Several years ago I repurposed an old WR743ND router as a Wi-fi Access Point. Now I’ve hit the limit of 15 wi-fi devices in its Station List, resulting in devices dropped from wi-fi and a flood of HA errors. So, time to get a new WAP … but which one ?
“Best” depends on an individuals requirements - and members here rush to recommend the specific product that they use … but I think a key factor to consider is wi-fi.
IoT devices are physically small, low cost, and only operate on 2.4GHz band. That is fine, they are low bandwidth - except that we are buying more and more of them. They probably use old wifi 4 (802.11n) technology, so can’t achieve anywhere near the claimed 574 Mbps for 2.4GHz 802.11ax. And consequently pointless paying extra for a high-end router.
The Wi-Fi router / Access Point manufacturers, on the other hand, are single-minded focussed on claiming the highest bandwidth for 8K TVs and gaming. The bulk of the improvements are in the 5GHz Wi-fi 6 area, and adding additional technologies (which have also to be available on client devices) to further boost their total to produce massive (theoretical maximum) speeds … despite adding small print advising these figures are meaningless in the real world.
The key spec for smart home builders therefore is, how many concurrent 2.4GHz IoT devices connections does the router / Access Point handle ? This is probably rephrased as “how many entries are there in the routers 2.4GHz wireless station list ?”.
The same issue applies to extra Access Points and Mesh systems - are we being tricked into paying a premium for all-singing high bandwidth wi-fi 6 … which IoT devices cannot use ?
I was floored by the prices of some of the current high end home WiFi setups. Dropping 10 to 20 Franklins on one of these setups is easy to do, ouch. TLDR, I just picked up a ‘refurbished by NetGear’, NightHawk mesh AX1800 WiFi6 with a base station (MR60) and two satellites (MS60) for one Franklin on US Amazon web site. Targeted this unit because they can be configured as a ‘mesh’ ‘bridged’ access point/points with an existing router/firewall setup (or they can be a router/firewall/access point all in). It is a mid-range setup of current generation devices. It is performing very well so far, but further testing is continuing. But for a 100 bucks, even a fail will be a inexpensive learning event.
I just did a rip and replace of my 2 year old Mikrotik access points due to some problems they appear to cause with some sbc devices I was experimenting with. These boards are Raspberry Pi level sbc’s. The one difference with these boards from current RPI’s and all most all current home automation mcu boards is that they have WiFi6 chips. While this issue does not seems wide spread, I did find others with similar conflicts while ‘googling’ the issue. What I am saying here, is to you point of ‘forced’ upgrades, I was forced by simply wanting to use this ‘new’ device on my LAN. No super speed or other requirements special requirements by me.
To what I think is your main point, yes the WiFi industry seems focused the requirement of providing high speed streaming to 4 or 5 devices. I get it, what parent wants to go to bed every night hearing constant complaints of ‘buffering’ from the 2.5 kids
And I also agree that the rapidly growing number of WiFi IoT devices in homes does not seem be on the radar or the WiFi vendors. I think having 50+ WiFi IoT devices in your house will be common soon. Most current generation WiFi6 systems seem to talk about 50 to 100 device capacity. But these numbers are often buried in the specifications, and it not clear to me how performance degrades (or stops) for even low bandwidth devices when you approach or hit this wall.
That is still very much true. For example let’s take the widely used esp8266 - “A cost-effective and highly integrated Wi-Fi MCU for IoT applications” as an example which has about the price of a chocolate . According to it’s data sheet:
3.4.1. Wi-Fi Radio and Baseband
The ESP8266EX Wi-Fi Radio and Baseband support the following features:
• 802.11 b and 802.11 g
• 802.11 n MCS0-7 in 20 MHz bandwidth
• 802.11 n 0.4 μs guard-interval
• up to 72.2 Mbps of data rate
• Receiving STBC 2 x 1
• Up to 20.5 dBm of transmitting power
• Adjustable transmitting power
only “speak” b,g,n with a theoretical maximum of 72.2Mbps and therefor
Coming back to your question:
Many - I can’t even believe that there is any device out there from the last 20 years which can only handle 15 devices.
Are you certain you don’t have your DHCP server restricted to only 15 leases and that causes you problems?
