DC calibration

Goal: lower middle "hill" in analyzer, Spectrum mode.

• Turn off osmo-trx.
• Set analyser to Spectrum, span 1kHz.
• Run host/build/utils/uhd_cal_tx_dc_offset_lms script from UHD (source) and note .csv file it writes:

uhd_cal_tx_dc_offset_lms --freq-start 940e6 --freq-stop 950e6

• Take a look at .csv file, "real" and "imag" columns. Values should differ between rows at no more than 10 or so. Calculate averages for these columns. Write them down as I ("real" column) and Q ("imag" column). Yes, it's not a typo.
• Edit umtrx_scripts/service/lms_tune.sh (source) so enable these:

send_lms_cmd 0x42 <I-value-in-hex>
send_lmx_cmd 0x43 <Q-in-hex>

• Run lms_tune.sh 1 and check "hill" in analyzer, so that it becomes lower. Play with I and Q values some more.
• Write resulting values to EEPROM. Some docs are at umtrx.org/hardware/calibration, but for single channel only. Commands to write values for the first channel of two channels:

cd /usr/local/share/uhd/utils
./usrp_burn_mb_eeprom --key tx1-vga1-dc-i --val <I goes here, in decimal>
./usrp_burn_mb_eeprom --key tx1-vga1-dc-q --val <Q goes here, in decimal>

• After the above changes uhd_usrp_probe should start showing values (because they now differ from default values).
• Run lms_tune.sh (commenting out 0x42 and 0x43 lines) to check that values from EEPROM are now getting used.
• Do the calibration for second channel. TBD.