Antenna Diversity vs MIMO discussion

I've been always asking myself this question -
If 2.4Ghz router is sold as MIMO capable and has 3 antennas and max bandwidth of 450mbps then what would happen if you switch off MIMO by enabling the antenna diversity.

As general rule of thumb 1 antenna is 150mbps (at least when used as MIMO),
although in real life, it is capable actually half of that speed (75mbps and that's most likely because it is actually 75mbps for output and 75mbps for input and marketing people sum them up to make it look better. But since speedtests don't do concurrent upload/download test we interest only what we see for single operation mode at a given time).

AFAIK there is option from OpenWRT to set each antenna tx or rx by choice in Diversity mode, thus this disables MIMO operation. (BTW is this still supported ?)

Antenna Diversity mode means each of the antennas do only 1 action.

Let's say I want 2 antennas to do transmission (tx) and 1 for reception (rx).
Question #1 - Would that mean I can have 2 * 150mbps (300mbps) tx speeds ?
Question #2 - Would that mean I can have 1 * 150mbps (150mbps) rx speeds ?
Question #3 - Would real life speeds are again split by 2 (total of 150mbps tx and 75mbps rx), or that total dedication onto rx or tx for a given antenna really does make it handle the advertised speeds.
Question #4 - Has anyone done a benchmark and can post the results between the two ?

And yes, MIMO is better in range and interferences, BUT if you do point-to-point WiFi link it might turn out that being able to set more tx antennas individually can give better bandwidth for the rx ?
Thank you !

additional antennas is not the same as additional radios.

Most home routers since years (except very compact USB Wifi sticks) are MIMO routers, where multiple antennas are connected to the same 1 radio component. The signal from the one radio is varied between the antennas using a mathematical formula, all antennas are fed with a synchronized signal at the same time from the 2 radio. In sum, the overall throughput is higher than with a single antenna, if the receiver also has the same number of antennas and can properly decode the same math formular.
(Mu-MIMO is a special corner case, where such a synchronous activity can happen between 1 sender and not just 1 but multiple receiver devices, if all involved devices support Mu-MIMO. This does not add yet more bandwith, but best case exploit of the max available bandwidth might improve)

But additional antennas or MIMO is not the same as having additional radios.

Separate from that, there is the concept of a few more expensive home devices, which have e.g. 3 independent radios, e.g. very common is 1x 2.4Ghz + 2x 5 Ghz radios. (And each radio usually having multiple MIMO antennas)


Yes, multiple antennas isn't the same as multiple radios, BUT if you have single radio, which is promoted as 450mbps and the 3 antennas are marketed as 150mbps each, then logically if you put a taps on two of the SMA ports, you only feed 150mbps to the board no matter how strong the signal is, isn't it ?
This is something that I don't know personally how it works.

There is channel mixer inside the board that combines all the antennas input for only stronger gain and that gain is the result for better throughput ? Again this means if you make antenna RX only for example this means all the bandwidth is reserved for Upload making your router faster on upload and slower on download. This is what I want to figure out.

"Diversity" for the wireless driver in the most cases is when a single-TX device has a diversity antenna and a RF switch inside. If there are more TX antennas is likely MIMO to be done by default, possibly doing some gain/beamforming/antenna selection if you have some redundant antennas or low signal level. Using the antenna as a real diversity gain RX antenna may not work as expected, but do only STBC processing, since some combiner methods could be optional or not implemented/enabled. It depends on the wireless radio and the driver implementation (and its options accepted/enabled), so you can check its datasheet, then check the driver. The radios do have both TX and RX chains, sometimes duplexed with external components over the PCB. Some devices do leak RF signal, so it is better the antenna to be removed than just disabled by software if unusable. Both TX and RX do share the same radio channel/spectrum, even if not sharing the antenna. You may check /sys/kernel/debug/ieee80211 for some more information by the wireless driver and the options being accepted.

Nowadays I do not think you need additional radios over the same band, unless going repeater mode. The channel width could be just increased. The current standards do define Preamble Puncturing / MRU Puncturing features, so the "bad" portions of the spectrum could be avoided and still having 160/320MHz bandwidth and mitigating the interference losses.

Here is an example a of Qualcomm Atheros 2.4/5Ghz abgn radio:


Thanks for your reply !
I will use the same image as it perfectly illustrates the router I'm testing at the moment.
This is the idea

People usually refer to
iw phy phy0 set antenna <bitmap> | all <txbitmap> <rxbitmap>
in order to setup the antennas in the requested mode but for me it doesn't work for TL-1043ND
as I get unsupported option notification message. I don't know how this would be doable.

160/320MHz bandwidth? I wish I could setup this to the atheros driver on 2.4 Ghz although this seems near to impossible to me..
I think it needs almost entire driver rewrite.. unless you have something in your pocket.

