40 Years of UK Telecoms Supply-Chain Evolution: From Cookie Cutter to Bespoke

1G: Cookie Cutter Networks

When Total Access Communication System (TACS) was launched in 1985, it comprised of ‘monolithic’ 3 cubic meter Base-Stations, and large rack-scale ‘Mobile Services Switching Centres’ which both filled rooms. It was a single frequency – 900MHz) and capacity was not an issue, supplying only basic voice services 

1G&2G: Few Building Blocks

When GSM became available around 1992, it launched in 900MHz and 1800 MHz, so there were 2 spectrum variants, and around 3 Combiner variants depending upon antenna type. There were indoor, outdoor and micro variants. 1, 4 and 6 Transceivers per rack were common configurations (1 TRX was ~8 voice calls though some capacity was given over to the new-fangled SMS.)

In the mid 1990s it was common to expand sites cabinet-by-cabinet , with payback periods as little as 1 month (with voice calls typically 1-50p per minute, average 10p per minute, adding 4 TRX x 8 calls/TRX generated £3.20 per hour. Even if the added capacity were only busy 20% of the time, that would be £28.5K revenue per month!).

For every ~20-200 Base stations there was a Base-Station Controller (BSC) (BSCs were scaled in Transceivers (TRX) and a large Base-Station might be 16/16/16 (i.e. 48TRX) so many 1024 TRX BSCs were needed.  

Eventually low bit-rate (I used 9.6Kbps in 1998) data services call General Packet Radio Services (GPRS) was added.

2G&3G: The Start of Complexity and Peak Volume

3G in 2000 added 2100 products, but there were 3G only with indoor and outdoor variants, and lower capacity micro variants. But each was essentially a single-line SKU with the cabinet, all the cables, backplanes and cards coming together as a single package. The radios and filters were – like GSM – still in separate units and sub-racks. SW upgrades saw speeds increase from 128Kbps, 384Kbps to 2Mbps per user.

For every ~100-200 Base stations there was a Radio Network Controller (RNC) with a throughput of approx. 300Mbps.

At the time of 3G launch, 1G had been recently retired, and a typical site saw an existing 2G BTS gaining a new 3G cabinet as a neighbour, with RNC being installed next to the BSC in Mobile Exchange Buildings.

What’s common to almost every one of the above examples is that the site was delivered by a single vendor, and a peak 2005 site might have 13 Racks of Radio Access Equipment (12x 2G 900/1800 plus 1x 3G 2100) plus antenna line equipment and a – typically 600x600mm – transmission rack. Large rooms or cabins were used to host this. We’re talking about several tonnes of RAN HW plus antennas, poles and cables, for just 2 or 3 bands of equipment.

During the 3G lifespan – in advance of 4G – Radio Units and Baseband Units became single units, allowing Main-Remote deployments of environmentally housed variants. New Bands 1300MHz, 2600 MHz became available. Around 2010 a combined BSC/RNC became available. GSM racks were taken out of service, to be replaced by new 3G racks ‘re-using’ the 2G spectrum.

2G&3G&4G: Simplicity borne out of Over Complexity

By the launch of 4G ‘Single-RAN’ was the watch-word, where 2G, 3G and 4G  could be delivered by radios which could operate on any of the 3 technologies, initially from technology-specific Basebands. It’s a little-known fact that the 4G launch in the UK was performed on ‘modernised 2G equipment’, where 6 racks of old Ericsson Equipment were replaced by a single rack of Huawei equipment which was later SW upgraded to add 4G in the same box.

In order to simplify sites in advance of 5G – with initial UK 5G services being launched from 3.5GHz so-called ‘massive MIMO’ Active antennas where the Radio and antenna are combined – main-remote has become the dominant form-factor for macro sites.

What was 3000 Kg in tri-band 2G/3G is now less than 10% of that, but is now 2G/3G/4G/5G; (700), 800, (900), (1300) 1800, 2100, (2300), (2600MHz) and 3500MHz Massive MIMO. The brackets indicate bands which not all operators have.

