Dahua IPC-HDW4231EM-AS review

Discussion in 'Dahua' started by wopi82, Sep 11, 2017.

Share This Page

  1. wopi82

    wopi82 n3wb

    Joined:
    Feb 9, 2017
    Messages:
    5
    Likes Received:
    20
    Location:
    Poland
    I've been playing with Dahua IPC-HDW4231EM-AS for a few months, so I've decided to share with you some thoughts about it. Maybe it will be helpful for somebody willing to buy a CCTV camera.

    First, some background. As a total beginner in the topic of security cams, before deciding on the camera model I went through dozens of pages describing hundreds models, different brands and different functions. In Poland there are around 40 brands available, 70% of which is Chinese low quality junk, being sold under different names. What drew my attention while going through gear specs, was common lack of information about image sensor used in certain models as well as rare information about lens brightness – especially varifocal type on the long end. These two parameters – together with image processor - are crucial for image quality, which was the main aspect for me. Hikvision says absolutely nothing about sensor type (apart from its size type). Dahua is very restrained in giving such info (at this moment the information on their website has been greatly improved, but with some errors). I had to dig into the world of surveillance image sensors, to find out the best solutions. Finding certain information here was sometimes hard to do. There are three main players in this field: Sony, Omnivision and ON Semiconductors (previously Aptina). After a few days of searching I knew Sony's Starvis technology (or Starlight, as Dahua used to call it) was my destination. Even Sony on their webpage has incomplete list of available sensors. They create really cool, advanced stuff like Digital Over Lap WDR, Dual Photodiode PDAF, onboard stabilization or SME-HDR but somehow are unwilling to advertise this properly. Instead of showing real life examples of how amazingly their new sensors perform, they show us some goofy graphics of wild cats sitting in the grass at night with photoshoped stars overhead. As somebody wrote on XDA-Developers webpage about SONY's policy:

    "...it really is a shame that it is so hard to access some of this information, even basic product information. When companies try to put information on their websites, it often can be rather inaccessible and incomplete, in large part because it is often treated as a secondary concern of the company’s employees, who are more focused on their main work. One dedicated person handling public relations can make a huge difference in terms of making this type of information available and accessible to the general public, and we’re seeing some people trying to do just that in their free time. Even on the Sony Exmor Wikipedia article itself, where over the course of a couple months a single person in their spare time laid most of the foundation to take it from a nearly useless 1,715 byte article that had been mostly the same for years, into the ~50,000 byte article which we see there today with 185 distinct editors. An article that is arguably the best repository of information about the Sony Exmor sensor line available online, and we can see a very similar pattern on other articles. A single dedicated writer can make a substantial difference in how easily customers can compare different products, and in how educated interested consumers are about the subject, which can have far-reaching effects."

    I did a list of Sony fullHD /4k sensor's sensitivity to rank them. There is also Sony's own list (SNR1s Image Sensor for Industry | Sony Semiconductor Solutions Corporation). IMX185 was the top one, due to its sensitivity + pixel size (at this moment it is IMX385 with twice sensitivity of IMX185, but not available yet on the market). Next was Starvis IMX290/291. I was hoping to find some performance examples on YouTube. No luck here. Only a few clips from Nayr and some chinese third-party producers. I found some tests from dashcam owners, like this one:

    or this one

    At that point it turned out that even having IMX291 onboard, doesn't mean it will perform as supposed. The chip, responsible for image processing (SoC) is the next crucial part. Warrior XM G1S DashCam is equipped with HiSilicon Hi3516D SoC while Viofo A119S DashCam is based on Novatek NTK96660SoC. Both equipped with IMX291 and yet they perform totally different. Dahua uses Ambarella SoCs. Finally I’ve reached this forum, where Nayr presented a review of few Dahua models and gave a link to a very useful document, listing all recent Dahua models with all necessary information. I've listed all models I've been interested in, and ended up with a choice between fixed lens IPC-HDW4231EM-AS and varifocal IPC-HDW5231R-Z. On AliExpress the price tag between these two models is somewhere around 70 bucks, however in my country, the varifocal version is almost twice the price of fixed lens. So finally I decided for HDW4231EM-AS. End of the story.

    Now for the camera itself. Its image quality is outstanding. I was literally blown away by its night time performance. The build quality is very good. All metal. As this is my first security camera, I have no reference to other models/brands, but this one seems really solid. I love the fact, to be able to set all image parameters individually, like in DSLR camera.

