Adding more speed to our home office / classroom networks

The last time we wrote one of these updates about boosting the performance of our home office networks we promised not to write again until we had made some substantial gains. Well, we did. Here’s why we have been working on that problem, and what we have done.

At the beginning of the COVID shutdown we suspected that it would last a lot longer than two weeks. Unfortunately we were right about that, and we were also correct in guessing that we would need to shift from our offices housing multiple people to multiple offices housing one person each. Our plan was to upgrade with particular emphasis on video conferencing, a service we had not employed to any great extent before COVID.

Early on, we picked the low hanging fruit. We have one house served by a telco fiber system. It streams at around 500Mbps down and 440 to 470Mbps up. The only problem with this location is the plaster walls. They seem to have far greater density than sheetrock / wallboard, at least for 5 GHz radio signals, so we were losing up to 60 to 70 per cent of our WiFi speed at one or two walls away distance. That problem was solved by running ethernet cable between rooms ( click here to read this story. ). Our second house project got a similar fix, with ethernet cable through conduit stuck on baseboards - (it looks a lot better than it sounds). We learned the value of powerline adapters and mesh WiFi in yet a third house. ( click here to read this story. ). Both of these houses have cable company - Comcast / Xfinity - internet service, the problem in the cable internet houses is relatively slow upload speeds ( click here to read our thoughts about telco vs. cable internet providers. ). We also did a lot of research and testing on how much speed is really required ( click here for that report. ). Then we got an in-your-face demonstration of speed requirements during a Zoom event.

Last week we had a Zoom courtroom hearing - as all Tennessee hearings are now conducted on Zoom or Webex. During this hearing, our client suddenly disappeared. After a quick phone call we discovered that a virtual student in our client’s house had logged on to attend class - and poof - the virtual courtroom was kicked off by virtual learning. Their WiFi / cable connection did not have enough upload speed to support two simultaneous HD appearances / conferences. Finally all our fooling around with WiFi and networking paid off and we were able to quickly give some advice to bandaid their system together for the duration of the hearing. We got the judicial decision we planned, and noticed that some judges have become surprisingly tolerant about plaintiffs or defendants suddenly disappearing from their Zoom courtrooms. But that’s a separate story, we lured you here to learn about our new and different network trick.

Our news is that we patched together two different types of systems - wireless and wired. In the jargon of networking, we backhauled a mesh system. We described the three types of wired connections available for home networks in an earlier chapter ( click here for that report. ). In the house with the new patch we have two levels / stories, cable internet service, and we have the potential of four simultaneous video conference users. Given our earlier articles you can probably guess we picked a powerline adapter to backhaul, or in effect carry the upload signals, to a first generation Google mesh WiFi 5 system, which could effectively devote all its efforts to downloading speed, or fronthauling, if you will.

Our results on the top floor were way, way better than we had hoped for. With three mesh units backhauled by three TP-Link AV2 spec AV2000 units, our iPhone XR running SpeedTest, we got 95% of the modem's output speed in the room directly above the modem / main mesh router. The other mesh puck upstairs gave us 90% of modem output speed - this is the same room where we had - zero - WiFi reach when we started this project. We placed that puck near a window, and it gave us 86% of modem speed in the middle of the backyard. Curiously a downstairs room on the other side of the house only gave us 83% of modem speed. We suspect some sheet metal HVAC ducts in the walls block this signal, but it could be the noise level generated by competing signals from neighboring houses as well.

To get this result, we stayed with WiFi 5 spec. We bought used Google mesh pucks on an eBay auction. Three packs of pucks were going for about $115 to $120 in August. The TP-Link AV2 wall wart boxes were $89 new and we found our last two at $59 as manufacturer refurbs. Our testing results encouraged us to buy one more used puck and those two extra refurb wall warts. After a few decades working at a producer of cable TV, powerline and radio devices - at an actual US electronics factory (which are now exceedingly scarce) - our network ‘guru’ says that any company that sells through mass merchants will generate lots of returns, mostly returned in damaged packaging, but mostly good products. The merchant has no incentive to ‘fix’ problems or argue with a customer when he can return the device to the maker for full credit. Most of these devices had no problem other than an end user who didn’t really want to read the owners manual. So buying refurbs is generally a good risk/return proposition if the price is right.

