CCTV Test: Low Light Performance (Part 1)
In recent times, an increasing number of camera manufacturers have made bold claims with regard to low light performance. These go beyond the normal sensitivity figures into numerous decimal places! The current trend is to suggest that proprietary technologies allow real-time colour images to be captured in previously difficult low light conditions. With so many established brands – and many of them credible ones – making such claims, Benchmark thought it worthwhile to take a closer look at the cameras on offer.
[dropcap]B[/dropcap]enchmark has carried out many low light tests in the past. In fact, it has become something of an annual ritual. People ask why we test for low light performance in the summer months, and the reason is because traditionally cameras have struggled with twilight conditions. They work well in daylight, and darkness can be dealt with by adding illumination. It’s the bit in between that often causes problems.
Despite what manufacturers claim, we’ve never had to work too late when testing low light performance. Most cameras – even the good ones – tend to struggle between 10 and 5 lux, before the image deteriorates and switching is required. Because of the increased pixel density on HD and megapixel models, the issues can occur towards the higher end of that range.
Despite the recent trend for adding proprietary ‘low light technologies’ and claiming that cameras will deliver real-time colour images in near darkness, we’ve yet to see this become a reality. Without giving too much away, we’re certainly getting closer to it. Are they really capable of delivering smooth colour footage in poor conditions? During the test, one tester remarked, ‘I hate these cameras; we’re going to be here all night!’
Many of the box cameras in the test were supplied with the manufacturers’ ‘recommended’ lens. Many of these are not supplied with the cameras. To deliver consistency in the results, all low light measurements were taken with control lenses, and all cameras were optimised to deliver an HD1080p stream.
Testing was completed using an open platform VMS, but the ease of installation ratings were based upon initial configuration, connection and set-up of the cameras using the supplied tools, processes and software.
The features and functions of the devices was also assessed. Whilst Benchmark has a policy of not publishing pricing information, the available features and functions are considered against the trade price of the units. This ensures that higher priced cameras are judged fairly against economical models.
The 2.0C-H3A-D01 is a networked day/night static dome HD1080p camera. It utilises H.264 compression, and features the manufacturer’s H3 platform, which is claimed to deliver enhanced resolution support and improved low light performance. Sensor size is 1/2.8 inch. The unit includes a 3–9mm P-Iris varifocal lens.
Sensitivity is quoted as 0.1 lux. The camera makes use of the manufacturer’s LightCatcher technology; this does not offer adjustability and is automated.
Other features include on-board analytics and SD-based edge recording. Power is PoE or 12V DC/24V AC.
We’ll get the gripe out of the way first! The camera is supplied with a hex key, fixings, a mounting template and a few pages of mandatory safety notes. There’s also a slip of paper advising the installer or integrator to visit a URL to access detailed installation guides.
In truth, most cameras don’t require detailed manuals. However, the 2.0C-H3A-D01 doesn’t even come with a quick start guide. Some of the headline features seemed to be absent, and we wondered whether they were only available if the camera was connected via Avigilon Control Center, so we went in search of the materials.
We were in possession of a very basic manual (which we had to download from the Avigilon website), but this didn’t cover the additional functionality.
The URL leads to a page which prompts the visitor to download an App for a smartphone or tablet. Manuals in PDF format can be a pain, so let’s hope the trend doesn’t move towards delivering them on a smartphone!
With the App loaded, we could download further documents. These were in an .epub format, so the next step was to find and download an .epub reader. When all of this was done, we discovered it was the same manual we’d already downloaded!
We’re sure that some people will love the idea of an app, but if someone at Avigilon reads this, please, remember those installers and integrators that just want the right information quickly!
Putting the wasted 10 minutes to one side, the installation of the 2.0C-H3A-D01 is very straightforward. With the camera up and running, and before we started any optimisation, it was clear that the processing is well implemented. The image was crisp and detailed with good colour fidelity. Motion was smooth and there were few signs of artefacting.
Whilst the camera does feature low light technologies, these work in the background, and the installer or integrator only has influence over exposure, iris and gain levels. These, however, seem to plenty to allow a good configuration!
As light levels fall past 10 lux, you expect to see the typical signs of early noise and the first indications of motion blur, but with the 2.0C-H3A-D01 they don’t arrive. The vibrancy of colour might be a little more muted than normal, but otherwise the image delivery is very good.