To give you an idea: I have around 100 () espressif devices on my 10 year old wifi AP which I bought used for $10 and just charged it with openwrt.
If any one wants long software support for there routers and wifi ap’s I guess it’s almost mandatory to get a device which is supported by openwrt - otherwise the hardware manufacture calls your device EOL after 2 or 3 (or x) years and you can’t do anything about it
If one does a little research () it should be possible to get new wifi ap/router with gigabit ethernet, 2.4/5GHz and MIMO for less than $30 which has openwrt support and is therefor future proof
Because Belkin has a marketing department and they were in charge of the wording?
1000 (one thousand) devices are also more then 25
I guess the more limiting factor is probably the memory of the device and at some point software limitations. I think openwrt for example can’t address more than 16000 (sixteen thousand) devices by default.
The reality (for me) is different. My router which is running openwrt and has 128MB memory only “consumes” about 30MB RAM with about 70 active devices connected…
So my marketing speech would be: “miniMUM 28O DEVICES with SUPER STABLE connection ”
In the end it’s probably not even predictable (and the technical department didn’t gave a number? Or simply a “guaranteed” minimum?) because one connected client which does some heavy 8k streaming for sure consumes more resources on a wifi AP than for example 50 esp(home) devices with the efficient native api because they just idle most of the time
WR743 is in Access Point mode (so everything else is turned off but still present in firmware); with DHCP provided by my DrayTek router which is currently set to 80 DHCP addresses, and currently using much less than that.
I had hoped to load WR743 with firmware which contains only WiFi Access Point functionality and thereby frees memory to make the Station List longer … but apparently my repurposed old WR743 is so old that OpenWRT is no longer available for it due to 4/32 issue. Even OpenWRT runs better with a faster CPU and more RAM / flash
Just received a reply from TP-Link (with my emphasis):
Yes, both HS110 and KP115 are support 802.11n standard.
Normally router have two clients number. One is maximum wireless clients number, one is recommended clients number.
The maximum wireless clients for Archer AX10, Archer AX55 is 64+64, for Archer AX20 is 128+128. The recommended wireless clients for them are all 30.
The maximum wireless clients for Archer C7 is 32+32, and the recommended clients number is 25.
For Access point, the recommended clients number is about 15. Since TL-WA1201 support dual-band Wi-Fi, you may connect your devices to this Wi-Fi separately if possible, not all connect to one band Wi-Fi.
I do appreciate that wi-fi is a b*tch to troubleshoot because there’s no way we can see
level of background electromagnetic iterference, including from microwaves and other electronic devices
how much of the radio frequency (wifi channel) is actually being used by our devices. Like 2-way radios, if 2 devices broadcast on the same frequency at the same time within range, the result is unintelligible and both have to try again when there is ‘clear air’.
Bandwidth is highly variable. an IoT sensor might send an update of maybe 32 butes every 5 minutes - but a camera broadcasts a steady stream of pictures, each frame may be 800*600 of 8-bit color (=480,000 bytes/frame)
how much of the radio frequency is being used by neighbours networks who also use the same wifi channel.
Marketing tell you the maximum (theoretical speed, number of clients, etc) in order to sell more and justify their big commission/salaries; but Support have to make it work and so their recommended figure is realistic. And the realistic figure for regular 802.11n devices is apparently 30. Fortunately IoT devices (with exception for cameras) send to send short messages (low bandwidth), so more should work OK.
Please be advised that most chips in both devices, IoT and Routers, support up to an IP Address Range, 256 devices.
However in real-world environments we recommend no more then 70 connections per Router. Most can handle up to 110 without issue, but then the performance and speed drop as the Wireless is a shared resource. Typically 70 items is the recommended so it does not impact too much on the router response and connections.
… which makes me wonder if they took the time to understand my question.
Your investigative reporting seems to reenforce my belief that nobody understand ‘the number’. So I am going with 42 . Or maybe as @orange-assistant said, only the marketing department knows .
I just read an article on WiFi7… the author predicts that WiFi7 will be the defacto in 12 to 18 months, supplanting WiFi6 and WiFi6e before they are even established…
That’s a real oldie indeed and on top of that (very) short on flash and memory.