Normally we would have gone to a Cisco or Netgear mesh router system, both good, trusted brands. A few things steered us off these two names. One is the crazy pricing because of the COVID supply chain disruption ( we posted a price chart of one mesh router here ) and the effect returned after a July / August lull / drop as back to virtual school re-stokes demand. Two is the fact that we do not expect to have economical WiFi 6 client units available until 12 to 24 months have passed, and so can wait for WiFi 6 routers until that time. Three is the fact that these two brands have started selling what were once included hardware / firmware settings as subscription services. No thanks, we’ll pass on that monthly parasite plan for now. When we select our WiFi 6 system, we will also have a bias against this new router service subscription model. Fourth is our observation of the used market pricing for these devices on eBay. We think the residual value of the devices we’ve purchased will make them fairly inexpensive rentals for 12 to 24 months.

You might wisely ask if there is anything we might have done to trick ourselves here. We say yes, there is. We started running these tests saying we’d use no test equipment, we’ll do things in a way that most anyone can duplicate. We only used the SpeedTest app, and a free Android app called WiFi Analyzer to pick the least used 2.4GHz channels — (less ambient noise on a channel and constant signal levels = better signal to noise ratios = better speed / performance). You should be aware that the results you get are also very dependent on the phone / tablet / laptop you use to walk around your house to measure speed - it’s easy to trick yourself. We’re doing a trial and error kind of science, but anyone facing the myriad of variables introduced with radio signals faces much the same problem, even with good test equipment, it’s far less than completely predictable. Our slow processor WiFi 4 radio Chromebooks struggled to do half the speed of our iPhone XR at the same locations. We’d be really depressed if we had only tested with these machines. The Chromebooks improved markedly when we plugged in USB WiFi 5 radios, but still did not come close to equalling the iPhones. Your results will vary in the same way, on both the remote and the router processor side of things. Also, wall composition, floor density, insulation types, the neighbor’s WiFi signal, even the iron content of the soil plus one hundred and one other different things in or around any house are likely to vary - and cause speed variance. And we must admit with some embarrassment that a few times we have read SpeedTest results in the Mega Byte, not the Mega Bit setting. (8 bits to a byte - a bit is 0 or 1 - a byte is 8 bits, the most info an early processor could grab off in one clock cycle.)

One more thing. We found a refurb TP-Link AX1500 router with WiFi 6 for $55.00 delivered - a manufacturer refurb unit. The Ubiquiti router unit we were considering as a bridge between the all-WiFi Google mesh system and a separate ethernet system is about the same price. The Ubiquiti offers PoE - Power over Ethernet - injection to power a remote WiFi point or security camera via easy-to-install, non - dangerous, ethernet cable to distant out building locations. The possibility of experimenting with WiFi was more appealing than PoE, and as you have probably guessed, some aspects of this project have become more avocational tinkering than a strictly business-like endeavor. We can use the DMZ setting on the AX1500 to feed the number one mesh puck. The puck thinks it’s looking directly at the internet, as DMZ is a straight pass-through. So neither of the two tries to control the other, they are two separate but equal routers, nether stepping on the other’s toes by double NAT.

So how much have we dumped into all these devices (toys) at this location? More than we originally anticipated. This last upgrade at this one location was about $450 in total. We have four mesh pucks, four powerline backhaul wall warts, four ports of conventional 1 Gigabit ethernet from the AX1500 router, plus a WiFi 6 box for experimentation. A new 2nd generation Google mesh three unit WiFi 5 system is now $468 on Amazon, a new Netgear 3 pack Orbi WiFi 5 system is $699 and a WiFi 6 Orbi 3 pack is currently $999 on Amazon. Will these pricey devices go substantially faster? No, not really. On the Comcast systems we are getting WiFi speeds that are only a 5 to 17% degradation from the modem’s nominal advertised speed. The actual direct ethernet connection speed at the Comcast modem / router is sometimes 15% faster than the nominal speed we’re paying for, e.g. if you buy the 200Mbps from Comcast, at some times of day you will see 230Mbps down. But if we enjoy an average of 90% of the modem speed to most parts of the house with our crazy used mesh / refurb powerline backhaul contraption , we see no compelling reason to pour 200 to 500 dollars more into a three location Orbi system. Up to twice the price for a return of a possible 10% bump up in speed simply does not excite us.

