Hikvision ColorVu + 4K (DS-2CD2087G2-L)

[ljw2k]

@Parley 2.8mm

 

[ljw2k]

@Parley Just looking again at your 6mm version and just something not right with the focus on that mate, I didn't like to say TBH. I have a 4mm Dahua in my yard and most of it is in focus close and far.

 

[Parley]

@Parley Just looking again at your 6mm version and just something not right with the focus on that mate, I didn't like to say TBH. I have a 4mm Dahua in my yard and most of it is in focus close and far.

I am going to relocate it to another more suitable location as soon as the Dahua 5442-Z12 comes out. At that time I will play with it some more.

 

[ljw2k]

The focus on my 2.8mm and 4mm was good at any range as seen in my last clips. I personally think it has a problem and wait your new clips from when you re position it.

 

[Dave Lonsdale]

My understanding is roughly the above but slightly different.

A sensor records the image as an electronic signal, in other words a photon of light striking the sensor induces an electric field in the sensor which the image processor then turns into a visible picture based on the electronic signal supplied. As Wittaj says, in addition to photons of light producing an electronic signal, you have "noise" which is transient current induced in the sensor by nearby electronic components. These are small electric signals that are much smaller in amplitude than eg the signal from a photon striking the sensor in daylight. As the amount of light reduces, so does the strength of the photon striking the sensor and so the amplitude of the electric current produced is smaller too. There becomes a point in low light where the signal from the photon striking the sensor becomes too weak for the processor to process, so the picture becomes black. In order to get around this or enhance a weak signal, the unit amplifies the signal it receives to make it strong enough to process. This is called the gain. However, the amplification process cannot distinguish between the actual signal and the electronic tranisent signals induced by nearby electric components as a weak signal in low light can have an ampitude not disimilar to that from the background noise of these transients. As a result, both get amplified and processed resulting in rogue pixels as the processor can't distinguish the genuine signal from the amplified background noise due oto their similar amplitude. These rogue pixels are seen in the picture as artefacts because they don't match the colour or shape of nearby pixels and so stand out as "snow". Now I don't completely understand the mechanism of ghosting but at a guess, I'd guess what is happening is the signal of the person who's moving in the picture is so weak due to their movement that the signal is intermittently too weak to be processed. As a result the person has missing pixels where the camera can't process the signal so these appear black. As these change from frame to frame, with pixels in some and not others at certain places in the image, the image takes on a transparent appearance as one frame the pixel is there and the next it's gone, then back etc. As a "video" image is a series of frames per second at a fast rate, if certain frames are missing, the object may adopt a less than solid look. I'm pleased to be corrected if anyone can offer a more accurate explanation.

I believe therefore the antidote to ghosting is to reduce the gain and if the subject can't be seen add more light. Again I defer to those with more experience if they have other successful solutions.

What an interesting assessment CCTVCam - thanks for that. Unfortunately, though, I don’t completely share some of your detail or your conclusion.
1. Clearly, noise changes the voltage presented to the a/d converter and you could argue that if you double the gain, so you double the error. However, the main signal is also doubled. So, with a higher gain the signal available for processing will be stronger and not weaker.
2. Ghosted pixels are not missing and black. They take on the average of nearby frames.
3. The character of ghosting is visually different to that of noise.
4. Ghosting has a characteristic trail. The effect you describe (noise) would be contained within each individual frame. A moving object with ghosting is exhibited simultaneously in separate frames.
5. A faster shutter (although reducing blur) makes the signal to noise ratio even worse.
6. What’s your assessment of ljw2k’s excellent example videos above? Is it your opinion that ghosting is more prominent when he increased the gain from 50% to 100%?

CCTVCam, please don’t be offended by my assessment being at variance with yours. Perhaps my view is far too superficial - there are many other factors and things going on in these cameras that I’m completely ignorant of.

 

[Dave Lonsdale]

@CCTVCam - very well said! It is amazing how much light cameras need. The sensors are small and need light.

