Cheap router with BEAMFORMING

Not if you use it as p-t-p link with a bridge to ethernet on both sides. That "cuts bandwidth in half" only applies to repeater situations which isn't exclusive to WDS. It will happen in IBSS mode, 802.11s, ap/sta and, well, ap/sta with WDS.

Signals in the mid -70's measured on a phone are usually quite usable on a router with its better antennas. When linked up, what is the MCS and radio Mb speeds?

2.4 gets interference not just from other WiFi but also cordless phones, microwave ovens, Bluetooth devices, poorly shielded USB3, etc. none of which are detectable with a phone app.

As discussed earlier, I found out, that my problem is not speed (alone) but stability of the signal. A GL-B1300 has significant less speed than a TL-WR1043N, but an UHD-video (mkv, H264)) was played without interruptions. This was at another (little bit better) place. So I need at least 2 other routers to go on with my tests and find a solution for my worst place.

I am doing now tests with a TP-Link Archer c7 v5.0 and a GL-B1300. I am testing with my Android mobile phone and see that the C7 has better dbm-values, about -5 to -10dBm better, but the GL has better Mbps. The GL-connections shows 86Mbps with 2.4GHz and with the C7 I get 72Mbps.

So which router is better or what is more important, dBm or Mbps?

The C7 doesn't have openwrt installed, but for measurements with a mobile phone, it shouldn't count, or? If I keep the C7 I will flash openwrt, but if I return it and keep the GL I wouldn't flash the C7.

Your best bet is to use the device in it's intended way and determine how well it performs, main issue is actual packet transfer speed and reliability. Do some speed tests and stream some movies etc. Which works better in practice, that's the most important thing.

You are right, the best comparison would be to flash Openwrt with both routers. But I do not have a localized OEM-firmware to flash back, Can I save the original firmware from the router before flashing openwrt?

This should be possible look at doc's on sysupgrade, I think it can create an archive, maybe

I searched for it, but I didn't find a solution. I found http://www.geektalks.org/how-to-backup-flash-from-tplink-wr703n which is for another router, but now I cannot get ssh-access with the stock-fw of the C7.

There is eg https://openwrt.org/docs/guide-user/installation/generic.backup But it doesn't help as long as I do not get ssh-access with the stock-fw.

Yeah, I guess that's right you need ssh access to stock, which doesn't happen.

You should be able to get a localized firmware from tp-link's website. what is your local?

https://www.tp-link.com/us/support/download-center

is the US link, but you can replace "us" with the two letter country code, like "de" or "it" or "fr" or "pl" or what have you for other locales. Since you mention you're in EU I'm guessing you can get what you need from there somehow?

The C7 was bought in Germany and the Germany site doesn't offer a download.

from my experience the best coverage are like this, if you want to go ac, then the best one is ubiquiti ac lite lr, or for the n, the best one is 1043nd-v4, they bot work great with openwrt, v4 can be found very cheap on second hand, didn't had the chance to test the v5, on v1 you could stop httpd service and compile only the needed options, maybe disable ipv6, if you use more flash then you also use more ram, maybe you can try the ath79 images on v1

I suspect this is an oversight on the part of TP-Link's web team or due to some kind of regulations in Germany. If you go to the French site, you get firmware, and it's called "Multi language" EU region. If you go to poland you also get the same firmware revision. I do suspect the EU region just gets a single firmware:

https://www.tp-link.com/fr/download/Archer-C7_V5.html#Firmware

https://www.tp-link.com/pl/download/Archer-C7_V5.html#Firmware

From the french site it links: https://static.tp-link.com/2018/201805/20180518/Archer%20C7(EU)_V5_180425.zip

From the polish site, same thing: https://static.tp-link.com/2018/201805/20180518/Archer%20C7(EU)_V5_180425.zip

Thank you again, I compared md5sum of the French and the Spanish version, they are they same.

In the meantime I flashed the Archer C7. Everything is fine, except that I do not understand with it gets only 72Mbps with my phone. There is no difference between the stock fw and openwrt. Maybe the C7 uses old hardware? With elder phones the shown speed is 65Mbps and 72Mbps at the same place. My newest phone shows 86Mbps. So there is a difference with the phones.

I do tests with a Netgear R7800 now too, which is planned as a master-wds-ap. but it uses stock-fw until now, I have it a few hours. With the R7800 I get 45Mbps with 2.4Ghz at my desired position, but close to the router 86Mbps, with 5Ghz it is -80dbM and 90Mbps (shown on my Android phone) at my 2nd worst place and unusable at my worst place. With 2.4GHz it is at the limit at my worst place, but the speed changes after a while significantly. So maybe beamforming is doing it's job?

I am not sure if the Netgear R7800 is able to use beamforming with openwrt 18.06.1?

These numbers are what, the connection speed or a speed-test speed? The connection speed just means what modulation they are negotiating, it doesn't have anything to do with the actual practical transfer speed really. When actively transferring, devices can dynamically renegotiate speeds through time, and etc. You have to actually test speed-test to understand how well things really work.

If you really need so much Beamforming, then you should pay close attention to the products Ruckus, in particular R500 (used on ebay ~$60-65).

I fhink the Ruckus cannot work with wifi bridging, it it is not possible to have an ethernet cable there.

The conneection speed are shown by my mobile phone and the values are theoretical values, but if I see bad values here, tests will even be more worse.

I found out, that with 5GHz I get 80Mbips and with 2.4GHz 85Mbps. So I think 5Ghz doesn't make sense. Testing is not so easy, before I flash the brandnew R7800 with Openwrt. Before flashing with Openwrt I will test a while if the hardware is ok with the stock-fw to get no problems with warranty, should it get broken.

I didn't find a master-wds-mode with the Netgear stock-fw, but it works somehow. If I check the devices with the client-wds-router (Openwrt Archer C7), I see all devices and I can surf with a mobile phone in the web, which is connected to the C7, but if I check the devices at R7800 with the stock-fw, I cannot see the devices, connected via LAN-ports at the C7. So WDS works in 1 direction only. Maybe the R7800 is misconfigured with the stock-fw, but I couldn't find options, which make sense to try. The R7800 was not changed very much after using, static IP, no dhcp, gateway.

Can you recommend Android-apps with iperf3 for testing. I know I need a server and a client with 2 phones, but I have problems with editing the options in the app, they seam to be fixed.

I use hurricane electrics Android app Network Tools has everything I want except ability to query DNS for srv records

Maybe I missed it while reading this thread, but which drivers (in OpenWRT) actually support beam forming??

I know e.g. that the MT7612, which would be in a lot of "cheap" routers, can do (at least it is referenced in the proprietary Mediatek driver), but its not (yet) implemented in the MT76 driver.

Do the Atheros (9K or 10K) drivers have this implemented?

I can't get it work with HE. Is the HE app a server or a client? I assumed it is a client.

But I found a solution. I installed termux on my Android phone and entered iperf3. Then I was told how to install it, very easy. Then I needed a keyboard, which is able to cancel the server with ctrl-c. I found AnySoftKeyboard, added the terminal layout and could change the keyboard with long pressing space and swiping to right.

Between 2 phones connected to the C7 (wds client, wireless bridge) I get about 20Mbit/s @2.4GHz.

Does this kind of speed test make sense?

You are right it only does client mode, I thought it could be switched.

I suspect phones have their own CPU limitations that might limit your speeds, but this test certainly makes sense as far as real world performance goes. You might look at UDP based tests in iperf3 which will detect lost packets and may offer latency measures as well, can't remember.