Welcome to EDAboard.com

Welcome to our site! EDAboard.com is an international Electronic Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

Register Log in

FPGA implementation of a CIC decimation filter for a single-bit sigma-delta modulator

Status
Not open for further replies.

bardia

Newbie level 5
Joined
Sep 14, 2010
Messages
8
Helped
0
Reputation
0
Reaction score
0
Trophy points
1,281
Activity points
1,378
Dear all,

I want to design a CIC decimation filter for a 3rd-order sigma-delta modulator with single-bit quantization.*As we know, CIC filter need's 2's complement arithmetic. Let's assume I need 16 bits internal word length for correct operation.

Now my question is, do I need to convert my 1-bit unsigned quantized data (at the the modulator output) to the equivalent 16-bit 2's complement world length at the input of the CIC filter?

In other word, do I need to convert 1b'0 to -1 = 16'hFFFF, and 1'b1 to +1 = 16'h0001.

Thanks for your help.
 

FvM

Super Moderator
Staff member
Joined
Jan 22, 2008
Messages
47,875
Helped
14,125
Reputation
28,507
Reaction score
12,809
Trophy points
1,393
Location
Bochum, Germany
Activity points
277,851
CIC don't particularly need two's complement, because signed/unsigned is only a matter of number interpretation for addition and substraction. You can implement it either with unsigned or signed arithmetic. The data coding can be converted at the output, e.g. from offset binary to two's complement by inverting the MSB.

In other word, do I need to convert 1b'0 to -1 = 16'hFFFF, and 1'b1 to +1 = 16'h0001
This way, you would multiply your input bit by two, which is usually unwanted, but possible. Otherwise sign extension at the input is required, giving input codes 0 and all 1.
 

bardia

Newbie level 5
Joined
Sep 14, 2010
Messages
8
Helped
0
Reputation
0
Reaction score
0
Trophy points
1,281
Activity points
1,378
Thanks FvM,

The reason I said signed number, is that as long as I know we leverage the 2's complement wrap around in integrator sections to avoid overflow problem. I don't know how can we do it with unsigned number.

Also if we do not convert 0 to -1, there will be no subtraction in integrator sections. because when input is 0, nothing will be added or subtracted.

Finally I dont understand why you say converting 1b'0 to -1 = 16'hFFFF, and 1'b1 to +1 = 16'h0001, is like multiplying the input bit by two.

Please correct me if I am wrong.
Thanks in advance.
 

FvM

Super Moderator
Staff member
Joined
Jan 22, 2008
Messages
47,875
Helped
14,125
Reputation
28,507
Reaction score
12,809
Trophy points
1,393
Location
Bochum, Germany
Activity points
277,851
According to my suggestion, you'll add 0x0000 and 0xFFFF for 0 and 1, or vice versa.

Both signed and unsigned arithmetic will overflow (modular addition) as required by the CIC design. As said, signed and unsigned adder are just the same on the gate level, only the number interpretation changes.
 
  • Like
Reactions: bardia

    bardia

    points: 2
    Helpful Answer Positive Rating
Status
Not open for further replies.

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Top