This is an interesting topic, I've looked at the chips used by (most/all?) IKEA devices lately, the Silabs ERF32MG1P series, they have a configurable output power of up to +16.5dBm. The EM3581 used in the HE C-5 has a configurable output of up to +8dBm. Using a CC2591 together with the CC2530 would give a rather high output power.
When I looked at if the power dropped on channel 26 with my SDR, I only looked at relative power drop from channel 20 to 25 and 26, I can't measure the actual output power well enough for it to be even remotely useful. Maybe this drop actually is due to antenna characteristics? I don't know enough about this to say one way or the other.
EDIT: This show the CC2530+CC2591 output power at different channels:
Assuming Ikea's ERF32MG1P have 16.5 dbm, I think this is right at the threshold of what is legal for US. I think this number is designed so that you can use the full power of the radio for US market with no limitation. There is no need to limit down the TX power for specific channel.
I am not sure about cc2591. But, with cc2592 which can push 20dbm, there is a code that lower the TX power for the later channel because cc2592 can push 20dbm (a bit more actually).
From the graph you attach, I think the power goes slightly down as the channel increase is not significant. I do understand that the dbm is not linear. But, I am guessing that you are talking about the drop at the tail end of the graph. I do not believe that it is due to TX power being limited. If the graph is taken with the firmware that is set for US, the blue line would show sharp drop to around 16dbm starting at channel 25 (not just the tail end of 26).
In addition, the graph is unlikely a test done using any firmware. This test could be done using TI SmartStudio where the TX will be set to continuously transmit and the frequency is sweep from channel 11 to 26.
Doing a test like what you have done using SDR is very hard to determine the real power the MCU pushing. There is a firmware that would perform modulation therefore the radio will essentially switching based on whatever Zigbee are modulating. This is hard to really measure the true power of the MCU pushing from the receiving end of a SDR.
I do not have code or insight of what manufacture are doing in term of the TX power limit. But, if my MCU TX power hard limit is 16dbm or less, what is the reason to lower channel 25 and 26 TX power specifically? They are already within spec. If the TX power is far below like 8dbm, I really cannot find any reason to limit specific channel.
Not being from the US, where can the, official, specs for what is allowed in the US for different bands be found? I'm sure I'll find "something" if I search for it, but what would be the official location?
This is an important thing in order to simplify the implementation. It can sometimes be a bit annoying in countries where we could have had a higher transmit power... Usually it doesn't really affect usability.
I've never tried that one, how well does it perform in general? Is it good and stable hardware?
This is normal when you apply the same amount of power but increase the frequency, but as you say, the difference is nominal and would probably have 0 real-life impact.
This is probably as you say a maximum output test done to show the chip maximum performance, not what would be allowed according to FCC.
Not just hard, I think with my setup it would be impossible. I can see a difference in power, but not how much, it is very much not precise.
This makes sense, and is probably as you speculate earlier why there are chips that cut off just at the maximum for those channels, but they do it for all channels.
That's interesting to know, I'm a software and MCU type of guy, don't know all that much about the rest of the hardware in terms of things like that.
Thank you both for some interesting information! Want to learn more about the hardware side of things, always good to have a deeper understanding.