Dahua Color 4K-X in-depth review

wopi82

IPCT Contributor
Joined
Feb 9, 2017
Messages
117
Reaction score
798
Location
Poland
Hi everyone, today I present you a review of recently announced Dahua Color 4K-X camera. This is a 4K resolution model with ultra bright f1.0 lens and huge 1/1.2” sensor. The combination of bright lens and such big CMOS, promises excellent image quality both in broad daylight as well as at night. The camera has been provided by Andy (Empire Technology) in exchange for honest review. 4K-X model is identical to Dahua HFW5849T1-ASE-LED.

The new 4K-X model is a step forward in development of 4K cameras. Onboard we can find Sony’s IMX485 image sensor, announced in 2020. With only one exception, it’s the biggest sensor used in Dahua cameras so far. In years 2016-2017 most Dahua 4K models were equipped with IMX274, a 1/2.5” sensor with a 1.62µm pixel size. This sensor was already labeled as Starvis by Sony, but its night time image quality was far from perfect. Than in 2018 a new 4K sensor has been implemented in Dahua cams. It was IMX334, a 1/1.8” CMOS with a 2.0µm pixels. The image quality was much better from previous generation of 4K cams, but it was still far from what we would expect. Compared to IMX290 Starlight models or even first Full Color models, the night time vision was still poor. Meanwhile, in 2018 Dahua has announced HF8835F a 4K bullet camera based on IMX294. It is a 4/3” CMOS with huge 4.63µm pixels. The camera is however intended for indoor, so you have to use separate housing to operate outdoor. I have no knowledge about its sale numbers, but have a feeling it wasn’t that much popular among homeowners. It’s now 2021 and we have the 4K-X model with pixel size of 2.8µm. Finally the 4K camera has reached the pixel size of first Starlight camera.

Having the same 2.8µm pixel size as IMX290 one might expect similar sensor performance for both CMOSs. This is however not really true. The technology has advanced significantly over recent years, and image sensor size is not the only factor affecting image quality of a camera. Producers like Sony, OmniVision, OnSemi or Smartsens have put a lot of effort to increase quantum efficiency of their sensors. It means that with every new generation, the sensors are more and more sensitive, even when pixel size is kept the same. It’s clearly seen in DSLR or phone camera market. The image quality difference between full-frame camera and APS-C is not so clear now. It’s also seen in surveillance camera market. Take for example a widely appreciated 54XX series based on 1/1.8” Omnivision’s 4 megapixel sensor. Its pixel size is 2.8µm, the same as IMX290, but their image quality is far from each other. That’s why I would expect the new IMX485 based camera, to be at least slightly better than 54XX series in terms of low light image.

camera-photos.jpg

The camera is very solid with most parts made of metal. Only the top part is plastic. At this moment it is only available as bullet. The design is very nice. At the bottom we can find a removable plate, where a memory card can be inserted or the reset button is. What is new to this model, is the presence of a speaker at the bottom. Yes… you can finally talk to a person standing next to the camera. If a speaker is set to 100% volume inside the Web Interface, your voice is well heard from 10-15 meters away. You can talk to a delivery guy. You can scare cats. Or deer. Or you neighbor. Just use your imagination :D

Next new feature are 3 screws for positioning the camera while mounting it. There is a separate screw for pitch, yaw and roll. This is very handy, and it makes the mounting process much easier. It also helps repositioning, as you can now modify one axis at a time.

The 4K-X is the first 4K camera I review, that has extended gain scale. But before I jump to the description of its potential I have to make a confession. I mislead you. It was not on purpose, but I did and I feel bad about it. When reviewing T5442TM-AS-LED I wrote:

As this is a Full Color model, I would expect it to reach much higher sensitivity, than regular models with ICR. At least this was a case with HFW4239T, which reached a stunning ISO200000.

This model stretches its sensitivity from ISO50 to around ISO40000. This is 1EV brighter than non-Full Color Dahua cam. This is not much in exchange for losing ICR.

