SG9665GC firmware updates and pre release access

[jokiin]

brightness levels haven't been adjusted yet

 

[Feitelijk]

But that's what the AE does, right?

Pixel captures foton, makes it into electron, pixel closes, number of electrons counted, look up AE value to convert this in to a count.

Or am I missing something and does the exposure time vary across the sensor?

 

[jokiin]

exposure time is the same for the whole sensor

AE readings will determine the level of brightness, but this is something we are still to adjust, we need more sample material before deciding on how much adjustment to do, also looking at (ideally) having some level of user adjustable options rather than just a one size fits all adjustment

 

[alexp999]

Yeah, this is on average as well: Any ideas why the G90 on the bottom seems to be sharper(look at sign and tileson the right)? And less overexposed?
Could be the exposure time right?


Exposure time is not individually set in the AE? That's an electronic gain right?


Is there a way to reduces the exposuretime to gain sharpness but loose some brightness?

Bottom one looks like it has been sharpened post process to me.

 

[Gary D]

Is there a way to reduces the exposuretime to gain sharpness but loose some brightness?

Globally? With the EV setting? I have my EV set to -2/3 on the theory that my observed video has more loss of detail in highlights than in shadows. However, I haven't been able to determine yet if that's having a detrimental impact on night (dark) video.

At least in a still camera, AE algorithms determine a target exposure, and that value is modified by the EV setting... with negative values indicating a darker image is preferred.

It's really very hard to determine if the setting is having a real impact, however. Every single day has different lighting, clouds, etc.

When playing with the EV once, I sat in my garage and kept changing the EV from 0 to -1 and back to 0 (recording after each change.) I noticed a difference in the resulting videos. On the other hand, when I did the exact same thing while parked on the road (at night), I couldn't tell the difference.

 

[Feitelijk]

I tried EV as well, but this still does not guarantee that it reduces the exposure time. It could also reduce the ISO (AE values)

I'd like to reduce motion blur, and that hangs on the exposuretime.

 

[thancam]

Yeah, this is on average as well: Any ideas why the G90 on the bottom seems to be sharper(look at sign and tileson the right)? And less overexposed?
Could be the exposure time right?


Exposure time is not individually set in the AE? That's an electronic gain right?


Is there a way to reduces the exposuretime to gain sharpness but loose some brightness?

I copy your pic in here to easy for comprare.
If SG fix the AE table, so in night footage, it'll be like this picture below.
You can see. Because the F770 dashcam balance light on centrel area. So when you have a car parking in red light just like this. You will see everything around this car very dark. Very hard to see anything even with HDR and WDR which I am sure the high end dashcam F770 must have.

Now, you may think it again. How many time you will go to the tunnel in one month ? Compare to: how many time you will park your car behind another car at night when you are waiting for green light ?
This is the link of video I taken a creenshoot:

 

[Gary D]

When playing with the EV once, I sat in my garage and kept changing the EV from 0 to -1 and back to 0 (recording after each change.) I noticed a difference in the resulting videos. On the other hand, when I did the exact same thing while parked on the road (at night), I couldn't tell the difference.

You know, I really need to re-do that test (where I change the EV values while parked on a road at night.) If it comes out the same (no visual difference), I can post the two videos here and nag @jokiin about it. Perhaps this weekend...

 

[jokiin]

You know, I really need to re-do that test (where I change the EV values while parked on a road at night.) If it comes out the same (no visual difference), I can post the two videos here and nag @jokiin about it. Perhaps this weekend...

https://dashcamtalk.com/forum/threa...ev-exposure-value-settings.13732/#post-179228

 

[thancam]

Globally? With the EV setting? I have my EV set to -2/3 on the theory that my observed video has more loss of detail in highlights than in shadows. However, I haven't been able to determine yet if that's having a detrimental impact on night (dark) video.

At least in a still camera, AE algorithms determine a target exposure, and that value is modified by the EV setting... with negative values indicating a darker image is preferred.

It's really very hard to determine if the setting is having a real impact, however. Every single day has different lighting, clouds, etc.

When playing with the EV once, I sat in my garage and kept changing the EV from 0 to -1 and back to 0 (recording after each change.) I noticed a difference in the resulting videos. On the other hand, when I did the exact same thing while parked on the road (at night), I couldn't tell the difference.

I repost these: (made by Niko)

 

[Gary D]

https://dashcamtalk.com/forum/threa...ev-exposure-value-settings.13732/#post-179228

Not the same as what I saw. At night (only headlights) in an area probably darker than @niko's "pitch black" example, I compared 0.0 to -2/3 and -1.0, and I saw no difference at all. I do see a difference in his images. It might be that I have a CPL installed, or it might be the different firmware version? I don't know.

It's even possible it was user-error. I'm still getting used to the camera and making the settings changes. (And as has been demonstrated by my erroneous reports in this thread, I've even completely missed some settings and reported it as a bug.)

What I found odd was that I noticed a large difference when I did the same thing in my (lit) garage with the headlights hitting a white wall 1 foot (umm.. 1/3 meter) in front of the car.

I'll try to remember to do the test again this weekend and will keep/share the raw files if I have the same results (or admit another error if I DO see a difference.)

 

[jokiin]

CPL will likely mask some of the effect in low light situations, the difference between each step is not huge so the differences can be quite subtle, when you see them presented like @niko did and you can see the progression I guess it's a bit more obvious

 

[funcarfun]

Is there a guide that explains how the EV and AE metering settings can be used?

