technical note regarding out of band emissions at 24.25 GHz from 5G and the need to protect earth exploration satellite services. The impact of setting such limits might mean that 5G cannot use below 26.5 GHz.
1. 1
Technical note on mm wave band issues of 5G / IMT 2020 and Beyond and Interference to Passive
Services (Sentinel)
Roberto Ercole, August 2018
Introduction
The next WRC will need to identify new bands for IMT/mobile broadband and is considering this
under agenda item 1.13. There were initially some 11 bands under consideration ranging from 24.25
to 86 GHz. The importance of bands above 20 GHz is that they will be able offer the very high data
rates required by 5G/IMT2020 and Beyond1
(in part because of the large amount of bandwidth
available). There may also be some advantages in terms of Massive MIMO2
, which uses a large
number of separate antennas to send and receive signals more efficiently.
The band that is supported by most regulators in ITU Region 1 (Europe, Africa, Middle East, and
Russia), is 24.25 to 27.5 GHz – known as the 26 GHz band. The 28 GHz band has been identified by
for next generation satellite services (offering high data rate services for homes and businesses in
mainly remote areas). If 5G were to use the 28 GHz instead of 26 GHz this could cause problems for
regulators who would need to decide between the two services.
However, a problem has arisen with regards to a passive satellite service just below 24 GHz used in
global warming monitoring. There have been discussions within CEPT about how to deal with the
issue, as the issue has raised concerns about the usability of this band for 5G/mobile broadband. If
the” spill-over” from IMT transmissions are reduced by too much (beyond what normal filtering can
achieve), then either the power may need to be turned down by a factor of 100 or more, or large
guard bands of many channels will be needed. For example, a 100 W base station will need to be
reduced in power to 1 W – which might make the band unusable for commercial wide area 5G. This
problem of spectrum spillage is referred to as the out of band emissions of a device (OOB) in
standards.
The setting of OOB limits for 5G in the 26 GHz (that are under discussion now) will be key to the
commercial viability of the band.
If the limits are not technically feasible (for commercial price points) then pressure will grow to use
the 28 GHz band.
The danger is that unless agreement is reached the commercial viability of the 26 GHz band may
force regulators in the future to licence other bands, such as the bands proposed for US, Korea, and
Japan – known as the 28 GHz band (26.5 GHz to 29.5 GHz).
1
See ITU-R Recommendation M.2083 on the IMT Vision on capabilities and technologies used.
2
Multiple Input Multiple Output. Typically, 100 or 200 plus antennas used to send and receive signals. At 20
GHz and above it is possible to have such a large antenna array not only on the base station but on the mobile.
2. 2
What is the interference issue?
Whilst the proposed 5G/IMT band is from 24.25 GHz upwards, and the passive satellite service used
by the European Space Agency (ESA), occupies 23.6 to 24.0 GHz there is a problem3
. Digital signals
of IMT “spill out” from their frequency band, and the higher the data rates used, the more spillage
there is likely to be. This is shown diagrammatically below.
Given that the bandwidth of the mobile system envisaged are 200 MHz (based on sharing study
parameters used in the ITU4
), we can see that there may be as little as just over one channel
separation between mobile and the ESA satellite system.
The problem is that the ESA satellite system (called Sentinel) is very sensitive and measures the
temperature of the sea and land to monitor global warming. The ESA satellite does not transmit
data, but effectively listens to large areas of the earth, and could in theory “see” entire cities such as
London. Hence it is a “passive” service. The other difficulty is that the satellite does not just
measure the temperature in Europe but globally. It is not a geostationary system, but a “non-
geostationary” satellite that passes over the earth every 20 days or so (depending on the latitude).
So even countries that are not in Europe will need to protect such systems.
It should also be bourn in mind that because this is effectively a satellite uplink band (the
interference is “seen” by the satellite), this makes national solutions very difficult. This is because
the ESA satellite will “see” all countries it passes over. This is not the case when it is satellite
downlink such as in the C band (3.4-3.8 GHz), where the receiving earth station could be interfered
with.
