number of bit of noise after division operation

[childs72]

Hi,

I am working on digital RTL design, however, I have to overcome this technical problem (which I believe it came under DSP subject) in order to hit the spec:
1. I have inputs from ADC (A & B), says, 10 bits data which consists 2 bits of random noise. I perform a division Z=(A*2^10)/B. In this case, what will be the number of bits that consists noise in Z?
2. From my understanding, if I perform an addition Y=A+B, I shall get Y with only 1 bit of noise, because noise in A & B are being kinda "cancel off each other" (I do not know the proper term for this, anyone who knows may correct me), is this correct?

Thanks.

 

[MagicianT]

Hi,

I am working on digital RTL design, however, I have to overcome this technical problem (which I believe it came under DSP subject) in order to hit the spec:
1. I have inputs from ADC (A & B), says, 10 bits data which consists 2 bits of random noise. I perform a division Z=(A*2^10)/B. In this case, what will be the number of bits that consists noise in Z?
2. From my understanding, if I perform an addition Y=A+B, I shall get Y with only 1 bit of noise, because noise in A & B are being kinda "cancel off each other" (I do not know the proper term for this, anyone who knows may correct me), is this correct?

Thanks.

1). sqrt( 2-bits).
2).No, not 1-bit, again sqrt( 2). To get 1-bit in noise reduction you need 4 samples.

 

[FvM]

Division is a nonlinear operation. You need to refer to propagation of uncertainty calculation methods to determine the effective result noise. E.g. if B is small, Z noise will be very high.