I've noticed that on TP-LINK WDR4900 and TP-LINK 1043ND with OpenWrt, the txpower regulation from LuCi or with CLI is totally useless. It saves the option but the radiated power is always the same at 0mw or 20mw or 25mw etc. Q1: I wonder if it is ath9k driver issue that was never resolved ?
Interesting part is that on the 1043ND I get in LuCi " - Current Power 25dbm" and it stays like that no matter the dropdown option. Q2: I wonder if that current power is true or not (possibly 20dbm by default?).
Q3: These routers come up with 3 (detachable) antennas. If the router was with just 1 antenna and set at 20dbm (100mw) it is obvious that 100mw are radiated trough this antenna. But what happens with router with more antennas at 20dbm (100mw) ? Is the power split to 33.3mw to each of the SMAs or we have 100mw * 3 (300mw in total radiated power) ?
I don't have your device and can't really speak to the first two questions with any knowledge. Except to say I suspect it's something that the firmware doesn't let you change.
Assuming 20dbm output, that is the total effective radiated power per unit time from the radio. There are two ways to use multiple antennae: Same signal over multiple antennae, or different signals on different antennae. When it's the same signal over multiple antennae it doesn't reduce the overall power, it just changes the radiation pattern. This is how beamforming works - it's similar to a phased-array radar. The radio tries varying which antennae are used to see which combination ends up sending a stronger signal to the remote. When it's a different signal on different antennae (mu-mimo for instance), then yes that 20dbm is spread over multiple signals.
This isn't quite true. Transmitting without an antennae is a really great way to burn out the power amplifier stage of a radio, due to power reflection. All the power gets reflected back to the amplifier which is really bad for it. There is generally a special protective circuitry built in to the final power amplifier which, when it detects a poor standing-wave ratio (SWR), will choke the power output. So you're not increasing the power to the one antenna that's connected, you're just cutting off the power going to the missing ones.
I have known some chipsets to be missing that power reflection detection choke circuit. So even though it's probably not unsafe 99% of the time, I would never get in the habit of running any WiFi device without an antenna.
These routers are MIMO, so practically this answers my question that each antennae has 33mW?
"About 1 antennae" I meant if I have another router that is with one antennae only (for example TP LINK 740N) that it would be logical to radiate the full 100mW trough it.
What do you mean ? The installed firmware on the routers is OpenWrt. Are you talking about the ath9k drivers that these devices use as my suspicion is also the same. I wonder if somebody else has some of these devices and test. If he gets the same connection RSSI with different dbm setting then these drivers were never adjusted for OpenWrt to manipulate the radiated power.
With my limited understanding of MIMO I would guess that the powerbudget can be dynamically divided between the antennas. So conceptually 100mW on antenna 1 and 0mW on antenna 2 and 3 would be an option as well, but see @VA1DER's comments above....
I meant the Ath9k's internal firmware. In general WiFi drivers when they talk to a WiFi device don't talk directly to the device. There is an intermediary layer - in most cases, the WiFi chipset has its own separate little computer with its own tiny RTOS all on their own. This internal firmware has it's own API/interface which lets the driver send it commands. Sometimes those internal firmwares don't expose all the functionality of the underlying chipset, or have their own logic to control the chipset in certain ways. My direct knowlege of Ath is limited, but from what you are reporting, it sounds like the underlying WiFi firmware is making its own power output decisions.