For example, in the C band, if IMT were deployed in France say, then the majority of “victim”
satellite receivers would be in France. Depending on exact satellite usage/configuration, it might be
that cross border interference could be limited to a few 10’s of km (although high sites could have
3
https://earth.esa.int/web/guest/missions/esa-eo-missions/sentinel-3
4
https://www.itu.int/en/ITU-R/study-groups/rsg5/tg5-1/Pages/default.aspx
3. 3
much larger ranges). This is because the 3.4-3.8 GHz band is used for satellite downlink (satellite to
receiving earth station), so only earth stations relatively near the IMT transmitter will be affected5
.
The other major problem is that whilst there is some 250 MHz separation between IMT and passive
frequency edges, the IMT service is intended for 5G, and will use much wider bandwidths than
traditional 4G systems. The ITU modelling is assuming 200 MHz, which is a little over one channel
away (as mentioned above). The choice of modulation scheme selected for 5G will be important as
this will affect the out of band emissions. Generally, the higher the data rate (more accurately the
faster the state transitions of the modulation scheme) the wider the occupied bandwidth, and hence
spillage. The current modulation scheme for 4G (OFDMA) leads to significant spectral regrowth/
spillage6
. This can be reduced by filtering the signals, but the effectiveness of filtering could be
limited by cost, as well as the impact on the performance of 5G devices (battery life etc).
There is the possibility of using a different modulation scheme to reduce the spectrum spillage/OOB
from 5G devices into ESA is under discussion. However as noted by Rhode and Schwartz (ref. 6) the
OOB advantages of the alternative candidates to OFDMA “more or less vanished when the signal is
amplified”.
ITU Rules
One of the main functions of the ITU Radio Sector is to control interference between the satellite
systems of different countries. This is done by new systems registering themselves with the ITU and
going through a coordination process to receive protection7
. The ESA Sentinel system has been
registered in the ITU Radio Bureau (BR) International Master Frequency Register, so effectively it has
protection under a UN Treaty, as do a number of other satellite systems in this band (see Annex 1).
The exact level of protection is determined by the appropriate technical recommendations (such as
ITU-R RS.1861) and how this should be modelled in sharing studies (such as those submitted by ESA
to TG 5-1 document 130). Because the current ITU Recommendations were developed by the
satellite community they are set at quite low levels (i.e. they accept very little spill-over).
As the ESA system is “filed” and approved in the BR registry, any new service that wants to deploy
will need to protect it from “harmful interference”. The test of harmful interference does not mean
no interference and would be based on the relevant ITU-R Recommendation as well as what impact
is measured on the Sentinel system. The exact outcome of any case would be hard to know but
would certainly introduce a large uncertainty for any potential investor in this band. It would be
extremely costly if a 5G operator deployed 1000 cells in a country and then found they all had to be
re-engineered to protect Sentinel.
5
It will depend in many cases if there is a line of sight between victim and receiver.
6
https://www.rohde-schwarz.com/us/applications/5g-waveform-candidates-application-note_56280-
267585.html
7
https://www.itu.int/en/ITU-R/space/plans/Pages/MIFR.aspx
4. 4
Current Negotiations
There are two main streams of activity on the development of OOB limits for 5G in the 26 GHz band,
namely:
1. WRC process and ITU-R TG 5-1 – conducting sharing studies for discussion at WRC-19;
2. The CEPT/ECC Decision (18)06 on Harmonised Conditions in 26 GHz8
.
The outcome of the WRC work would in theory be used to determine the limits that would go into a
new Radio Regulations Articles via a “Footnote” for IMT in the band (for example see the one of the
C band 5.430A). However, work on getting agreement has been slow, and past experience has
shown that a strategy of delay by incumbents can sometimes be effective in blocking new
allocations. In any case if no agreement is reached then the limits already specified in the ITU-R
would be used as the basis for deciding an interference complaint. According to ESA inputs to the
ITU (5-1/130) this would set a limit of -166 dBW/200 MHz as a threshold. This limit would then need
to be used in sharing studies to determine if “harmful interference” is caused. A major uncertainty
would be in how to use the limit correctly in any sharing study.
Ideally the new Footnote in the Radio Regulations would specify a trigger limit to be used, that
would remove some of the uncertainty. The exact level and how the trigger limit would be specified
is for discussion at the next WRC. In advance of the WRC the CEPT has been consulting on an ECC
Decision on use of the 26 GHz band for 5G mobile services.