I do not think you can configure it. For the Atheros driver the diversity means exactly the diversity receiver of a single TX antenna device β€” for example WR740N β€” The diversity antenna is clearly visible on the PCB. You can check the driver state in /sys when some antenna gets TX disabled and the diversity configuration option in place (it's on by default.).

You can use the antenna setting is to test the TX and the RX levels or each antenna (the chip cannot do VSWR test). Other possible use is to disable an unusable antenna - for example you have a directional antenna and the remaining one adds only interference which is then picked as a signal, because the combiner method used. It is possible the radio to become unstable if you left only one antenna with both TX and RX enabled and it's not the first one and the diversity being disabled.

The bitmap should be

111 = 0x7
101 = 0x5
010 = 0x2

and so on where a "1" is an enabled antenna.

If you remove an antenna, it will reduce the PHY speed, so you need the same number of antennas at both ends. If having a device with less antennas it is possible that the driver could use the redundant antenna for diversity gain automatically, which depends on the combiner method supported or the receiver will just reconstruct some diversity data via STBC. Some of these chips do support TxBF - please do not ask me if it's hardware-only feature in this generation or should be visible to the driver and advertised.

Comparing these two statements say the following to me - > If you have directional antenna (yes, I do and plan to use one for Point-to-Point), you should plug it to SMB (connector) 1 and disable SMB 2 and SMB 3 with the "iw phy ph0 antenna" command.
But after you disable them you lose PHY from 450mbps to 150mbps.

What is the alternative to do not lose PHY ?
I imagine putting this

This means that signal from the directional antenna will be split to 3 and going into all RP SMA ports.
Now MIMO will not work as usual as the source signal isn't from 3 sources but 1 and that STBC won't work ?
Anyway I will remain my PHY rate (as well as throughput) ?
(Of course I must expect really high gain from this antenna because the dbi will be split to 3, but that's not an issue here).

Thanks !

You will have only one propagation way, so the additional features which benefit from the multiantenna propagation will not work. The way to retain the rate is the channel width or the modulation to be increased. The radio in your case is unlikely to support anything over 64QAM and may not have a sufficient SNR in order use it.
This adapter on the picture looks like a scam to me. You can equally combine/divide (easily) an even number of RF paths with impedance matching, so to be done properly you need a quad combiner and a load for the remaining port. So if you have, such a case it's better to combine only two ports and the third to remain for polarization diversity. The signal level when divided may go below the sensitivity threshold of the receiver, so there could be no benefit. Maybe there could be something if the RX port goes to the more-powerful port of the RF splitter/combiner. In the real life you should consider some device intended for the purpose with good enough external LNAs and PAs.


.. unless there's external power amplifier ? If I put power amp between the antenna and splitter,
(and I suppose it should be.. PA RF in <--> splitter (the combined port), PA RF out <--> Antenna )
will I remain my troughput / PHY rate, as the case is needed for good transmission speeds.
And all this MIMO limitations confuse me how it would work in such scenario.
In Point-to-Point should I care at all of MIMO STBC if the signal is so strong that all the data is received perfectly from 1 antenna without data combinations (from multiple antennas) for stronger throughput ?

The power amplifier will introduce additional noise and distortion and you will need a duplexer or biderectional ampifier. The case of the low power levels of the WiFi makes it impractical as a dedicated device.
In case of point-to-point you can easy do polarisation diversity.

Ok to keep it simple.
I already have this AMP

What am I supposed to do to connect it to external antenna and the router as well as remain at least the TX speeds ? (that would mean good download speeds for connected client to the router).

My best bet was to use one SMB port for TX only (as in my first picture) , connect the AMP directly and the AMP connected to the antenna, that's 150mbps which is not ideal but acceptable.
Now the other 2 antennas are connected normally to usual 5dbi antennas and they can help a bit too.
Since this isn't possible, I guess I have to use the splitter option.. but how exactly ?


You can dedicate one antenna as TX-only, the other one to be externally duplexed, the third to be for polarization diversity. On the far end you need two TX amplifiers and one RX port matching the remote TX polarization. In this case you will have up to 300Mbps PHY rate, let's say in the download direction and up to 150Mbps in the uplink direction, when using a 40Mhz channel width. You should use dual-polarized directional antennas. Using omni antennas over a point-to-point link will likely ruin its performance. If using a splitter, you will likely burn the RX part of the device and will loose 3.5dB of the TX power.

:warning: Using external amplifiers may result power levels outside the regulatory limits, possibly you could be counted liable for this by the local regulator, especially if there are neighbour complaints or the device is not good enough and you are jamming some telecom/radar equipment in the area. The latter may lead you to criminal prosecution. In the real life for the 2.4Ghz I would suggest you to go to the CPE210.

How can I do this as the router is MIMO ?
As in my first picture that would be 75mbps TX only ! If there's no some kind of additional configuration (which I still cannot find).

