I describe the implementation of a simple algorithm to optimize the buffer frequency location to maximize the number of resonators.

Motivation

Since the spread of resonator frequencies (especially in the upper band) exceeds the bandwidth (490 MHz) we have available in the SDR readout, it is important to pick the buffer location so as to maximize the number of tones that can actually be used.

Method and Results

The algorithm starts with a first-guess for the LO frequency which is simply the mean of all of the resonator frequencies in the device list (for either the lower or upper band). It also excludes any resonators within +/- 10 MHz of the LO frequency. The number of available resonators is determined. The routine then shifts the buffer both up and down by 50 MHz spacings (the number of shifts in each direction depend on how far the initial buffer edges lie from the edges of the resonator distribution). At each step comparison is made with the initial number of available resonators; if the new number is larger, the current parameters are adopted as the new benchmark.

Conclusions

For the lower bands, this isn't too much of an issue, as the resonator spread is not much larger than the available bandwidth. For the first 6 X 12 devices (Xa and Xb), 36/49 and 37/44 good resonators can be fit into the buffer in the lower band; these numbers are not improved by shifting the buffer. For the upper band, the initial values are 25/65 for Xa and 23/54 for Xb; these are improved to 32/65 and 29/54, respectively.

The routine also generates plots showing the location of the LO, the buffer bandwidth edges, and which resonators (both good and bad) fall within the useable bandwidth.

devXa_part1.pdf devXa_part2.pdf devXb_part1.pdf devXb_part2.pdf

Further Work

The buffer generation code needs to be adapted to incorporate this.