    It's a pity the fixed lens version is 1 stop darker than varifocal HDW5231R-Z. It seems to me illogical, as producing bright fixed focal length lens is easier and cheaper than producing same brightness zoom version - at least that's how it works in DSLR world. Nayr explained to somebody on this forum, that the producer do that to keep as much in focus as possible. It is of course true that brighter lens means shallower depth of field, but due to sensor size, at these focal lengths (2.8mm/3.6mm) the difference between f1.4 and f2.0 is negligible. See my test image from Canon 6D below. From 2m ahead, there is no difference. Interestingly HDCVI variant of this camera is distinctly brighter - f1.5.

    porownanie-GO2.jpg

    At this moment Dahua has already presented a whole new range of cameras with extended PoE. The HDW4231EM-ASE is now equipped with f1.6 lens, which makes it around 2/3 more sensitive to light, compared with f2.0 model. Dahua HDW4431EM-ASE is now also equipped with f1.6 lens but the specs on the webpage are incorrect. They did not replace the sensor (most likely OmniVision) so there is no Sony Starvis inside. Sony has no 1/3 inch, 4 megapixel sensor on their list. Thus the sensitivity is at best 0.05lux at 1/3s and not 0.007lux as stated on the webpage.

    As the GAIN scale was a little bit unclear for me, I've compared it with ISO rating, which I am much more familiar with. Below you can see a diagram of GAIN translated to ISO scale.

    iso-gain.jpg
    zestawienie ISO.jpg

    This camera reaches ISO 20000 at GAIN 100! Combined with very efficient noise reduction, it produces same or even better quality video than my Canon 6D DSLR. Here is an image of night sky.

    noc (0-31-01-07)b.png

    These are stacked frames from recorded footage, between 1am and 5am. You can see star trails, flying planes, some satellites and even faint trails of falling stars. The camera was set to 1/3s exposure, GAIN 80 and noise reduction: 50. Of course setting security camera to 1/3s exposure makes no sense from a security point of view due to motion blur – if something is moving in the frame. But for somebody willing to observe sky phenomena, distant storms or whatever hard to catch with human eye, this device might be perfect. Add to this motion detection, and you can for example catch every shooting star during meteor shower ;)



    Just for fun, here is a comparison of HDW4231EM-AS with GoPro HERO 5 Black in fullHD video mode. Dahua was set to 1/50s - due to brighter lens, no noise reduction, sharpness 0, all other parameters were set to 50. GoPro was set to protune, 1/25s, high sharpness and high bitrate.

    GOPR2589 b small.jpg

    Here is daytime and nighttime comparison with all other devices I had available at the moment of testing. This one is just for picture detail comparison. All devices were set to maximum performance with all enhancements turned on.

    DZIEN - porownanie rozpietosci tonalnej.jpg

    NOC.jpg

    During moonlit night, when set to black and white mode it can see almost like during day. Sunlight reflected from moon is reach in infrared so in BW mode with IR filter removed it looks almost like if the sun was shining. The far part of the meadow behind the house on the example below has no artificial light reaching. Only pure full moon light. Notice the difference between color and BW mode. It was almost total darkness with naked eye.

    porownanie przy ksiezycu.jpg

    WDR mode works very well. The two exposures blend almost seamlessly. At rare occasions, when fast moving objects appear in frame (fast moving car, bird) some dark smears can appear around object. The WDR scale ranging from 1 to 100 works slightly strange. From 1 to 44 the image is just brighten up, with no real WDR. From 45 real WDR DOL function is activated. 45-50 range is most effective in my opinion. From 50 up, the image gets flat and starts to look surrealistic. Ultra series have SSA (Smart Scene Adaptive) function, which can further enhance the final image, bringing shadows up and lowering highlights. Nayr gave an example of this function, somewhere on this forum.

    test hdr.jpg

    IR led can reach objects as far as 25-30 meters at 1/25s exposure and full gain. Of course stepping down to 1/3s will vastly increase that range, but again, there is no point in setting such shutter speed for security porpoise. The house in the example below is 17 meters ahead of the camera. The problem is that IR led lights up every single insect in the air. Even tiny pollution like seeds or dust becomes visible. This in turn triggers motion detection and fills up memory card with unnecessary footage. Thus I never turn IR on.

    porownanie IR.jpg

    Noise reduction is based on frame stacking. It is extremely effective. In fact I have never seen any noise reduction software so powerful as this one. I guess it is the same algorithm as used for deep space photography, when stacking multiple exposures for clear picture. Yet this one is realized live.

    noise reduction comparison.jpg

    So far I have no NVR. Everything is recorded on microSD memory card inserted inside camera. With motion detection turned on, H264 compression (4Mb/s), 16GB card gives me 6-8 day archive, depending on amount of motion detected.

    I will not discuss network stuff, as I have absolutely no knowledge in that part. The camera is connected straight to my router. I use web interface and it works fine. I can operate the camera through IE and Chrome. Downloading recorded footage through the interface is slightly annoying and slow. I was trying to change recording destination to external USB drive, connected to the router, but with no luck. Maybe someone can help me with this.

    Here is a night footage of 4 foxes playing on the street in the early morning. The camera was set to 1/50s and limited to maximum GAIN 80.



