You are on page 1of 47

Redefining Research August 31, 2022 Vol. 18, No.

The Greatest
Show on Earth!

VoNR
VOLUME 27: BEHIND THE VoLTE
CURTAIN, PART 7
PART OF “THE MOTHER OF ALL NETWORK BENCHMARK TESTS” SERIES OF REPORTS
YOUR ATTENTION
PLEASE
EFFECTIVE IMMEDIATELY, ANY UNAUTHORIZED USE OF OUR RESEARCH MATERIAL WILL RESULT IN THE NON-REFUND-
ABLE CANCELLATION OF YOUR SUBSCRIPTION. We also reserve the right to post your company’s name, with logo, to the
“SRG Wall of Shame.” If you received this issue from someone outside of your organization and it did not come directly from
SRG then the licensing terms for our research are being violated. If you forward this research to external organizations, either
in whole or in part, or if you share the contents of the report beyond the authorized allocation within your organization then
the licensing terms for our research are being violated.
If you value the information and insight that we provide then I strongly urge you to respect our hard work and livelihood and
subscribe to our research. If you do not have a platinum license or a global license, you may want to upgrade your license so
that you can share this issue across your entire organization with our blessing.
If you or your organization is interested in distributing this report to outside organizations, please feel free to contact us to
discuss licensing terms and fees.
If you would like to leverage a quote from this report and you have at least a global license, please contact us for permission
and we will be happy to provide it.
1.0 Executive Summary

Key Highlights from this Study


SRG just completed its 27th 5G NR benchmark study. For this study, we conducted
an independent evaluation of VoNR (Voice over New Radio) and how it compares and
contrasts with VoLTE (Voice over LTE). We analyzed voice quality (MOS – Mean Opinion
Score), call setup times, network resource utilization requirements, and current drain. In
addition to comparing VoNR and VoLTE, we also looked at VoNR performance in Band
n41 versus Band n71.

We did the testing in a T-Mobile 17-site test cluster in southern California, or the same
test cluster that we used for our 3CC Signals Ahead report that we published last month.
In addition to supporting three component carriers (3CC), including both FDD and TDD
bands, the network supported the Standalone (SA) network architecture and, most
importantly, VoNR. Interestingly, at least to us, Band n41 is an important part of the
operator’s VoNR strategy and by testing in this cluster, which is part of its commercial
network, we were able to look at the performance differences of VoNR between the two
5G bands.

VoNR and VoLTE use the same voice codec (EVS 13.2 kbps) so all things being equal one
should expect similar voice quality between the two voice technologies. In general, we
can confirm this hypothesis, however, it must be balanced against the maturity of the
two networks. T-Mobile launched LTE in March 2013 and it launched 5G (n71) in December
2019, while it wasn’t until April 2020 that it launched its first 5G network in Band n41. The
further buildout of new 5G sites as well as network optimization/interference manage-
ment will help improve signal quality (SINR) and signal strength (RSRP) and subsequently
reduce the variability of the voice quality we observed.

For consumers, the single biggest advantage of VoNR is that it allows the smartphone
to remain on 5G instead of falling back to LTE with VoLTE. This feature can be huge if
the consumer is concurrently using data in the background during the call. Longer term,
consumers will benefit when it allows the operator to migrate its LTE spectrum to 5G
usage, while SA mode has other compelling attributes that can improve performance
across the board. We doubt DSS will ever play much of a role in the operator’s 5G migra-
tion strategy. Current drain can be an issue, but it may also hinge on which 5G band
the smartphone is using. We also anticipate/hope for further 5G related features and/or
vendor optimizations that will minimize network resource requirements, not to mention
indirectly improve voice quality.

Thanks to Accuver Americas (XCAL-M and XCAP) and Spirent Communications (Umetrix
Voice and Umetrix Data) for the use of their products and platforms to conduct this
study. Both companies have been valued partners for more than a decade.

We’ll be attending MWC Las Vegas 2022 later in September (Tuesday afternoon through early
Friday morning). If you are attending and you would like to meet up, then please drop us a line.

In August 2014, Signals Research Group (SRG) published the industry’s first independent bench-
mark study of VoLTE. For that study, we used the AT&T LTE network in Minneapolis-Saint
Paul to compare VoLTE performance with 3G circuit switched voice and Skype, addressing many
of the same test scenarios and performance parameters that we covered in this study. Fortunately,

3 August 31, 2022 | Signals Ahead, Vol. 18, Number 7


CSFB (Circuit Switched Fallback) is no longer a consideration. To put things into perspective, for
that VoLTE study we used the Galaxy S4 mini smartphone for the VoLTE calls and Alcatel Lucent
was the infrastructure supplier. The study was also done at a time when consumers paid for data in
bundled packages and there was “real concern” that consumers would gravitate to an OTT applica-
tion like Skype instead of using VoLTE. One noteworthy outcome of that study is that SRG HQ
packed up its bags and moved from California to Minnesota less than two years later.
Since that first VoLTE study, we’ve published an additional five reports under the “Behind the
VoLTE” umbrella with this report marking our seventh report in the series. In remembrance to
our first VoLTE report, and because we’re inherently cheap and uncreative, we simply refreshed the
original report cover for this report.
This report on VoNR marks our 27th 5G NR benchmark study. For this study, we leveraged the
same 17-site test cluster in the T-Mobile network (Ericsson infrastructure) that we used for our last
report on 3CC. We actually did this testing prior to testing 3CC, but since we knew it would take a
lot more analysis and time for this study, we elected to publish the other report first. As a courtesy,
we provided the operator with a pre-brief of our findings just prior to releasing this report. There
were also a few observations that triggered the need to get some feedback from the infrastructure
vendor, so we took that step at the eleventh hour.
We kindly remind our subscribers that Signals Ahead is a subscription-based service. This bench- We kindly remind our readers
mark study, like all our studies published in Signals Ahead, is entirely supported by these subscrip- that Signals Ahead is a
tions. We expect our subscribers to respect the amount of effort and expense associated with doing subscription-based service
and that you must abide by
these studies. Although companies with a global license may share these reports internally within
the terms of the subscription.
their organization, we do not allow any external sharing of this report, either in whole or in part.
Companies with a group subscription cannot share the report, either in whole or in part, beyond the
individuals signed up to receive these reports. If you feel compelled to share these reports, then you
should also feel compelled to upgrade your company’s subscription. If you are not sure what type
of subscription your company has then please reach out and ask us. We also note that we provide a
report preview of each Signals Ahead report. You can share the report preview, which is available for
download on our website or by contacting us.
Like the earlier VoLTE report we analyzed several different aspects of VoNR and VoLTE perfor-
mance. We looked at how VoNR/VoLTE impacted concurrent downlink or uplink data transfers,
not to mention the impact of data transfers on call quality. We did this portion of the study primarily
with stationary testing to maintain consistent RF conditions throughout the tests. We also analyzed
the downlink and uplink bandwidth requirements of the two voice technologies to determine the
relative impact of these voice calls on overall network capacity. Another key aspect of the study was
to determine voice quality and how it was impacted during drive tests involving dozens of handovers
and varying RF conditions from good SINR and favorable RSRP to low SINR and unfavorable
RSRP. Of course, no VoNR study would be complete unless we also included VoNR in the two
possible 5G band combinations: Band n71 (600 MHz) and Band n41 (2500 MHz). We also included
the impact of VoNR and VoLTE calls on the battery life, focusing on the measured current drain
during the call and how it compared with the current drain in idle mode. Lastly, we looked at the
call setup times for the two voice technologies, based on a lengthy drive test with dozens of voice
calls placed between the two smartphones.
The conclusion of our study is that while VoNR is promising, we think it could be a while before Although VoNR is promising,
VoNR becomes the preferred voice technology in T-Mobile’s network. We would be shocked if it we think it could be a while
happened this year and we would be pleasantly surprised if it occurred in 2023, especially across large before VoNR becomes the
preferred voice technology
swaths of the operator’s network versus token launches in a few hard-to-reach markets. Portland and
in T-Mobile’s network.
Salt Lake have VoNR, but it isn’t the preferred voice technology, meaning more than likely a voice
call would default to VoLTE unless LTE wasn’t available and 5G was present. There is a reason
DISH is playing musical chairs with its infrastructure vendors. VoNR isn’t easy, SA isn’t easy, and

4 August 31, 2022 | Signals Ahead, Vol. 18, Number 7


it doesn’t help that the DISH network is just being rolled out, the cell density is far from adequate,
and it leverages Open RAN/AWS for its network.
For those readers that didn’t digest our DISH Las Vegas report (SA 06/23/22, “Hold ‘em or Fold
‘em”), we want to make it clear there was a world of difference in the VoNR experiences on the
two networks. With DISH, the RF-related problems were far worse than what we witnessed in
the T-Mobile network plus there were non-RF-related problems that had a dramatic impact on the
VoNR voice quality in the DISH network even when the RF conditions were good, and we were
testing from a stationary position.
We tested VoLTE in AT&T’s network in July 2014. We considered the network quasi-commercial
when we tested it since you needed to have a smartphone with VoLTE functionality (we borrowed
one from the operator), but at least we tested VoLTE across all of the Twin Cities (or pretty much
close to it). By the end of 2015, the AT&T VoLTE network covered 295 million Americans.
T-Mobile, we believe, has rolled out VoNR functionality to most, if not all, its cell sites, but it isn’t
as simple as flipping a switch and turning it on. Furthermore, although we tested in a large cluster,
it wasn’t as if we tested across greater Los Angeles.
We also believe there was a more compelling argument to move from circuit switched voice to We believe there was a more
VoLTE than there is to move from VoLTE to VoNR. Call quality with the narrow band codec used compelling argument to
on UMTS was pretty poor, especially compared with VoLTE, which at the time used the AMR move from circuit switched
voice to VoLTE than there is
wideband codec. Additionally, you could use a sand dial to measure the time it took for a smartphone
to move from VoLTE to VoNR.
to move from LTE to 3G to place/receive a voice call and to then return back to LTE once the 3G
call was finished. With VoNR versus VoLTE, the benefits are more nuanced, especially since the
call setup times with either option are very similar, and the voice quality can be comparable since
both voice technologies use EVS – the voice codec that followed AMR. Worth noting, VoLTE
introduced packet-based voice traffic with priority scheduling over best effort data packets and that
hurdle wasn’t easy to clear. VoNR simply leverages that well-established capability.
To be fair, when we revisited our results from the first VoLTE study, we saw some of the same VoNR actually performed
issues we saw when testing VoNR. In fact, setting aside the different sizes of the two networks, better in the T-Mobile test
VoNR actually performed better in the test cluster than VoLTE performed in the Twin Cities. cluster than VoLTE performed
in the Twin Cities when
We attribute this outcome primarily to the EVS codec which delivered higher overall MOS over
we first tested it in 2014.
VoNR than the AMR codec did over VoLTE. The voice quality on both networks faltered when
the RF conditions were less favorable and, in our view, the degradation of the voice quality back in
2014 was more substantial than it was when we tested last month in Los Angeles. We’ve collected
RF measurements on the T-Mobile network in at least a few markets over the last year, so we have
a reasonable sense of where things stand and how they compare with the test cluster. Our humble
opinion is there is ample work ahead for the operator and its vendors to further build out its 5G
network and to then optimize it for voice services. Given what we have documented with Band
n71 and Band n41 RF conditions in the markets where we have collected data, we also suspect that
when the day finally arrives, in those areas where there is both Band n71 and Band n41 coverage, the
higher frequency will become the dominant provider of VoNR services.
That being said [we’re just full of caveats today], T-Mobile seems to be aggressive with SA and
VoNR will play a critical role in this strategy, especially if the operator wants to merge VoNR with
SA-based data activities. There is also the distinct possibility that the operator will leverage EPS
fallback (EPSFB) so that it can make SA the priority network architecture but allow smartphones
to fallback to NSA when placing/receiving a call. Ultimately, T-Mobile will need to balance its
aspirations for nationwide SA as the dominant network architecture with the decision to move from
a relatively safe VoLTE strategy to a VoNR implementation that is less proven. EPSFB could play a
role in the move to a nationwide SA network with priority over NSA and it could even be the case
that T-Mobile hasn’t made the final determination on whether or not to go that route.

5 August 31, 2022 | Signals Ahead, Vol. 18, Number 7


Cutting to the chase, we observed that:

➤ VoNR and VoLTE voice quality were impacted by poor RF conditions – low SINR and low
RSRP – but that, in general, the RF conditions were better, or at least more consistent with
LTE.

➤ Depending on the 5G band, VoNR bandwidth consumption requirements were much higher
than VoLTE requirements. This issue can be addressed through vendor enhancements, and it is
not specific to VoNR versus VoLTE.