I suspect that ghosting is probably present even at lower gains, but they just are not registered or visibly seen until the gain is cranked up.

I have not touched noise reduction and will see ghosting just by cranking up the gain.

And we do not know the algorithms of these cameras - some or all could change the noise algorithm based on a setting and maybe because gain amplifies the noise, the algorithm automatically adjust the NR to a different scale, even if we can control the NR. So a NR at say 30 and gain at 40 may be a different NR scale for a gain at 80 and even though it may show 30 as the NR level because you didn't touch it, you probably have to crank it down as a result of the higher gain?

Mmm. Good point wittaj. Maybe these clever Chinese engineers find ways to get the results their bosses demand.

 

[HomieTheClown]

These cameras are pretty hard to come by in the states.B&H was all I could find. Seemed pricey...does anyone anticipate the turret versions being better and worth waiting for? Is Andy still selling these. Doesn’t seem to have any

 

[Parley]

The focus on my 2.8mm and 4mm was good at any range as seen in my last clips. I personally think it has a problem and wait your new clips from when you re position it.

I was just thinking that the tree and wall being so close to the camera might have an effect on it. The new location will not have those problems. We shall see.

 

[ljw2k]

I can get good captures @ 4ft - 16ft with the 2.8mm

View attachment Hikvision.20210223_170003_1.mp4

 

[ZeeCam]

For this particular case 2.8mm works good. For me it's useless. I would either capture a bush or a garage door.

 

[CCTVCam]

What an interesting assessment CCTVCam - thanks for that. Unfortunately, though, I don’t completely share some of your detail or your conclusion.
1. Clearly, noise changes the voltage presented to the a/d converter and you could argue that if you double the gain, so you double the error. However, the main signal is also doubled. So, with a higher gain the signal available for processing will be stronger and not weaker.
2. Ghosted pixels are not missing and black. They take on the average of nearby frames.
3. The character of ghosting is visually different to that of noise.
4. Ghosting has a characteristic trail. The effect you describe (noise) would be contained within each individual frame. A moving object with ghosting is exhibited simultaneously in separate frames.
5. A faster shutter (although reducing blur) makes the signal to noise ratio even worse.
6. What’s your assessment of ljw2k’s excellent example videos above? Is it your opinion that ghosting is more prominent when he increased the gain from 50% to 100%?

CCTVCam, please don’t be offended by my assessment being at variance with yours. Perhaps my view is far too superficial - there are many other factors and things going on in these cameras that I’m completely ignorant of.

No Offence taken. I don't claim to be an expert. A few thoughts:

1. Yes both signals are doubled but the signals will remain close in amplitude after doubling so whilst processable, amplification probably does little to hlep with separation. This is where the noise reduction maybe has issues in deciding between signal and junk.

2. I think you're correct as the software is probably guessing although on a really low light video, a lot of surrounding pixels maybe black / near black.

3. Yes I agree and different to blur. However, there have been videos in other threads where people have appeared hollow. I can't find the thread but I remember a video of someones back garden where a figure walked across and the face / figure disappeared / reappeared almost frame by frame and was translucent most of the time. That suggests to me maybe some data is incorrect / missing on some of the frames.

4. I would agree, a faster shutter reduces the available light so will make the non noise proportion of the signal weaker. However, there's a difference between ghosting and blur. Something is generating a persistence of image between frames. Maybe as compression guesses the change in pixels from 1 frame to a next where the info is missing it's guessing wrongly based on the pixels in the previous frame. I don't know, I'm not a camera or electronics expert.

5. I find it very hard to conclude anything about ljk's videos as I think the light is sufficient to make the effect marginal. I think there maybe some improvement from 1/60 @ 100 gain to 1/60 @ 50 gain. However, 1/175 @ 50 seems to make the picture slightly less sharp and introduces a bit more blur / ghosting! The exact opposite of what you might expect.

I did a bit of googling and came across this, although explanations seem to vary from website to website. It seems the general concensus is it's a mix between gain, shutter and noise reduction.

