Linrad Notes
Source Issues

Calibration

Calibration builds a digital filter whose response is the inverse of the analog hardware's, so the data entering the DSP has flat amplitude and linear phase. Because absent frequencies cannot be recovered, a desired response is also applied so the total filter does not have extreme gain or Q. For direct-conversion hardware (two audio channels per HF channel) a separate step corrects I/Q amplitude and phase imbalance. From z_CALIBRATE.txt:

The calibration procedure adds a digital filter with a frequency response that is the inverse of your hardware's response … the desired frequency response must be chosen for the total digital filter to not have extreme gain or Q.

Calibration

Filter correction

A broadband calibration pulse is passed through the hardware and captured. Linrad triggers on the pulse, averages many pulses to get the hardware's amplitude and phase response, inverts it, multiplies by the desired passband shape, and stores the result as fft1_filtercorr[]. The number of points is reduced (removing spur-induced discontinuities and noise) to a size matched to the hardware's Q. The correction is applied in the first FFT as a per-bin complex multiply, giving flat amplitude and linear phase across the band.

In the source

Pulse capture and averaging: cal_filtercorr() — calibrate.c:376 (triggering via trig_power, START_PULSES / INIT_PULSENUM, caldef.h:4). Final reduction/normalization: final_filtercorr_init() — calibrate.c:50; RAM update cal_update_ram() and make_cal_fft1_filtercorr() in calsub2.c. The per-bin complex multiply by fft1_filtercorr is in fft1.c (one and two channel). fft1_desired holds the target response. The CALAMP bit of fft1_calibrate_flag is defined at globdef.h:341.

I/Q balance (direct conversion)

When each HF channel is fed as two audio channels (I and Q) from two mixers, gain and phase mismatches produce an image. Linrad measures the imbalance across the band and stores a fold-correction (iq_foldcorr) that cancels the image; it is applied to the raw I/Q. This step exists only in direct-conversion setups.

In the source

I/Q balance in caliq.c: measurement cal_iqbalance() — caliq.c:403; correction build update_iq_foldcorr() — caliq.c:222; fold-correction expand/contract expand_foldcorr() / contract_foldcorr() — caliq.c:40; write write_iq_foldcorr() — caliq.c:152. The CALIQ bit is at globdef.h:340. Calibration modes are the CAL_TYPE_* states in caldef.h:10.

Per-mode files

Calibration is done per mode; the files can be shared by copying or symlinking. Each mode has a filter-correction file dsp_<mode>_corr and, for direct conversion, an I/Q file dsp_<mode>_iqcorr (modes: wcw, cw, hsms, ssb, fm, am, qrss, txtest, test, tune). The full list is in z_CALIBRATE.txt.

The two-RF-channel amplitude/phase balance between front-ends (pg.ch2_amp / pg.ch2_phase) is a separate adjustment on the polarization graph; the corrected fft1 spectrum feeds the whole signal chain.