➤ One great benefit of VoNR over VoLTE is what happens with background data transfers that
can be occurring during a voice call. Over time, we expect an even greater separation in the user
experience for uplink data transfers, and for a reason we discuss in the report.

➤ Current drain tended to [considerably] favor VoLTE over VoNR, especially when compared
with VoNR on Band n41. Then again, complaining about high current drain with a voice call
and ignoring what happens with high bandwidth applications or the display backlight on full
brightness is akin to complaining about inflation by noting a penny gumball machine now only
accepts nickels and it doesn’t dispense change.
We collaborated with two test and measurement (T&M) companies – ironically, the same two We collaborated with
companies that we collaborated with for the first VoLTE study. Accuver Americas provided its Accuver Americas and
XCAL-M drive test solution, as well as its XCAP post-processing tool. We used XCAL-M to log Spirent Communications
for this benchmark study.
the chipset diagnostic messages coming from the smartphones on the T-Mobile network and we
used XCAP to post process and analyze all the chipset-related data we collected.
We used Umetrix Voice for Spirent Communications for the voice quality testing. Umetrix Voice
leverages the industry-standard POLQA algorithms to measure voice quality and assign perceived
voice quality with an objective numerical score (Mean Opinion Score (MOS). We also used the
company’s Umetrix Data platform to generate high bandwidth downlink and uplink data transfers
over sustained periods of time. We used the HTTP protocol in this study with two-minute data
streams (continuously repeated) so that we could perform extended drive testing as well as relatively
short HTTP downlink, uplink, and simultaneous downlink/uplink data transfers for some of our
comparative tests.
We’ve collaborated with both companies over the last decade in numerous benchmark studies for
Signals Ahead or commissioned engagements.
Chapter 2 contains the key findings from this study. Chapter 3 provides detailed results and
analysis for our stationary and drive tests. Finally, Chapter 4 provides our test methodology.

6 August 31, 2022 | Signals Ahead, Vol. 18, Number 7


IN CASE YOU MISSED IT:
IN CASE
SIGNALS YOU BACK
AHEAD MISSED IT:
ISSUES
SIGNALS AHEAD BACK ISSUES
➤ 8/1/22 “5G: The Greatest Show on Earth! Vol 26: Three’s ➤ 6/23/22 “5G: The Greatest Show on Earth! Vol 25:
Company” SRG just completed its 26th 5G NR benchmark Hold ‘em or Fold ‘em” SRG just completed its 25th 5G NR
study. For this endeavor we collaborated with Accuver Americas benchmark study. For this endeavor we collaborated with Accuver
and Spirent Communications to conduct an independent bench- Americas and Spirent Communications to conduct an indepen-
mark study of 5G 3CC, including Band n71 (FDD – 15 MHz), dent benchmark study of the DISH Open RAN network (mobile
Band n41 (100 MHz – TDD) and Band n41 (40 MHz – TDD). data and VoNR) in Las Vegas.
Highlights of the Report include the following: Highlights of the Report include the following:
Our Thanks. We did this study in collaboration with Accuver Our Thanks. We did this study in collaboration with Accuver
Americas (XCAL-M and XCAP) and Spirent Communications Americas (XCAL-M and XCAP) and Spirent Communications
(Umetrix Voice and Umetrix Data). SRG is responsible for the data (Umetrix Data and Umetrix Voice). SRG is responsible for the data
collection and all analysis and commentary provided in this report. collection and all analysis and commentary provided in this report.
Our Methodology. Testing took place over a three-day period in Our Methodology. Testing took place over a two-day period in
early July. We had access to a T-Mobile 17-site test cluster that late May. We leveraged up to two Motorola Edge+ smartphones
is part of its commercial network. Using a Galaxy S22 with pre- on the DISH network and up to two Galaxy S20 smartphones on
release software that supported 3CC and SA we did comparative the T-Mobile network (primarily n71 for comparison purposes).
drive testing with a Galaxy S21 smartphone that was limited to We tested full buffer downlink/uplink/simultaneous transfers with
2CC. We also forced this phone to operate in NSA mode. HTTP and UDP while stationary and mobile. We also tested voice
Four Areas of Focus. We looked at 3CC peak performance in an services, specifically VoNR on DISH and VoLTE on T-Mobile.
“empty network,” mapped SINR and RSRP to the distance to the We also did latency/jitter stress tests using a low bit rate UDP data
serving cell(s) for both n71 and n41, documented the incremental transfer to the Umetrix Data server,.
benefits of using Band n71 FDD as the anchor carrier with Band Degree of Difficulty. Prior to sharing a summary of the results, we
n41 serving as the secondary cells, and evaluated the performance note the high degree of difficulty - entirely new network, 5G SA
attributes of the Galaxy S22 and Galaxy S21 smartphones. with carrier aggregation, VoNR, and the use of AWS. However,
FDD-TDD Advantage. We once again observed the benefits of using consumers don’t know/don’t care about the network architecture.
FDD as the anchor band to help extend and improve Band n41 They only want to have a good and consistent user experience.
coverage and performance. For us, the challenge remains finding Consistently Inconsistent. End user data speeds could be well
locations in the network where the feature is necessary. within expectations, excluding HTTP uplink which meaningfully
n71 Versus n41. By knowing the exact locations of the serving cell lagged expectations. However, the seemingly on-again / off-again
sites we were able to show the relationships between Band n41 use of carrier aggregation was problematic, as was RF coverage,
and Band n71 RSRP and the distance to the serving cell. We show which is mentioned in the next bullet. Voice services (VoNR) could
that in a network designed for capacity (not simply coverage) the deliver very high voice quality with MOS well above 4.0, only to be
penalty for using mid-band frequencies for 5G are overstated followed by very poor voice quality, to the point of being unintel-
(although real). Further, we once again show that low frequencies ligible (below 2.0 or no MOS score achieved).
may be great for coverage but the mid-band frequencies actually RF Coverage was Suboptimal. We documented a significant differ-
deliver overall better network quality. ence in Band n71 coverage (RSRP) and quality (SINR) between
S21 Versus S22. For us, this analysis proved to be the most inter- the DISH and T-Mobile networks, and largely due to a significant
esting. Although the two phones performed largely the same over difference in the number of unique cells (PCI counts). More often
all network conditions, there were obvious differences in how they than not, a drive test resulted in the phone on the DISH network
achieved their respective results. Further, one smartphone tended reverting to the AT&T network.
to perform much better with more favorable RF. We’ll let you guess Opportunities Abound. DISH needs to improve and optimize
which one. network coverage, AWS may need to focus on reducing the very
long latency tail and VoNR consistency, and the Open RAN
vendors should continue working on improving VoNR and sched-
uling efficiency. The 5G clock is ticking.

7 August 31, 2022 | Signals Ahead, Vol. 18, Number 7


Contents
1.0 Executive Summary………………………………………………………………………………………………………………………………………… 3
2.0 Key Observations…………………………………………………………………………………………………………………………………………… 10
3.0 VoNR Results and Analysis……………………………………………………………………………………………………………………………… 14
3.1 VoLTE and VoNR Stationary Tests with Background Data Transfers…………………………………………………………… 14
3.1.1 VoLTE……………………………………………………………………………………………………………………………………………… 14
3.1.2 VoNR……………………………………………………………………………………………………………………………………………… 16
3.2 VoLTE and VoNR Drive Test Results…………………………………………………………………………………………………………… 19
3.3 n71 Versus n41 Drive Test Results…………………………………………………………………………………………………………… 27
3.4 VoNR Freewheeling Test with Background Data Traffic…………………………………………………………………………… 35
3.5 VoLTE and VoNR Current Analysis and Call Setup Times………………………………………………………………………… 39
3.5.1 Current Analysis…………………………………………………………………………………………………………………………… 39
3.5.2 Call Setup Times……………………………………………………………………………………………………………………………… 41
4.0 Test Methodology…………………………………………………………………………………………………………………………………………… 42
5.0 Final Thoughts…………………………………………………………………………………………………………………………………………………45

Index of Figures & Tables


Figure 1. VoLTE Voice Quality……………………………………………………………………………………………………………………………………… 14
Figure 2. Background Data Traffic During a VoLTE Call – Time Series……………………………………………………………………………… 15
Figure 3. Background Data Traffic During a VoLTE Call – Median Values………………………………………………………………………… 15
Figure 4. VoNR Voice Quality…………………………………………………………………………………………………………………………………… 16
Figure 5. Background Data Traffic During a VoNR Call – Time Series……………………………………………………………………………… 16
Figure 6. Background Data Traffic During a VoNR Call – Median Values………………………………………………………………………… 17
Figure 7. Background PUSCH Data Traffic with HTTP Uplink Only During a VoNR Call – Time Series………………………………… 18
Figure 8. Background PUSCH Data Traffic with HTTP Uplink Only During a VoNR Call – Comparative Results…………………… 18
Figure 9. Uplink MCS and Modulation Scheme Allocations During a VoNR Call – Time Series………………………………………… 18
Figure 10. VoNR 5G Band Usage………………………………………………………………………………………………………………………………… 19
Figure 11. VoLTE 5G Band Usage………………………………………………………………………………………………………………………………… 19
Figure 12. VoLTE and VoNR Voice Quality………………………………………………………………………………………………………………… 20
Figure 13. VoLTE MOS……………………………………………………………………………………………………………………………………………… 20
Figure 14. VoNR MOS………………………………………………………………………………………………………………………………………………… 21
Figure 15. VoLTE MOS Versus LTE SINR – Time Series…………………………………………………………………………………………………… 21
Figure 16. VoLTE MOS Versus LTE RSRP – Time Series………………………………………………………………………………………………… 22
Figure 17. VoNR MOS Versus 5G Frequency Band – Time Series………………………………………………………………………………… 22
Figure 18. VoNR UE MOS Versus Average SINR – Time Series……………………………………………………………………………………… 23
Figure 19. VoNR Terminating UE, Originating UE and Average SINR – Time Series………………………………………………………… 23
Figure 20. VoNR UE MOS Versus Average RSRP – Time Series…………………………………………………………………………………… 24

8 August 31, 2022 | Signals Ahead, Vol. 18, Number 7


Figure 21. VoNR Originating UE MOS Versus UE SINR Correlations……………………………………………………………………………… 24
Figure 22. VoNR Terminating UE MOS Versus UE SINR Correlations…………………………………………………………………………… 25
Figure 23. VoNR UE MOS Versus UE RSRP Correlations – Band n71 only……………………………………………………………………… 25
Figure 24. VoLTE Originating and Terminating UE MOS Versus Average SINR Correlations…………………………………………… 26
Figure 25. VoLTE Originating and Terminating UE MOS Versus Average RSRP Correlations…………………………………………… 26
Figure 26. n41 and n71 VoNR MOS…………………………………………………………………………………………………………………………… 27
Figure 27. n71 and n41 VoNR MOS…………………………………………………………………………………………………………………………… 27
Figure 28. Band n71 VoNR MOS and SINR – Time Series…………………………………………………………………………………………… 28
Figure 29. Band n41 VoNR MOS and SINR – Time Series…………………………………………………………………………………………… 28
Figure 30. Band n41 and Band n71 VoNR MOS Versus Average RSRP Correlations………………………………………………………… 29
Figure 31. Band n71 VoNR PDSCH and PUSCH Throughput………………………………………………………………………………………… 29
Figure 32. Band n41 VoNR PDSCH and PUSCH Throughput………………………………………………………………………………………… 30
Figure 33. Band n41 and Band n71 VoNR PDSCH Throughput – Time Series………………………………………………………………… 30
Figure 34. Band n41 and Band n71 VoNR PUSCH Throughput – Time Series…………………………………………………………………… 31
Figure 35. VoLTE PDSCH and PUSCH Throughput………………………………………………………………………………………………………… 31
Figure 36. VoLTE, VoNR n71 and VoNR n41 Median Throughput……………………………………………………………………………………32
Figure 37. VoLTE and VoNR n71 Throughput Relative to n41 Throughput…………………………………………………………………………32
Figure 38. n71 and n41 Downlink and Uplink Throughput – Time Series………………………………………………………………………… 33
Figure 39. n71 and n41 Downlink and Uplink Throughput – Average……………………………………………………………………………… 33
Figure 40. n71 and n41 Downlink and Uplink Spectral Efficiency…………………………………………………………………………………… 33
Figure 41. n71 and n41 Relative Bandwidth Consumption…………………………………………………………………………………………… 34
Figure 42. Originating and Terminating UE 5G Band Selection…………………………………………………………………………………… 35
Figure 43. Originating and Terminating UE Voice Quality…………………………………………………………………………………………… 35
Figure 44. Freewheeling MOS………………………………………………………………………………………………………………………………… 36
Figure 45. Originating UE PDSCH Throughput…………………………………………………………………………………………………………… 36
Figure 46. Terminating UE PUSCH Throughput……………………………………………………………………………………………………………37
Figure 47. VoNR Originating and Terminating UE MOS and Average SINR – Time Series…………………………………………………37
Figure 48. VoNR Originating and Terminating UE MOS Versus Average SINR Correlations…………………………………………… 38
Figure 49. VoLTE Current Drain with 5G Disabled……………………………………………………………………………………………………… 39
Figure 50. VoLTE Current Drain with 5G Enabled……………………………………………………………………………………………………… 40
Figure 51. VoNR Current Drain with 5G Enabled………………………………………………………………………………………………………… 40
Figure 52. VoLTE and VoNR Call Setup Times……………………………………………………………………………………………………………… 41
Figure 53. VoNR Call Signaling Messages……………………………………………………………………………………………………………………… 41
Figure 54. XCAL-M………………………………………………………………………………………………………………………………………………… 43
Figure 55. Umetrix Data Architecture……………………………………………………………………………………………………………………… 43
Figure 56. Umetrix Voice………………………………………………………………………………………………………………………………………… 44