A smaller firmware will free up space in the flash but not essentially save much (ram) memory if the firmware does things right . If you are adventures you could try to build a minimal openwrt firmware for your device and need. I’m very confident that the hardware allows (much) more than only 15 devices on it’s wifi radio
A properly shielded microwave shouldn’t have a negative impact on wifi perception.
Indeed a dense network can cause re-transmissions. But that is nothing one need to worry to much about as this is gracefully handled.
That’s true. About IoT stuff we are talking the lower line of transmitted data here.
That can be quite easily analyzed. As easy as throwing a app on your phone and have a little site survey.
I would say they just can’t.
Depending on the sh*ttines of the manufacture firmware many routers even nowadays suffer from bufferbloat for example. It can be easy enough with one wifi or wired connected client and some heavy traffic to “clog” the whole device and even cause a significant degradation of service (down to throughput in worst case) for all other connected devices (even for the wired ones) .
If one is lucky not only owning the hardware but also has full control over the software they can make use of sqm to avoid bufferbloat.
Thanks for that link, I had not seen that page (probably because it’s under “Installing additional software” ). With talk about reusing old Wifi routers as WAPs I had assumed others would have had the same idea, and thus hoped to find a prebuilt device-specific firmware. Looks like another steep learning curve for me, so wait until it’s not my only WAP.
I use a different Wi-Fi Analyser app, which shows signal strength - but not how much data is being broadcast on that frequency. If you have a couple of neighbours streaming high quality video (because the sales droid told them that since the media centre and TV sitting next to each other both have wi-fi so no need to buy a cable), then there isn’t so much “free air time” available; and yes your devices are programmed to re-transmit … up to a point. I see the key question being how much free air is available … and even with a scanner that is sniffing packets that changes minute by minute.
I agree - there are plenty of factors that are impossible to see and/or difficult to quantify … making wi-fi more of a black art than a science. And as you point out there are also issues like bufferbloat which the end user is rarely aware of.
About the amount of clients, i can’t find a number but I can assure you it’s working fine with my 45 devices at the moment. Where more that half of those devices are 2.4ghz devices.
I really enjoyed reading your discussion as I share the same questions on this topic. I am simply looking for an inexpensive wifi router that can manage multiple IoT devices (30-50+). I’m not looking for a high performance router, I’m not into gaming, I’m not into streaming 4k/8k, I just want to build an environment with Home Assistant and a lot of ESPHome sensors, smart switches, etc. For example, I will also have 12 x ESP32-CAM, 6 x ESPresense, 4 x BT-Proxies, etc.
Thank you @frits1980 for suggesting EAP653. Anyone has other suggestions or I’ll go with EAP653
I’d add in, consider coupling the WiFi base station with a pfsense router this way you can isolate, protect the IOT devices and see who and what data they are sending to the cloud.
Are you considering a AX3000 WiFi 6 device for use with esp’s only? That’s like getting a racing car with 300km/h top speed but staying on unpaved roads only and never go beyond 50km/h
While the device itself looks nice an probably offers great performance (specially for 5Ghz/Wifi6 devices - so not esp’s!) I couldn’t spot it as a supported device for openwrt which would make it a no go for me as I can’t fully own/control the device and future updates (including security fixes) are 100% depended from the manufacture. Beside advertising “cloud management” beside everyone knows that this is probably the worst idea to control a local network - ubiquiti could sing about it
If you need a only a WIFI AP for esphome devices I would check for a used device that is capable of running openwrt. For example a TP-Link N600 can be available used for $10 and is capable running the latest openwrt release
Beside esphome devices don’t talk to a cloud (when not configured so) it’s not even necessary to have a extra appliance for VLAN’s as openwrt supports this out of the box
Unfortunately these are totally irrelevant for IoT on 2.4GHz band I agree with orange-assistant that a more basic model, or even second hand should do the IoT job just as well. I personally opted for a new router which can run openwrt to keep my options open.
IoT devices are generally low bandwidth (only send short messages); however I would suggest you look closer at the bandwidth required for cameras; specifically video resolution x color depth x frame rate. For example 800x600 x 16bit colour x 10 frames/second = 9.6Mbps per camera just for the image.
