I was not sure if I should post my “TFO and TrFO introduction” here as there are probably only a handful of people on the globe who are interested in this topic. Also, some prior knowledge is required on how GSM and UMTS networks handle circuit switched calls and how things change with the introduction of 3GPP Release 4 BICN functionality. A good intro on this topic can for example be found in my book (take a look on the left). Most of my blog entries require no such in-depth prior knowledge but this one definitely does. I decided to do it anyway as I didn’t find any other usable overview on the web.
Tandem Free Operation (TFO) and Transcoding Free Operation (TrFO) Overview
Voice calls from mobile phones today always involve several voice transcoding steps in GSM and UMTS Release 99 networks. This is done mainly to transport a voice call, which uses sophisticated compression in the radio network, through the core network with the standard G.711 PCM (Pulse Code Modulated) codec over 64 kbit/s circuit switched links. Furthermore, transcoding a voice data stream to the 64 kbit/s G.711 PCM codec is necessary in order to call fixed line phones. Another advantage of voice transcoding in the network is to be able to process the voice data stream and to compensate for unwanted effects such as echo in the speech path.
With the introduction of the Bearer Independent Core Network functionality in the 3GPP Release 4 standards, core network connections are based on high bandwidth ATM or IP links rather than on 64 kbit/s TDM links. In such networks it is therefore possible to transmit a voice data stream in the core network with other codecs than 64 kbit/s G.711. This also has the additional benefit of removing transcoders from the speech path which reduces speech path delay and improves speech quality to some degree as the speech signal is no longer degraded by the transcoding process. Furthermore, not having to convert to the narrow voice band G.711 codec allows the use of AMR Wideband (AMR-WB) codecs. These codecs are not backwards compatible to G.711 without loosing their much better audio quality which they achieve by encoding twice the frequency range as previous Full Rate, Enhanced Full Rate, AMR and G.711 codecs. A general downside of not using a transcoder in the network is that it is no longer possible to process the voice data stream to remove effects such as echo.
Transcoder Free Operation (TrFO)
In UMTS Bearer Independent Core Networks (BICN), PCM links which require the use of the G.711 codec are no longer used. Instead, connections to RNCs in the radio network and connections to other Media Gateways are based on ATM and IP. Therefore, the MSC servers involved in a call can establish a bearer path without activating voice transcoders in the media gateways between two mobile phones. For a mobile to landline call the connection remains transcoder free up to the media gateway which connects the wireless network to the fixed network. Here, the transcoder is used to convert the compressed audio signal to G.711.
In mobile to mobile calls, both handsets report their codec capabilities to the MSC servers involved in the call as part of the signaling before the bearer path is established. Thus, an AMR-WB codec is automatically used if supported by both terminals. This greatly improves the voice quality of the call. If not supported on both ends, a standard narrow band AMR codec will be used for the connection instead. This is called Transcoder Free Operation (TrFO) and the out of band transcoder control performed by MSC servers is described in 3GPP TS 23.153.
While TrFO works well while both ends of the connection remain in the UMTS network, it has to be deactivated if one of the two parties is handed over to a GSM network during the connection. This happens, for example, if the subscriber roams out of the UMTS coverage area or in case the UMTS coverage is temporarily weakened by obstacles while GSM network coverage remains acceptable. This can happen for a number of reasons like because of different locations of the UMTS and GSM base station sites or due to better GSM in-house coverage. If AMR-WB was used for the call, deactivating TrFO also requires a fallback to AMR which results in an abrupt voice quality degradation of the call in case the GSM network does not support Tandem Free Operation which is described next (see also 3GPP 23.153, chapter 6.5).
To communicate with services in the networks such as a voice mail system, or a prepaid top up voice server, DTMF (Dual Tone Multiple Frequency) tones are used to send passwords and commands to the system. If a call is established from a UMTS terminal, DTMF tones are sent as out of band signaling messages to the MSC Server. The MSC Server is then responsible to forward the message. If the service platform which analyzes the DTMF tones is BICN compatible, the tone is delivered as a message. If the service platform is still using PCM links and the G.711 codec, the DTMF tone is inserted into the voice connection at the Media Gateway at border to the PCM network. If a call is established from an external PCM network to a service in BICN network, DTMF tones are not sent as messages but inside the voice bearer. In this case, 3GPP 23.153, 5.9 requires the gateway at the border of the BICN network to discover the inband DTMF tone, remove it from the voice stream and send a DTMF signaling message to its MSC Server which then forwards the DTMF message out of the voice band to the destination.
Tandem Free Operation (TFO)
For GSM networks, TrFO can not be used even if the core network uses BICN media gateways and MSC-servers. This is because the A-Interface, which connects the media gateway to the BSC/transcoding unit is based on PCM. Thus, voice calls from GSM radio networks always arrive at the media gateway as a PCM G.711 64 kbit/s circuit switched stream. Unlike in UMTS, where the transcoding is a functionality of the media gateway, GSM uses a dedicated transcoding unit between the BSC and the media gateway. This is necessary in order not to change the radio network architecture when the core network is upgraded to BICN. As a consequence, a mobile to mobile GSM call will always have two PCM links in the connection, one to and from each transcoder and thus a GSM call is always established using the G.711 codec. The two transcoders of the connection are also called a tandem. In order to remove the transcoder tandem from such a connection, in band codec negotiation between the two transcoders can be attempted once the call is established. Instead of sending signaling messages via the MSC-servers through the network, the two transcoder units send their information embedded in the G.711 link. This is done by using a special bit pattern as part of the voice stream which is recognized by the other end as a transcoding control message and not as a part of the voice signal. If both transcoder units support at least a single common codec like AMR 12.2 or an AMR-WB codec, the 64 kbit/s G.711 connection is used to tunnel the compressed and encoded voice stream. As the audio stream is compressed most of the bits of the transparent 64 kbit/s stream are not used. This effectively removes the transcoder tandem from the voice connection and is thus called Tandem Free Operation (see 3GPP TS 23.053 and 3GPP TS 28.062). While audio quality and delay times can be improved this way, no bandwidth savings can be achieved in the core network unlike with TrFO.
Some events in the network can temporarily interrupt TFO and the speech path in the core network automatically falls back to standard G.711 encoding. This is the case for handovers for example in which the call is handed over to a cell which is controlled by a different TRAU. This is the case for an inter-BSC handover for example. Once the handover is complete, TFO has to be negotiated again between the two TRAUs (see 3GPP TS 23.053, 6.5.7). TFO is also temporarily interrupted when the MSC inserts DTMF tones or announcements. As the MSC is not aware of TFO it overwrites the compressed speech information in the circuit connection. This is detected by the terminating transcoder and transcoding elements are automatically reinserted into the speech path. Especially for AMR-WB, TFO has to be started again as soon as possible in order to avoid speech quality degradation.
Conference bridges are another network feature which automatically disable TFO. This is because standard conference bridges mix G.711 voice signals from several parties to produce a combined signal which also overwrites the TFO data stream. This again forces the TRAUs to introduce a transcoder in the speech path again. While this has little impact on narrow band speech calls from a voice quality perspective, it is not possible to use AMR-WB when a conference bridge is inserted unless the conference bridge also supports TFO and the AMR-WB codec.
There we go, if you have made it up to here, please consider leaving a comment :-)