Digital video units usually installed in a surveillance system are designed around a video encoder, whose job is basically to compress and transmit video streams to Video Management System (VMS). A digital raw (uncompressed) video takes a huge data rate, so video stream needs to be compressed in order to lower the amount of traffic (bitrate) introduced in the network, as well as store it on VMS side in a feasible way.
Video encoders use compression algorithms (H.264, MPEG4, MJPEG) that are able to reduce correlated information contained inside same frame or inside blocks of frames (e.g. steady elements or areas); the general idea is: instead of sending a full picture twice, it's better to send the picture once, then send the differences compared to the original picture.
Also, the algorithm may cut unnecessary information, such as small details or small colour variations, which would be useless for the sake of the application.
That's why such compression schemas are called lossy: a small amount of information gets lost to reduce the quantity of data to be sent.
Encoders allow to set also a maximum bitrate for generated video stream; easily, the lower the bitrate is, the higher will be the information that algorithm needs to cut, reducing the general stream quality.
Constant Bit Rate (CBR) or Variable Bit Rate (VBR) streams can also be produced.
In CBR streams, a maximum bandwidth is configured as a parameter, and the bitrate is kept within this no matter how complex the scene is; that's useful in case of channels with limited capacity (e.g. wireless), but on the other side the configured bitrate could not be enough to transmit those complex sections with an acceptable quality.
Using VBR, the encoder can allocate more bandwidth when the situation requires it, to stay inside the configured quality parameters; a more complex area is going to be encoded using a larger amount of bits, while a steady region will be more compressed on fewer data.
As a thumb rule, CBR focuses on bandwidth occupation, VBR on quality.
Today's cameras are all able to produce good quality videos during the day time, when the illumination level is high, so that the CCD sensor can capture a good amount of light. The more the light hitting the sensor is, the higher the video signal is going to be, without being amplified really much.
During night time, situation changes a lot: less illumination means a lower level of light captured by the sensor; electrical signal at the CCD's exit needs to be amplified, otherwise it wouldn't be correctly managed by the camera's electronics.
Unfortunately, amplification always introduces noise. Those salt&pepper images that get on the screen during the night are due to a high amplification level (or gain) indeed.
How does this impact with the video stream bitrate?
Actually, the salt&pepper pixels are detected by the compression algorithm as image details; moreover they are not stable at all, so it interprets them as moving targets as well.
If the encoder is using VBR, it is driven to allocate more bandwidth in order to compress what it thinks to be a really complex image with the configured quality level: bitrate could be 10 times more than day time one.
If CBR is being used, the overall quality will be cut to stay within the maximum bitrate: probably real details on the image are going to be lost.
There are not-so-many solution to such occurrences. Said that final aim is to produce and store an acceptable quality video (far from being useless at least), the camera should be put in the condition to catch enough light, so that a lower level of signal amplification is needed.
Night Mode/IR Filter should be enabled on the camera during night time. But a suitable IR illumination should be also installed together with the camera itself.
CBR or VBR should be also selected keeping in mind the major system constraints, in terms of channels' capacity and storage capacity.
If the system has been designed considering the day's figures only, using VBR with inadequate illumination will drive to network congestion, camera disconnections, buffering, or unexpected disk occupation during the night.
On the other hand CBR could not deliver the expected stream quality, losing important scenes and details that would important to have during alert situations.