Then add overhead and consider congestion … and that is what you know about because you are putting it on your LAN. Goodness knows how much interference is being generated by neighbours using the same frequency.
You might be better off considering several cheaper WAPs wired back to your router/switch to spread the load, or maybe a “mesh” system if it uses a separate channel for backhaul.
Whatever, you can almost guarantee that your mileage will vary from whatever the manufacturers and salesdroids tell you - it is your job to make sure it works for you in your environment.
As for EAP653 - I am renting, so didn’t even consider any ceiling mount devices.
To make a long story short, this project is not for a house, I will rent a 2400 Sq.Ft. floor in a mall where I will transform food products. It’s an old restaurant that will be converted into a 25% industrial floor + 75% commercial floor. I jumped into Home Assistant 1 year ago and started doing more and more ESPHome projects. I want to convert that old restaurant into a small but smart production factory. I will monitor everything and create dashboards accordingly, and thus I will have a lot of 2.4 Ghz micro-controller boards around with some smart switches and more.
When it came to choosing a router, I didn’t want to choose an overkill router with a high bandwidth performance, I want a router that will simply be able to manage and connect multiple IoT devices. I asked on different platforms for suggestions to help me choose a router that would meet these requirements and I ended up with 6 models. Since this thread was precisely talking about what I was wondering, and that there are some active folks here who took time to share their ideas, I thought I could at least take time to thank you and share the final decision.
First, I didn’t want a POE setup, it would have been too costly. Secondly, being able now or later to upgrade to OpenWRT was important. Finally, I looked at Mbps on the 2.4 Ghz band.
What a fascinating project you have there Vincent. I wish you every success with it (and hope you write it up as a case study when/if you have time).
I am a little surprised, given that you’re looking at a commercial premises where ceiling-mount APs would be ideal, and you may need multiple APs to cover that area. But it probably comes down to whether there are already spare power sockets in the ceiling cavity.
If you have to pay an electrician to run power cables through the ceiling cavity, you will probably find that the extra cost of POE devices and a POE hub is the cheaper alternative, given that you have to install the APs and run ethernet cable anyway. And I always recommend wired connections where feasible for speed and reliability, leaving wi-fi for those connections where wires are not feasible.
This sounds to me like marketing doublespeak. They do love to mix things up and misinterpret theoretical figures as though they were real-world benchmarks.
It also leaves the question of whether having faster speed means they have allowed more devices to be connected concurrently - which I believe is an arbitrary number of entries in a table in the AP software. I found all manufacturers very reluctant to give hard figures … which in turn seems suspicious
I had a quick look at the alternative firmwares, but none seemed to promote >32 wi-fi devices as a feature, even when promoting reuse of old APs - which I consider a real benefit in our IoT world.
Would be interesting to know the shape. For example it could be 15 by 15 meters which should be easily cover-able with one AP on the right spot.
Indeed. Probably dual stream (isn’t supported by esp’s afaik) as well as a “maximum” theoretical allowed per stream of 400 Mbps by the 2.4Ghz wifi clients (esp’s are limited to 72 Mbps afaik).
While multiple streams are always useful the claimed theoretical maximum throughput will never be reached by any esp (probably not even 1/10th of it).
A little bit like
Having chosen that particular device (over others) only because of the claim it has the highest theoretical throughput on 2.4Ghz wifi was clearly what netgear marketing wanted.
Still there is a high chance other devices on your list could (even advertising less theoretical maximum throughput) actually perform better for your particular scenario For example the same priced RT-AC68U is a 3 stream MIMO device that (in theory) should outperform your chosen 2(?) stream device.
In any way it should be more than enough for your use case
. According to it’s 
) espressif devices on my 10 year old wifi AP which I bought used for $10 and just charged it with 
) it should be possible to get new wifi ap/router with gigabit ethernet, 2.4/5GHz and MIMO for less than $30 which has openwrt support and is therefor future proof 
STABLE 
connection 
”
throughput in worst case) for all other connected devices (even for the wired ones) 
but staying on unpaved roads only and never go beyond 50km/h 
I agree with orange-assistant that a more basic model, or even second hand should do the IoT job just as well. I personally opted for a new router which can run openwrt to keep my options open.