9 August 31, 2022 | Signals Ahead, Vol. 18, Number 7


2.0 Key Observations
We offer the following key observations based on our analysis of the data.

VoNR voice quality could be slightly better than VoLTE, but it was also suscep-
tible to greater variance in the voice quality, largely due to the 5G RF conditions.
We used Umetrix Voice to measure voice quality for VoNR and VoLTE while stationary and while
mobile. We also looked at potential differences in VoNR voice quality between the two bands and
with and without concurrent data transfers occurring in the background during the voice call. The
quick synopsis is that all things being equal, VoNR voice quality in the two 5G bands was similar to
VoLTE (largely Band 2) while background data transfers had no obvious impact on call quality. A
statement that is true for both VoLTE and VoNR.
In a comparative lengthy drive test in which we allowed the VoNR smartphones to operate in In our VoLTE testing that we
either n71 or n41, the median VoNR MOS was 4.0 while the median VoLTE MOS was only 3.75. did in 2014, we rarely / almost
In a separate VoNR drive test in which we band locked the smartphones to either n41 or n71, the never observed a VoLTE MOS
higher than 4.0, which was
median VoNR MOS was 4.18 and 4.26, respectively. We note band locking the smartphones forced
still arguably much better
them to potentially operate in a less-than-ideal band, based on the RF conditions present along than the 3G CS MOS of 2.49.
the drive route. Nonetheless, the median voice quality slightly favored VoNR, or at least VoLTE
voice quality clearly was not superior to that of VoNR. And at the 90th percentile, the VoNR MOS
(unlocked phones) was a very strong 4.3, compared with 3.9 for VoLTE. In our VoLTE testing that
we did in 2014, we rarely/almost never observed a VoLTE MOS higher than 4.0, which was still
arguably much better than the 3G CS MOS of 2.49. At the time, VoLTE used the AMR wideband
codec while 3G CS used a narrowband voice codec.
Median values only tell part of the story. Looking at the MOS standard deviations, the results Virtually all poorer VoNR
favored VoLTE with a standard deviation of 0.35 compared with 0.6 for VoNR. VoNR MOS at the results (sub 3.25) occurred
10th percentile in this particular drive route was 3.26 and for VoLTE MOS it was 3.19. In chapter 3 when the SINR was below
-5 dB and/or the RSRP
we provide multiple plots that look at voice quality as a function of signal quality (SINR) and signal
was below -105 dBm.
strength (RSRP). It isn’t feasible to directly compare results between the two voice technologies
since SINR and RSRP are calculated differently on the two technologies, plus the two technolo-
gies have different means of coping with high interference. However, suffice it to say that VoNR
and VoLTE voice quality suffered when there was higher interference and/or low signal strength.
Virtually all poorer VoNR results (sub 3.25) occurred when the SINR was below -5 dB and/or the
RSRP was below -105 dBm.
Along this lengthy drive route, the RSRP at the 10th percentile was -105 dBm and the SINR at
the 10th percentile was -4.3 dB. Put another way, over 10% of the area during the drive test the voice
quality would potentially be noticeably worse than what consumers are used to with VoLTE. In our
testing of T-Mobile in other markets, we’ve observed that Band n41 typically provided far more
favorable RF conditions for VoNR than VoLTE, at least in those areas where Band n41 coverage
existed. Worth noting, when we tested the DISH network for our recent Signals Ahead report on
Open RAN, we determined the T-Mobile Band n71 SINR at the 10th percentile was -3.7 dB while
for Band n41 SINR at the 10th percentile it was -3.1 dB. This value, however, includes portions
along the drive routes where the operator hadn’t deployed Band n41 in any meaningful capacity, so
the data point is somewhat skewed to the low side. Therefore, we expect Band n41 to play a much
greater role in the operator’s VoNR strategy.

The most compelling factor of VoNR is that the data session remains on 5G instead
of falling back to LTE with VoLTE. We already knew from our inadvertent testing of VoNR
last August that when a VoNR call is placed, background data traffic moves from 5G NR to LTE,
and that it returns to 5G NR when the voice call is finished. In this recent study, we reconfirmed

10 August 31, 2022 | Signals Ahead, Vol. 18, Number 7


this observation. We did stationary testing in which we ran through a series of data transfer tests
with two pairs of Galaxy smartphones: the Galaxy S20 and the Galaxy S21. We executed a HTTP
full buffer download test, followed by a HTTP full buffer upload test and a HTTP full buffer
download/upload test, in which data transfers occurred simultaneously in both directions. We then
established a voice call and repeated the data transfer tests.
In the first set of data tests with the Galaxy S20, the data transfers occurred on 5G and LTE
since the phones were operating in NSA mode. The S20 achieved an average downlink throughput
of nearly 1 Gbps and an average uplink throughput of 150 Mbps. The throughput with the simulta-
neous data transfers was just under 1.1 Gbps. When we initiated the VoLTE call and repeated the
test, the data throughput dropped to 72 Mbps, 32 Mbps, and 80 Mbps, respectively, since all the
data traffic occurred over LTE.
With the Galaxy S21, the downlink throughput was just over 1.2 Gbps without VoNR and just
over 1.1 Gbps with VoNR. In both cases the smartphone was using n41 carrier aggregation with
100 MHz in the P Cell and 40 MHz in the S1 Cell. As previously noted, all data traffic went over
5G. For the uplink, the throughput was just under 75 Mbps and 33 Mbps with VoNR. The uplink
throughput dropped considerably with VoNR for reasons discussed in the last section, even though
the data transfer remained on 5G. For the simultaneous data transfer test, the total downlink/uplink
throughput was just under 1.2 Gbps with and without VoNR.
The measured data speeds we obtained for these tests were greatly influenced by the radio condi-
tions and network loading. We can’t claim these results are representative of a typical user experi-
ence – the LTE throughput could be higher and the 5G throughput could be lower. However,
the advantage clearly goes to VoNR and 5G, especially when smartphones can support VoNR and
3CC. Keeping the data sessions on 5G also frees up LTE capacity for other subscribers since we’ve
observed that typically the n41 network is far less capacity constrained than its elder sibling network
and its respective frequency bands.

Bandwidth consumption on VoNR was higher than VoLTE, but it could be a near-
term phenomenon. We measured how much bandwidth the smartphones used when placing a
VoNR call and when placing a VoLTE call. We also tested the smartphones with VoNR in Band
n41 and in Band n71 to examine relative differences between the two 5G bands. The bandwidth
advantage clearly went to VoLTE (Band 2), followed by VoNR in Band n71 with VoNR in Band n41
trailing by a considerable margin.
With the traffic activity pattern used for the MOS tests, VoLTE generated a median down- The VoNR median downlink
link throughput of 0.004 Mbps and 0.014 Mbps in the uplink. The VoNR median downlink throughput in Band n41
throughput in Band n71 (0.055 Mbps) was nearly 14x higher than VoLTE while the median uplink was 77.5x higher than
VoLTE and the median
throughput (0.029 Mbps) in Band n71 was 2x higher than VoLTE. Finally, the VoNR median
uplink throughput was
downlink throughput in Band n41 (0.31 Mbps) was 77.5x higher than VoLTE and the median 10x higher than VoLTE.
uplink throughput (0.14 Mbps) was 10x higher than VoLTE. Comparing the downlink and uplink
throughput within each voice technology/band the median uplink throughput on VoLTE was 3.5x
higher than the downlink throughput while for VoNR the median downlink throughput exceeded
the median uplink throughput, or 1.9x higher for Band n71 and 2.2x higher for Band n41. We
address this point in the last discussion topic in this chapter since we believe there are opportunities
to reduce the bandwidth requirements of VoNR without impacting the voice quality.

Current drain favored VoLTE while call setup times were very similar between
the two voice technologies. We measured the current drain during a VoNR call and during a
VoLTE call and compared it with the baseline idle mode, as well as airplane mode. In all cases, the
backlight display, which can have a great impact on the battery life, was turned on.
With VoLTE, we observed a ~100 mA increase in the current drain relative to the S20 smartphone
in idle mode. In absolute terms, the current measurement in idle mode was approximately 200 mA

11 August 31, 2022 | Signals Ahead, Vol. 18, Number 7


while during the voice call it was approximately 300 mA. With VoNR, the current drain on the S21
smartphone was just under 400 mA, compared with approximately 200 mA with the smartphone
in idle mode. It isn’t appropriate to compare absolute values between the two sets of results since we
used two different phone models, and the display backlight brightness was likely slightly different as
well. This VoNR call occurred with the S21 smartphones using n41. We didn’t separately test with
n71 (regrets). In some limited testing T-Mobile hasn’t observed much impact on the battery life with
VoNR versus VoLTE and it believes our higher current drain was largely the result of using n41 and
its wider channel bandwidth than anything else. We don’t disagree but we would like to collect more
additional data on our own to confirm the statement.
We did a lengthy drive route to look at the VoNR and VoLTE call setup times. In this study we In this study we found the
found the call setup times were nearly the same, albeit slightly favoring VoLTE (6.3 seconds) over call setup times were nearly
VoNR (6.6 seconds). We doubt consumers would ever notice the difference since user behavior plays the same, albeit slightly
favoring VoLTE (6.3 seconds)
a big role in the process. The same could not be said for CSFB when we vividly remember waiting
over VoNR (6.6 seconds).
patiently for the ring tone to occur and/or for the call to return to LTE mode after finishing the 3G
voice call. In our testing in Vegas, we documented a meaningful difference in the call setup times
between VoNR (DISH) and VoLTE (T-Mobile). Although we attributed the differences to VoNR,
in hindsight it is likely other factors came into play, such as the relative locations of the two networks
and network loading. For its part, T-Mobile did not expect to see any differences in the call setup
times since the two networks use the same IMS core network.

There are opportunities to address various aspects of VoNR performance to


improve upon the performance we observed in our recent testing. There are four
opportunities for improving VoNR performance. Two of these opportunities are specific to VoNR
and the other two opportunities are more general in nature since they can be observed in the absence
of VoNR, and they have benefits that transcend improving voice services over 5G.
When testing VoNR with simultaneous background data transfers we observe an abnormally high
about of data traffic in the downlink direction, even though the background data only involved
HTTP full buffer uplink data transfers. Specifically, we observed sustained downlink data speeds
approaching 30 Mbps when the only known traffic in the downlink direction was the ACK/NACK
messages associated with the uplink data transfer, as well as the modest amount of data required
to support the VoNR voice activity. From testing other networks, such as T-Mobile (Nokia), we
know the ACK/NACK messages only require 0.5 to 2 Mbps of bandwidth, or far less than what we
encountered in our most recent testing. We observed a similar phenomenon when testing Verizon
Band n77 in Minneapolis, which is also an Ericsson market. Based on feedback from both T-Mobile
and Ericsson we can attribute the high downlink data traffic to the use of Resource Allocation Type
0, whereby downlink scheduling of resource blocks is done at the resource block group (RBG) level,
meaning it isn’t possible to interject more granularity and only allocate one RB for the data traffic (or
ACK/NACK messages) if only one RB is needed.
In addition to scheduling downlink RBs at the RBG level, Ericsson uses high transport block
sizes for what are essentially small ACK/NACK data packets. This situation means the RBs are
packed with padding bits, thereby accounting for the high/unnecessary downlink throughput. The
use of Resource Allocation Type 0 also resulted in much higher resource requirements for VoNR
than for VoLTE. As noted earlier in this section, VoNR in Band n41 utilized 77.5x more bandwidth
in the downlink direction than VoLTE in Band 2. Resource Allocation Type 1 supports much better
granularity of RB scheduling – down to the RB level – but it also requires that 5G chipsets support
dynamic switching between the two resource allocation types. This capability doesn’t exist today, as
we understand. Resource Allocation Type 0 also has its benefits, including with MU-MIMO since
it makes concurrent scheduling of two UEs much easier to accomplish.
Ericsson and T-Mobile didn’t seem too concerned about the high/unnecessary bandwidth associ-
ated with the use of Resource Allocation Type 0, noting that mid-band 5G networks are lightly

