Rootmetrics found surprisingly little difference between indoor and outdoor performance in mid-band 5G at 3.5 GHz.mmWave gets clobbered by many kinds of walls and windows. Apparently, that's much less of a problem at 3.5 GHz. A team conducted hundreds of tests in Korea, the best independent study yet. The data on 5G downloads:

 

  • Indoor speed (out of 30 locations): 204 Mbps to 401 Mbps (median speeds)

  • Outdoor speed: 160 Mbps to 457 Mbps

  • Drive speed: 163 Mbps to 431 Mbps

Rootmetrics has done the best survey yet, but like all studies there are limits.

To begin with, wireless performance varies enormously, so a huge sample is required. I was particularly interested in the indoor results, which have not previously been reported. The result through multiple walls in an office building will be different than next to the window of a McDonalds. Suzanth generously explained their methodology and how they strive for accuracy. I was convinced the work is good. 

We've all seen reports of 700-1000 megabit results. Midband 5G with 100 MHz of spectrum puts out as much as 2000 megabits shared. 1000 Mbps is definitely possible in the right position when few others are on the network. That's why it's so important to get average speeds.

The 204 Mbps to 401 Mbps number originally confused me. Rootmetrics has several data sets because they are trying to compare the performance of the phones and the networks. One cohort had a 204 Mbps median and another a 401 Mbps median. 

The median latency was 32 ms, although 5% of the tests found 19 ms or below, That's consistent with data from around the world, usually within 25-35 ms. 

Open Signal, working with data from thousands of phones running speed tests, found a median 5G speed of <200 Mbps. (The discrepancy is why I looked so closely.) The difference could be that people with poor performance were more likely to run speed tests. I'll be looking closely.

Since the variation is so wide, I've been saying typical speeds of mid-band are 100-400 Mbps, If I see more data like this, I may be accurate saying 100-500 Mbps

  

 

dave ask

Newsfeed

CFO John Stephens says AT&T is going to cut capex soon.

Bharti in India has lost 45M customers who did not want to pay the minimum US#2/month. It's shutting down 3G to free some spectrum for 4G. It is cutting capex, dangerous when the 12 gigabytes/month of use continues to rise.

Huawei in 6 days sold 300,000 5G Mate 20s. Delivery begins on 8/16. 

China has over 50,000 upgraded base stations and may have more than 200,000 by yearend 2019. The growth is astonishing and about to accelerate. China will have more 5G than North America and Europe combined for several years.

5G phone prices are down to $580 in China from Oppo. Headed under $300 in 2020 and driving demand.

No one believed me when I wrote in May, 90% of Huawei U.S. purchases can be rapidly replaced and that Huawei would survive and thrive. Financial results are in, with 23% growth and increased phone sales. It is spending $17B on research in 2019, up > 10%. 

5G phones spotted from Sharp and Sony

NTT DOCOMO will begin "pre-commercial service Sept 20 with over 100 live bases. Officially, the commercial start is 2020.

 More newsfeed

----------

Welcome  1,800,000 Koreans bought 5G in the first four months. The demand is there, and most of the technology works. Meanwhile, the hype is unreal. Time for reporting closer to the truth.

The estimates you hear about 5G costs are wildly exaggerated. Verizon is building the most advanced wireless network while reducing capex. Deutsche Telekom and Orange/France Telecom also confirm they won't raise capex.

Massive MIMO in either 4G or "5G" can increase capacity 3X to 7X, including putting 2.3 GHz to 4.2 GHz to use. Carrier Aggregation, 256 QAM, and other tools double and triple that. Verizon sees cost/bit dropping 40% per year.

Cisco & others see traffic growth slowing to 30%/year or less.  I infer overcapacity almost everywhere.  

Believe it or not, 80+% of 5G (mid-band) for several years will be slower than good 4G, which is more developed.