?

If I could pick a user adjustable setting I would go for the exposure time: give the user the option of putting emphasis on the exposure time of the iso. I have the impression that the GC is geared toward longer exposure times (I see motion blur) to get the night images bright. But I am not able to read any licensplates at night due to the blur.

The player I use is set at default brightness and contrast (and I can't even change it...)
 
It feels as if the Gamma curve on the GC is way too high and gamma is basically a different concept than brightness or contrast........or exposure.

Gamma is a measure of how tonal changes in the scene map to tonal changes in the output image. When Gamma = 1.0, you get accurate tonal reproduction: For every one stop increase in light intensity in the original scene, you should get 1 stop more intensity (within a limited range, of course).

A higher gamma increases the contrast of the scene - it increases the spacing between tones.

Lower gamma values compress a large scene contrast range into a smaller image contrast range.

What we are seeing in the SG9665GC appears as if we are ending up with values that are too high.

I'm not exactly sure how this works or is controlled and programed in digital SoC cameras like the GC but I believe in a video camera it has to do with the camera's ability to accurately set black levels and gain in reference to how they are output.

Edit: Gamma is a term often used in connection with viewing images on a display, however if the corresponding levels are not present in the original, they cannot be recovered later.
 
Last edited:
not sure at this stage if Gamma is something we can adjust or it's fixed, it's something that will be looked at further though
 
not sure at this stage if Gamma is something we can adjust or it's fixed, it's something that will be looked at further though

Gamma is a difficult concept to understand. It took me a long time to even grasp the basics and I'm not even exactly sure if it is useful here or how it would actually be applied. I am merely trying to describe what I am seeing in the footage. Gamma is a "power law" (a functional relationship between two quantities) which in the case of video is a function relating light intensity with voltage. (which is why it is different than traditional concepts of contrast or brightness).
 
not sure at this stage if Gamma is something we can adjust or it's fixed, it's something that will be looked at further though

In my minimal understanding of how this works, the sensor would inherently have certain characteristic gamma values and the DSP needs to properly map those values for the desired output. I believe it is accomplished with a LUT similarly to the AE LUT. For some reason that seems to be a difficult task with this particular sensor/DSP combination for whatever the reasons, as if the camera is properly mapping the exposure values but not properly mapping the tonal values.
 
In my minimal understanding of how this works, the sensor would inherently have certain characteristic gamma values and the DSP needs to properly map those values for the desired output. I believe it is accomplished with a LUT similarly to the AE LUT. For some reason that seems to be a difficult task with this particular sensor/DSP combination for whatever the reasons, as if the camera is properly mapping the exposure values but not properly mapping the tonal values.

yeah I'm not sure what adjustment is available at this stage, I have another model based on a different solution that we've been working on and the engineer made a gamma adjustment as that is an available option and it made a total mess of the output, that could have been a mistake on the engineers behalf though, not entirely sure at this stage, sometimes these things get lost in translation so it could just be a communication error that lead to the end result, I'm back in China in a few weeks time and will be working on this, I find I get a lot further with this kind of thing when I'm in the same room as the engineers
 
yeah I'm not sure what adjustment is available at this stage, I have another model based on a different solution that we've been working on and the engineer made a gamma adjustment as that is an available option and it made a total mess of the output, that could have been a mistake on the engineers behalf though, not entirely sure at this stage, sometimes these things get lost in translation so it could just be a communication error that lead to the end result, I'm back in China in a few weeks time and will be working on this, I find I get a lot further with this kind of thing when I'm in the same room as the engineers

If your engineer knows how to adjust the gamma but doesn't understand the concept of how it actually works he could easily make it all worse. It is confusing, in part because it is an inverse curve intended to bring the values back to linearity.
 
If your engineer knows how to adjust the gamma but doesn't understand the concept of how it actually works he could easily make it all worse. It is confusing, in part because it is an inverse curve intended to bring the values back to linearity.


This makes me worried...

Well, it could well explain all the problems here. To be perfectly frank about the whole thing, I have speculated as to whether a different group of engineers might possibly do a better job with this camera.


you might have missed what I meant as I didn't explain clearly, the gamma adjustment was on a different product, different solution, this is not done by the same engineer, I gave instructions to make a change and he came back with an update that included a gamma adjustment and he totally screwed the result, quite likely he did know how to adjust it as it's something available on the other platform (different chipset we are evaluating for a future product) but you are quite right that he clearly didn't understand the implication of the adjustment, it was truly very bad, had him undo what he did and just do what I requested and it's back on track though, I have no idea how exactly it was adjusted but if this is the type of mess they can make of it then it's going to need a much more measured approach that's for sure
 
I switched off WDR, and AE is center.
Guess this is what Dashmellow means with the canopy and darkness.

 
I know, low sun and all, but this should really be better. (AE center WDR off)

 
Testing, anyways, you told me it was only beneficial for night time, I can switch it back on if you think it'll be better.

Another one, I can't read the licenseplate even though she's move slowly and lighting is optimal, though from the right.
With the original file here: https://drive.google.com/open?id=0BxJHQyAct4u4dTVSamtJNDJXd3M
 
Back
Top