12 August 31, 2022 | Signals Ahead, Vol. 18, Number 7


loaded today. We would agree and the impact on voice capacity is trivial, however, it is a bit discon-
certing to witness 20-30 Mbps in the downlink direction terminating on our smartphone when it is
largely unnecessary. Furthermore, when 5G networks due start to become congested, minimizing
unnecessary resource block allocations will become far more important so that those RBs can be
allocated to real data traffic.
In addition to supporting Resource Allocation Type 1, Ericsson is working on means to minimize
the padding bits that are generating the abnormally high downlink throughput. We would expect
these features to be in use in commercial 5G networks in the very near term. Although these features
will minimize the observed downlink data speeds, we do not believe they would have any impact on
how resource blocks get allocated with type 0 scheduling, meaning “a chunk” of RBs would still be
used for ACK/NACK messages even though far fewer RBs could accomplish the same task. Still,
implementing these features should help improve network capacity with reduced interference and
reducing the downlink throughput should have a positive impact on the smartphone’s battery life.
We believe all infrastructure vendors are using Resource Allocation Type 0 today although some
vendors may have already introduced means of minimizing the high bandwidth while still using
more than necessary RBs.
We also observed that the smartphone was limited to 16QAM in the uplink with an active VoNR We observed that the
call. When testing with full buffer uplink data transfers and VoNR, the smartphone stopped using smartphone was limited
uplink-64QAM as soon it started a VoNR call, reverting instead to uplink-16QAM. As soon as the to 16QAM in the uplink
with an active VoNR call.
voice call terminated, the smartphone returned to using uplink-64QAM. The impact of limiting
the uplink modulation scheme to 16QAM resulted in 45% lower uplink throughput (background
data transfers) with VoNR than without VoNR. We note that when we tested with VoLTE, the
smartphone continued to use the higher modulation scheme to support the full buffer uplink data
transfer that was occurring during the voice call.
According to Ericsson and T-Mobile, this early and more conservative link adaptation was imple-
mented to ensure a good voice call. With uplink-64QAM the higher burst rate of the voice packets
could degrade the voice quality if the packets are not successfully received at the distant end. We
believe Ericsson will eventually support the higher modulation scheme, just as it has done with
VoLTE, and that it will then be up to the mobile operator to decide how aggressive to make its link
adaptation when scheduling uplink data packets during a VoNR call.
VoLTE has been around since 2014 so vendors have had ample time to implement new features Slot aggregation can
and enhancements. VoNR is still in its infancy. Over time, we would expect vendors to support help improve VoNR
more features to help improve the overall VoNR experience. TTI bundling/slot aggregation is one uplink coverage.
example of a feature that we believe is currently missing with VoNR and which exists with VoLTE.
Slot aggregation can help improve VoNR uplink coverage and it will be very useful when operators
deploy voice services in mid-band spectrum.
Lastly, there is the “simple matter” of optimizing the deployed 5G network by filling in coverage
where necessary and minimizing interference everywhere across the network. Ericsson indicated
it has and is working on enhancements to minimize PDSCH-related interference. One example,
which we believe is already implemented, is NR lean carrier, whereby the network doesn’t broadcast
NR CSI (channel state information) unless it is needed for scheduling purposes. In Chapter 3 we
show an obvious correlation between low SINR/high interference and poorer voice quality. The
phenomenon isn’t unique to 5G since it also exists with VoLTE and LTE, but anything and every-
thing that can be done to improve 5G coverage and signal quality needs to be done to ensure the
best possible user experience.

13 August 31, 2022 | Signals Ahead, Vol. 18, Number 7


3.0 VoNR Results and Analysis
We’ve divided our analysis into five sections. In each section we explore a different facet of VoNR
or incorporate a different test scenario. In the first section, we provide results from a comparative
stationary test in which we had background data traffic occurring during the VoLTE and VoNR
calls. In the second section, we look results while drive testing around the test cluster. In the third
section we show results from two different drive tests involving the same route – in one test we
locked the smartphones to Band n71 and in the second test we locked the smartphones to Band n41.
In both tests we had background data transfers occurring alongside the VoNR calls. In the fourth
section, we provide results from a drive test in which both smartphones were enabled to support
Band n71 and Band n41 while also generating background data traffic in the downlink or uplink
directions. Finally, in the last section we show the impact of VoNR on current drain and compare it
with VoLTE. We also include our analysis of call setup times between the two voice technologies.

3.1 VoLTE and VoNR Stationary Tests with Background Data


Transfers
We did a series of tests while stationary to evaluate the impact of a voice call (VoNR or VoLTE)
on the data transfer speeds over 5G or LTE. Conversely, these tests also allowed us to observe how
background data transfers impacted voice quality. We provide the VoLTE results in the first subsec-
tion and the VoNR results in the second subsection.

3.1.1 VoLTE
For this stationary test we used Umetrix Data to execute a series of short throughput tests: HTTP
Full Buffer Downlink, HTTP Full Buffer Uplink, and HTTP Full Buffer Downlink/Uplink. For
the last test, the downlink and uplink data transfers occurred simultaneously. Figure 1 shows the
median and standard deviation for the voice quality (MOS). Since we only captured the voice quality
when running the three aforementioned tests, the test time was relatively short, which helps explain
the slightly higher standard deviation compared with our overall experiences for VoLTE. Although
it isn’t shown in the figure, we never observed the simultaneous data transfers impacting the VoLTE
call quality. The same cannot be said for VoLTE’s impact on the data transfer speeds.

Figure 1. VoLTE Voice Quality


3.78 3.71

0.30
0.07
Originating Terminating
MOS (Median)
Standard Deviation
Source: Signals Research Group

14 August 31, 2022 | Signals Ahead, Vol. 18, Number 7


Figure 2 provides a time series plot of the downlink and uplink data speeds during this test. The
HTTP downlink and uplink tests were ten seconds long, so we repeated the downlink test three
times and the uplink test two times. This approach explains the periodic dips in the throughput,
which occurred during the very short time gaps between each test. The full buffer downlink/uplink
test was approximately twenty-five seconds long, so we only ran this test once without and then with
VoLTE. At approximately 200 seconds into the test, and after executing all three test scenarios, we
started the VoLTE call. Prior to the VoLTE call occurring, all the data traffic went over 5G (n41
carrier aggregation) but as soon as we started the VoLTE call all the data traffic switched to LTE.
We’ve observed this behavior numerous times in the past, so it wasn’t unexpected.

Figure 2. Background Data Traffic During a VoLTE Call – Time Series


Median (Mbps) HTTP DL HTTP UL HTTP DL/UL HTTP DL HTTP UL HTTP DL/UL EVS
1000
5G n41 PDSCH

800 13.2 Kbps

600

400 Phone uses LTE Only with VoLTE


LTE S1
Median (Mbps) HTTP DL HTTP UL HTTPCell
DL/UL EVS DL
HTTP DL EVSHTTP
UL UL HTTP DL/UL EVS
1000 PDSCH
200
5G n41 PDSCH LTE P Cell PDSCH
5G n41 PUSCH
LTE PUSCH
8000 13.2 Kbps

0 20 40 60 80 100 120 140 160 180 200 220 240 260 280 300
600 Time (sec)
Source: Signals Research Group

400
Mbps Phone uses LTE Only with VoLTE
LTE S1
1200As shown in Figure 2 and emphasized in Figure 3, there was a dramatic Cell drop in the EVS
totalDL down-EVS TheUL simultaneous downlink/
link/uplink throughput when the data traffic switched from 5G to LTE. PDSCH For example, the simul- uplink throughput was
200
LTE PUSCH 38.3
1000 LTE PUSCH 0.1
taneous downlink/uplink throughput was 13.5 5G
times higher over 5G (no VoLTE) than
n41 PUSCH LTEitP Cell over 13.5
wasPDSCH times higher over 5G
LTE VoLTE)
PUSCH than it was
LTE PDSCH 148.1 (no
LTE 0 (with VoLTE).
LTE PDSCH Another
122.4 observation, which we will examine in more detail in a subsequent 13.5x Higher
800 5G n41 PUSCH 76.7 over LTE (with VoLTE).
0 20 n41 PUSCH
5G 40 0.7 60 80 100 120 140 160 180 200 220 240 260 280 300
Time (sec)
600
Figure 3. Background Data Traffic During a VoLTE Call – Median Values
Mbps
400 5G n41 PDSCH
1200 5G n41 PDSCH
836.1
816.5
200
LTE PUSCH 0.1 LTE PUSCH 38.3
1000 LTE PUSCH 37.1
LTE PDSCH 148.1 LTE PUSCH 31.7
5G n41 PUSCH 70.7 LTE PUSCH 32.5
0 LTE PDSCH 122.4 5G n41 PDSCH 43.2 LTE PDSCH 71.7 LTE
13.5x PDSCH 1.6
Higher LTE PDSCH 47.3
800 HTTP DL HTTP UL 5G n41 PUSCH
HTTP DL/UL 76.7 HTTP DL HTTP UL HTTP DL/UL
5G n41 PUSCH 0.7 No VoLTE with VoLTE

600

400 5G n41 PDSCH


5G n41 PDSCH
836.1
816.5
200
LTE PUSCH 37.1 LTE PUSCH 31.7
5G n41 PUSCH 70.7 LTE PUSCH 32.5
0 5G n41 PDSCH 43.2 LTE PDSCH 71.7 LTE PDSCH 1.6 LTE PDSCH 47.3
HTTP DL HTTP UL HTTP DL/UL HTTP DL HTTP UL HTTP DL/UL
No VoLTE with VoLTE
Source: Signals Research Group

15 August 31, 2022 | Signals Ahead, Vol. 18, Number 7


section in this chapter, is that there was unexpectedly high PDSCH (downlink) throughput occur-
ring during the HTTP uplink only tests.

3.1.2 VoNR
The VoNR stationary test occurred at a different location from the VoLTE test that we discussed in
the previous section. We also did this test while parked very close to the cell site and with relatively
ideal RF conditions (SINR = ~19 dB). These factors explain the very favorable voice quality results
we obtained (reference Figure 4).

Figure 4. VoNR Voice Quality


4.29 4.32

0.04 0.04
Originating Terminating
MOS
Standard Deviation
Source: Signals Research Group

The test scenario was very similar to what we did with the VoLTE tests. However, just by looking
at the time series plot of the throughput in Figure 5, it is evident we only did the HTTP downlink
tests twice (versus three times with VoLTE) and that we did the HTTP downlink/uplink tests twice
versus only once with the VoLTE test from the previous section.

Figure 5. Background Data Traffic During a VoNR Call – Time Series


Mbps HTTP DL HTTP UL HTTP DL/UL HTTP DL HTTP UL HTTP DL/UL EVS
1400 (w/VoNR) (w/VoNR) (w/VoNR) Activity
Total PDSCH

1200

1000
P Cell PDSCH
800

600

S1 Cell PDSCH
400
EVS Codec Activity
200
P Cell PUSCH
0
0 50 100 150 200 250 300 350 400
Time (sec)
Source: Signals Research Group

16 August 31, 2022 | Signals Ahead, Vol. 18, Number 7


Although it is somewhat hard to visualize in the figure, there are three important points worth The downlink throughput
mentioning. First, the downlink throughput remained largely unchanged with VoNR as compared remained largely unchanged
to without it. Second, there was a noticeable drop in the uplink throughput with VoNR than without with VoNR as compared
to without it, while there
it. Lastly, during the HTTP uplink only tests there was once again much higher than expected
was a noticeable drop in
downlink data traffic. These three points are more apparent in Figure 6, which shows the median the uplink throughput with
values for each test. Figure 7 and Figure 8 provide even more visibility into the results. In Figure 7, VoNR than without it.
it is evident there was nearly 30 Mbps of downlink data traffic taking place during a test that only
involved uplink data traffic. We would expect a very minor amount of downlink throughput due
to ACK/NACKs, but more likely in the range of 0.5-2 Mbps, at most, based on testing we have
done with other infrastructure suppliers (e.g., T-Mobile Band n41 and Nokia from our testing in
Minneapolis). We attribute this phenomenon to Ericsson since we also observed it when testing
Verizon’s Band n77 network. We note the vendor is using Resource Allocation Type 0, as discussed
in the previous chapter, and this decision forces the use of a group of resource blocks (16 RBs), even
though fewer RBs may have sufficed. With Resource Allocation Type 1, it will be possible to apply
a more granular selection of RB usage, which should minimize the unnecessary throughput that
we currently observe. Figure 8 shows the median downlink and uplink throughput which occurred
during the HTTP Uplink only tests, both with and without VoNR. In addition to the high amount
of PDSCH traffic due to VoNR, it is also obvious the PUSCH background data traffic dropped
considerably with VoNR. Specifically, it dropped by 45% with the introduction of the 5G voice call.

