Already back in 2009 (!), I wrote a blog post on LTE’s TTI-bundling feature. It was supposed to be used to improve cell edge scenarios when a device’s uplink power would not be sufficient anymore. In particular it was seen to improve VoLTE speech quality. Quite a bit of thinking ahead because we were far away from VoLTE in practice back then. Over the years I haven’t seen TTI Bundling in the wild until recently.
Just a quick recap on how TTI bundling works: Instead of one transmission, TTI bundling transmits 4 different redundancy versions of a packet in the uplink direction, i.e. from the mobile device to the network. This way, the network can combine the four transmissions and thus cope with more transmission errors.
Over the years I haven’t specifically checked if anyone used TTI bundling but it didn’t seem so. But when I was recently doing some drive testing in the US in very rural areas, I noticed that TTI-bundling actually got activated occasionally. And it was direly needed indeed as the base station sites were so far apart from each other that handovers regularly occurred at an incredibly low signal level of -126 dBm. To set that into context, in urban environments, LTE handovers are done somewhere between -95 and -100 dBm, i.e. when the signal at the cell edge is still 1000 times (30 dB) stronger!
In practice, the network can activate TTI bundling at any time for a connection with an RRCReconfiguration message. For details see the specification I linked to in the original post back in 2009. Here’s how the relevant part of the RRCReconfiguration message looks like:
radio_Resource_Config_Dedicated mac-Main_Config ul-SCH-Config maxHARQ-Tx: n12 periodicBSR-Timer: sf20 retxBSR-Timer: 320 ms .1.. .... ttiBundling: True
Interestingly, TTI-bundling was activated without a measurement report being sent by the UE first. On second thought, that makes sense, because measurement reports from the UE inform the eNB about the downlink signal quality. TTI-bundling gets activated when the signal level in the uplink direction gets too low, which is measured by the eNB itself. Also, I noticed that TTI-bundling was used sparingly and was switched-off again pretty quickly.
Back to the VoLTE call. So despite the handovers regularly occurring at very low signal levels, my VoLTE call did not drop over the 45 minutes I had it running. I couldn’t check how good the uplink speech path was but at least in the trace I could see occasional RRCReestablishment requests due to the UE loosing network coverage. While the VoLTE call staid up, browsing the web at such low signal levels was not possible. Not pretty so it definitely doesn’t work wonders.