In order to better understand how disturbances affect the user experience of IPTV, and what role the selection of STB has on the IPTV service when it comes to network impairments, we tested how a few different set-top boxes (STB) handle disturbances such as packet loss and jitter.
With all providers and STBs the IPTV channels showed similar degradation when losses were introduced. The same is true for delay and jitter events. There were however one exception. One of the set top boxes had significant problems handling HD streams when losses were introduced. The channel completely froze, and sometimes the STB rebooted, also at quite low level of packet loss. This was shown to be a known problem that might be related to the low processing power of this particular STB.
Since the STBs proved to handle packet loss differently, and one of the STBs was much more sensitive than the others, it is of highest importance to test the robustness of the STB you plan to use. Choosing the wrong one might be disastrous for you profitability, due to high support costs.
It is important to measure the loss level of your network, to quickly detect and fix quality issues. Netrounds can be used for this purpose, it is easy to use, and you can even get started without cost. See https://www.netrounds.com/solutions/video-delivery-monitoring.
Between the STB and the IP network a network simulator was placed. The network simulator was acting as a bridge forwarding traffic between the network and the STB. The NTools network emulator program was used to introduce network a controllable level of network impairments.
During these tests a loss burst was introduced, when a number of consecutive packets were dropped. The burst length went up from 1 packet to 1024.
The main conclusion of the test is that even a single frame loss can be noticed, depending on the actual TV content, whether there is much movement at the time, and which frame is lost. A frame loss usually appears as a small glitch in the picture: a small part of the picture could freeze for a split second, or show pixelation, or some other artifact. Audio is not that much effected, very rarely a single frame loss can produce a disturbing hiss, similar to a loud white-noise signal. It seemed as one of the STBs had a bit better tolerance for losses.
Effect of a burst of packet losses (length in packets):
- 1 Even a single loss can cause small disturbances in the picture. Audio not affected that much. Maybe 1 out of 10 losses can produce a noticeable audio problem.
- 2 Almost every burst produces a smaller, sometimes a bit bigger video glitch.
- 4 Almost every burst produces a smaller, sometimes a bit bigger video glitch.
- 8 Almost every burst produces a smaller, sometimes a bit bigger video glitch. Video sometimes can be pixelation for half a second.
- 16 The whole picture can freeze for a moment if there is movement on the video. Audio can also freeze, or produce a disturbing hiss.
- 32 The whole picture can freeze for a moment if there is movement on the video. Audio can also freeze, or produce a disturbing hiss.
- 256 Picture breaks up, and can be frozen for half a second.
- 1024 Picture freezes for a second.
During these tests continuous loss was introduced to the video streams.
Same conclusions apply as for the bursty loss tests. Here one of the STBs (the same as above) showed a bit better results, 0.01% loss was basically imperceptible.
Effect of loss (in %):
- 0.01 Sometimes even a single frame loss can be noticed. Quality mostly excellent, every 30 sec or so a glitch can be noticed on the video. In audio there can a very disturbing hiss every few minutes or so.
- 0.05 Noticeable picture problems every 20 sec or so. Audio glitches/hiss can be every minute
- 0.1 Noticeable picture problems frequently, every 10 sec or so. Audio glitches/hiss can be every half minute.
- 0.5 Continuous picture and audio problems.
- 1 Picture is broken up continuously. Continuous audio glitches and hisses every sec or so. The picture could freeze for short time.
- 5 Picture broken up. Inaudible.
- 10 Picture broken up, and freezes for a second, only fragments of messy audio can be heard.
Delay, continuous jitter and jitter spikes
During these tests continuous delay, jitter or a transient jitter spike was introduced to the video streams. Results
These impairments are grouped together as their effects were identical on the streams. The general behavior is that when x ms delay or jitter is introduced the stream freezes for x ms, and then continues without much problem. Delay/jitter below 100 ms is not very disturbing. Also, once the STB increased its adaptive buffer, further jitter events cause no problems at all.
There were two exceptions. One of the STBs showed no artifacts at all even with 1 s delay/jitter with an HD channel. The reason could be that it worked with a big buffer from the beginning. For another STB a delay/jitter of 700 ms caused audio glitches, and at 1000 ms audio just went away.