With taxes and fees all in our new ARRIS modem will pay for itself in ~9 months vs. monthly modem/gateway rental, so we exclude it from our cost calculations. Our various dongles serve as backups and enhancements, and also keep us from running out to buy new tablets and laptops with higher speed WiFi when our current machines have a lot of useful life in them. We will be tempted to buy WiFi 6 dongles when they become available, but are not tempted to spend any more WiFi 5 devices. Our thought now is that we should upgrade the speed level on at least one of the Comcast houses. Their 600 down and 15 up or their 1 Gig down and 35 up level might do, or instead adding the AT&T 50 DSL service, with its 50 down and 10 up, which is available at this house. We could patch this together with our current Comcast service using a $50 switch on fail / load balancing ethernet router which might give us reasonable margins of safety and backup....but..that's another $50/mo. service plus tax assumming we BYO modem. Buying into DSL now is about as exciting as buying tubes for tires..it's a dying tech. In a pinch now, we can use our the iPhone XR hotspot from Verizon. We measure 120Mbps down and 55Mbps up at one house - or easily four Zoom sessions - if all else fails. The only problem is how to distribute that speed throughout the house, and we expect Verizon's LTE will soon re-institute data caps or overage charges. We recall one student regularly bumping our plan into overage data charges with Spotify, so that's not an optimal solution. Our telco 500/450 location has us spoiled. We would love to buy that system at our Comcast locations, but Ma Bell doesn’t seem to want to take our money, and no Shelby Co. muni group seems to make much effort at cajoling them into a faster fiber rollout.

Our search for alternatives to Comcast / AT&T would not be complete if we did not investigate 5G cellular. Verizon has info on a home cell / internet service for $70/mo. for current subscribers. When we look at their service in Chicago the numbers are very good - 1Gbps down and ~100Mbps up. The frequencies they use, in Chicago, for example, mean you will need an unobstructed view of the cell site before it will give you those spectacular numbers. Those who have had satellite TV will understand perfectly - a verdant tree branch or heavy rain storm takes out most of your service. A building between you and the cell site means you get zero service. The T Mobile version of 5G has more promise as it uses former UHF TV frequencies. But that means less bandwidth and therefore slower speeds. Their performance in NYC is roughly equivalent to DSL speeds - good for a backup but we've been spoiled by > 100Mbps speeds. The typical suburban house / classroom / office is best served by the sort of service telco fiber offers. Even the 1Gbps / 35Mbps cable service become 'iffy' in a household of four, each doing HD video conferencing in a neighborhood where all other households are doing the same.

That is our current, and admittedly verbose report. We hope at least one idea in all this helps those who are dealing with much the same school / work streaming situation daily, especially those doing home K-12 classrooms. The adjustment has not been particularly easy, but there are ways to get things… at least the computer hardware and networking… performing well enough to be tolerable. We’ll be back with another report when we discover and try something new.

We use SpeedTest apps by Ookla. You may have a different preference, but remember to use the same test consistently - make apples-to-apples comparisons.

Exploring the magic of mesh WiFi - good gains for minimal effort, just pay the price.

Sorting out a Comcast modem / router and its WiFi - then replacing it.

Stringing ethernet? Using cable TV cable, or house wiring?

How much internet speed do we really need?

Sending our internet connection into the electric breaker box.

Big gains by simple means.

Bringing our slowest devices up to speed.

Our WiFi radio is way too slow.

Our WiFi home network has become a hinderance.

Why this information from a law firm?

How, you might ask, does a law firm have any bona fides in this subject matter. A staff member whose first encounter with the digital world was Fortran on a Univac, followed by IMSAI CP/M S-100 bus, manufacturing using 3870 Mostek microcontrollers (2 kilo bytes of onboard ROM!) then 6805 & HC11 Motorola, Intel 8051, etc. etc... for years... plus a few US patents involving RF devices along the way in these areas is our credential. When the isolation happened this year, we realized we could and should help our clients stay current with their health & safety isolation and distancing needs while we worked with them on their family, estate and tax law matters. So you see above the kind of --additional -- advice and service we have been dispensing over the past months.