    If you wonder, how this camera would perform, compared to other Dahua models, I made a simulated comparison of sensitivity, based on minimum illumination information on Dahua's web page. Unfortunately Dahua is really inconsistent with this parameter. For example, for every camera, they give two values - for 1/3s exposure and 1/30s exposure. If the value for 1/3s is 0.007lux than for 1/30s is should be 0.07lux, so one tenth of first value. The values on their webpage are often very different. The camera with f1.4 lens gathers twice as much light as a camera with f2.0 lens, assuming all other parameters are identical. Thus if the first one is described as 0.007lux than the second one should be 0.014lux. That’s not always the case on Dahuas’s webpage.

    For comparison I took following cameras:

    IPC-HFW8232E-Z
    Sensor: 1/1.9” 2MP SONY IMX185
    Lens: f1.5
    Min. Illumination: 0.002 for 1/3s exp.

    IPC-HFW8231E-Z
    Sensor: 1/2.8” 2MP SONY STARVIS IMX290
    Lens: f1.4
    Min. Illumination: 0.005 for 1/3s exp.

    IPC-HDW4231EM-ASE
    Sensor: 1/2.8” 2MP SONY STARVIS IMX290
    Lens: f1.6
    Min. Illumination: 0.007 for 1/3s exp.

    IPC-HDW4231EM-AS
    Sensor: 1/2.8” 2MP SONY STARVIS IMX290
    Lens: f2.0
    Min. Illumination: 0.009 for 1/3s exp.

    IPC-HDW4631EM-ASE
    Sensor: 1/2.9” 6MP SONY STARVIS IMX326
    Lens: f1.6
    Min. Illumination: 0.06 for 1/3s exp.

    IPC-HDW4431EM-AS
    Sensor: 1/3” 4MP most likely - OmniVision OV4689
    Lens: f2.0
    Min. Illumination: 0.08 for 1/3s exp.

    porownanie jasnosci.jpg

    This is a simulated comparison of maximum brightness the camera is able to deliver. It’s not to compare image quality or any other parameter. I set the parameters quite precisely, but afterwards I found out that the street lamp light is fluctuating very gently. So treat it with a grain of salt :)

    Hopefully this description will be helpful for somebody, willing to buy a new security camera. If you have any questions, or think I did something incorrectly with my testing, please let me know.
     
    Arjun, zeoclang, fvdwerdt and 13 others like this.
  2. mat200

    mat200 Pulling my weight

    Joined:
    Jan 17, 2017
    Messages:
    485
    Likes Received:
    199
    Wow! Excellent, very nicely done. Thanks wopi82!
     
    lgRich likes this.
  3. yuhnam

    yuhnam n3wb

    Joined:
    Jan 16, 2017
    Messages:
    1
    Likes Received:
    0
    Excellent and impressive review! Thank you, I've just confidently ordered 2 of them.
     
  4. pcdo78

    pcdo78 Young grasshopper

    Joined:
    Aug 6, 2017
    Messages:
    32
    Likes Received:
    6
    Thanks for the review. This is the camera I've most been interested in for budget reasons. Just to clarify, the HDW4231EM-ASE has a f1.6 lens and the same Sony STARVIS that's in the non-ePOE version?
     
  5. wopi82

    wopi82 n3wb

    Joined:
    Feb 9, 2017
    Messages:
    5
    Likes Received:
    20
    Location:
    Poland
    Exactly. The difference won't be huge, but it's always better to have brighter lens.
     
  6. pcdo78

    pcdo78 Young grasshopper

    Joined:
    Aug 6, 2017
    Messages:
    32
    Likes Received:
    6
    Good to know. I saw the ePOE version for the same price as the non-ePOE version so I agree go for the better lens.
     
  7. TVT73

    TVT73 Getting the hang of it

    Joined:
    Aug 29, 2016
    Messages:
    337
    Likes Received:
    85
    Location:
    Germany
    Hi,
    you did many work in your review. Especially it´s nice to see a total different way of reviewing. We need more guys like you :goodpost:

    I have some minor problems as you compare several devices together like your DSLR and GoPros, which in my opinion can´t be really compared together due to different scope of applications.

    Some parts of your review based on examinations between your DSLR Cam, like your ISO/Gain table. You did it just by view comparison between those pictures? I think your ISO table is not comparable to other devices, the gain factor is an software based picture manipulation, and depends on the camera and firmware.