It seems the general snow you see is something called Shot Noise and is caused by insufficient photons hitting the sensor.

An article on IPVM blames shutter speed. However this looks to me to be more like blur than ghosting:

Camera Slow Shutter / Ghosting Tested

Reolink have an article discussing ghosts and cctv:

CCTV Security Camera Ghost: Scientific Explanation – Reolink Blog

Could the ghost on CCTV cameras be real or fake? Read this to discover why the CCTV security camera ghosts should occur and learn how to fix the CCTV ghosting problems.

reolink.com

This seems closer to an explanation in my opinion. About 1/2 way down the page, this is their conclusion / suggestions:

3. Improper Camera Settings
Ghost on CCTV cameras can also be caused by slow shutter speed, high gain, high DNR, high DWDR and smoothing settings.

Taking the noise reduction for example, the higher the noise reduction is, the worse the CCTV camera ghosting effect becomes.

Usually around 35-40 noise reduction is a decent balance between less noise and too much security IP camera ghosting.

So they blame a combination of 5 settings - slow shutter speed, high gain, high DNR, high DWDR and smoothing settings.

It would seem that the explanation may lie somewhere between all our conclusions therefore Dave and tweaking those settings relative to each other maybe the solution.

 

[Dave Lonsdale]

What excellent information CCTVCam! On my side, I would have to confess not to have googled for information - instead just spouting what seems right to me. There’s a string of links, some of which I’ve now read. I think my mental definition of ghosting has been too narrow and when broadened, then many camera settings are seen to contribute. Also, one particularly interesting characteristic I had not considered from the Reolink article was, “As the heat is far less transient than light, it will still remain when the source of heat is removed.”

Thanks again for broadening my understanding. I will be taking delivery later this week of one of the cameras that’s the subject of this thread. It will be far better to “practice instead of profess”, especially in my case. We shall see...

 

[Dave Lonsdale]

Good evening CCTVCam. Relaxing in the armchair this evening after a bottle of red, I’m picking up on your valid comment that ljw2k’s video at 1/175, 50% was less sharp than the one at 1/60, 50%. I have a theory. Could it be that this camera has an auto iris and when fixing the shutter at 1/60, the camera has too much light and needs to close the aperture down, thereby increasing the depth of focus (the distance to the camera is far less than the specified minimum of 7m). However, at 1/175, the available light is reduced and the camera automatically opens the iris fully out to f 1.0. So this time, the Sainsbury’s delivery guy becomes more and more out of focus as he approaches the camera. To support this theory, I would otherwise expect the 1/175 video to be darker.

Just thought of another theory - more in line with our key discussion topic. At 1/175, the noise level increases. So, although not severe, the noise reduction algorithm averages the pixels from previous frames causing blur in a moving person.

It‘s a shame that wopi82 no longer visits this forum. For sure, he would know the actual reason.

A technical footnote that I really would like feedback on is about the brightness of an image. I think wopi82 once pointed out that the perception of light to the eye is logarithmic. Does this mean, for example, when it’s nearly dark and the shutter time is lengthened by 300% (1/175 to 1/60), the eye only perceives a change corresponding to log3 - i.e. is only 50% brighter, even though the signal to noise ratio would be improved by a factor of 3? And, of course, when going the other way, the noise becomes three times worse even though the image is still half as bright. I really struggle with this one.

 

[CCTVCam]

I wouldn't know Dave, sorry. Do CCTV cameras even have an adjustable aperture? My technical knowledge is limited.

 

[geezernerd]

Yes, I have one.

Hi Parley - I'm looking to buy a DS-2CD2087G2-LU with a 6mm lens. Is this the model and lens option you purchased and if you don't mind me asking where did you purchase it from? I just installed a DS-2CD2087G2-LU yesterday that has a 4mm lens, but I really need a 6mm but I can't seem to find anyone selling it. I've messaged Andy "EMPIRETECANDY" to see if he sells it, just waiting to hear back from him.

