Analog Charge-Diffusion Filtering Circuit for Adjustable Gaussian Bandwidth
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing analog filtering methods require large components like resistors and capacitors for slow signals, and digital filtering is inefficient for analog signals, while Gaussian filters are difficult to implement in digital circuits.
Innovation Solution
A method and circuit that uses charge diffusion on capacitors to achieve Gaussian and Gabor-type filtering, allowing adjustable bandwidth and efficient energy use, using an analog circuit with capacitors and resistors to simulate RC transmission line behavior.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If analog filtering uses large resistors and capacitors to create large RC time constants for filtering very slow signals, then filtering performance for slow signals is improved, but chip area increases significantly
Solution Approach 1:
The patent segments the filtering function into multiple discrete stages using cascaded common-source amplifiers with resistive loads. Each stage contributes to the overall filtering response, allowing the system to achieve large effective time constants without requiring单个 large RC components. This segmentation enables distributed filtering across multiple smaller components.
Solution Approach 2:
The patent replaces traditional passive RC filtering mechanisms with an active filtering approach using MOSFET amplifiers. The resistive loads in combination with the amplifiers create an active RC network that achieves the desired filtering characteristics with much smaller physical components, substituting the mechanical/passive RC time constant approach with an active circuit topology.
2Adaptability or versatility
If digital filtering is used to process analog signals, then filtering flexibility and programmability are improved, but conversion losses and additional complexity are introduced
Solution Approach 1:
The patent creates a hybrid filtering system where the analog circuit performs the core filtering function using passive RC networks and active amplifiers, while a digital processor provides configuration and control. The analog portion handles signal processing natively without requiring A/D conversion, and the digital portion provides programmable filtering parameters, allowing each domain to serve its strengths.
Solution Approach 2:
The patent designs a universal filtering architecture that can operate in multiple modes - purely analog filtering for high-speed paths, purely digital filtering for flexible programmable applications, and hybrid mode combining both. The same hardware infrastructure supports different filtering strategies depending on the application requirements, providing multi-functionality.
3Measurement precision
If Gaussian filters are implemented in digital circuits, then filtering precision is improved, but implementation difficulty increases significantly
Solution Approach 1:
The patent implements Gaussian filtering using an analog circuit topology where the transfer function naturally approximates a Gaussian response. The cascaded common-source amplifiers with resistive loads create a frequency response that matches Gaussian characteristics without requiring complex digital computation. This substitutes the computationally intensive digital Gaussian filter with a simpler analog circuit implementation.
Solution Approach 2:
The patent achieves precise Gaussian filtering by adjusting circuit parameters such as resistor values, capacitor values, and amplifier characteristics rather than through complex digital algorithms. By changing physical component parameters, the system achieves accurate Gaussian response with simpler implementation compared to digital methods that would require extensive computation and precision components.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Enables efficient Gaussian filtering with adjustable bandwidth, requiring less energy and suitable for both timely and spatial signals, with applications in neural networks and various signal processing tasks.
Implementation Method 1
The signal values SV0 to SVn are stored in form of charges in a series of first capacitors C0 to Cn
Implementation Method 2
The filtering is then performed by diffusing the charges on each first capacitor over neighboring capacitors of said series of first capacitors by connecting a resistor for at least partial charge equalization between each pair of consecutive first capacitors
Implementation Method 3
Analog filtering on the other hand often requires large resistors and/or large capacitors and thus a lot of chip area in order to create large RC time constants for the filtering of very slow signals
Data Source
Figure 1
Figure 2
Figure 3
AI summary
In a method of filtering a signal, the signal is sampled to obtain a series of signal values SV0 to SVn. The signal values are stored in form of charges in a series of capacitors C0 to Cn, said series of capacitors corresponding to said series of signal values SV0 to SVn. The charges on each capacitor are then diffused over neighboring capacitors of said series by connecting a resistor for at least partial charge equalization between each pair of consecutive capacitors Ci, Ci+1 (i=0 to n-1) in said series of capacitors for a time interval selected dependent on a desired filtering result. The remaining voltages across the capacitors of the series after said diffusion step represent filtered signal values corresponding to a filtered signal. With this method a Gaussian filtering with an analog circuit is realized in which the band width of the filtering can be easily adapted.