This is not true. I have no idea how it happened, but the top value is very wrong. Maybe it was temporary malfunction of my brain or maybe one of the firmware updates, but the real value oscillates around ISO 200000.

The worst thing about it, is that meanwhile I did the Low Light Dahua test and didn’t even notice something was wrong. All the values in this video are correct, but this should force me to notice that something was not right. But it didn’t. Even worse, I did the T5442TM-AS quick review as well as recent T5442T-Z review based on the values from T5442TM-AS-LED review by just comparing their brightness. There is a video from T5442T-Z review, that clearly shows, that both 5442 as well as 4239 full color camera reach similar brightness when dealing with moon light.


But it again didn’t make me suspicious. Only now, when examining 4K-X model, I did new calculations to find out I was so wrong about 5442TM-AS-LED as well as all other cams from this series. I’m really sorry for this inaccuracy. Hopefully no one died due to this error :)

So what are the values for 4K-X model? My new measurements indicate that it starts at ISO100 for 0% gain and reaches values of around ISO 320000 or even ISO 400000 at 100% gain. To be clear, I only asses these values, based on comparison with DSLR camera. The process includes shooting an image with DSLR – with certain ISO, shutter and f values - to match brightness level of that from the camera, and then, by comparing details of both images I asses the ISO scale for the CCTV camera. These values are not super exact but quite precise.

I’ve also recalculated values for all other Full Color cameras I have, that is the 4239T-ASE and 5442TM-AS-LED. Here is the result:

ISO-chart.jpg

The 4239T-ASE goes head to head with 4K-X starting at ISO100 and reaching ~ISO400000 in color mode. The 5442TM-AS-LED starts at ISO100 and reaches ~ISO200000 in color mode. The difference in brightness for 5442 in the image below comes from both, lower top ISO value (1EV darker) as well as darker lens – f1.6 vs f1.0. The black and white mode is a bit different story, but more on that later. You can already see from the diagram that the values for BW are different.

ISO-testing.jpg

Generally these numbers are very high. At max gain, the street looks almost totally over burned. Only the far right side can be recognized, as there is far less light reaching this place at night. At top gain and NR at 40%, the 4K-X noise looks very pleasant, resembling fine digital grain. It’s not a crazy mess as in 4239T at 80-100% gain.

At daytime there is no doubt, the 4K-X wins, competing with the two other Full Color cameras. Resolution makes the difference. The image is crisp and detailed with no visible noise reduction smearing or soft edges of moving objects - as in case of HFW1831E. When the WDR is turned on at 50%, the amount of details grows significantly. The codec has hard time with WDR turned on. I found it difficult to get a clear image - with no compression artifacts - in this mode. What worked for me with other 4K cams, didn't work here. I was rising the bitrate to some 15000kbps with no visible improvement. Finally I changed iFrame interval from 25 (one per second) to 150. And this was the solution. The WDR image instantly become clear, even at 9000kbps.

iframes.gif

In color and BW mode this is not an issue. I've spotted however something else. While in color mode, some parts of the image are soft and deprived of fine detail. Look at the image below:

blurry-COLOR.jpg

This is a crop with Region of Interest (ROI) turned on. The left image is from color mode but with saturation at 0%. The right one is a BW mode. The paving on the road blend together in color mode while in BW there is plenty of detail. There were many other areas of the image, where this phenomenon could be seen, like for example distant trees, grass detail or other fine textures. In the image below I was testing resolution in COLOR/BW/WDR mode with test chart.

detail.jpg

Notice the areas marked with arrows. The BW mode is more detailed. WDR introduces a little bit of noise on grey tiles but otherwise is as detailed as BW. Color mode is fuzzy. It has nothing to do with bitrate (used ROI for all modes), or brightness level. It was every time the same difference, whatever setup I used.