The first signs of noise appear at around 2.5 lux, although colour detail is still more than usable, and motion blur is so minor as to be negligible. The image remains usable without any significant degradation, until at around 0.8 lux the camera switches to monochrome mode.
Once switched to monochrome mode, the image is clean and stable, and retains a high level of definition. Contrast is good, and fast motion is handled well.
Axis Communications: Q1635
The Q1635 is a networked day/night box-type HD1080p camera with a frame rate of up to 50fps. It utilises H.264 and M-JPEG compression. The camera features a 1/2 inch sensor to deliver improved low light performance, and sensitivity is quoted as 0.1 lux (with a 25fps stream).
The camera makes use of the Axis Lightfinder technology to deliver colour images in very low light. Other features include intelligent video and SD-based edge recording. Power is PoE or 12V DC.
Axis Communications have, in our experience (and we’ve been testing products for decades now), always had a straightforward and reliable installation utility. It used to stand out from the crowd, and whilst many manufacturers have now caught up and deliver similar simplicity, configuring an Axis product is always a straightforward and intuitive process.
In line with other Q series cameras we’ve tested, at initial log-in you need to select between HD1080p/50 without WDR or HD1080p/25 with WDR; there is also a megapixel option. This can be changed at a later time if required, but it will result in all configurations bar the network settings being returned to default.
If low light performance is your ultimate goal, then we recommend using the 25fps configuration.
As with a few of the other cameras on test, the initial image is very impressive, even before all optimisation has been completed. The quality is high with a good degree of visual detail, colours are accurate, motion is smooth and there are no signs of any digital artefacts. It is true that the Q1635 isn’t a cheap camera, but it’s actually not that expensive either, because you really do get what you’re paying for.
The camera features Lightfinder, the Axis proprietary low light technology. It doesn’t require configuration and is working in the background. However, you can adjust shutter speed, set a ceiling for maximum gain and balance the priority between low noise and low motion blur. We opted to maximise reduced blur, and to be fair noise wasn’t too great of an issue.
With regard to day/night switching, the Q1635 does offer an adjustable level, and this gives a real scope for maximising performance. With many cameras, all too often adjusting the switching point isn’t that effective, but in the Q16 series Axis has delivered a decent and usable degree of flexibility.
The first signs of noise appear at around 3 lux, and whilst you might think it’s the beginning of the end, nothing is further from the truth. The image remains usable with good colour detail, minimal motion blur and without any significant degradation until below 1 lux.
If you want the camera to switch earlier, it will with a few changes to switching configurations, but pushing it to its limit reveals a usable and clean image until the camera switches at just below 0.5 lux.
Once switched, the image retains its sharpness and detail. Performance is good in terms of visual quality and motion remains smooth with no dropped frames or other artefacts.
Bosch: Dinion IP Starlight 8000
The Dinion IP Starlight 8000 is a networked day/night 5 megapixel camera. It utilises H.264 and M-JPEG compression and features a 1/1.8 inch sensor. Sensitivity is quoted as 0.012 lux (30IRE).
The camera’s other features include intelligent video analytics, defogging, intelligent auto-exposure and edge recording. Power is PoE or 12V DC.
When Bosch first launched the Dinion IP Starlight 8000, it was tested by Benchmark. At the time, despite being a new launch (and products at that point of their lifecycle often display teething troubles) it achieved Outstanding status: one of only fourteen products to have ever done so in the past decade. It was going to be interesting to see if other cameras had made up the gap, but we didn’t expect Bosch to have slid backwards.
One thing noted in the previous test was that Bosch had improved its configuration process, and the supplied Bosch Video Client (BVC) software included useful utilities that delivered benefits for installers and integrators.
Expecting a repeat of the quick and easy set-up, we sought out the BVC CD, but it wasn’t there. Instead we got a QR code from which we could download the software. Given that the 8000 isn’t a low cost camera, it feels like a negative move to exclude a necessary CD containing installation utilities.
Whilst we wouldn’t swear to it, it also felt as if the interface of the video client had changed. The process didn’t feel as smooth, and checking back over notes revealed that the earlier experience was somewhat different.
With the camera connected, there are three operational modes and you prompted to select one. These are IP Starlight 8000 megapixel, IP Starlight 8000 megapixel 4:3 or IP Starlight 8000 HD1080p. The mode can be changed at any time, but if you switch between these once the camera has been set up, then all settings aside from the network configuration will return to default!
