MUROS Goes VAMOS And Becomes Interesting!

Back in 1991 when first GSM networks were launched and still today in many networks, the GSM world is simple when it comes to voice: 1 timeslot = 1 user. Some networks use half rate codecs in some situations, basically splitting a timeslot and use every alternate timeslot occasion for two users. This in effect doubles voice capacity in networks at somewhat the expense of voice quality. Back in 2008 I reported  about a new scheme coming up called MUROS (Multiple Users Reusing One Slot) which can cram up to 4 users in one timeslot.

At the time I perceived it as an interesting idea but thought it wouldn’t probably go very far as, and that was my assumption, MUROS would only work with new mobiles with special transceivers. Well, it turns out that is not the case as MUROS, now standardized in 3GPP and referred to as VAMOS (Voice services over Adaptive Multi-user channels on One Slot) also works with a significant number of mobiles already in the field.

In essence VAMOS works as follows: In a first instance, VAMOS can be combined with half rate channels, i.e. the starting point is using the AMR half-rate or AMR-Wideband half rate speech codecs which already results in two users sharing one timeslot. Nothing new here. VAMOS then extends this scheme by transmitting the combination of two signals at the same time over the same channel, each with a different sequence of training bits in the middle of the timeslot which is used by the receiver for channel estimation. Each of the two mobiles that receive the data stream at the same time use their knowledge of their individual training sequence to reconstruct their own part of the signal, effectively filtering away the second data stream as noise. Instead of today’s GMSK (Gaussian Minimum Shift Keying) modulation, VAMOS uses QPSK modulation in the downlink direction that approximates a GMSK signal when the “noise” (i.e. the second signal) is filtered away.

The scheme works best if the mobile has a Single Antenna Interference Cancellation (SAIC) receiver. I have found two papers on the Internet that describe this in more detail, one from NSN here and one from Ericsson here. From an implementation point of view the important thing is that for several years now, many mobiles have been equipped with SAIC receivers to improve their resistance against inter-cell interference, i.e. not even with VAMOS in mind. These receivers are also able to filter out local interference, a nice side benefit. The NSN paper, which must date back from 2009 mentions that there are already over 1 billion devices shipped with SAIC receivers. In other words, when VAMOS gets deployed one does not have to wait for special VAMOS capable devices to reach a critical mass before the benefits can be seen.

VAMOS in 3GPP has also specified extensions to mobile chipsets so they can be informed of the training sequence of the other signal not intended for them which further improves their ability to filter out the second signal. Also, one paper mentions that it is even possible to pair non-SAIC and SAIC/VAMOS mobiles on the same timeslot if the non-SAIC mobile gets more of the signal energy than the SAIC mobile, i.e. there is less noise from its point of view. And finally, there could be two users sharing a timeslot of legacy half rate operation and one using a full rate VAMOS channel which in effect allows three users sharing the timeslot or two half rate VAMOS channels and one full rate VAMOS channel.

Another way to look at this is that VAMOS exploits the fact that no matter how good or bad signal conditions are for a mobile device at any particular time, the voice channel always uses the same power and same modulation, thus in effect wasting spectrum. VAMOS doubles the number of bits sent per time and thus doubles the efficiency of spectrum usage. That means that VAMOS does not work everywhere in the cell. Especially in cell border areas where the signal strength is weaker and interference from neighboring cells is higher compared to the center of the cell the connection might have to fall back to normal GMSK modulation. In effect, VAMOS adds the channel adaptivity that was introduced many years ago with GPRS and then EDGE for packet switched data.

In the uplink direction, and that’s again an interesting twist, VAMOS uses the existing GMSK modulation scheme. In other words, no new transmitter elements are required in mobile devices. On the base station side, two antennas are required to tell the two GMSK transmissions from two devices apart, using Multi User MIMO (Multiple Input Multiple Output) algorithms.

In addition, the radio network has to monitor the signal quality in the uplink and in the downlink direction very carefully and change the timeslot configuration to single use in case the bit error rate becomes too high on a VAMOS channel for one of the users.

Altogether, a very interesting scheme. I can very well imagine that VAMOS is going to be used in emerging markets to expand capacity without additional hardware in the base stations. It should be noted, however, that additional calls on the air interface requires additional backhaul capacity, so activating VAMOS requires an upgrade of backhaul links. The papers also note that another benefit of using a timeslot for more than one or two users is the overall power reduction of the base stations per call. That is perhaps a good reason for countries in which diesel generators, solar panels and wind mills supply the electric power for base stations and every watt counts. In countries where network operators are measured in terms of voice quality, VAMOS might have a more difficult stance. This probably depends how much speech quality is impacted.

In a somewhat more distant future when GSM service could be much reduced in countries having deployed later generation systems and only serves M2M traffic (data), roamers and other devices not being 3G and perhaps LTE voice capable, one might perhaps even live with a somewhat degraded speech quality in return for freed-up bandwidth that can then be put to good use for UTMS and LTE.

So I think it’s quite likely that in both emerging and developed markets, this is not the last we have heard of VAMOS.