High-pass dither generator and method
View Patent ↗A method and circuit for generating an M-bit digital dither signal with a substantially uniform probability density function and high-pass spectrum are disclosed. The circuit includes a linear feedback shift register (LFSR) with N storage elements where N>M, and a high-pass filter. The method involves sampling at least M storage elements of the LFSR with each clock cycle to form an M-bit LFSR output and high-pass filtering and the M-bit LFSR output to provide the M-bit dither signal.
1. A method of generating an M-bit dither signal comprising:
(i) with each clock cycle, sampling at least M storage registers of an LFSR of order N comprising N storage elements and where N>M, to form an M-bit LFSR output, and advancing said LFSR; and
(ii) high-pass filtering said M-bit LFSR output to provide said M-bit dither signal.
2. The method of claim 1 , wherein N=28, and M=4.
3. The method of claim 1 , wherein said high pass filtering comprises subtracting sequential values of said M-bit LFSR output.
4. The method of claim 1 , wherein said high pass filtering comprises subtracting from each current one of said M-bit LFSR output, a previous value of said M-bit LFSR output.
5. The method of claim 1 , wherein said high pass filtering comprises sampling M+K storage elements of said LFSR.
6. The method of claim 5 , wherein N=28, and M=4 and K=1.
7. The method of claim 1 , wherein said high-pass filtering comprises normalizing said M-bit dither signal.
8. The method of claim 7 wherein said normalizing comprises mapping said M-bit dither signal to a range of 0 to 2 M −1.
9. The method of claim 1 , wherein said high-pass filtering comprises inverting the most significant bit of said M-bit LFSR output.
10. A method of generating a high-pass M-bit pseudorandom dither signal having a period of 2 N −1 and a substantially uniform PDF, wherein N>M, said method comprising:
(i) generating a binary pseudorandom sequence, using an order N linear feedback shift register (LFSR) comprising N storage elements that change with each clock cycle;
(ii) with each clock cycle, forming an M-bit binary number from M of said N storage elements; and
(iii) high-pass filtering a sequence of said M-bit binary numbers, to form said high-pass M-bit pseudorandom dither signal.
11. The method of claim 10 , wherein said high-pass filtering comprises, with each said clock cycle
forming a second M-bit binary number from a second subset of said N storage elements of said LFSR, wherein at least one bit of each of said M-bit number and said second M-bit number are formed from the same storage element of said N storage elements; and
subtracting values of said M-bit binary number and said second M-bit binary number to form said M-bit dither signal.
12. The method of claim 11 , further comprising normalizing said M-bit dither signal to the range [0 to 2 M −1].
13. The method of claim 10 , wherein M=4 and N=28.
14. A circuit for generating an M-bit dither signal with a period 2 N −1 wherein N>M, said circuit comprising:
(i) A linear feedback shift register (LFSR) comprising,
a) N memory elements, each with a data input, a data output and a clock input accepting a clock signal, said memory elements interconnected sequentially, said memory elements accepting an input signal and providing an output signal at each cycle of said clock signal; and
b) a feedback network interconnecting data outputs of predetermined memory elements and feeding a data input of one of said memory elements to produce a pseudorandom sequence of bits in said N memory elements; and
(ii) a high-pass filter interconnecting a subset of said memory elements to produce said M-bit dither signal.
15. The circuit of claim 14 , wherein said high-pass filter comprises,
(i) a subtraction block receiving an M-bit binary number and a second M-bit binary number from said subset of said memory elements and outputting the difference between said M-bit binary number and said second M-bit binary number; and
(ii) a level-shifter block adding a predetermined value to said difference to form a level-shifted output; and
(iii) a normalization block comparing said level-shifted output to a predetermined number to output the lesser of said level-shifted output and said predetermined number.
16. The circuit of claim 14 , wherein each of said memory elements comprises a flip-flop.
17. The circuit of claim 14 , wherein M=4 and N=28.
18. The circuit of claim 17 , wherein said LFSR is connected in feedback using a connection polynomial P(x)=x 28 +x 3 +1.
19. The circuit of claim 14 , wherein said high-pass filter comprises, an inverter that inverts the most significant bit in said subset of said memory elements.
20. The circuit of claim 15 , wherein said normalization block performs modulo operation.
21. A quantizer with a fixed quantization step, comprising the circuit of claim 14 , wherein said M-bit dither signal is added to an input signal of said quantizer.
22. The quantizer of claim 21 , wherein said quantization step is 2 M .
23. A graphics adapter comprising the quantizer of claim 21 .
24. A printer comprising the quantizer of claim 21 .
25. A dither generator for generating an M bit dither signal, comprising:
(i) an N-bit linear feedback shift register (LFSR) connected in feedback using a connection polynomial of order N; and
(ii) a high-pass filter to receive the contents of M registers of said LFSR, in each clock cycle of said LFSR, where M<N, to filter said contents to produce said M-bit dither signal.
26. A graphics adapter comprising the dither generator of claim 25 .