In HD1080p mode, the 8000 shows a good level of detail, high colour fidelity and smooth motion without any blur. Even as light levels fall, the quality remains high with no obvious degradation in the image.
The camera’s pre-defined scene modes are somewhat generic with regard to specifications, but they give a starting point for specific needs.
As light levels fall, the video processing which delivers cleaner images has no detrimental effect on the quality. The implementation of the processing engine is good, and at no point does the image show signs of the background work that’s going on.
The quoted specification of 0.00825 lux is for a 30IRE image, so you’re unlikely to get close to that with a decent quality image!
The camera has two settings for controlling day/night switching: day-to-night switchover and night-to-day switchover. This allows hunting to be eliminated and ensures that things remain stable.
As light levels fall towards 2 lux, the 8000 shows a relatively clean image, with no real blur. Even below the point, the 1/1.8 inch CCD works well to deliver a sharp and detailed picture. Some artefacting can be seen, but it’s not to a negative degree.
In a side-by-side comparison, the Dinion IP 8000 doesn’t have the edge that a few of the best performing cameras do with regard to delivering smooth colour images in very low light. That isn’t to say it’s inferior. If anything, the 8000 is more like a traditional camera, allowing switching early enough to preserve video integrity.
From 2 lux downwards, you can either work with the processing parameters to get a little more from the colour image, or allow the unit to switch to night mode. The flexibility is there for either approach, and both will suit certain requirements
Once switched and under IR illumination, the 8000 retains a sharp and clean image with plenty of detail, and motion is smooth and glitch-free.
The VB-H730F is a networked day/night HD1080p camera with an integral 2.8–8.4mm lens which delivers a wide angle view of 112 degrees. The camera utilises H.264 and M-JPEG compression. It features a 1/3 inch sensor, and sensitivity is quoted as 0.2 lux (50IRE).
Other features include integral video content analysis, smart shade control and SD-based edge recording. Power is PoE or 12V DC/24V DC.
If Canon is famed for one thing, it’s optics, and so you have every reason to expect that this element of the VB-H730F will be excellent. The Canon installation utility, however, isn’t the best. It works; it’s just a bit slow and clumsy. Given that Canon has acquired Axis, it might be nice to think that they’ll send someone along to ask, ever so nicely, if they’ll show them how to make a simple and reliable installation utility. That isn’t a complaint; you can’t be good at everything, and no doubt there will be some traffic the other way to talk optics!
The Canon set-up GUI has one frustrating element to it: the settings menus are predominantly textual and don’t have any visual display. If you want to see what effect changes have, you’ll need to save them, leave the menus and open the viewer, check the results, and then return to the menu.
Whilst the camera is capable and boasts a good degree of functionality, these little configuration quirks do inevitably put some installers and integrators off the products. Interestingly, most of the cameras in this test allow changes to be viewed directly in the menu screens.
It is a shame, because the VB-H730F is actually a promising camera. Canon’s optics are really very good, and the camera’s processing power is none too shoddy either.
Image detail is sharp and clean, colour fidelity is high, motion is smooth and the overall image is very impressive.
The only downside is that latency is a little higher than the other cameras, but it is consistent so shouldn’t present any problems for operators.
With regard to low light set-up, there is the ability to limit gain. In our experience the common belief that in low light you maximise gain is often untrue. A balance of gain and noise reduction, with sensible shutter control, delivers cleaner images for longer. The Canon unit allows the flexibility to achieve this.
Day/night switching can also be adjusted. Someone at Canon had the brainwave of naming the settings in a way that makes sense (darker, slightly darker, standard, slightly brighter, brighter) so hats-off to them for that!
As light levels decrease, so the VB-H730F begins to impress. It might lack the outright clean colour images of the more expensive models, but it certainly is no slouch.
As light levels fell to below 2 lux, noise started to creep in. We’d set the camera to the ‘Brighter’ switching point, so we expected the unit to go to night mode. However, it didn’t. Noise became more significant, colour detail faded, and still it didn’t switch. Once the image became unusable, we decided that further investigation was needed.
The camera refused to switch regardless of the settings. If the external input was used to send a telemetry signal, then the camera would switch, but with the configuration set to Auto it would remain in colour mode, even in pitch darkness!