An ECC decision9
was adopted 6th
July 2018, for base station and mobile OOB limits of -42 dBW/200
MHz for a base station and -38 dBW/200 MHz for the user equipment (UE). These limits would
specify the OOB/spectrum spillage that could fall within the receive band of the satellite system
(23.6 to 24 GHz).
What is the relationship between the two streams?
In theory the Radio Regulations take precedence over any national or regional agreement on levels.
For example, if a country decided to claim protection for its satellite service with the ITU, then as
currently understood, the CEPT level has no legal authority over the ITU/UN Treaty obligations. If
the level agreed by WRC-19 was more stringent than the CEPT limit, then the CEPT should amend its
limits to comply with the radio regulations.
However, in practice the fact that ESA and France agreed to the level in the CEPT makes it likely that
such a level would be agreed at WRC-19. This assumes that ESA is the most affected system, and
one of the other satellite filings (such as by China or the USA) does not start to make higher claims
for protection. So far this does not appear to be the case. It does appear that some regions (ATU
and ASMG) may ask for more relaxed limits at WRC-19, because of concerns about how easily 5G can
meet the CEPT limits.
8
https://www.ecodocdb.dk/download/5e74d0b8-fbab/ECCDec1806.pdf
9
https://www.ecodocdb.dk/download/5e74d0b8-fbab/ECCDec1806.pdf
5. 5
What would the -42 and -38 dBW/200 MHz mean in practice for 5G at 26 GHz?
The problem for 5G/IMT is how easily it can meet these proposed levels, and that depends on how
far these are from what the current 3GPP standards imply.
3GPP Equivalent* CEPT level** Difference dB
(absolute)
Base Station -13 dBM/MHz -20 dBW/200 MHz -42 dBW/200 MHz 22 dB (@160x)
User
Equipment
-13 dBM/MHz -20 dBW/200 MHz -38 dBW/200 MHz 18 dB (@65x)
*converting from dBm/MHz to dBW/200 MHz.
**As agreed in ECC Decision (18)06.
Main Source: TG 5/1 Feb. 17 meeting, Liaison statement from 5D (Temp/265.)
The absolute difference (highlighted above) shows that there is a significant gap between what is
suggested in the CEPT process and what is in the 3GPP standard.
The table shows that for a base station the emission limit is some 22 dB tighter (a factor of around
around 160 times less), and for a mobile some 18 dB less (a factor of around 65 times less).
These levels could cause difficulties for mobile services in the 26 GHz band. In the minutes of the EC
meeting where the draft Decision was agreed (noting many limits were in square brackets) the
mobile industry raised concerns stating the proposed limits “will have serious and negative
consequences for the deployment possibilities for 5G” as well as “severely reducing throughput, high
insertion loss for filters and very large guard bands (more than 1 GHz)”, and that this might threaten
the successful development of 5G in Europe.
One major question is how much “give” has been put into the 3GPP OOB limits and how easily they
can be improved in the future. The maximum power of a base station is assumed to be 34.5 dBm (in
200 MHz), which equates to 4.5 dBW/200 MHz. Digital signals tend to decay slowly, in terms of
what “spills” into adjacent channels10
, so single channel guard band is likely to have little impact.
This means it is likely that say the next 5 channels (i.e. 1000 MHz away) will have similar spillage (or
adjacent channel leakage as it is sometimes called). So, to meet the OOB limit requires the base
station to ensure the energy leaking into the passive band (below 24 GHz) by reducing it by nearly 50
dB below the carrier signal level (50 dBc). This could be feasible.
In theory for a base station, there is no limitation in space (to physically site a filter) and any loss in
power because of the filter attenuating the signal (like sunglasses making things fainter) can be
overcome (with more power). There may be impacts on costs of filters as well as energy
consumption. It is also of course possible to reduce the transmitter power (back-off) as well as
reduce the data rate. If a guard band were to be employed, then given the wide band nature of 5G
signals (200 MHz) then it is likely that the benefit would be limited unless the guard band were say
at least 5 channels wide (1 GHz).
10
It is normally assumed that OOB limits apply 250% of the necessary bandwidth of the signal. If we assume
90% of the 200 MHz is needed, then the OOB limit applies some 450 MHz away from the lowest 5G frequency
channel. Before this limit a higher figure is used normally based on ACLR.