Amp has only 1 RF in and OUT.. 1 tx only antenna and 1 externally duplexed into 1 amp connector ? I miss the link here.

The power levels in most countries is reallly really huge ones ! This is not so important but the Gain happening between 2 links !
2W is nothing compared 40 to even 120W 3600mhz 5G towers.
It is about the gain and huge role for the gain is the antennas.

Yes, if having a commercial license.

Your directional antenna is likely to have only two ports, so you must combine the TX and the RX somehow after the amplifier, without burning the receiver.
It would be 2T1R AP connected to a 1T2R STA. It could be unsupported by the driver.

You have both TX and RX driver antenna mask.

It could be up to 150Mbps if 40MHz is enabled.

Hmm, I don't think you can have commercial license upon WiFi which is non-commercial.
Microwave is 2.4Ghz and typically 800W and if what you're saying is true it is illegal to turn on the microwave. As long as there's no antenna attached to my microwave I think it is all OK.
I just added this clarification because I think it is beneficial for the others as well.

My directional antenna has only 1 port. I guess this makes it easier ?
But again I must somehow connect the Amp from the one end to 2 SMB router ports.
If it is with splitter that's 50% reduction in gain. But actually this amp should act like pre-amp and on the RF out to the antenna it would be 2*100% (probably over 100%), without added noise ?

The RF out of the amp is easily connected to the single port antenna.
To clarify - Dish reflecting to a biquad antenna. And biquad aint melting from 2W :grin:


The AMP will return to the receiver, so you should not concern about the output power, but the maximum RX input power. There is a switch after the LNA/PA for a reason. If you have an external AMP, you must somehow control the switch of a different radio chain if using a splitter, but not some directional RF device. Or you may go to board level and connect the amplifier somehow before the switch if the switch is external. In this case if you already have some crappy internal PA - just replace it with some more decent with the same pinout. Or you may go with dedicated antennas for TX and RX or go to some device intended for the purpose like Ubiquiti Bullet M XW.

Actually back the days of the wireless ISPs this was a common practice of the wired ones.

AFAIK the reason for this switch is just to switch between RX and TX very quickly in order to have MIMO antenna. Otherwise it won't be MIMO but dedicated to RX or TX. So nothing to do with the power.

It would be really difficult for me to replace the internal PA to this external PA, although it makes sense.

I'm 40Mhz already and it is 75 in one direction (per antenna). This is the marketing trick they use as initially mentioned to calculate the throughput.
This router is advertised as 450mbps but in fact it is 225 at the very best with 40Mhz (in 1 direction).

The maximum achievable PHY rate is 150, this includes all the overhead, so on application level you will get less, but lets say something over 90Mbps for TCP with the default settings. However, you will probably get 40Mbps or less if passing an RFC2544 test.

You may possibly go to Wifi 6, 5Ghz or a microwave in a band not requiring a spectrum use license if the speed is the most important. The microwave link offers more options like 2+0 XPIC or space diversity over longer links.

Introducing more RF devices over the antenna path will introduce RX signal loss if there is no external LNA and noise as well, so I don't get the point increasing the power by 3dB, but reducing the SNR over the RX path if using the same antenna. If you think the speed can be increased this way - better the antenna gain to be improved and the VSWR to be reduced than adding more power, more noise and distortion.

I don't know what exactly is but I used iperf3 with udp at 30cm distance between router and laptop with high end wifi6 card.
Average speed 150mbps... Top speed 200mbps in the Openwrt Realtime graphs-> Traffic wlan0. Let's say there's 25mbps of overhead which exactly matches my calculations. Iperf3 reports 110mbps which means that on protocol level the things become even worse.. And by that I believe you can use the PHY rate only at half one way and the other half the other way in MIMO mode. That's why I started hating MIMO :smiley: especially if you want to dedicate all the bandwidth for RX.

LNA can be applied to the distant receiver.
The main point here is to preserve and have max RX bandwidth for receiver and TX for sender router.

5Ghz and over is very prone to any objects and it is practically half the distance of what 2.4G does.
And moreover 5Ghz chips are with less sensitivity. If 2.4G chip is capable of -79dbi, 5Ghz chip is capable of -55 - 65dbi sensitivity.
If the receiver signal is -65dbi. I don't know what will perform better as throughput but I think if 5Ghz is edging between working or not then obviously it would be 2.4G.

The frequency does not exactly going to reduce the distance when point-to-point.

Your speed is probably limited by the number of the antennas in use and/or the CPU.
Go to the CLI and monitor the CPU idle via top for example.

The only meaningful way the speed to be increased without replacing the device is the antenna to be modified to suport dual polarity.

The MIMO does not reduce the speed in case of duplex traffic, but the CSMA/CA algorithm. If you have the need of full-duplex, you may do a second link over a different channel and adjust the routing to make them unidirectional.