Figure 6. Background Data Traffic During a VoNR Call – Median Values


PUSCH 1.1
PUSCH 0.4 PUSCH 35.2
PUSCH 80.1

PDSCH PDSCH
1210.4 PDSCH PDSCH
1114.3 1115.7 1182.7

PUSCH 73.5 PUSCH 32.7


PDSCH 21.8 PDSCH 21.6
Full Buffer Full Buffer Full Buffer Full Buffer Full Buffer Full Buffer
HTTP Downlink Only HTTP Uplink Only HTTP Downlink/Uplink HTTP Downlink HTTP Uplink Only HTTP Downlink/Uplink
Only (w/ VoNR) (w/ VoNR) (w/ VoNR)
Mbps (Median)
Source: Signals Research Group

We had originally speculated that the lower uplink throughput was somehow related to sched- We attribute the lower uplink
uling grants and/or slot allocations. We were, of course, wrong. Instead, the sharp drop in uplink throughput to the absence
throughput was due to the smartphone no longer using uplink-64QAM during a VoNR call. This of uplink-64QAM and the
more conservative uplink
fact is evident in Figure 9, which provides a time series plot of the uplink MCS allocations and
link adaptation approach
modulation schemes. During the first ~200 seconds of the call the smartphone always used uplink- Ericsson is currently using.
64QAM when the Umetrix Data session was active – the periodic drops indicate the end/start of
each test. Likewise, the uplink MCS value was almost consistently MCS 27 when there was uplink
data traffic. With VoNR (the last 200 seconds of the figure), the uplink modulation scheme was
limited to 16QAM and the uplink MCS was limited to MCS 16. As discussed in the last chapter,
we attribute this limitation to the more conservative uplink link adaptation approach that Ericsson
is using with VoNR. A more aggressive approach with high bursty traffic could materially degrade
voice quality during a VoNR call. We would anticipate the scheduling to become less conservative
over time and we note that VoLTE does not have this same limitation.

17 August 31, 2022 | Signals Ahead, Vol. 18, Number 7


Figure 7. Background PUSCH Data Traffic with HTTP Uplink Only During a VoNR Call – Time Series
Mbps
Mbps
90
90
P Cell PUSCH
80 P Cell PUSCH
80
70 Nearly 30 Mbps of PDSCH Traffic during
70 Nearly 30 Mbps
HTTP ofOnly
Uplink PDSCH Traffic
Data during
Transfer
HTTP Uplink Only Data Transfer
60
60
50
50
40
40
30
30 Total PDSCH
20 Total PDSCH
20
10 P Cell PDSCH
10 P Cell PDSCH
S1 Cell PDSCH
0 S1 Cell PDSCH
0
0 50 100 150 200 250 300 350 400
0 50 100 150 200 250 300 350 400
Time (sec)
Time (sec) Source: Signals Research Group

Figure 8. Background PUSCH Data Traffic with HTTP Uplink Only During a VoNR Call –
Comparative Results

PUSCH throughput was


PUSCH throughput
45% lower was
with VoNR
45% lower with VoNR

PUSCH
PUSCH
73.5
73.5 PDSCH throughput
PDSCH
was 66%throughput
of the
PUSCH was 66% throughput
PUSCH of the
PUSCH
32.7 PUSCH throughput
in an HTTP Uplink
32.7 inOnly
an HTTP
test Uplink
Only test

PDSCH PDSCH
PDSCH
21.8 PDSCH
21.6
21.8 21.6
Full Bu er HTTP Uplink Only Full Bu er HTTP Uplink Only (w/ VoNR)
Full Bu er HTTP Uplink Only Full Bu er HTTP Uplink Only (w/ VoNR)
Median (Mbps)
Median (Mbps) Source: Signals Research Group

Figure 9. Uplink MCS and Modulation Scheme Allocations During a VoNR Call – Time Series
MCS Value Modulation Type
30 P Cell UL MCS

64QAM
VoNR call is up

20 Uplink Modulation
16QAM

10
QPSK

0
0 50 100 150 200 250 300 350 400
Time (sec)
Source: Signals Research Group

18 August 31, 2022 | Signals Ahead, Vol. 18, Number 7


3.2 VoLTE and VoNR Drive Test Results
This section provides results for a drive test along a large loop within the test cluster. During one
loop, we tested VoLTE with the phones unlocked so that they could use any LTE band. During
the second loop, we tested VoNR with the phones unlocked so that they could use n71 or n41.
We also ran a continuous HTTP full buffer downlink transfer on the originating smartphone and
a continuous HTTP full buffer uplink data transfer on the terminating smartphone. We aren’t
showing the throughput performance in this section since we cover this aspect in other sections.
We know, however, from our testing that the data transfers didn’t impact the voice quality. Figure
10 shows the drive loop and how the two smartphones used the two 5G bands. Interestingly, there
were some obvious differences in band usage between the two phones with the terminating phone
making much greater use of n41. Figure 11 shows how the two smartphones use the various LTE
bands during the VoLTE drive test. Both phones used Band 2 for much of the drive route but there
was some Band 12 and even some Band 71 along with a few inconsequential seconds of Band 66.

Figure 10. VoNR 5G Band Usage

n41 n41
n71 n71

Originating UE Terminating UE
Source: Signals Research Group

Figure 11. VoLTE 5G Band Usage

Lime = Band 2 Lime = Band 2


Red = Band 12 Red = Band 12
Purple = Band 66 Purple = Band 66
Yellow = Band 71 Yellow = Band 71

Originating UE Terminating UE
Source: Signals Research Group

19 August 31, 2022 | Signals Ahead, Vol. 18, Number 7


Figure 12 shows the median MOS and its standard deviation for the two sets of tests. The VoNR
voice quality was slightly better (higher MOS), albeit with a higher standard deviation.

Figure 12. VoLTE and VoNR Voice Quality


4.0 4.0
3.8 3.7

0.6 0.6
0.4 0.3

VoNR Originating VoNR Terminating VoLTE Originating VoLTE Terminating


Median (MOS)
Standard Deviation
Source: Signals Research Group

Figure 13 (VoLTE) and Figure 14 (VoNR) provide geo plots of the voice quality along the drive
route. Since the originating and terminating smartphones had comparable voice quality over the
drive route, we’ve plotted the voice quality for both smartphones in each figure. Conceptually, there
shouldn’t be any differences in the voice quality since both smartphones were the Galaxy S21 and
they were placed adjacent to each other in the vehicle. The “originating” nomenclature merely signi-
fies which smartphone started the call. Once the call was established the voice activity went in both
directions, thus allowing us to measure the voice quality on each end of the call.

Figure 13. VoLTE MOS


> 4.25
> 4.0
> 3.75
> 3.5
> 3.25
> 3.0
<= 3.0

Source: Signals Research Group

20 August 31, 2022 | Signals Ahead, Vol. 18, Number 7


Figure 14. VoNR MOS
> 4.25
> 4.0
> 3.75
> 3.5
> 3.25
> 3.0
<= 3.0

Source: Signals Research Group

The next several figures look at the relationships between SINR (signal quality) and MOS (voice The lower MOS values tended
quality). Figure 15 provides a time series plot of the VoLTE MOS as a function of the SINR. In to coincide with lower SINR,
this figure we are showing the results for both smartphones. For the SINR in this figure and in although we also note there
were periods with low SINR,
subsequent figures, we used a running ten second average that was subsequently offset by ten seconds
but relatively good MOS.
relative to the MOS value. We took this approach to align the radio conditions with the voice
activity that occurred prior to Umetrix Voice determining the voice quality. The Average SINR line
is simply an average of the SINR values for both smartphones (running ten second averages, offset
by ten seconds). We can make two observations. The SINR for the two smartphones was largely the
same over the drive test, not surprising since over most of the drive route the two smartphones used
the same frequency band at each location along the route. Second, the lower MOS values tended to
coincide with lower SINR, although we also note there were periods with low SINR, but relatively
good MOS.

Figure 15. VoLTE MOS Versus LTE SINR – Time Series


SINR (dB) MOS
30 5

Originating MOS Terminating MOS


4
20

3
10
2

0
1
Terminating SINR Average SINR
−10 Originating SINR 0
0 200 400 600 800 1000 1200 1400 1600
Time (sec)
Source: Signals Research Group

21 August 31, 2022 | Signals Ahead, Vol. 18, Number 7


Figure 16 shows a similar time series plot but instead of showing SINR we are showing RSRP
and its impact on voice quality. Comparing the two figures, it is evident there was a strong correla-
tion between SINR and RSRP. Further, it is evident that dips in RSRP could result in poorer voice
quality.

Figure 16. VoLTE MOS Versus LTE RSRP – Time Series


RSRP (dBm) MOS
−55 5
−60 Originating MOS Terminating MOS
−65
−70 4
−75
−80
3
−85
−90
−95
2
−100 Average RSRP
−105
−110 Terminating RSRP 1
−115 Originating RSRP
−120
−125 0
0 200 400 600 800 1000 1200 1400 1600
Time (sec)
Source: Signals Research Group

Switching to VoNR, Figure 17 shows a time series plot of the MOS for the two phones along with
their 5G band usage. All the really low MOS values occurred when one or both smartphones were
using n71, but since one of the smartphones used n71 for much of the drive route, we can’t infer
much from this observation. Further, we believe/know that SINR played an important role with
voice quality. That being said, we also know SINR was generally more favorable in n41 than in n71
so we would expect better voice quality in n41, excluding n41 coverage constrained areas where both
the RSRP and SINR could be unfavorable.

Figure 17. VoNR MOS Versus 5G Frequency Band – Time Series


Frequency MOS
5
VoNR MOS (Originating UE) VoNR MOS (Terminating UE)

n41 3
Frequency (Terminating UE)

2
n71

Frequency (Originating UE) 1

0
0 200 400 600 800 1000 1200 1400 1600 1800 2000
Time (sec)
Source: Signals Research Group

22 August 31, 2022 | Signals Ahead, Vol. 18, Number 7


Figure 18 provides a time series plot of the MOS values for both smartphones as well as the average
SINR, as measured between the two smartphones (10 second binning and offset by 10 seconds).
Figure 19 shows the measured SINR for each smartphone along with the average of the two SINR
values. Along most of the drive route, the measured SINR on the two phones was very similar. The
figure shows a big disparity between 1400 and 1600 seconds with the terminating smartphone’s
SINR being much higher than the originating smartphone’s SINR. During this portion of the drive
route, the terminating smartphone was using Band n41 and the originating smartphone was using
Band n71. The originating smartphone also dropped the call during this portion of the drive route,
thus explaining why there is a gap in time where there are not any MOS values shown in the figure
(the area within the rectangle).

Figure 18. VoNR UE MOS Versus Average SINR – Time Series


SINR (dB) MOS
20 5
VoNR MOS (Originating UE) VoNR MOS (Terminating UE)
SINR
15 (dB) MOS
4
20 5
VoNR MOS (Originating UE) VoNR MOS (Terminating UE)
10
15 3
4
5
10 2
0 Avg SINR 3
5
1
−5 2
0 Avg SINR
−10 0
1
−5 0 200 400 600 800 1000 1200 1400 1600 1800 2000
Time (sec)
−10 0 Group
Source: Signals Research
0 200 400 600 800 1000 1200 1400 1600 1800 2000
SINR (dB) Time (sec)
30
Figure 19. VoNR Terminating UE, Originating UE and Average SINR – Time Series
SINR (dB) Terminating SINR
30
20

Avg SINR Terminating SINR


20
10

Avg SINR
10
0

Originating
SINR
0
−10
0 200 400 600 800 1000 1200 1400 1600 1800 2000
Originating
SINR
−10 Time (sec)
0 200 400 600 800 1000 1200 1400 1600 1800 2000

Time (sec) Source: Signals Research Group

23 August 31, 2022 | Signals Ahead, Vol. 18, Number 7


Figure 20 shows a time series plot of the average RSRP between the two phones and the MOS for
each phone. There is a similar trend in the data with lower RSRP generally, but not always, resulting
in unfavorable voice quality.