6. 6
The most problematic issue is likely to be user devices (terminals). As 5G will most likely be TDD,
then mobiles can operate close to the 24.25 GHz lower 5G/IMT band edge. The ability to use extra
filtering in a UE would be much more difficult because:
1. Space in mobile devices (UE) is very limited – so extra filtering could be a problem;
2. The UE power budget is very sensitive and so losing say 1 or 2 dB in a filter (which might be
assumed initially) could severely reduce battery life. 1-2 dB extra loss might suggest 20-40%
extra power consumption when transmitting.
This could however be offset by the current traffic asymmetry we see today for mobile services, with
traffic mostly favouring downlink (from 1:4 to 1:9 currently)11
. So, if most of the traffic is downlink
and mobiles transmit only represent 10% to 20%, the adverse impact on 5G may be more limited.
But this depends on what may happen to traffic asymmetry in the future.
Conclusion
There is a lot of uncertainty on this issue, and the past track record at WRCs (with 3.6-3.8 GHz for 5G
at 2015) does not make for confidence on the mobile side. Two visions for the future are possible:
1. 5G can meet the out of band emission limits required by CEPT/WRC-19 and the
26 GHz band becomes the main capacity band for 5G, once C band is fully used .
2. 5G cannot practically meet the limits required to protect passive, because of
device cost and/or performance issues. Hence operators will not invest in 26
GHz services.
It has been suggested that a compromise may be possible with a device tuning across the two ranges
(26 GHz and 28 GHz). This ignores the fact that radio devices have a limited tuning range. This is
because the gain and performance of radio devices drops with the bandwidth (especially things like
filters and power amplifiers). Exactly how wide a tuning range can be is a complex trade-off
between performance, and cost. At some point it becomes more cost effective simply to have two
radios (one for 26 and one for 28 GHz).
An oft quoted figure is a range of 10% of the centre frequency, which in this case is 2.4 to 2.6 GHz.
Assuming some improvements in the future this might be 3 GHz. Trying to widen this too much more
might make it impossible to meet OOB limits required or require larger guard bands or power back-
off. Also, it has been widely suggested that to make investments in such high bands attractive,
operators will want close to 1 GHz of spectrum each making large guard bands problematic.
Under scenario 2 (and considering tuning ranges) this makes it likely that operators will want devices
that tune across around 26 to 29 GHz (or 26.5 to 29.5 GHz). This could allow for 1 GHz of spectrum
to be used in the 26 GHz band, but this might be too little to be commercially attractive in markets
with 3 or more operators. Why would you invest in 26 GHz if you only have 300 MHz, when you
might have 100 to 200 in C band (which has much better propagation characteristics)?
11
https://www.itu.int/dms_pub/itu-r/opb/rep/R-REP-M.2370-2015-PDF-E.pdf
7. 7
If this were to happen then one could see a future where spectrum below say 26.5 GHz is seen as
secondary in value to 27.5 to 29.5 GHz where you get global economies of scale with US, Japan and
Korea. Indeed, device availability below say 26.5 may make it commercially of little value for 5G.
In the end this will depend upon cost and performance considerations when major investments
come to be made in the 20-30 GHz band range for 5G.
8. 8
Annex 1
Other Bands considered outside the WRC-19 Process
There are also bands outside of the 26 GHz range that are being implemented by USA, Korea, and
Japan, known as 28 GHz. These bands are all “MOBILE co-primary” globally in the Radio Regulations,
so do not require WRC-15 approval or changes to be implemented and be protected from other
primary services (once filed with the BR).
USA: 27.5 – 28.35 GHz
Likely to move fast12
and have
widespread low-cost kit.
Auction start Nov. 18
Korea: 26.5 – 29.5 GHz
Intention to have pre-
commercial 5G trials 2018
winter Olympics – auctioned
July 18
Japan: 27.5 – 28.28 GHz
Early system trials from 2017 in
Tokyo
ECC Band Plan – from ECC/EC/(18)06
Possible Frequency Arrangements for 24.75 – 27.5 GHz (TDD with 200 MHz block size)
Satellite Filing in ITU Register between 23.6 and 24 GHz that are Notified (N) – hence protected
12
https://apps.fcc.gov/edocs_public/attachmatch/DOC-340310A1.pdf