Sunny Day. B40 / A118 vs Street Guardian SG9665GC

niko

Well-Known Member
Retailer
Joined
May 1, 2013
Messages
10,668
Reaction score
6,561
Location
Dublin
Country
Ireland
Dash Cam
Street Guardian UK / EU / EEA Distributor
Please Note Firmware version below. Developers may release new firmware in future with different video performance.

In this test you may see both dashcams performance in different conditions of Sunny Day ( low direct sun, back sun, "underground car park", near trees, etc )

Both units have default EV value "0", WDR / HDR On.

Raw / Original video files for both dashcams can be downloaded HERE

Top: Street Guardian SG9665GC. FW: SG20150210.002. SONY EXMOR IMX322 CMOS
Bottom: Tiesfong B40 / A118. FW: 20141204. Aptina AR0330 CMOS
Sound ( Microphone ) is from SG9665GC

Part 1.


Individual videos



Part 2.


Individual videos





 
SG9665GC has some pretty clear issues metering. It looks to be 2/3 EV over-exposed (confirmed in Photoshop). It definitely is sharper though.

The B40 has better metering, but the colors are super-saturated and fake looking and overall softer throughout the entire frame.

Somewhat disappointed by the dynamic range of the Sony sensor in the daylight to be honest. It redeems itself at night.

I wonder if both of these could benefit from being angled downward more. And what all dashcams really need is a configurable spot metering system instead of overall average frame metering (which it seems like they are doing).
 
Last edited:
SG9665GC has some pretty clear issues metering. It looks to be 2/3 EV over-exposed (confirmed in Photoshop). It definitely is sharper though.

The B40 has better metering, but the colors are super-saturated and fake looking and overall softer throughout the entire frame.

Somewhat disappointed by the dynamic range of the Sony sensor in the daylight to be honest. It redeems itself at night.

I wonder if both of these could benefit from being angled downward more.

we're still doing some work on daylight performance, this is just our baseline, anything we do is a balance though between night and day performance so we need to be careful how we adjust it, will take some time to get the best out of it
 
we're still doing some work on daylight performance, this is just our baseline, anything we do is a balance though between night and day performance so we need to be careful how we adjust it, will take some time to get the best out of it

I honestly have no idea how complicated the firmware is, but is there a way to configure the exposure value as a response/function of the shutter speed?

For example, when exposure is less than 1/200 sec (1/200 sec - 1/inf), underexpose by 2/3 EV? - Brighter conditions will then get better exposure across the frame, including the sky.
When exposure is longer than 1/60sec (1/60 sec - 1/30 sec), overexpose by 1/3EV? - This will give the image more grain, but also more brightness across the frame.
 
some may prefer to knock the EV back a touch, it's user adjustable, we do like to get the default setting as close to ideal as possible though so it is something we'll be working on, I have samples going out to a bunch of people so we can widen our test group for firmware betas, stuff like this needs to be tried out in the field as what things look like in a lab environment don't always represent real world results too well
 
@niko, we can always count on you to challenge these dashcams. I like the ending of the side by side video too. With all the driving you do, I hope your car gets good mileage. ;)
 
SG9665GC has some pretty clear issues metering. It looks to be 2/3 EV over-exposed (confirmed in Photoshop). It definitely is sharper though.

I wonder if both of these could benefit from being angled downward more. And what all dashcams really need is a configurable spot metering system instead of overall average frame metering (which it seems like they are doing).
I noticed the same over-exposure in earlier daylight videos. This is something that will have to be tweaked via trial and error. Angling the lens down for these cameras would certainly help in the case when you are headed into direct sunlight.

In the end, it is a compromise. For example for accidents, most of the action I want to capture is down low. The video might look a bit odd but as long as you still capture traffic lights, what more do you really need?

Agree on the sharpness of the new lens. There are several places in the video where you can clearly see the texture of the road, particularly after the car park. If it looks that good on YouTube ....
 
Thanks for the vids Niko. There is a clear difference in sharpness, to the SG's advantage.
 
Thanks.
P.S. - this test here must be considered as an "Extreme daytime conditions", because in February we have very low sun and I havnt seen yet dashcams that could perfectly cope with those conditions without affecting a night video IQ.
Regarding angling more down, - yes, it might help a little bit, but I do believe 40/60 ( @horizon ) is perfect golden middle how dashcam should be aimed. If to aim even more down, then it might not capture some nearby road signs and traffic lights, etc.
I agree with some point regarding overexposure. As @jokiin mentioned, - it is very hard to find a perfect balance between day and night IQ ( within hardware used ). I havn't seen yet ANY dashcam performing outstanding IQ at night and same time at day. There are always a compromises must be made, but IQ firmware tune-up is ongoing and with future firmware upgrades we should see some changes, thats why I have mentioned in the beginning of each side by side test @ " Please Note Firmware version below. Developers may release new firmware in future with different video performance."
My goal is to keep updat people with present stage product development as is, with all its pros and cons, so with upcoming firmware adjustments we could see a difference of "before and after". First and most important priority was to make firmware 100% stable, which took many month of test n try, second stage is night IQ, and now is trying to balance between night and day IQ with all it compromises.
 
