A regular conversation has two lines of subtitles with a duration of 6 seconds. In this calculation we are using 5 seconds duration and subtitles back to back, i.e. no gap in-between.
We have for this calculation used a font size of 24 and a bit plane of 16 colours. The size of the bitmap is related to the number of colours used for rendering the DVB bitmap. The more colours used the better anti-alias is performed. See pictures below.
Two colours bitmap with a size of 5 KByte
16 colours bitmap with a size of 10 Kbyte
For calculations within this document we will use 7 Kbyte for a typical two row subtitle. Note: Measuring of subtitles in the UK DTT network shows that subtitles are between 3 and 7 Kbyte, which indicates that only a few colours are used for rendering bitmaps.
Normal speech subtitling
The subtitles are received on the incoming stage of the encoder in teletext subtitling format. With a Video delay of 3 seconds (Video delay of Grass Valley encoder) we have approximately 2.7 seconds for rendering and pre-sending of subtitles to the decoder, assuming that the STBs need 0.3 seconds for rendering in the box. The time for receiving, render and TS wrapping in the encoder is less than 0.2 seconds which gives us 2.5 seconds for transmission time of a subtitle.
The example below shows 5 sec duration subtitles with a zero gap, each subtitle has a size of 7 Kbyte (56 Kbit). We have a present duration of 2.5 seconds which gives a peak bandwidth of 56 Kbit/2.5 = 22,4 Kbit/s.
The peak bandwidth used in this case is 22,4 Kbit/s but is only used during 50% of the time. The remaining time, the non-used bandwidth can be allocated to the common statistical multiplexing pool giving an average bandwidth over time of approximately 12 Kbit/s. Note: With bigger sized subtitles, e.g. more colours used, the bit rate will be higher.
Live subtitling with progressive subtitles
Let us assume in a Live subtitling scenario that subtitles are sent word by word and each word is a separate bitmap object with a size of 5 Kbit (a total of 16 words for a two line subtitle). A Live subtitler can type 5 words/second which gives 5 x 5Kbit = 25 Kbit/s.
Each word is sent in a separate object in a progressive mode adding objects to the region as they are written.
The bandwidth allocation is dependent on the rate of the text, the number of colours used (quality of text) and the size of the font. This document has analysed the required bandwidth based on the assumptions above. With a fixed allocation the bandwidth requirement would be 25 Kbit/s. It is also estimated that the peak allocated bandwidth is not used for more than 50% of the time. This spare capacity can be allocated to the common statistical multiplexing pool of 9 encoders and will thus improve the picture quality.
Total 25 x 9 = 225 Kbit/s for 9 subtitle streams, but with statistical multiplexing the mean bandwidth over a period of time will be much less. An estimation would be 50 %, i.e. 115 Kbit/s.