I've been in Rome recently for a week and I noticed that on Vodafone and TIM's 3G networks the experience on my mobile phone was quite bad. Quite often when clicking on a link the page would not load in any reasonable amount of time. When switching to their GPRS networks page load times with Opera Mini where good so my problem likely resulted from some air interface issues. On Tre's 3G network, my device performed flawlessly so they must do something different to Vodafone and TIM. To see where my problems came from I therefore decided to take a closer look at how the radio network state changes where configured. Here's the result:
TIM:
DCH timer: < 5 s
FACH timer: 75 s
Final state: idle
Vodafone:
DCH timer: < 10 s
FACH timer: 45 s
Final state: cell-pch
Wind:
DCH timer: < 3 s
FACH timer: 75 s
Final state: idle
Tre (3IT):
DCH timer: < 5 s
FACH timer: 60 s
Final state: idle
When compared to network settings in other countries such as Germany, for example, I was quite surprised about the very long FACH timers. In Germany, those timers are much shorter, and in the range of 15-20 seconds to conserve battery power in mobile devices. Beyond 30 seconds, they are a huge energy drain and really, Fast Dormancy is a mandatory self defense mechanism against such settings…
Concerning Fast Dormany, I am at a loss when it comes to Vodafone Italy's settings. Why is there a 45 seconds Cell-FACH phase when the network then transitions to Cell-PCH instead of Idle. Cell-PCH combines the advantages of low battery consumption with fast data transfer resumption with less signaling in the network to reestablish the connection so such a long Cell-FACH phase seems very unnecessary (for details see the Fast Dormancy link above).
On Wind's network I found the Cell-DCH timer of 3 seconds or perhaps even a bit less quite surprising. In practice this means that the connection frequently changes between DCH and FACH, resulting in an inferior web browsing experience, as each time the state is changed, the transmission is interrupted and packets have to be queued. I noticed this when surfing on my pad as pages loaded much slower than they usually do, especially if they contained content that took a bit of time to be downloaded. Wind as furthermore set the thresholds in a way that the DCH is not kept if only little data flows. So a default "ping" will not keep the connection in DCH state. Only a ping packet size of around 500 bytes had the desired effect. Again, I am wondering why they are doing this!? Are they having problems with the number of concurrent connections in DCH state? It surely can't be to conserve power on the UE side. Time to buy some more DCH licenses guys instead of crippling the performance of your network!
While all of this is very interesting it does not explain why web pages are often not correctly loading after pressing a link on Vodafone and TIM. I therefore suspect that it has something to do with the UE and networks having an interoperability issue when changing between the different states and/or perhaps carrier frequencies, since both have two 5 MHz carriers deployed. Difficult to tell without a deep drill down. So during my stay, Tre.it became my favorite roaming network in Italy and I am glad about manual network selection.
Vodafone’s cell-pch final state is probably a configuration bug, but Wind’s RCC settings are ok. As you know there is a tradeoff between minimizing the carrier’s signaling activity (long T1, T2) and minimizing the power consumption of the UE (short T1, T2).
If you only send a short ping I would hope to transition to FACH and then idle very quickly. Assuming 1W in DCH, 0.6W in FACH, 0.1W in cell-pch, and 0W in idle a minimal transfer is
– TIM, 5*1+75*0.6=50J
– Vodafone, unbounded
– WInd, 48J
– Tre, 41J
Using Nokia Energy Profiler I’ve observed
– Finland 3G T1=2s T2=2s, min 3.2J (2s lag when browsing, but low UE power)
– ATT 3G California T1=10s T2=10s, min 16J
– ATT 3G microcell T1=240s T2=0s, min 240J (excellent UE responsiveness, but battery empty after 4h).
So the carriers’ different RCC settings result in 3J, 16J, ~50J, or 240J for a minimum transmission. IMHO handsets are better served with lower energy rather than lower latency, given that a typical battery only holds about 17.5kJ. A plugged-in laptop with 3G dongle would benefit from different RCC settings.
Hi Martin,
Very interesting..
Can I ask how you were able to get those timings ? Were you able to measure them somehow (and if so how ?) or is it because you know the right people ?
Thanks
Robin
Edit…. scratch my last message…. Nokia Energy Profiler no doubt. Durrrr ! 🙂
Just a comment regarding Fast Dormancy.
As implemented in phones pre-release 8, many operators are requesting to disable FD due to signalling congestion in their network. I’m aware of many cases so far. Please refer to this GSMA paper: http://www.gsm.org/documents/TS18_v10_TSG_PRD_Fast_Dormancy_Best_Practices.pdf
I’ve recently started to use 3G again on my Blackberry (which makes very efficient use of EDGE and even GPRS) and noticed how much better battery life is compared to what I remember it being before I sidelined 3G. I presumed this was due to an increase in the number of ‘base stations’ and not having to continually hunt for new 3G signals when out of range, but I’m also now wondering if my most used network (Belgium: BASE) has changed its settings such as FACH. Either way, it has restored my faith in 3G and made it more usable.