    The simulated comparison of maximum brightness the camera is able to deliver, your lux pictures were all taken with the hdw4231em-as? I don´t think that this is real live representative and it´s based on suggestions. It looks nice but sorry not realistic. The different lens systems,fov , sensor sizes and post processing parts, this not possible to simulate in that way for realistic results.
    The difficulty in comparisons is to see the behaviour in low light motion frames. So the only realistic way is to parallel record the same motion stream of the same scene with all those types of cams with the same settings. If you took still frames of moving objects under this conditions, it would show a representative difference between them.
    Still pictures can´t show the problems of security cams. You used sometimes mjpeg, why? This would not deliver the same as h265, which is a much better stream codec, especially for ip cams. When you use h265 you can also reduce 3dnr. mjpeg is outdated.

    Some hints for real life:
    - prevent to use too high noise reductions, it causes motion bluer the sony starlight series needs under normal condition a maximum on 20-30
    - same as wdr, too high rates like 50 (normaly is 20-30 enough) causes negative looking pictures under bad weather conditions.
    - fixed shutter rates are only for special purposes useful, here are some firmware limitations because shutter range using has bugs as I noticed at my weathercam setup
     
  8. wopi82

    wopi82 n3wb

    Joined:
    Feb 9, 2017
    Messages:
    5
    Likes Received:
    20
    Location:
    Poland
    Hi TVT73,

    I’m glad you like my review and you’ve responded to it.I’ll try to address all your concerns and explain why I did things, the way I did.

    I’ve compared this security cam to devices such as GoPro or Mobius, simply because I have no other security camera to deal with. This comparison is however not so meaningless. All of these devices are small action cams or dashcams equipped with sensors of similar size and similar optics. Mobius and Xblitz cams are based on Aptina AR0330 1/3" sensor and NTK96650 SoC. GoPro 4 and 5 are equipped with 12 megapixel 1/2.3" sensor and Ambarella SoC. This, by common sense, would generate more noise due to smaller pixel size, but fullHD video is downsized from almost whole sensor area, thus the noise is averaged slightly and should become less visible. All other cameras are also based on small sensors. The goal was to show, that despite this Dahua cam is cost friendly (I’m deliberately not using the word “cheap”) it can deliver image quality significantly better than some other devices, some of which are twice or three times the price of Dahua.

    In topic of ISO vs GAIN, I’m sorry but I cannot agree with you. GAIN and ISO are in fact same exact thing. They both describe amplification of a signal from the sensor. This amplification is executed internaly by onboard amplifiers, after the shutter is closed, both in DSLR’s and cameras. It is not software based (most commonly). ISO scale is consistent across all digital photography equipment of all brands. Whether it is a small point-and-shoot camera from Sony, a bridge compact from Canon or professional DSLR like Nikon D5. Look at below example, I did last night.

    zestawienie.jpg

    Canon 6D and Panasonic FZ1000 are completely different cameras, but if I set same parameters on both, the image will have almost exactly same brightness. I even set similar field of view to match images. At high ISO, noise starts to influence the final look. Panasonic gets a blue tint. But the overall brightness is still equal between the cameras. GAIN is a bit different story. This scale is used for example in professional video equipment like Sony CineAlta, Arri or RED and is expressed in decibels. As a base it uses native sensitivity of given sensor, so it may vary from camera to camera. Even my old Canon HV20 camcorder has GAIN scale. However I work mostly with photo equipment and my brain works in ISO. That’s why I wanted to translate GAIN to ISO. And yes, I did it on organoleptic basis ;) Having such translation, lets me or any other person simulate, how this camera would perform in given light conditions. Instead of unplugging it from a fixed mount, dragging a network cable to a certain place, I can take my DSLR or compact camera, set the parameters and see if it will see anything at given place during the night (omitting the issue of noise level).

    You’ve asked, why I used MJPEG codec. There is a very certain reason for that. It is indeed an old codec, from 90’s but it doesn’t mean it’s totally bad. Unlike very efficient h264 and h265, which are interframe codecs, MJPEG is an intraframe codec. This means it compresses every frame separately, one by one. Thanks to this, it can cope very well with grainy footage, when the noise reduction is turned off. I’ve made a quick example of that below (it’s scaled up to 4K to minimize Youtube compression – for comparison, watch it only on 4k mode).



    With h264/265, which have a minimum of 25 frames between each I-frame, the grain looks like a underwater scene with wobbly blocks. Lot of detail is lost. Of course we are talking only about situation, when the noise reduction is off, and noise level is high. With NR turned on, or during the day, the MJPEG actually sucks, and looks really blocky, compared to h26X codec. Otherwise, who would want to use a codec with 40mbits bitrate. It would eat all memory in a very short time. But for testing porpoise it’s good.

    Lastly, you’ve mentioned the brightness comparison of different Dahua models. As I wrote in the description, it’s only a simulation, and of course it was done only on the camera I own. This is only a comparison of maximum brightness the camera can deliver in comparison to other models. It is NOT comparing noise level or detail level. And it’s only valid, assuming Dahua is not misleading us with minimum illumination values on their webpage.
     
    mat200, Kroghy and Probird79 like this.