The lens of the camera is very sharp and introduces almost no flare at night. The edges of the image are very sharp. With WDR turned on, a little bit of chromatic aberration is visible (see the image above), but that's not an issue. I've noticed though one thing in terms of lens quality. It's some kind of softness or lower contrast in certain area of the image. This is very subtle, and I'm not sure if all of you will get the difference. Not a big issue, but I can see it. To compare it, I’ve rotated the camera to grab the same scene with different areas of the camera frame. Again, all below was done using ROI to avoid compression influence.

softness.jpg

Ok. Now, about the night time performance. Before I recorded the material for night time comparison I thought it would be cool to know, what is the exact light level at the scene. I was always curious, how these LUX values correlate with real world. So... I took a professional studio light meter, and walked on the street to create a light intensity map. Below you can see a night time top view photo of the scene, with LUX rating for incident light.

LUX_map.jpg

The lowest LUX value is at the end of the road. It's a place, where a lux meter showed 0.13 lux reading. Going deeper into the dark area didn't give me any lower value than that.

lux-photos.jpg

According to Wikipedia 0.13 lux is between a full Moon, clear night sky and quarter Moon, clear night sky. It was overcast that night, but in my place, there is plenty of city lights bouncing of the clouds, so there is practically never completely dark.

Before you watch the video comparison, note two things. One: the 5442TM-AS-LED has a 2.8mm lens, while the 4239T and 4K-X are 3.6mm. This work slightly in favor for the last two. I kept the scale difference in video. And two: I recorded this material before I found out the solution for WDR compression artifacts in case of 4K-X – the iframe time increase. So the WDR material from 4K-X looks slightly worse than it could be.


I wonder if you have the same observations as I do.
To me the 4K-X wins in all modes. But…

Out of all modes, color is the worst one. It is quite noisy even though it has the NR set to higher value than other cams. Then comes the WDR, which is much less noisy, and helps bringing up dark areas as well as healing overburned area under the street lamp. The WDR was always said to reduce noise at night, but many older camera models had problems with WDR at night. Often the image became too just dark. Starting from HFW1831E and later the 54XX series the WDR started to look very nice at night, and I personally use it that way. The trick to make it look good at night is to limit higher gain values to something like 60-70%. It will than look very nice. The downside – if there is blackout, the camera will be limited by this setting, and will never boost the brightness to its maximum.

Ok. Let’s leave these boring color modes. Did you just see what happens in BW mode??? The 4K-X transforms into a Terminator. It just shreds the competing cameras. I had a hard time getting any visible noise at 1/50s. I even lowered NR to 30%. Still nothing. So I jumped to 1/100s and the image still looked very nice. Even at 100% gain the image is totally usable. I hate to say that, but this Full Color model works the best in black and white mode. I have never ever seen such a good quality image at night before!

The trick here is probably channel subtraction. The chroma is of course absent, but I think Dahua is also reducing one of the Red/Green/Blue channels. Maybe the blue one, but I’m guessing here. That’s why the 4K-X looses almost 2EV of sensitivity in BW mode, reaching “only” ISO100000. The same thing happens in 5442TM-AS-LED but only by 1EV, from ISO200000 to ISO100000 (refer to the diagram in the middle of review). And so the effect of noise reduction is not so visible. In case of 4239T-ASE only the chroma is being subtracted, thus no brightness change is visible between color and BW mode. Interestingly, at lowest gain there is practically no brightness difference between color and BW. So gain scale in case of 4K-X and 5442 is not shifted but rather squeezed. That's why, when shifting from color to BW in day light doesn't give such brightness difference, because the cameras are operating mostly on lowest gain values. Again, look at the image below, from 4K-X.

bw-vs-color.jpg

At lower gain value, the difference between color and BW is clear. At top gain value, the noise reduction seems to be more aggressive in color mode – even when kept at same value. But still, the BW is more usable.

Summarizing, the 4K-X is a very good camera. You definitely want to have one. Hopefully the noisy (at night) and slightly fuzzy color mode can be fixed by another firmware update. The WDR and BW mode are the best of all cameras I’ve reviewed so far, not only 4K but all of them. It is a pleasure to watch a night time image produced by this cam. Thank you Andy for providing this equipment for reviewing. If anyone has questions about this camera, or the settings I used, I'll be pleased to answer in the conversation below.

