Tested: Lowering Bandwidth at Night is Good
Bandwidth spikes in low light are commonplace, with increases of 300 to 500% over daytime normal.
IPVM recommends lowering bandwidth levels at night, relative to those spikes. For example, if your camera consumes 2Mb/s during the day, but 10Mb/s at night, you can easily cap it to 5Mb/s (or even less) without any quality loss. This can reduce overall storage costs by 50%.
In this report, we share our test findings of multiple cameras set to average compression levels and then to high compression level to answer the following questions:
- What visible difference does high compression video have in low light?
- Are human subject or text details more difficult to discern at higher compression levels?
- How did this increase in compression impact bandwidth?
There are two key points from this test:
- Increasing compression (i.e. quantization) from 28 to 35 had little effect on low light (~0.25 lux) image quality at even high PPF levels (80+) while massively decreasing bitrates by an average of 87%.
- Only at extremely high PPF (150+) were moderate differences in details seen, with less definition in subject features and decreased legibility of fine text.
Why This Happens
Even the 'best' surveillance cameras struggle with handling low light. When they produce a 'decent' image, this is almost always because of applying heavy image processing / gain control. This gain control increases visible noise which is much harder to encode / compression than daytime video, spiking bandwidth and storage costs. Though advancing digital noise reduction is bringing this dow somewhat, bandwidth consumption is still far higher than the comparable daytime scene. For background, see: Testing: Gain / AGC Impact on Surveillance Video and Camera DNR (Digital Noise Reduction) Guide.
Recommendation: Use MBR
These test findings reinforces our recommendation of MBR as best practice for CODEC configuration, originally discussed in our CBR vs VBR vs MBR report. Bitrate reductions realized by this streaming method while maintaining similar image quality provide the best balance of performance. Low light image quality issues such as increased digital noise and artifacting impact image quality far more than the increase in quantization seen when cameras near their VBR cap.
Compression Levels of Cameras
For background on how compression is set and measured, see our IP Camera Manufacturer Compression Comparison and How to Measure Video Quality / Compression Levels reports.
The comparisons below shows each camera at average compression levels (Q 28) and at high ones (Q 35) that reduce bandwidth consumption by 80-90%.
Facial details and chart legibility are similar at both quantization levels, with no major differences. Only in one camera (C, below) are facial details moderately worse at Q35 compared to Q28 and that is because of an extremely high 150ppf scene / FoV.
Increasing quantization from 28 to 35 resulted in significant bitrate reductions, 87% on average, seen in the chart below. Higher compression reduces bitrate increases from digital noise typical in low light images, as noise "macroblocks" become larger and less complex, requiring less bandwidth.
Greater Impact On Daytime Video
Because of these drastic bandwidth decreases, it may be tempting to increase compression on the camera full time, even during the day. However, high quantization has much greater impact on daytime video than low light examples seen above, so this is not recommended.
For example, the image below shows the effects of increased quantization in full light (~160 lux). Facial details become more difficult to discern, with macroblocking making the subject's face look pixelated and blurred/smeared. The test chart becomes more difficult to make out at lines 4/5 and beyond, as well.
Minimize changes / increases to compression settings in general, as these can materially worsen daytime video. However, check if your camera's bandwidth consumption spikes at night (most do, even with IR). For these cameras, implement a VBR 'cap' that is greater than your day time average consumption but a fraction of what the uncapped stream is consuming at night.
By doing so, you can save 50% or more bandwidth at night while having negligible impact on video quality.