I'm using the beta 20 firmware, but the recorded video seems a bit too bright, especially white cars in front of me.

Thank you.

 

[jokiin]

Is there a guide that explains how the EV and AE metering settings can be used?

I'm using the beta 20 firmware, but the recorded video seems a bit too bright, especially white cars in front of me.

Thank you.

this thread demonstrates the differences in EV settings https://dashcamtalk.com/forum/threa...ev-exposure-value-settings.13732/#post-179228

AE you will probably do best with default Average setting right now, you might want to look at how the camera is aimed as well as this can affect the results

 

[funcarfun]

I've made sure its pointing a bit downwards. That thread gave me some ideas to try, I'll start by lowering EV a little bit and see how that goes. My next problem that I have to tackle, is the dashboard reflection, which is really horrible, maybe a strip of 3M film on the lower windscreen would help.

Thank you for your help.

 

[Street Guardian USA]

My next problem that I have to tackle, is the dashboard reflection, which is really horrible.

https://dashcamtalk.com/forum/threa...ns-filter-for-street-guardian-sg9665gc.14256/

 

[funcarfun]

I've seen that, but that reduces light.

I prefer to find something to eliminate glare entirely. I'll start by testing some cloth materials over the part of the dashboard that causes the reflection (about 15cm at the edge) and also try a 3M transparent film at the bottom of the windscreen.

 

[jokiin]

CPL filter or something to cover the dashboard will be most effective, the 3M films I don't think are that suitable

 

[normm]

I've seen that, but that reduces light.

I prefer to find something to eliminate glare entirely. I'll start by testing some cloth materials over the part of the dashboard that causes the reflection (about 15cm at the edge) and also try a 3M transparent film at the bottom of the windscreen.

Yeah, I've installed an anti-glare dash mat on my dash. Still shows some reflections as can be seen in the screenshot of Registrator viewer a couple of pages back but MUCH better than when I didn't have it.

I also use a CPL as well so a little darker at night but still really good. I'd say with the CPL the GC is still better at night than my old A118C was.

 

[Dashmellow]

I've mentioned several times that I've been very pleased to finally see the excellent progress and improvement regarding the serious exposure and contrast issues with the SG9665GC I first tried to bring to everyone's attention in the "High Contrast/Dynamic Range Flaw" thread I started last March. The new AE tables have made a major difference in the performance and usability of the camera for anyone who regularly must deal with high dynamic range scenarios. I've been testing out my new number 5 replacement unit that arrived last week and things have been looking pretty good with Beta 19 and the camera has been performing well overall.

Yesterday, I decided to experiment with tilting the camera down slightly to more closely match the standard and oft recommended 60/40 ratio of road/vehicle hood vs sky/horizon.

As many here know, I use a four camera set-up which includes the SG9665GC up front and three Mobius cameras facing the sides and rear. Tilting the camera down to match what has always been normal practice for me with other brands of cameras provides better coverage on either side of the front of my vehicle and achieves the desired overlap of coverage from the two side facing and rear cameras as to more fully capture 360 degree coverage around my vehicle.

Unfortunately, tilting the camera down apparently causes the entire new AE table improvement efforts to completely fall apart. Depending on the conditions, things are almost as bad as they were as demonstrated in my other thread on this subject. Rather than having brief blow-outs of bright glare in the upper end of the tonal range that quickly recover, the camera now manifests fairly sustained blooms and glare which render the video capture highly problematic. With the new AE tables it seems that the angle of aim over a dark vehicle hood is much more sensitive than it was previously. Before the new firmware update changing the tilt of the lens made a relatively a minor difference with the contrast/exposure problems.

The camera is running Beta 19 in the following video clip with no CPL but with a black dash mat. Camera is set to "Average" exposure. (Will flash Beta 20 shortly, but for AE testing it should make no difference with v.19 from my understanding.)

Download

For a frame of reference comparison of lens module tilt, here is a screen shot of the previous camera aim. Obviously, unless or until we see further AE table improvements I will be going back to the previous camera aiming scheme with slightly less hood in the FOV which generally seems to work much better. Still, I'm happy for the current AE improvements which make the camera much more viable considering my usual driving conditions. The question for me is, why does the SG9665GC struggle with this when other cameras apparently do not as demonstrated in my other thread on this issue? Seems to be a Sony sensor thing.



Please note: The rattling noise you hear in the video has nothing to do with the camera. Yesterday, as an experiment, I installed an indoor/outdoor digital thermometer to monitor the ambient temperatures inside my vehicle as well as directly at the camera itself. I have the wired outdoor temperature probe positioned directly on top of the camera near the plug ports and the indoor LCD module mounted on my dashboard. After reviewing yesterday's footage I can see that I will need to do a better job of securing the probe near the camera at the top of my windshield to prevent it from making any noise.

Also, you may notice an unusual amount of macro blocking in the YouTube video. While it may be difficult to see in the video but as the rattling temperature probe will attest to, the road I am traveling on is in quite bad condition and is rather bumpy and this is one of the reasons the video is not quite as sharp as it may be capable of being on smoother pavement. Interestingly, towards the end of the clip I start to pass into a different, more wealthy township where the road is kept in better condition. Shortly after the clip ends the road becomes much smoother and more recently paved.