Oh… by the way… on 29th of June 2021 Sony announced a direct successor of IMX485. It is IMX585, which is the first sensor to be labeled as STARVIS 2. It has a dual gain feature, enabling a WDR mode in one exposure. It also has a 70% better NiR quantum efficiency – at 850nm. Hopefully we can see this sensor in Dahua camera sometime in the future.
 
Last edited:

CCTVCam

Known around here
Joined
Sep 25, 2017
Messages
2,660
Reaction score
3,480
Excellent review. One point to note though, you had the 4239T at 200% scaling vs 140% on the 5442. Another 60% of digital zoom will degrade the picture considerably. I apporeciate why you did it, to equalise the focal lengths to give an en equal view in each camera. However, that 60% will increase noise, blur and softness in the comparison shots.
 

EMPIRETECANDY

IPCT Vendor
Joined
Nov 8, 2016
Messages
8,259
Reaction score
23,719
Location
HONGKONG
@EMPIRETECANDY when will this camera be back in stock on aliexpress?
I put some stocks on the aliexpress, but can shipped later next week, ready date is Oct.5th.
Chips supply is a big headache, so the stocks just come in small lot:(
 

NightLife

Getting comfortable
Joined
Sep 10, 2021
Messages
490
Reaction score
1,096
Location
Canada
@wopi82 thanks for the thorough review, and candour where you recognized any oversights. And thanks @EMPIRETECANDY (Andy) for steering me towards this Color4K-X camera. I'm really looking forward to getting it installed once it arrives. I think the results will be nothing short of stellar where my specific needs are concerned. After seeing this review a few details clicked which make me confident in my selection.


Cheers
 

EMPIRETECANDY

IPCT Vendor
Joined
Nov 8, 2016
Messages
8,259
Reaction score
23,719
Location
HONGKONG
@wopi82 thanks for the thorough review, and candour where you recognized any oversights. And thanks @EMPIRETECANDY (Andy) for steering me towards this Color4K-X camera. I'm really looking forward to getting it installed once it arrives. I think the results will be nothing short of stellar where my specific needs are concerned. After seeing this review a few details clicked which make me confident in my selection.


Cheers
Get some new stocks in, so next week can ship out them. China and HONGKONG is at public holidays, so shipment has some delay.
 

wopi82

IPCT Contributor
Joined
Feb 9, 2017
Messages
117
Reaction score
798
Location
Poland
@CCTVCam You are right. I did it because I compared these camera as a whole, not just sensors itself. Thus all parameters matter. Resolution also. If this was only sensor comparison, I would have to omit for example lens brightness. The fact that 4K-X is better at night than 5442 and 4239 having 4 times the resolution of 4239 and 2 times of 5442 makes it really good. The BW mode is unbeatable.

@Parley DH-SD8A820WA-HNF is the one I wasn’t aware of :D Thanks for the info.

@alwaysoff That is true. It depends on what you are “observing”. For typical human traffic it will be ok. For LPR it will fail. I just like the WDR image at night, because it keeps all area properly illuminated (no burned out or too dark areas) and reduces noise.
 

CCTVCam

Known around here
Joined
Sep 25, 2017
Messages
2,660
Reaction score
3,480
Thanks for adding the extra tests. It looks very much like the dual lens cam testing in the other thread is going to be the best one at night if they can sort the misalignment of the 2 images. I'm not convinced the 4k-X was the winner in colour above unless you assess purely for minimum light viewing. The 5442T seemed to have less noise and softness. In fact the 4k-X actually looked to me to maybe the worst of the 3 for picture quality under low light colour.

One other point, the gains are set very high - 75% on the 5442 and 60% on the 4K-X. This may affect motion and ghosting with faster walking under some light conditions. Without other tests I'm nto sure if these are relasitic figures that could be set for daily use or not. However, they do beak the 50% rule that most generally follow although it's known some cameras can go above this.
 

alekk

Pulling my weight
Joined
Aug 13, 2018
Messages
114
Reaction score
126
Location
Republic of Boulder
GREAT review @wopi82 - as a Canon shooter, I especially appreciate that comparison and your using the exposure triangle to make apples-to-apples comparisons.

That IMX585 sounds pretty impressive and the "70% better NiR quantum efficiency – at 850nm" should make for even cleaner B&W imaging, but obviously not as much for color.

LOL in your "when the lights go out" video I count at least three cars with red "security" lights.
alek
P.S. I posted over in the Hikvision forums that my "wish list" would be a vari-focal camera with F/1.0 lens and (latest technology) sensor of 1/1.2" (or bigger!) that is "only" 2MP (so HUGE pixels), since you can zoom the field-of-view to be exactly what you want. Now that would be a serious "light sucker" in the dark!
 

wopi82

IPCT Contributor
Joined
Feb 9, 2017
Messages
117
Reaction score
798
Location
Poland
@CCTVCam Extra tests? I haven't added anything else since publishing this review. I've expressed myself wrong. I don't think the 4K-X was the worst on COLOR out of all three cams, but it was definitely not the best one. Yes, the sentence "To me the 4K-X wins in all modes." shouldn't be there. The 5442 looks better (in COLOR), even having less resolution. I haven't heard about any 50% rule. To me, if the camera can go higher than 50% gain, with usable image, why not using it? These are values set by the camera with auto gain. I just checked what the values are, by setting the same brightness in manual mode. And I don't think these are too high. The light level on the street dictates such settings.

@alekk Such sensor already exists. It is Sony IMX482 announced in 2020. Unfortunately I haven't heard of any camera using it so far.
 

alekk

Pulling my weight
Joined
Aug 13, 2018
Messages
114
Reaction score
126
Location
Republic of Boulder
Thanks for the info @wopi82 and I don't mean to hijack your (excellent) camera review thread, but you got me looking around and I found this Sony page that shows their various security sensors with links for more info - I'm sure you already know about it.

That IMX482 appears to actually be a 8MP sensor that only outputs 1920x1080 as they are using pixel binning (?) ... gotta love the effective 5.8uM pixels. The 585 (and the 485) are also 8MP and provide 3840x2160 output ... but also appear capable of doing 1920x1080 via (the same?) pixel binning ... which presumably is better than downsampling an 8MP image.
 

Corvus85

Getting the hang of it
Joined
Aug 18, 2021
Messages
495
Reaction score
79
Location
Australia
My Color4K-X has been giving me some strange trouble recently.
After tweaking my settings recently (shutter speed/gain,etc), I've noticed that it intermittently switches BLC on/off throughout the day/night.

I've got BI to schedule the camera to go into night/day mode according to the sun rise/set - however I can confirm it's not switching profiles. The only thing it's switching is BLC on/off.

It was working fine for weeks before I tweaked my image settings last night. Ideally, I'd prefer to leave BLC on, but I have no idea how to do this.

Below, you'll see a screenshot of what it looks like when it switches off BLC. As you can see, the image is horrible. It looks washed out, with a yellow tint.
View attachment 143845

When I log into the camera, I see that it thinks BLC is still 'on'. However, when I change it manually to 'off', no difference is noticeable in the image (because it is already indeed off).

When I change it back to BLC again manually, the image is restored to its intended quality, as seen below:

View attachment 143846

However this only lasts about an hour or two before the camera switches BLC off again. Sometimes, I'll notice that it switches it off and then on again over several hours. This also occurs during the day, however it's less noticable unless it's a bright sunny day and there are hard shadows (when BLC switches itself off, the shadows look extremely dark).

Here is my current image settings:
View attachment 143847
View attachment 143848
View attachment 143849
View attachment 143851


Can you please tell me what I might be doing wrong? It was working fine before I tweaked settings last night.
 
Last edited:
Top