Interestingly, the menus won’t allow a conflict between manual and automatic switching. For example, if a relay is set to trigger switching and the mode is changed from manual to automatic in another menu screen, the camera won’t allow the change and flags it as an error.
As a final resort we employed the old IT approach, powered off the camera, ignored it for a while, and then rebooted it. The switching worked! Attempts to replicate the problem failed, so we can only put it down as one of those things. The variable switching points do deliver a degree of flexibility.
Once switched, video quality is good. Detail is clear and images are sharp with plenty of contrast. Motion is smooth with no signs of blur or artefacting.
The DF5200HD-DN is a networked day/night camera which delivers HD1080P streams. It utilises H.264 and M-JPEG compression, and multiple streaming is supported. The camera includes an integral 4.5–10mm varifocal motorised lens and features a 1/1.9 inch image sensor. Sensitivity is claimed to be 0.0002 lux.
Other features include edge storage, P-Iris support and 120fps with an optional upgrade. Power is PoE or 24V AC.
Installation is straightforward. The camera has a static IP address. It may necessitate changing the address of the server for set-up, but that’s a quick and simple task, and the camera is responsive. You are prompted to load some ActiveX elements. The process is quick and relatively straightforward.
An initial scout through the menus shows a clean and ordered environment, and everything you expect to find is where you’d expect it to be. It’s a sign that someone with a surveillance mind has written the GUI!
The DF5200HD-DN is supplied with a factory-fitted lens, and this cannot be changed in the field. It is a varifocal motorised unit, and won’t present any problems to installers and integrators.
It might be unkind to say, but it does look a bit odd. Based purely on looks, we didn’t hold out great hope for the final image quality. However, it goes to show how wrong you can be; the DF5200HD-DN delivers an image that is sharp, detailed, colour accurate and very well focused. It might (and will) be argued that in terms of definition and detail, the DF5200HD-DN delivered the best image in this test.
As light levels fall, the DF5200HD-DN continues to impress. Colours remain faithful and the image is relatively noise-free, even in conditions down to 1 lux. There is no obvious motion blur, and although it is obvious the processing is working hard, there’s no visible sign that it is. There are no artefacts, missed frames or image rebuild issues.
As light levels drop below 1 lux, the camera retains a decent image until it switches. There is a good degree of flexibility with regard to the settings. The interface is a little odd, but you’ll soon get used to it. We opted to switch at around 1 lux, and this ensured noise did not become an issue. Switching is easily controlled, allowing a good level of flexibility.
Once the image switches, the monochrome detail is pin sharp, and the optics and compression on the unit are very good indeed. Whilst retaining a somewhat goofy look, the DF5200HD-DN is a serious heavyweight performer and that’s a welcome surprise.
Previous low light tests with HD1080p cameras have typically seen noise become evident at around 5 lux, with most cameras either switching or delivering poor quality images shortly after. While manufacturers’ claims have indicated significant improvements were afoot, this is the first test where we’ve seen evidence of it in the real world, across the board.
The 2.0C-H3A-D01 from Avigilon is very impressive. Colour detail remains usable, and the image shows no real impact from motion blur, right down to levels of o.8 lux. Indeed, the camera switches long before you feel it is reaching its limit.
The Q1635 from Axis is verging on the spectacular in low light. It’s true that the camera has a larger sensor, and it carries a raised price tag because of that, but the performance is noteworthy. If you consider it against cameras from just 12 months ago, then it shows just how far low light performance has developed.
The Dinion IP Starlight 8000 is a very capable camera, and until recently it had established itself as a ‘go to’ camera for troublesome environments. While other units struggled to deliver colour images in very low light, the Dinion made a fair fist of doing so. While other cameras are now equalling its performance, it remains a solid choice.
Canon’s VB-H730F is a frustration. It’s clearly a very decent unit, but it’s the only camera in the test without a smooth transition between the menus and a view of how changes are affecting the video. As the browser view is only used for set-up, it’s an odd decision for the manufacturer to make. It almost feels like the software engineers thought people might operate the camera via the browser. That aside, the optics are very good and performance is high.
The Dallmeier DF5200HD-DN is a camera that will make most installers and integrators smile. The image is pin sharp, low light performance is good, and because the lens is factory configured, it is easy to achieve the best quality. Processing is well implemented too!