Hikvision DS-2CD2387G2H - ok at night or better with larger sensor?

garmcqui

Getting the hang of it
Apr 25, 2017
57
33
Hi all,

Currently using a Dahua IPC-HDW5541H in 2.8mm turret form to watch over our cars, hooked up to Blue Iris. Great during the day but not brilliant at night as it's a 1/2.7" sensor and F1.6 lens, so keep the exposure at up to 1/25, which means motion can be blurred. Would rather not use Infrared as it causes lots of issues with spiders.

Current Images:
Day:
IMG_0949.JPG

Night:

IMG_0950.JPG

I want a replacement that gave better resolution during the day, but was also as good as possible in colour at night. I ordered an 8MP Hikvision DS-2CD2387G2-LU, which has a nice bright F1.0 lens and a large 1/1.2 sensor. However, the suppliers have replaced it with the newer hybrid model DS-2CD2387G2H-LISU, which I notice has a smaller 1/1.8" sensor.

Before I open it, will this be as good at night as the older model with larger sensor? Or should I return and try to obtain the 1/1.2" version?

Thanks,
G
 
The 1/1.2" sensor model will be better at night.

That said, ANY $100 2MP or bigger camera practically regardless of sensor, can look good during the day.

At night, if you want color, you will have to supply a LOT of light to get good non-blur, non-noise images for ANY camera. There is no magic. Its 100% about light.
 
That is why I was worried. The supplier assures me that the newer software and fast lens ensure this is almost, if not as, good as the older one with larger sensor, which is becoming increasingly difficult to get hold of now.

This is the 2nd time I’ve tried to buy the G2-LU version, but have ended up with the G2H model
 
I dont own either one so YMMV. But in general yes the bigger sensor on same MP camera will do much better at night.

Most folks just don't really comprehend how much light is needed to get good quality color images at night regardless of camera/sensor.

To freeze an average human walking at night to get a clear, focused, no-blur image, you need to run at least 1/120 exposure. I would argue 1/250. That takes a but load of light
 
I have both G2-LU and G2H-LIU, I noticed G2-LU perform better as it have better sensor. But now i can see they are pushing G3 cameras that will use AI to accelerate quality, color and enhancing sharpness in motions.
 
I have three 2387G2-LU's that are fantastic at night with very little light. That said, way too many firmware bugs so I'd not recommend them. In fact I'm about this close to ripping them down and putting my old 2347G1's back up there. They worked perfectly for 5 or 6 years without a single problem, but I just had to get a newer model to see what it could do. I regret the purchase.
 
I have three 2387G2-LU's that are fantastic at night with very little light. That said, way too many firmware bugs so I'd not recommend them. In fact I'm about this close to ripping them down and putting my old 2347G1's back up there. They worked perfectly for 5 or 6 years without a single problem, but I just had to get a newer model to see what it could do. I regret the purchase.

About the only thing worse than bricking a camera because OCD causes you to update the firmware for no reason is buying a new camera that doesn't work as well as the previous one LOL.
 
Ok, so went ahead and fitted the DS-2CD2387G2H-LISU hybrid model with the 1/1.8" sensor. This is the full colour night image it produces at 1/100 exposure:

Cars 2025-02-26 03.00.05.153.jpg

Which I am more than happy with, especially compared to the camera it replaced, and it doesn't require the external lights to be left on.

Thanks for help.
 
Be interested to see movement at night with that, like a car or human on the street or driveway...
 
+1 above.

Depending on the model, Hikvision either doesn't let you adjust gain or it is so hidden people can't find it.

That looks high gain to me.

A 1/100 shutter with gain at 100 is just as bad as a slow 1/12 shutter and the movement could be complete ghosts like this where the upper body is missing:

1740587558939.png
 
  • Like
Reactions: JDreaming
Yeah you need to lower the gain.

That high gain is what is giving that 'pixelated" look and causing the blur.

It means the image will get darker, but the primary focus of this camera should be to get good ID of the perps door checking your vehicles.
 
Impressive. Though I think you have a shitload of ambient light to work with
 
The more I look at it the more it resembles my 4K 1/1.2" Dahua camera. Same issues.

They simply need more light than we have, thus the pixelization and artifacts
In many cases I get a tiger stripe effect as well. (see snap)

***Direction of light makes a huge difference. In both of our cases the majority of light is coming from the opposite side of the subject from the camera, a classic backlit scene.

192.168.1.110_4K-XCorner_main_20250226202830_@1.jpg 192.168.1.110_4K-XCorner_main_20250226202831_@1.jpg

View attachment 192.168.1.110_ch5_20250226202822_20250226202835.mp4
 
I also have a similar Dahua camera, the IPC-Color4K-T in 2.8mm, covering some of the same area but looking from a different direction. It is interesting to compare it and the Hikvision - so far I wouldn't say either is significantly better.

The tiger stripe effect is strange, I haven't seen that on mine (yet!).

Here is the same scene of someone walking their dog I posted above:



BTW - unsure of how to post video without resorting to youtube...
 
  • Like
Reactions: bigredfish
Yeah you have better ambient light. Even though I have a street light, it’s from the turn of the 19th century and uses candles :(
 
Here's an example from last night. Gain is at 65

What gain setting are you adjusting? I've got 2 x 2387G2-LU and haven't come across the gain that controls the amplification of the light picked up by the sensor.

They seem to use the same firmware for both.

I don't have many street lights in my area (one- on the opposite side of the road and one house down.) I have to supply additional white light. Even then, 1/100 can be a struggle and you run into issues with uneven light spread across a scene causing exposure problems.
 
Last edited: