"BxBChan" Channelizer
"BxBChan" Channelizer
Bit by Bit Signal Processing (BxB) is proud to intrudce the BxBChan -- an amazing high-speed digital channelizer. It is targeted at processing the largest signal bandwidths with the highest accuracy in the fewest resources in FPGAs and ASICs.
Polyphase Filter Bank (PFB) channelizers are highly efficient processing engines with many applications. They break a signal with large bandwidth into small pieces of bandwidth that can be processed individually at lower clock rates. They do this in a highly efficient way, exploiting multiple types of symmetry to reduce filter multiplies, to reduce mixer multiplies, and then to further reduce processing through the efficiency of a Fast Fourier Transform.
This can make many types of signal processing both immensely more efficient as well as immensely more accurate.
Channelizer performance can be measured with various performance graphs. The most important performance graph is one like the example below, which shows response of output channels (bins) 511, 512, and 513 of a 1024-point channelizer to input frequencies. This graph gives information about channel oversampling, crossover frequencies between channels, channel response at the crossover point, and aliasing levels. These parameters are things that are negotiated prior to core delivery. BxBChan filters are top-grade, but customers can also choose to use their own filters to customize their performance, for example in systems using matched filtering.
Note that this graph shows performance of a typical channelizer, and not the limits of what can be achieved. If you have specific needs, contact us for a channelizer targeted at those needs.
Another test is to put a sine wave through the channelizer, to see the response across all channels. An example of this test is below. This verifies aliasing levels across the entire spectrum, to a sine wave that's slightly off-center of any bin. Typical FFT-based spectrums would show sinc ripple in such a spectrum. A PFB channelizer shows no such behavior, giving in this case 100dB of dynamic range, right down to quantization noise levels.
Tests such as these are included with every BxBChan release, along with detailed explanations of the tests and methods to reproduce them.
For an engineer to truly understand some of the benefits of a PFB channelizer, it's important to understand the theory of operation. A derivation of the PFB from first principles can be found here. This shows how the PFB is equivalent to a bank of basebanding mixers and filters, which is a standard structure and a great way to understand its algorithmic performance. This basebanding operation can be made arbitrarily close to perfection by choosing the right filter quality.
PFBs allow arbitrary selection of processed bandwidth from large input bandwidths, and thus make great radio front ends. PFBs have an inverse, which can reconstruct arbitrary bandwidths from the individual channels. Bandwidth increases as more individual channels are used in the reconstruction. This allows signal processing with precisely-controlled variable bandwidths.
Large PFBs can divide bandwidth into very small channels, which enables effective frequency-domain processing on the bandwidth by multiplying each channel by a different phase coefficient. This gives low-cost application of arbitrary filter functions to a processed bandwidth, with application to linear system compensation and beamforming.
When there is sufficient input data, PFBs are superior to FFTs for spectral analysis and for computing Power Spectral Densities (PSDs). FFTs are typically windowed and averaged for these applications, but the windowing is limited to the length of the FFT. This limits out-of-bin spectral rejection of the FFT and increases mainlobe width. PFBs have no such limitation, since they can be viewed as windowed FFTs but with the window being a near-ideal filter that can extend beyond the length of the FFT. This allows a PFB to provide any desired level of out-of-bin spectral rejection without increasing mainlobe width.
PFBs are also superior to FFTs for handling streaming time signals. FFTs are sometimes used to process time signals with an overlapping approach, but this is non-ideal and introduces processing errors. PFBs provide an invertible transform from a stream of high-bandwith input samples to multiple streams of low-bandwidth output samples. This approach is as ideal as desired, based on the selection of PFB filter. So the transform can be performed continuously without introducing any errors.
The BxBChan is designed for digital signal processing applications where the sample rate is many times the FPGA clock rate. It has multiple applications:
The BxBChan has many important features, a significant number of which can't be found in other channelizers:
The BxBChan in its initial release has near-immediate delivery for Xilinx Ultrascale/Ultrascale+ devices. The BxBChan also supports implementation in other FPGA families and in ASICs, but these are have longer delivery times and in some cases may have higher costs.
The BxBChan is based on the BxBFFT, and inherits most of its advantages. Rather than listing them all here, see them on the BxBFFT's web page, here. One BxBFFT feature not currently included in BxBChan releases is support for C++ or Matlab models. These seemed to have little customer need. If these are important to you, contact us.
BxBChan pricing is intended to make channelizers available for all professional uses at reasonable cost. If you think prices are unreasonable for your project, send an email with a justification for a different pricing scheme, and we'll discuss it.
BxBChans are based on the BxBFFT. Customers who have already purchased equivalent BxBFFTs will get a discount on pricing.
The BxBChan is available for small academic projects for US$2000 per BxBChan. License terms will require that the BxBChan is cited in papers to which the BxBChan contributed, and that Bit by Bit Signal Processing should receive copies of any performance measurements made that are related to the BxBChan. Distribution rights are not included with academic pricing. Bit by Bit Signal Processing will have rights to use information from academic projects to make BxBChan advantages known for marketing purposes. Support for academic projects is at a lower priority than commercial jobs.
Commercial companies can get access to the entire range of BxBChans as well as BxBFFTs, with binary distribution rights and support for both, for US$25000 per year. Rights are purchased 3 years ahead, so the first-year cost is US$75000, and then it is US$25000 each year thereafter. Support ends after payments cease, and distribution rights end 3 years after payments cease. (These prices may be increased periodically to match inflation.) A wide range of power-of-2 BxBChans and BxBFFTs is immediately available after purchase. Non-power-of-2 versions are generated at customer request with modest lead time, since there are too many to have them all pre-generated.
Alternately, BxBChans for a specific channelizer size and parallelism can be purchased individually for commercial development or distribution. BxBChans in a selected FPGA family without distribution rights are US$4000 each, with 1 year support. The price drops to USE$3000 each for 3 to 8 channelizer sizes, and US$2400 each for 9 or more BxBChan sizes.
With never-ending distribution rights for unlimited products in the selected FPGA family, BxBFFTs are US$30000 each for 1 or 2 sizes, US$20000 for 3 to 8 sizes, and US$16000 each for 9 or more sizes.
These prices reflect a discount for allowing Bit by Bit Signal Processing to publicize the relationship for marketing purposes. Purchases that must be kept secret will have slightly higher cost.
Other arrangements are possible to match your business needs. If you would like to propose an alternate arrangement, please do so.
Purchases that could see applications with non-U.S. militaries will need to be reviewed for compliance with U.S. export law. Otherwise, this is the same as commercial applications.
The BxBChan is an excellent Polyphase Digital Channelizer, and although extensive comparisons haven't been done, it's believed that it provides higher numerical performance, lower power, lower resource utilization, and higher clock speeds than other options. It is unmatched in supported features that are enablers for some designs and also speed development. It is cross-platform, supporting both Xilinx and Altera FPGAs, with a path into ASICs.