2G&3G&4G&5G: Grey Box Overload

In order to fit this, we’re now seeing dual-band and tri-band radios becoming dominant. But as above – with some operators having extra bands – it’s difficult to create radios which have bands which all customers want. So now vendors have to create 7/8/900MHz and 18/21/2600MHz tri-bands, 7/800MHz, 8/900MHz, 18/2100MHz dual-bands etc as well as single-band radios. Antennas similarly went from 900MHz 1-port antennas to 7/8/9/18/21/26/35 28-port antennas.

Antenna poles which used to support 30KG of antennas, now need to support 100KG of antennas and 80Kg of radios, and potentially 100Kg of Active Antennas. You can see this new-build site to the right (ready for 3500 massive MIMO but not yet installed) that the radios have been stayed on the ground in the ventilated cabinets, but this site offers dual-operator, tri-sector 1Gbps 2G/4G/5G rather than the single operator 1x42Mbps single-sector 3G site that it replaced to the left, and will be 2Gbps+ when the 5G mMIMO is added.

Clearly these sites are extremely large and expensive, and they are increasingly being supplemented by small cells for capacity. While Small Cells typically focus upon higher frequencies where there is more capacity than the more coverage focused lower frequencies (O2, for example has 2x10MHz of L800 spectrum for coverage, 40MHz of L2300 spectrum and 80MHz of NR3500 spectrum for capacity) multi-band cells are typically only dual-band so that are even more variants of small cells (e.g. 1800, 2100, 2600, 18/21, 21/26, 18/26) plus there are indoor (lowest power 100mW, 25oMW and 1W ) and outdoor (low and medium power 5W, 10W and 20W) variants to develop.

Compare this with ‘TACs 900MHz’ less than 40 years ago, and delivering ‘Total Portfolio’ RAN is an increasingly difficult target. When adding in global delivery of similar – but requiring different products such as 600, 850, 1900, 17/2100 plus 29GHz mmWave – radios for global vendors, it’s a challenge to get all the radios that an operator needs from a single vendor.     

Supply and Demand: When combinations go mad!

While basebands are available in indoor and outdoor variants, and differing capacities, there are far fewer combinations.

All the products above are created in a quasi demand-centric manner. The highest demand products are developed first – or those where lower demand is compensated for by good customer pricing with a firm commitment – with later products arising when R&D capacity is available commensurate to their demand.

So a customer may have to wait 1 or 2 years – or forever – to receive a product if the vendor lacks R&D capacity and the customer is reluctant to commit. E.g. Low capacity outdoor 4G/5G BBU to support small cells. And their demands matched with ‘lowest common denominator’ if pooled with requirements from another requestor of a similar product.

This is because the SW and HW are single vendor, and cannot be used with another vendors’ radios.

Imagine the frustration – as an operator – in managing several hundred site combinations, but being unable to deploy valuable ones, because e2e Vendors A and C make your baseband but Vendor B is the only e2e vendor who supplies the radio you want. And even though the standards support standalone CU and DU, your vendors only supply combined BBU (CU+DU)

Best of Breed

Open vRAN (Open AND Virtualised Radios Access Networks) can be deployed anywhere. 100% on site (including Core), only radios on site, baseband deployed in 2 hierarchies, baseband in the cloud or all of the above!

In the example above you simply need to buy Vendor B’s radios and the baseband can be any vendors CU / DU SW, and any DU HW. It may be that 79% of the network (RUs, DU&CU SW) is from Vendor A, 20% is Cloud Infra from Vendor X, and only 1% of NW is the Vendor B radios, but you’re now not limited by Vendor A’s R&D priorities. And if best of breed/specialist Small Cell RU Vendor K happens to only make the perfect Radio because that’s all they do, even better.

By moving to a pooled cloud structure, baseband HW utilisation can be maximised, site visits reduced, 

In this picture, the rack being configured could be a 5G Core, a vCU stack, a vDU build, or all 3! It could be Dell, HPE, or there could be no rack and it’s running on AWS, Azure, GCP etc.

By moving the baseband away from the site, the rental on the whole grey cabin below becomes redundant if the transmission and PSU/PDU is simply placed on the stub tower.

Instead of the CapEx and OpEx of supplying, delivering, maintaining and paying rental upon thousands of cabins, and visiting each of them for Baseband upgrades, simply rent rack-space in a few dozen Edge DataCentres, put your DU/CU/RIC there (and UPF, MEC and applications) and deliver low-latency C-RAN services to your customers.