==== Installation

Use source code in this tarball. Compile, install, and operate Yate and YateBTS
based on these instructions:

https://github.com/Nuand/bladeRF/wiki/Setting-up-Yate-and-YateBTS-with-the-bladeRF#Building_and_Installing_Yate__YateBTS

==== Running

Before starting Yate, load the FPGA. Do not rely on autoload feature.
Find the appropriate RBF for your device's version and size, for example the
bladeRF 2.0 micro xA4 can be loaded as follows:

bladeRF-cli -l hostedxA4.rbf
sleep 1
sudo yate

If yate hangs or freezes, try to exit the program and run
`sudo killall -9 mbts yate`. Reload the FPGA, and rerun yate.
If problems persist plug the device into another USB port.

==== Fine tuning

Yate's architecture is to generate as many future TX bursts as possible
until the TX fifos start blocking. Essentially RadioInterface::send() is
supposed to block until there is enough buffer room between
libusb->FX3->FPGA to fit another 1024 sample buffer.

Yate's expects to generate TX burst roughly 2 subframes before their
RF transmission. To calculate how many samples 2 subframes correspond to,
multiply the following:

   2 subframes
   8 timeslots / subframe
   156.25 symbols / timeslot
   8 IQ samples / symbol

   2*8*156.25*8= 20,000 IQ samples

24 buffers * 1024 samples is about 24576 samples, which (factoring in
some processing lag) corresponds to Yate's expectation of sending TX
bursts 2 sub frames before RF transmission.

The buffer size, and num buffers, num_transfers numbers were chosen
based on this methodlogy, however they might need to be tuned depending
on the USB host controller, CPU, and OS settings.

HOW TO TUNE:
Define '#define BRF_DEBUG' at the top of ybladerf.cpp and recompile and
install yate.


Strong deviations from 20,000 samples is not good, however issues really
only arise (and they do so visibly) if the time difference number drops
below 5,000 (or goes negative) TX bursts will not make it to the FPGA in
time to be sent over the SMA port. In this case, consider increasing the
"num buffers" and "num transfers" in unison.

If CPU throughput is the issue, multiply buffer_size by a factor of 2,
and divide "num buffers" and "num transfers" by the same number.

Increasing "num transfers" is helpful if USB timeout errors appearing.
Care must be taken to not increase this number past 50% of "num buffers"