Figure 20. VoNR UE MOS Versus Average RSRP – Time Series


RSRP (dBm) MOS
−55 VoNR MOS (Originating UE) 5
−60 VoNR MOS (Terminating UE)
−65
−70 4
−75
−80
3
−85
−90
−95
2
−100
−105
−110 1
−115 Avg RSRP
−120
−125 0
0 200 400 600 800 1000 1200 1400 1600 1800 2000
Time (sec)
Source: Signals Research Group

The next three figures provide correlation analysis between the SINR and MOS. Figure 21
provides the relationships between the MOS of the originating smartphone and three different
SINR values – the originating smartphone’s SINR, the terminating smartphone’s SINR and the
average SINR. Figure 22 provides comparable information for the MOS of the terminating smart-
phone. In both cases, there is an obvious relationship between the two metrics – the MOS values
were less favorable with lower SINR and more favorable with higher SINR. This relationship is more
evident in Figure 21.

Figure 21. VoNR Originating UE MOS Versus UE SINR Correlations


MOS (Originating UE)

5.0

4.5
Originating SINR
4.0
Terminating SINR

3.5

3.0

2.5

Average SINR
2.0
−20 −15 −10 −5 0 5 10 15 20 25 30

SINR (dB) Source: Signals Research Group

24 August 31, 2022 | Signals Ahead, Vol. 18, Number 7


Figure 22. VoNR Terminating UE MOS Versus UE SINR Correlations
MOS (Terminating UE)
5.0

4.5
Terminating SINR
Originating SINR
4.0

3.5

Limited Data Points


3.0

2.5
Average SINR
2.0
−20 −15 −10 −5 0 5 10 15 20 25 30
SINR (dB)
Source: Signals Research Group

Figure 23 shows the correlation between voice quality and RSRP for those times when both Virtually all of the poor
smartphones were using Band n71. We did not include the results for Band n41 since there were not voice quality scores (MOS
enough data points to produce a statistically meaningful chart. We do, however, show correlation < 3) occurred when the
RSRP was below -105 dBm.
results with n41 in a subsequent section of this report. As evident in the figure, virtually all of the
poor voice quality scores (MOS < 3) occurred when the RSRP was below -105 dBm.

Figure 23. VoNR UE MOS Versus UE RSRP Correlations – Band n71 only
MOS

5
Terminating UE (n71)
4

Originating UE (n71)
2

0
−120 −110 −100 −90 −80 −70 −60 −50
Average RSRP (dBm)
Source: Signals Research Group

25 August 31, 2022 | Signals Ahead, Vol. 18, Number 7


With both VoNR and VoLTE,
Figure 24 and Figure 25 provide the same information for the VoLTE drive test. There is a similar if the radio conditions are
correlation between lower SINR/RSRP and poorer voice quality, with the biggest difference being reasonably good (favorable
that the SINR/RSRP values didn’t drop to the same levels as obtained with 5G, hence the MOS SINR and RSRP) then
values didn’t drop as appreciably. We note that directly comparing SINR and RSRP between LTE the voice quality should
and 5G is not accurate since the parameters are calculated differently and the two technologies be reasonably good.
have different means of dealing with interference. However, the observation still holds that if the
radio conditions are reasonably good (favorable SINR and RSRP) then the voice quality should be
reasonably good.

Figure 24. VoLTE Originating and Terminating UE MOS Versus Average SINR Correlations
MOS
5.0

4.5

4.0

Terminating MOS
3.5

3.0

Originating MOS
2.5

2.0
−20 −15 −10 −5 0 5 10 15 20 25 30
Average SINR (dB)
Source: Signals Research Group

Figure 25. VoLTE Originating and Terminating UE MOS Versus Average RSRP Correlations
MOS
5

Originating UE (VoLTE)
4

Terminating UE (VoLTE)
3

0
− 120 −110 −100 −90 −80 −70 −60 −50
Average RSRP (dBm)
Source: Signals Research Group

26 August 31, 2022 | Signals Ahead, Vol. 18, Number 7


3.3 n71 Versus n41 Drive Test Results
The results in this section stem from two separate drive tests. In one drive test we locked both smart-
phones to Band n71. In the second drive test we locked both smartphones to Band n41. Since we
didn’t want the call to drop due to poor n41 coverage and since we forced the smartphones to operate
in an unnatural manner, we purposefully picked a loop where we felt comfortable the smartphones
would remain attached to the network with either frequency band.
Figure 26 shows the median MOS values and standard deviations for the two tests. Figure 27
provides a geo plot of the MOS values for both smartphones. From these test results it is apparent
that both bands delivered comparable performance with respect to voice quality. This outcome
doesn’t negate the observation that SINR and RSRP played an important role in determining the
voice quality.

Figure 26. n41 and n71 VoNR MOS


4.20 4.16 4.28 4.24

0.67
0.53
0.40 0.30

n41 Originating n41 Terminating n71 Originating n71 Terminating


Median (MOS)
Standard Deviation Source: Signals Research Group

Figure 27. n71 and n41 VoNR MOS

> 4.25 > 4.25


> 4.0 > 4.0
> 3.75 > 3.75
> 3.5 > 3.5
> 3.25 > 3.25
> 3.0 > 3.0
<= 3.0 <= 3.0

Band n71 Band n41


Source: Signals Research Group

27 August 31, 2022 | Signals Ahead, Vol. 18, Number 7


Figure 28 (n71) and Figure 29 (n41) provide time series plots of the VoNR MOS and SINR
values. Once again, we took a running ten second average for the SINR values and offset them by
10 seconds relative to the MOS values. For most of the time during the two tests, the SINR values
between the two smartphones were very similar within that test. Further, most of the lower MOS
values correspond to points in time when the SINR was also low. In Figure 29, we’ve highlighted
a couple of lower MOS values. In one instance there were several handovers occurring and this
action could also explain the lower MOS. In the other instance, the ten second average MOS values
understate really low (< -10 dB) SINR values that occurred for a few seconds. We theorize these very
short and precipitous dips in the SINR could help explain the lower MOS values although we note
there are also several points in the figure where the SINR was just as low as these examples, yet the
MOS values were not impacted nearly as much.

Figure 28. Band n71 VoNR MOS and SINR – Time Series
SINR (dB) MOS
30 5
Originating MOS (n71)

4
20
Terminating MOS (n71)
Avg SINR (n71) 3
10 Originating
SINR (n71) 2

0
1

Terminating SINR (n71)


−10 0
0 100 200 300 400 500 600 700 800 900
Time (sec)
Source: Signals Research Group

Figure 29. Band n41 VoNR MOS and SINR – Time Series
SINR (dB) MOS
30 Terminating SINR (n41) Originating MOS (n41) 5

4
20

Terminating MOS (n41) 3


Avg SINR (n41)
10
Momentarily Very LOW SINR
2

0
1
Originating SINR (n41)
Frequent HOs
−10 0
0 100 200 300 400 500 600 700 800 900 1000
Time (sec)
Source: Signals Research Group

28 August 31, 2022 | Signals Ahead, Vol. 18, Number 7


Figure 30 provides the relationships between the MOS and RSRP for both bands, stemming from
the two separate drive routes. Above RSRP = -105 dBm, the voice quality was generally good, but
with lower signal quality, the voice quality was frequently degraded.

Figure 30. Band n41 and Band n71 VoNR MOS Versus Average RSRP Correlations
MOS
5
Terminating UE (n71)

4
Terminating UE (n41)
Originating UE (n71)
3

Originating UE (n41)
1

0
−120 −110 −100 −90 −80 −70 −60 −50
Average RSRP (dBm)
Source: Signals Research Group

In this section we also examine the network resource requirements of VoNR, how it compares With both smartphones
between the two 5G bands and with VoLTE. Figure 31 shows the PDSCH and PUSCH throughput the downlink throughput
for the originating smartphone on Band n71 as well as the median throughput values for both smart- (PDSCH) on Band n71
was approximately 1.8x
phones during this test. With both smartphones the downlink throughput (PDSCH) due to VoNR
higher than the uplink
voice activity was approximately 1.8x higher than the uplink throughput (PUSCH) due to VoNR throughput (PUSCH).
voice activity. We attribute this outcome to the use of Resource Allocation Type 0.

Figure 31. Band n71 VoNR PDSCH and PUSCH Throughput


Mbps
0.6
0.055
Originating DL (n71)
0.5 0.049

0.4
Originating UL (n71)
0.3 0.029
0.026

0.2

0.1

0
0 50 100 150 200 250 300 350 400 450 500 550 600 650 700 750 800 850 Originating Originating Terminating Terminating
DL (n71) UL (n71) DL (n71) UL (n71)
Time (sec)
Mbps (Median)
Source: Signals Research Group

29 August 31, 2022 | Signals Ahead, Vol. 18, Number 7


Figure 32 provides similar information for Band n41. In this case, the downlink throughput was
just over 2x higher than the uplink throughput. The more interesting comparisons are between
frequency bands and with VoLTE. Figure 33 plots the downlink throughput from the two tests/
frequency bands and Figure 34 shows the uplink throughput from the two tests/frequency bands. It
is very evident in both figures that the VoNR bandwidth requirements with n41 were much higher
than they were with n71.

Figure 32. Band n41 VoNR PDSCH and PUSCH Throughput


Mbps
3.5
0.305
3.0 0.295

2.5 Originating DL (n41)

2.0

1.5 0.144 0.143


Originating UL (n41)
1.0

0.5

0
0 50 100 150 200 250 300 350 400 450 500 550 600 650 700 750 800 850 Originating Originating Terminating Terminating
DL (n41) UL (n41) DL (n41) UL (n41)
Time (sec)
Mbps (Median)
Source: Signals Research Group

Figure 33. Band n41 and Band n71 VoNR PDSCH Throughput – Time Series
Mbps
3.5

3.0
Originating DL (n41)
2.5

2.0

1.5
Originating DL (n71)
1.0

0.5

0
0 50 100 150 200 250 300 350 400 450 500 550 600 650 700 750 800 850

Time (sec)
Source: Signals Research Group
Mbps
0.5

0.4
Originating UL (n41)
Originating UL (n71)
0.3

0.2

0.1
30 August 31, 2022 | Signals Ahead, Vol. 18, Number 7

0
0
0 50 100 150 200 250 300 350 400 450 500 550 600 650 700 750 800 850

Time (sec)
Figure 34. Band n41 and Band n71 VoNR PUSCH Throughput – Time Series
Mbps
0.5

0.4
Originating UL (n41)
Originating UL (n71)
0.3

0.2

0.1

0
0 50 100 150 200 250 300 350 400 450 500 550 600 650 700 750 800 850
Time (sec)
Source: Signals Research Group

Figure 35 shows the downlink and uplink bandwidth requirements for a VoLTE call. These results
stem from a drive test over the same route. With VoLTE, the uplink bandwidth requirements were
higher than the downlink bandwidth requirements (3.5x), and largely because the throughput at the
10th percentile was 5x higher in the uplink direction than in the downlink direction.

Figure 35. VoLTE PDSCH and PUSCH Throughput


Mbps
0.20

VoLTE DL
0.15

0.10
VoLTE UL

0.05

0
0 100 200 300 400 500 600 700 800 900 1000
Time (sec)
Source: Signals Research Group

31 August 31, 2022 | Signals Ahead, Vol. 18, Number 7


Figure 36 and Figure 37 brings everything together. Figure 36 shows the median downlink and The downlink throughput
uplink throughput for VoLTE and VoNR in the two 5G bands. Figure 37 compares the downlink/ with VoNR in Band n41 was
uplink throughput for VoLTE and VoNR Band n71 to VoNR Band n41. It is very apparent in both 77.5x higher than VoLTE
while the uplink throughput
figures that VoNR, especially when using Band n41, required significantly more bandwidth than
with VoNR in Band n41 was
VoLTE. The downlink throughput with VoNR in Band n41 was 77.5x higher than VoLTE while 10x higher than VoLTE.
the uplink throughput with VoNR in Band n41 was 10x higher than VoLTE.

Figure 36. VoLTE, VoNR n71 and VoNR n41 Median Throughput

0.31
0.31

0.14
0.14

0.055
0.055
0.029
0.014 0.029
0.004 0.014
0.004DL
VoLTE n71 VoNR DL n41 VoNR DL VoLTE UL n71 VoNR UL n41 VoNR UL
VoLTE DL n71 VoNR DL n41 VoNR DL Mbps (Median) VoLTE UL n71 VoNR UL n41 VoNR UL
Mbps (Median)
Source: Signals Research Group

Figure 37. VoLTE and VoNR n71 Throughput Relative to n41 Throughput
1.000 1.000
1.000 1.000

0.207
0.177 0.207
0.177 0.100
0.100
0.013
0.013DL
VoLTE n71 VoNR DL n41 VoNR DL VoLTE UL n71 VoNR UL n41 VoNR UL
VoLTE DL n71 VoNR DL n41 VoNR DL VoLTE UL n71 VoNR UL n41 VoNR UL
Relative Throughput
Relative Throughput
Source: Signals Research Group

To demonstrate these results were not an anomaly, we are including two additional figures stem-
ming from another VoNR drive test when the smartphones were not band locked, thus allowing
them to move between the two frequency bands. As shown in Figure 38, this smartphone started
off on Band n41 for the first ~500 seconds of the drive test before switching to Band n71 for the
remainder of the test. When the smartphone changed bands, there was a noticeable drop in the
downlink and uplink throughput. Figure 39 shows the average bandwidth requirements for VoNR
in the two frequency bands during this test and Figure 40 shows the implied spectral efficiencies.
For these calculations, we grossed up the average throughput values in Figure 39 by assuming the
full allocation of resource blocks (RBs). For Band n41, we assumed an 78/22 split in the downlink

32 August 31, 2022 | Signals Ahead, Vol. 18, Number 7


and uplink time slot allocations. One could rightly note VoNR was more spectral efficient in n41, but
the point remains that VoNR in Band n41 required far more bandwidth than VoNR in Band n71.

