Integrate EVM - Githubissues

harrison-caudill / pylink

Python Link Budget Tools

BSD 3-Clause "New" or "Revised" License

41 stars 10 forks source link

Integrate EVM #29

Open harrison-caudill opened 5 years ago

harrison-caudill commented 5 years ago

Need to find a reference, and do the math on this one first. Initial thoughts:

An EVM distribution can be modeled as AWGN at the transmitter...I think...
We are then transmitting amplified noise
The amplified noise can probably be reduced by FSPL just like the signal (a la friis cascade)
The size of the decision-region of the modulation is modulation-dependent, but if it can be modeled as AWGN, that won't matter...I think...
Not sure whether or not realistic EVM's have non-trivial impacts upon system performance so it may not be worth inclusion in the overall budget
Not sure if EVM distributions are even provided by VSA's
If the EVM distribution is not very gaussian, would it be appropriate to integrate the distribution for a target BER/decision-region based upon the selected modulation?

harrison-caudill commented 5 years ago

I ran a quick test as follows:

The transmitter was modeled as a perfect sum of signal power and noise power
I used the pre-existing power infrastructure to compute received signal and noise power
I updated the received signal power to reflect the correct portion which is signal power from the transmitter vs amplified noise from the transmitter
I computed the received amplified noise power from the transmitter by taking the noise portion of the received power at the demodulator
That received noise power was spread across the allocation (not the occupied BW because of the issue with cycles there)
The definition of EVM used was fairly simple: EVM = P_T,N / P_T,S, or noise power over signal power.
The test at 5% EVM on a 2W transmission with realistic parameters yielded a 1dB difference in overall link margin.
My mentor considers that to be a reasonable estimate.

harrison-caudill commented 5 years ago

Max bitrate should be calculable for a given modcode/cn0...just have to switch to linear space.

https://photos.app.goo.gl/BBumB8HxBfAczLog7

harrison-caudill commented 5 years ago

With a defined modcode, which can be computed with the above formulation for max possible bitrate, we can spread the amplified noise from the transmitter across the receive bandwidth and augment the noise spectral density to incorporate both the thermal noise and amplified noise from the transmitter.

This one is REALLY making me think it might be better to switch the entire computational chain over to a friis cascade starting with the transmitter.