I am using an SXT 2 (the 2.4ghz)
Trying to understand the FCC limits for united states.
If I selected United States, regulatory-domain, 802.11, 2Ghz-only-N, channel width 20/40mhz eC It sets the TX power to 17dbm with total tx power 20dBm.
If I go to superchannel then I get Tx power 29dBm and total power 32dBm.

I thought EIRP for 2.4Ghz in the United states was 36dBm? So why does it limit the default to (20dBm)? I must be misunderstanding something....
Thanks!

I'm not sure if country limits, burned in ROS running on SXT 2 are special. However, on my RB951G running ROS 7.7, country limits for "united states" says that EIRP is limited to 30dBm on 2.4GHz band. SXT 2 antenna has 10dBi gain meaning that Tx power, output by Tx power amplifier, is limited to 20dBm (or 17 dBm per chain). Which is what you're getting.
Keep in mind that EIRP equals sum of Tx power output (on device antenna connectors), loss on antenna cable (so negative when expressed in dB; 0 if the antenna is attached directly to device) and antenna gain.

When disregarding country limitations (e.g. by setting to superchannel), Tx power is limited by device capabilities. For SXT 2 absolute maximum total output power is 32dBm (which on SXT 2 translates to EIRP of 42dBm).

So if that EIRP limitation to 36dBm is a fact (and you can present some official document that says so ... and in which conditions), you should open trouble ticket with MT's support so they can fix it.

Thanks for the clarification!

Thanks for the clarification!

And the gain of the antenna where go???

You have 2 streams 10dBi each...

10dBi x 2 = ... 13dBi ..... +20dBm = 33dBm (2W) ... + 3dB (penalty of 3dB for missing the ATPC) = 36dBm, the precise EIRP allowed value, 4W

But for point-to-point device, the max EIRP is based on antenna gain...
With 2 SXT 2 you must be allowed (but RouterOS do not allow) to set the device on this way:
TX power ch0+ch1 = 27dBm / 500mW (24dBm per chain) for obtain one EIRP of 40dBm, yes 10W, on USA are allowed this rule for point-to-point link...
RouterOS can not do that because is missing the ATPC and have a penalty of -3dB for that,
and 40dBm - 3dB is still around 36dBm (4W), 3dB more than the 33dBm max for multtipoint.

And the gain of the antenna where go???

You should only count number of Tx streams once ... and usually that's counted in total Tx power of device, which in my previous post is worded as "output by Tx power amplifier, is limited to 20dBm (or 17 dBm per chain)". So your equation should read "10dBi x2=13dBi .... +17dBm" (the last number being Tx power per chain) ... and missing TPC penalty should rather be substracted from other numbers (because it reduces allowed Tx power). BTW, the TPC consideration is only necessary for 5GHz radios and we're talking about 2.4GHz in this thread so far.
But yes, if equation is flawed (as it often is when written by people who don't understand how logarithmic calculations go), then one easily comes to higher limitations. Which have nothing to do with regulatory documents, but seem logical to laymen and soon enters into moden folklore.

As I already mentioned, country limitations, baked in ROS, are the general ones because they are safe and legal for all uses. If device is used in regime for which country limitations are more lax, then ROS allows to do it (superchannel), but device admin has to know the limits to properly configure device. After all, it's device owner who has to abide the law, device manufacturer can do only as much to help at it (i.e. by offering set of configuration values which are very likely legal to use).