Figure 38. n71 and n41 Downlink and Uplink Throughput – Time Series
Time Freq (MHz)
2.0 3000
DL Tput
2500
5.4
1.5
2000

1.0 1500

2.2
1.9 1000
0.5 1.3 Frequency
500

0 n71 Tput
UL n41 n71 n41 0
0 100 Downlink
200 300 400 500 Uplink 600 700 800 900 1000
bps/Hz Mbps
Source: Signals Research Group

Figure 39. n71 and n41 Downlink and Uplink Throughput – Average
5.4
0.48

2.2
1.9
0.19
1.3

0.06
0.02
n71 n41 n71 n41
n71 n41 n71 n41
Downlink Uplink
Downlink Uplink
bps/Hz
Mbps (Average)
Source: Signals Research Group

Figure 40. n71 and n41 Downlink and Uplink Spectral Efficiency
0.48

0.19

0.06
0.02
n71 n41 n71 n41
Downlink Uplink
Mbps (Average)
Source: Signals Research Group

33 August 31, 2022 | Signals Ahead, Vol. 18, Number 7


Since the comparisons between n71 and n41 involve different channel bandwidths, not to mention
different duplex schemes, we wanted to come up with a comparison that normalized the two sets of
results. To normalize the data, we took the average throughput for VoNR (downlink and uplink)
and divided it by the corresponding RB normalized throughput (downlink and uplink) for each
band. The simplistic way of looking at it is that this analysis shows the maximum number of voice
calls possible, based entirely on RB availability without any consideration for other limiting factors
(signaling, latency, etc.). For example, the downlink bandwidth requirement of VoNR in Band n71
was 0.06 Mbps. With RB normalization, which assumes all RBs were assigned with the same spectral
efficiency, the RB normalized throughput would be 32.9 Mbps for a spectral efficiency of 2.2 bps/
Hz (32.9 Mbps/15 MHz = 2.2 bps/Hz). If we divide the bandwidth requirements of a single VoNR
call (0.06 Mbps) by the maximum throughput possible assuming the same spectral efficiency (32.9
Mbps) then we obtain a relative bandwidth consumption of 0.19% as shown in Figure 41. With this
metric a lower percentage is better because it indicates the VoNR call required less bandwidth relative
to what was available in the band. This means, for example, that although VoNR in Band n41 was
more spectral efficient in the uplink than in Band n71, VoNR in Band n41 still had a greater impact
on available uplink network resources in Band n41 than VoNR had on the uplink in Band n71.

Figure 41. n71 and n41 Relative Bandwidth Consumption

0.46%

0.19%

0.13%
0.11%

n71 n41 n71 n41


Downlink Uplink
Relative Bandwidth Consumption
Source: Signals Research Group

34 August 31, 2022 | Signals Ahead, Vol. 18, Number 7


3.4 VoNR Freewheeling Test with Background Data Traffic
The results in this section are based on a lengthy drive test in which we established a VoNR call
between the two S21 smartphones which were not 5G band locked. Additionally, we ran a full
buffer HTTP downlink session in the background on the “originating” smartphone and a full buffer
HTTP uplink session in the background on the “terminating” smartphone. Figure 42 shows the
5G band usage of the two smartphones, which was largely comparable over the drive test. Figure 43
illustrates the voice quality for the two smartphones. To our eye, the results look very comparable
between the two phones, which isn’t surprising since they were frequently using the same frequency
band/serving cell during the drive.

Figure 42. Originating and Terminating UE 5G Band Selection

n41 n41
n71 n71

Originating Terminating
Source: Signals Research Group

Figure 43. Originating and Terminating UE Voice Quality

> 4.25 > 4.25


> 4.0 > 4.0
> 3.75 > 3.75
> 3.5 > 3.5
> 3.25 > 3.25
> 3.0 > 3.0
<= 3.0 <= 3.0

Originating Terminating
Source: Signals Research Group

35 August 31, 2022 | Signals Ahead, Vol. 18, Number 7


Figure 44 shows the median voice quality and standard deviation for the two smartphones.
Although the median MOS values were quite favorable, the standard deviation was also high, indi-
cating a lot of variability in the results.

Figure 44. Freewheeling MOS


4.02 3.98

0.67 0.71

Originating Terminating
Median (MOS)

Standard Deviation
Source: Signals Research Group

One significant advantage of VoNR is that any background data traffic remains on 5G versus One significant advantage of
falling back to LTE. To this point, it is also more advantageous if the VoNR call used Band n41 VoNR is that any background
than Band n71 due to the wider channel bandwidth/higher data speeds possible in the mid-band data traffic remains on 5G
versus falling back to LTE.
spectrum. Figure 45 shows the PDSCH throughput during the test as well as frequency band being
used by the originating smartphone. The figure also includes the average throughput for the two
bands. Figure 46 provides similar information for the terminating smartphone, although in this
case the focus is on the uplink throughput since this smartphone was concurrently doing a full
buffer HTTP uplink data transfer along with the VoNR call. The figure also includes results for the
downlink to emphasize the high amount of downlink data traffic that was getting generated with
the uplink data transfer. We emphasize that the Umetrix test profile we used for this smartphone
only generated HTTP data traffic in the uplink direction. We also point out the smartphones were

Figure 45. Originating UE PDSCH Throughput


Mbps 5G Band
1400
572.6
Total Downlink Tput
1200

1000 5G Band
n41
800

600
n71 188.5
400

200

0
0 200 400 600 800 1000 1200 1400 1600 1800 2000 2200 2400 P Cell = n41 P Cell = n71
Time (sec) Average (Mbps)
Source: Signals Research Group

36 August 31, 2022 | Signals Ahead, Vol. 18, Number 7


Figure 46. Terminating UE PUSCH Throughput
Mbps 5G Band
100

Uplink Throughput
80
5G Band
n41
60

40 Total Downlink Tput


n71

20

0
0 200 400 600 800 1000 1200 1400 1600 1800 2000 2200 2400
Time (sec) Source: Signals Research Group

using carrier aggregation in the downlink direction so the throughput shown in the figure includes
30.1
both component carriers.
Finally, the last two figures show the relationships between voice quality and MOS. Although If the SINR could remain
the two smartphones changed frequency bands periodically during this test, they generally made above 0 dB, then the VoNR
the switches at the same time, so the SINR reported by the two smartphones was very similar voice quality would remain
very favorable with far less
throughout the test. Once again, many of the lower MOS values occurred with low SINR. Figure
variability in the MOS values.
48 shows the correlation of these two metrics. Based on these
12.5results, one could conclude that if the

8.4
Figure 47. VoNR Originating and Terminating UE MOS and Average SINR – Time Series
SINR (dB) MOS
30 1.3 5
VoNR Terminating
VoNR Originating
Uplink Downlink Uplink Downlink

P Cell = n41 P Cell = n71 4


20
Average (Mbps)
Originating
3
SINR
10
2

0
1

Terminating SINR
−10 0
0 200 400 600 800 1000 1200 1400 1600 1800 2000 2200 2400
Time (sec) Source: Signals Research Group

SINR could remain above 0 dB then the VoNR voice quality would remain very favorable with far
less variability in the MOS values.

37 August 31, 2022 | Signals Ahead, Vol. 18, Number 7


Figure 48. VoNR Originating and Terminating UE MOS Versus Average SINR Correlations
MOS
5

Terminating MOS
4

Originating MOS
2

0
−15 −10 −5 0 5 10 15 20 25 30

Average SINR (dB) Source: Signals Research Group

38 August 31, 2022 | Signals Ahead, Vol. 18, Number 7


3.5 VoLTE and VoNR Current Analysis and Call Setup Times
In this section we provide results from measuring the current requirements of VoLTE and VoNR
(Section 3.5.1) and the call setup times (Section 3.5.2).

3.5.1 Current Analysis


For the current analysis we placed a voice call (VoNR or VoLTE) between the two smartphones
without Umetrix Data or XCAL-M since we didn’t want to introduce extraneous factors which
would negate the validity of the tests. For example, a USB cable attached from a phone to a PC
would introduce a trickle charge that would make it impractical to measure the current drain. We
did, however, need to take certain steps to expose and record the real-time current drain for post
analysis. We took a similar approach to what we have several done in the past. We used a third-
party application called AccuBattery (downloaded from Google Play) to display the current drain
– reported once every two seconds. We then used a display capture application called AZ Screen
Recorder to record the phone’s display and hence the real-time current measurements. Lastly, during
the analysis phase we played the video back on our PC in slow motion so that we could write down
the current measurements. It’s a pain in the butt, but it works.
We recognize the screen capture application had some impact on the current drain, but it occurred
throughout the test (with and without the voice call). Further, in previous studies we’ve calculated
the impact this application has on the current drain, and we’ve concluded its impact is insignificant
relative to the current drain with most applications. If anything, the biggest limitation of this test
is that the backlight needs to be on for the screen capture application to see the real-time current
measurement reports. Once again, we can also account for the display with reasonable accuracy.
Worth noting, we realized during this testing that it is crucial to have the adaptive brightness feature
disabled since this feature introduces all sorts of spurious effects, including huge spikes in the current
when the phone is adjusting the display’s brightness, even modest adjustments that are not evident
to the eye.
Figure 49 shows the current drain due to a VoLTE call. In this test, we disabled 5G on the smart- VoLTE increased the
phone since even though the call was going over an LTE radio bearer the 5G radio could still have current consumption
been making measurement reports which impacts the total current usage. Comparing the results by approximately 130
mA with 5G disabled
between 4G idle and 4G VoLTE, we can conclude VoLTE increased the current consumption by
in the smartphone.
approximately 130 mA.

Figure 49. VoLTE Current Drain with 5G Disabled


mA
1200 330.5

1000

Galaxy S20 4G VoLTE


800
198.0
600

400 Galaxy S20 4G Idle

200

0
0 20 40 60 80 100 120 140 160 Galaxy S20 4G Idle Galaxy S20 4G VoLTE
Time (sec) mA (Median)
Source: Signals Research Group

39 August 31, 2022 | Signals Ahead, Vol. 18, Number 7


In the second test, we once again did a VoLTE call, but we kept the 5G radio turned on. This
scenario is more representative of what consumers would encounter today since very few people disable
5G and 5G low-band coverage is fairly ubiquitous on the T-Mobile network. With this scenario, the
current drain increased by approximately 80 mA, or 90 mA compared with airplane mode.

Figure 50. VoLTE Current Drain with 5G Enabled


mA
1500
286.5

1000 209.0
197.5

Galaxy S20 5G UC Idle


Galaxy S20 5G UC VoNR Galaxy S20
Airplane
500

0
0 20 40 60 80 100 120 140 160 Galaxy S20 Galaxy S20 Galaxy S20
Time (sec) 5G Idle 5G VoLTE Airplane
Source: Signals Research Group
mA (Median)

Finally, Figure 51 shows the current drain with a VoNR call. During this test the smartphone With VoNR, the current
was almost certainly using Band n41 during this call, or at least the phone’s display was showing drain increased by nearly 200
“5G UC.” With VoNR, the current drain increased by nearly 200 mA compared with 5G idle. Put mA compared with 5G idle,
although the use of a 100
another way, the VoNR current drain was slightly more than 2x higher than the VoLTE current
MHz channel likely played
drain. According to T-Mobile, the higher current drain was entirely due to the smartphones using a critical role in the higher
a wider channel bandwidth for the call. If we had tested VoNR on Band n71 by band locking the current measurement.
phone then the operator believes from some of its testing that the current drain would have been
comparable to VoLTE.