Great experiment Niko,it really is a good test.Now,from what i saw b40 is sharp when the car is not moving,you can see the asphalt textures but when you are driving is looses a little sharpness,same thing happends i think with the SG but because it has better and and sensor and FW tailored to them it's not that visible.Can you tell us what's the frame rate for the B40?
 
I don't know what goes into firmware development, but to get the best day and night image, is it possible to configure settings for both daytime and nightime, with a user menu setting that allows the nightime setting to be enabled based on a user defined time?

For example, the menu setting would allow me to select a time that the nighttime setting would be enabled and disabled. Optimum nightime setting would then be used during that specified time frame, and revert back to daytime settings when outside the specified time.

Just a thought, may not be possible.
 
Great experiment Niko,it really is a good test.Now,from what i saw b40 is sharp when the car is not moving,you can see the asphalt textures but when you are driving is looses a little sharpness,same thing happends i think with the SG but because it has better and and sensor and FW tailored to them it's not that visible.Can you tell us what's the frame rate for the B40?

Frame rate should be the same, that is 30 frames per second. Given that both cameras have similar aperture values, the only thing changing here is the shutter speed for each camera.

What you think you may be seeing is likely motion blur - which is what occurs with nighttime videos in particular. Nightime videos at high speeds are even worse. The camera doesn't have enough light so the sensor takes longer to resolve each individual frame (long shutter speed/extended exposure). All this time the camera is moving, and the image being recorded is moving as well.

However, in Niko's test the B40 is underexposing, so the B40 taking shorter exposures than the SG. Shorter exposures tend to result in sharper, crisper images. All things being equal, the B40 should be sharper across the board than the SG, but it isn't. Regardless, there is enough light in both of these videos to allow for maximal clarity at the speeds being driven.

Long story short, the B40 lens and sensor combination cannot compete with the clarity of the SG equipment.
 
SG9665GC has some pretty clear issues metering. It looks to be 2/3 EV over-exposed (confirmed in Photoshop). It definitely is sharper though.

The B40 has better metering, but the colors are super-saturated and fake looking and overall softer throughout the entire frame.

Somewhat disappointed by the dynamic range of the Sony sensor in the daylight to be honest. It redeems itself at night.

I wonder if both of these could benefit from being angled downward more. And what all dashcams really need is a configurable spot metering system instead of overall average frame metering (which it seems like they are doing).

The SG9665GC looked like it's on center metering ( usually the default setting for most cameras). The 322 can collect more light but lacks the dynamic range ( I prefer average metering for my G90) . Not much they can do except playing around with their fake "WDR/HDR" processing.

The Nokia N8 cell phone has a mechanical ND filter built in ( basically to reduce the dynamic range) !



A with/without ND filter comparison ...



My G90 in" stacked/or staggered" HDR ( semi-real) mode with a Fotga ND2 filter ( ~ 50% less light) ...

 
Last edited:
That G90 has really been a workhorse for you mate and it seems to have handled every type of filter you've thrown at it!

Just double checking - is that the A7LA30 or the A7LA70 version please?

On paper the Sony sensors are double the dynamic range of the OmniVision, but they do have a tendency to use that to really brighten up scenes which other sensors are getting close to darkness on.
 
That G90 has really been a workhorse for you mate and it seems to have handled every type of filter you've thrown at it!

Just double checking - is that the A7LA30 or the A7LA70 version please?

On paper the Sony sensors are double the dynamic range of the OmniVision, but they do have a tendency to use that to really brighten up scenes which other sensors are getting close to darkness on.

I was trying the non-variable Fotga ND2 filter on the g90-A7LA30 ( the video was rescaled to 1440p using Kdenlive 0.9.10) and I didn't see any obvious distractions (same for the variable one). I don't mind leaving it on the camera because over exposed images bother me more.

Maybe its "real" WDR function is disabled? If the IMX322 is "stacked" , it suggests HDR processing is done within the cmos ? Any CPU involvement? Here is an example of how HDR is processed ...



Now this is "HDR" ...

 
I was trying the non-variable Fotga ND2 filter on the g90-A7LA30 ( the video was rescaled to 1440p using Kdenlive 0.9.10) and I didn't see any obvious distractions (same for the variable one). I don't mind leaving it on the camera because over exposed images bother me more.

Maybe its "real" WDR function is disabled? If the IMX322 is "stacked" , it suggests HDR processing is done within the cmos ? Any CPU involvement? Here is an example of how HDR is processed ...



Now this is "HDR" ...



Vosonic v747W uses Aptina AR0331 CMOS with true wide dynamic range of up to 120db SN.
For many years it has been a bench mark unit for most Russian forums well known true enthusiasts.
 
How come more people haven't used the AR0331 sensor mate? Is it cost or are there other reasons like power usage / heat generation or something?
 
How come more people haven't used the AR0331 sensor mate? Is it cost or are there other reasons like power usage / heat generation or something?

it's a support issue really, limited SDK support so you're on your own for a lot of issues
 
Back
Top