Wrong values while calibrating LMS6002D for TX

Dear all,

We have a LMS6002D evaluation board (EVB) connected to an Altera DE4 development board (as BB card) through the HSMC port . At the time, we can generate “0” baseband and SSB in order to calibrate, and we control the EVB independeptly though USB and the provided Windows Client application. We are following the document https://github.com/myriadrf/LMS6002D-docs/blob/master/LMS6002D_TRX_Calibration_Using_RF_Loopback-v3.pdf. In this process, we faced some critical problems:

  • In the “Calibration procedure” section (P. 4), you first enable RF loopback to LNA2, and then set TX VGA2 at full gain (25 dB); however as you did not turn PA2 off you are still appllying this gain to PA2 instead of AUX PA. Is it intended to keep RF otuput to measure with spectrum analyzer? If we turn off PA2, we are effectively applying gain to AUXPA and we see signal at the loopback, but there is no RF output so we could not measure with spectrum analyzer AND FFT at the same time as you suggest in the document.

  • When we apply 25 dB to PA2, leds LD4 and LD2 begin to blink, which could mean the the power supply cannot provide enough current. We feed both DE4 and EVB with a 250W switching power supply. We checked maximum ratings (the EVB requires 280 mA max fot TX, the HSM provides 3A max, and the power supply can provide more than 1A at any of the voltages), so it seems to be OK. The only problem could be that the FPGA is leaving too few current available to the EVB, so we are considering moving SW3 and testing with a separate supply. Any suggestions about this problem?

  • Even at lower power levels, when everything works OK, we were not able to achieve LOFT or undesired sideband rejections higher that ~25 dB, when pages 4 and 5 of the document mentions about -60 to -70 dBc. We carefully followed the recomendations: first we apply 0 baseband and perform automatic calibrations. Then we apply SSB and minimize LOFT by adjunsting I and Q LO leackage DACs (which must be driven to -16 mV to get any result). Finally, we apply gain and phase correstions at BB to minimize undesired sideband (LSB). For reference, we include a capture of this and the .rmv file with the status of all LMS6002D registers. For thsi case, VGA1=-10dB and VGA2=19 dB, and baseband processor is applying a cos and a sin to I and Q respectively with full 12-bit 2-complement excursion (+ 2047 to -2048), even of we also tried with lower excursions with similar results. We appreciate any suggestions, since these are quite bad values compared with the documentation.

Caoture of TX output with direct connection to Signalhound SA44B through DC block + 10 dB attenuator:

Register status:


Thank you, kind regards.

Carlos Zerbini
Digital Communications Research Lab
National University of Córdoba