Figure 51. VoNR Current Drain with 5G Enabled


nA
3000
396.0
2500

2000

1500
199.5
Galaxy S21 5G UC VoNR
1000

Galaxy S21 5G UC Idle


500

0
0 20 40 60 80 100 120 140 160 180 200 220 Galaxy S21 Galaxy S21
5G UC Idle 5G UC VoNR
Time (sec) Source: Signals Research Group
NA (Median)

40 August 31, 2022 | Signals Ahead, Vol. 18, Number 7


3.5.2 Call Setup Times
We did a lengthy drive test with Umetrix Voice to evaluate the call setup times for the two voice The VoNR and VoLTE
technologies. Based on the results of this study, the call setup times were very similar between the call setup times were
two technologies, as shown in Figure 52. This outcome is consistent with T-Mobile’s expectations largely comparable.
since both voice technologies used the same IMS core network. When we tested call setup times
in Las Vegas, we observed VoNR had a much faster call setup time but that could have been due
to other factors since we were testing on two different operator networks, only one network was
supporting a large amount of commercial voice and data traffic, and the two IMS/core networks may
have been located at different locations/distances relative to the radio access network.

Figure 52. VoLTE and VoNR Call Setup Times


6.55 6.55
6.32 6.32

VoLTE Originating VoLTE Terminating VoNR Originating VoNR Terminating

Setup Time (sec)


Source: Signals Research Group

Finally, Figure 53 shows the downlink and uplink signaling messages that occurred when estab-
lishing a VoNR call.

Figure 53. VoNR Call Signaling Messages

Source: Signals Research Group

41 August 31, 2022 | Signals Ahead, Vol. 18, Number 7


4.0 Test Methodology
SRG collaborated with Accuver Americas and Spirent Communications to conduct this study. SRG SRG collaborated with
conducted all the tests, and we take full responsibility for the analysis and commentary provided Accuver Americas and
in this report. Spirent Communications provided its Umetrix Data platform to generate the data Spirent Communications
to conduct this study.
transfers, using one of the company’s high-bandwidth servers located in California, and we used
the company’s Umetrix Voice test platform for the voice quality measurements. Accuver Americas
provided its XCAL-M drive test solution to capture 5G NR and LTE network parameters, and we
used the company’s XCAP post-processing software to analyze the data. We’ve worked with both
companies for more than a decade and the use of their suite of tools has been invaluable to our 5G
benchmark studies.
We did the testing over a period of two days in mid-July. For this study, we used two Galaxy S21
smartphones for the VoNR calls and we used two Galaxy S20 smartphones for the VoNR calls.
Depending on the test, we allowed the phones to select their preferred 5G band (n71 or n41) or we
locked the phones to a specific 5G band. For the VoLTE testing, we always allowed the smartphones
to select their preferred LTE band.
We used the Umetrix data platform to generate full buffer downlink and uplink data transfers
using either HTTP or UDP. For purposes of the VoNR drive tests, we used 2-minute HTTP
downlink or uplink tests that continuously repeated so that we could do lengthy drive tests. For the
stationary tests, we used short data transfer tests, including HTTP downlink, HTTP uplink and
HTTP downlink/uplink in which the data transfers occurred simultaneously in both directions.
We did a mix of stationary and drive tests as part of this study. The drive routes we used were
entirely random with our only criteria being that we wanted to remain within the test cluster. Most
of the cell sites supported n71 and n41 although a few sites only supported n71.
For our voice tests, we used the Umetrix Voice test platform to analyze the voice quality, call setup
time, and the call completion rate. Umetrix Voice can also automatically originate and terminate
voice calls between the two phones on each network and we used this feature to automate the testing
process while driving around the city. For the voice quality measurements, we used a super wideband
(SWB) audio file, and we used a 16 kHz sampling rate when analyzing the voice quality with the
POLQA algorithm. We connected the smartphones to Umetrix Voice via Bluetooth since most
consumers use Bluetooth when talking on their phones. Further, by using Bluetooth we were able
to avoid potential grounding issues associated with using audio cables. Such issues can sometimes
generate background hum which influences the MOS measurements.
For the data analysis, we merged the XCAL-M log files from both phones to do comparative
analysis. In terms of our data analysis, we looked at PDSCH/PUSCH throughput, resource block
allocations, RB normalized throughput, spectral efficiency, MCS assignments, modulation schemes,
MIMO rank, and PUSCH transmit power, to name a few. We also merged in Umetrix Voice data
that was also geo tagged with latitude/longitude information and GPS time stamps. For compara-
tive purposes, we time shifted the Umetrix Voice data by ten seconds to more closely align the voice
quality measurement (MOS) with the radio conditions occurring at the time.
For the current analysis we placed a voice call (VoNR or VoLTE) between the two smartphones
without Umetrix Data or XCAL-M since we didn’t want to introduce extraneous factors which
would negate the validity of the tests. For example, a USB cable attached from a phone to a PC
would introduce a trickle charge that would make it impractical to measure the current drain. We
did, however, need to take certain steps to expose and record the real-time current drain for post
analysis. We took a similar approach to what we have several done in the past. We used a third-
party application called AccuBattery (downloaded from Google Play) to display the current drain
– reported once every two seconds. We then used a display capture application called AZ Screen
Recorder to record the phone’s display and hence the real-time current measurements. Lastly, during
the analysis phase we played the video back on our PC in slow motion so that we could write down

42 August 31, 2022 | Signals Ahead, Vol. 18, Number 7


the current measurements. As we’ve proven in previous tests, the impact of the screen recorder was
very minor compared with the current requirements of the display light and the voice/data activity
occurring on the phone.
Figure 54 shows a screen shot of the XCAL-M GUI. The two images show an instance during
a VoNR drive test when one smartphone was using Band n71 and the other smartphone was using
Band n41. Figure 56 shows a picture of Umetrix Voice.

Figure 54. XCAL-M

VoNR on Band n71 VoNR on Band n41


Source: Accuver Americas

Figure 55. Umetrix Data Architecture

Source: Spirent Communications

43 August 31, 2022 | Signals Ahead, Vol. 18, Number 7


Figure 56. Umetrix Voice

Source: Spirent Communications

44 August 31, 2022 | Signals Ahead, Vol. 18, Number 7


5.0 Final Thoughts
We have lots of fun and interesting stuff in store for the remainder of the year. Until next time, be
on the lookout for the next Signals Ahead….

Michael Thelander
Michael Thelander is the President and Founder of Signals Research Group (SRG), a US-based
research consultancy that offers thought-leading field research and consulting services on the
wireless telecommunications industry.
Its flagship research product is a research product entitled Signals Ahead, which has attracted
a strong following across the entire wireless ecosystem with corporate subscribers on five conti-
nents. SRG’s Signals Ahead research product and its consulting services are technology-focused
with a strong emphasis on next-generation networks and performance benchmarking.
In his current endeavor, Mr. Thelander is the lead analyst for Signals Ahead and he guides a
team of industry experts that provide consulting services for the wireless industry, including
some of the largest mobile operators, the top equipment OEMs, trade associations, and financial
institutions. He has also served as a member of an industry advisory board for one of the world’s
largest wireless infrastructure suppliers.
Mr. Thelander earned a Masters of Science in Solid State Physics from North Carolina
State University and a Masters of Business Administration from the University of Chicago,
Graduate School of Business.

Emil Olbrich
Emil Olbrich is currently VP of Networks with Signals Research Group. Prior to this he was
head of LTE research, development, test and evaluation for the Public Safety Communications
Research Program where he deployed the first and most diverse Public Safety 700 MHz
LTE test lab in the world with over 70 participating vendors and commercial carriers. He
was responsible for the specifying, deploying and maintaining the entire ecosystem of LTE
which included devices, air interface, transport, radio access network, evolved packet core,
IP networking, IMS core and application servers. He also led the team efforts, which include
standards work, test case development and test case execution.
Mr. Olbrich has over 20 years of experience in the field of wireless telecommunications. He
has worked primarily in R&D at some of the largest telecommunication companies in the world
- such as Motorola, Qualcomm and Ericsson. His scope of work includes deploying and oper-
ating LTE infrastructure (RAN, EPC and IMS) from numerous Tier 1 vendors; testing new
LTE mobile devices from multiple suppliers; testing, deploying and operating some of the first
commercial CDMA networks; serving as Lead Project engineer for the 2002 Salt Lake City
Winter Olympics and as the Project Manager for the China Ministry of Information Industry 3G
testing in China; and supporting the early development of HDR (EV-DO and EV-DO Rev A).
He has been a speaker at events such as the GSMA Mobile World Congress, LTE North
America, 4G World, International Wireless Communications Expo and LTE World Summit.
Mr. Olbrich has a B.S. degree in Electrical Engineering Technology from Southern Illinois
University.

45 August 31, 2022 | Signals Ahead, Vol. 18, Number 7


ON THE HORIZON: POTENTIAL SIGNALS AHEAD/SIGNALS FLASH! TOPICS
We have identified a list of pending research topics that we are currently considering or presently working on
completing. The topics at the top of the list are definitive with many of them already in the works. The topics toward
the bottom of the page are a bit more speculative. Obviously, this list is subject to change based on various factors and
market trends. As always, we welcome suggestions from our readers.

5G Standardization
➤ 5G from a 3GPP Perspective (ongoing series of reports – published quarterly or as warranted)

Thematic Reports
➤ Mobile Edge Computing and the impact of data caching at the cell edge

Benchmark Studies
➤ 5G NR mmWave Fixed Wireless Access with IAB

➤ Over-the-Air 5G NR smartphone performance benchmark study (FR1)

➤ SRS versus codebook beamforming benchmark study

➤ Over-the-Air 5G NR smartphone performance benchmark study (FR2)

➤ Multi-operator benchmark study based on ETSI 103.559

➤ Mobile Edge Computing

➤ Open RAN network performance benchmark study 1 – RF performance

➤ Open RAN network performance benchmark study 2 – User Experience

➤ Open RAN network performance benchmark study 3 – Scheduling Efficiency

➤ FR1 + FR2 EN-DC network performance benchmark study

➤ MU-MIMO benchmark study (FR1)

➤ MU-MIMO benchmark study (FR2)

46 August 31, 2022 | Signals Ahead, Vol. 18, Number 7


Signals Ahead Subscription
The Signals Ahead newsletter is available on a subscription basis. We offer three distinct packages that have been
tailored to address the needs of our corporate users. The Group License includes up to five users from the same company.
The Global License is the most attractive package for companies that have several readers since it is offered to an unlimited
number of employees from the same organization. Finally, the Platinum package includes the Global License, plus up to five
hours of analyst time. Other packages are available.

Corporate Rates (12 issues)


❒ Group License ($3,995)
❒ Global License (Price Available upon Request)
❒ Platinum (Price Available upon Request)
❒ Gold Pass (Price Available upon Request)

Payment Terms
❒ American Express ❒ Vias ❒ MasterCard Credit Card # Exp Date / /
❒ Check Check Number
❒ Purchase Order PO Number

Name: Title:
Affiliation: Phone: ( )
Mailing Address:

Mailing Address
Signals Research Group – ATTN: Sales
5300 Painter Creek Green
Independence, MN 55359

Alternatively, you may contact us at (510) 273-2439 or at information@signalsresearch.com and we will contact you for your
billing information.

Terms and Conditions: Any copying, redistributing, or republishing of this material, including unauthorized
sharing of user accounts, is strictly prohibited without the written consent of SRG.

please note disclaimer: The views expressed in this newsletter reflect those of Signals Research Group and are based on our understanding of past and current events shaping the wireless industry.
This report is provided for informational purposes only and on the condition that it will not form a basis for any investment decision. The information has been obtained from sources believed to be
reliable, but Signals Research Group makes no representation as to the accuracy or completeness of such information. Opinions, estimates, projections or forecasts in this report constitute the current
judgment of the author(s) as of the date of this report. Signals Research Group has no obligation to update, modify or amend this report or to otherwise notify a reader thereof in the event that any matter
stated herein, or any opinion, projection, forecast or estimate set forth herein, changes or subsequently becomes inaccurate.
If you feel our opinions, analysis or interpretations of events are inaccurate, please fell free to contact Signals Research Group. We are always seeking a more accurate understanding of the topics
that influence the wireless industry. Reference in the newsletter to a company that is publicly traded is not a recommendation to buy or sell the shares of such company. Signals Research Group and/or
its affiliates/investors may hold securities positions in the companies discussed in this report and may frequently trade in such positions. Such investment activity may be inconsistent with the analysis
provided in this report. Signals Research Group seeks to do business and may currently be doing business with companies discussed in this report. Readers should be aware that Signals Research Group
might have a conflict of interest that could affect the objectivity of this report. Additional information and disclosures can be found at our website at www.signalsresearch.com. This report may not be
47
reproduced, copied, distributed or published without the prior written authorization of Signals Research Group (copyright ©2022, August 31,reserved
all rights 2022 by | Signals Ahead,
Signals Research Vol. 18, Number 7
Group).

You might also like