 

[Parley]

Hi Parley - I'm looking to buy a DS-2CD2087G2-LU with a 6mm lens. Is this the model and lens option you purchased and if you don't mind me asking where did you purchase it from? I just installed a DS-2CD2087G2-LU yesterday that has a 4mm lens, but I really need a 6mm but I can't seem to find anyone selling it. I've messaged Andy "EMPIRETECANDY" to see if he sells it, just waiting to hear back from him.

I bought it from Andy. It was his last one. I am waiting for him to restock as I want to buy the 4mm version for another location.

 

[Dave Lonsdale]

I’ve bought the 4mm version of this camera (from ljw2k, it was surplus to his requirements). I’m extremely pleased with it. To get the very best image quality possible, I‘ve used the ‘highest’ Video Quality setting with a constant bit rate of 16,384kbit/s (the maximum). However, with a detailed but static image, the actual main stream bitrate is approx 10 times lower (it reads roughly 160kB/s in the General tab of Blue Iris). It is the same bitrate as this when set to VBR. Is there another parameter in the camera that I’m not aware of that when set prevents CBR or is it a firmware bug?

The bitrate doesn’t seem to increase much when motion is present, when I want as many data bits in the frames as possible. To get the bitrate up to anywhere near 16Mbit, I have to turn the noise reduction completely off at night with a short exposure time.

Has anyone else on the forum checked this out?

I‘m astonished how the image of a 8MP camera can be so good with such a low bitrate, even when the image is static (looks perfect to me). I say this in light of how blotchy the image quality of Dahua’s 5442 family can be with VBR, even though the bitrate only ever falls to roughly 50% of maximum with a static image.

Another small issue that has been mentioned by ljw2k is that the manual LED brightness control does not function and is permanently on maximum (although this is my preferred setting anyway).

 

[alastairstevenson]

(it reads roughly 160kB/s in the General tab of Blue Iris)

With the caveat that I'm not a BI user --- that's bytes per second, as opposed to bits per second in the camera bitrate settings.

 

[ljw2k]

Here is my Bitrate settings in BI in the camera the bitrate is set to 16384

 

[spammenotinoz]

I’ve bought the 4mm version of this camera (from ljw2k, it was surplus to his requirements). I’m extremely pleased with it. To get the very best image quality possible, I‘ve used the ‘highest’ Video Quality setting with a constant bit rate of 16,384kbit/s (the maximum). However, with a detailed but static image, the actual main stream bitrate is approx 10 times lower (it reads roughly 160kB/s in the General tab of Blue Iris). It is the same bitrate as this when set to VBR. Is there another parameter in the camera that I’m not aware of that when set prevents CBR or is it a firmware bug?

The bitrate doesn’t seem to increase much when motion is present, when I want as many data bits in the frames as possible. To get the bitrate up to anywhere near 16Mbit, I have to turn the noise reduction completely off at night with a short exposure time.

Has anyone else on the forum checked this out?

I‘m astonished how the image of a 8MP camera can be so good with such a low bitrate, even when the image is static (looks perfect to me). I say this in light of how blotchy the image quality of Dahua’s 5442 family can be with VBR, even though the bitrate only ever falls to roughly 50% of maximum with a static image.

Another small issue that has been mentioned by ljw2k is that the manual LED brightness control does not function and is permanently on maximum (although this is my preferred setting anyway).

As pointed out by others the cam is kbits and bi is bytes.
1 bit = 0.125 BytesI don't have this cam, butI also feel the CBR is a targeted bit rate, and with the efficiency of H.265 it often won't reach the target bitrate.
This isn't so much a flaw, but a positive of the high efficiency of compression.

Even with CBR, even at a low 15fps, I am disappointed in my 5442. I get slight ghosting (and no it's not shutter speed). Complete speculation but it's like Dahua skimped out on the CPU and the cam can not keep up with the work when there is a lot of motion.
If I drop to H.264 it improved, but the 5442 wasn't a cheap cam.