0
$\begingroup$

I want to estimate the SNR of a bipolar signal. I know that each bit increases the SNR at 6 dB. Do i have to subtract one bit for sign, so that for bipolar signals the SNR is (n-1) x 6 dB ?

$\endgroup$
1
$\begingroup$

Your question appears to be about the signal to quantization noise ratio for uniform quantization. You have to count each bit, but that generally doesn't mean that the SNR is $6N$ dB, where $N$ is the number of bits. You normally have a constant offset:

$$\text{SNR}_{\text{dB}}\approx c+6N\tag{1}$$

where the constant $c$ depends on the Crest factor of the signal.

To give you an example, assume that the signal is in the range $[-x_p,x_p]$, and assume uniform quantization with $N$ bits. Then the quantization step size is

$$q=\frac{2x_p}{2^N}\tag{2}$$

If we further assume that the quantization error is uniformly distributed in the interval $[-q/2,q/2]$, the quantization noise power is

$$P_n=\frac{1}{q}\int_{-q/2}^{q/2}\xi^2d\xi=\frac{q^2}{12}=\frac13\frac{x_p^2}{2^{2N}}\tag{3}$$

With $x_{\text{RMS}}$ the RMS value of the signal, the signal to quantization noise ratio is given by

$$\text{SNR}=\frac{x^2_{\text{RMS}}}{P_n}=3\frac{x^2_{\text{RMS}}}{x^2_p}\cdot2^{2N}=\frac{3}{C^2_x}\cdot2^{2N}\tag{4}$$

where $C_x$ is the signal's Crest factor. From $(4)$, the SNR in dB is

$$\text{SNR}_{\text{dB}}=10\log_{10}\left(\frac{3}{C^2_x}\right)+20N\log_{10}2=10\log_{10}\left(\frac{3}{C^2_x}\right)+6.02\cdot N\tag{5}$$

which has the form of Eq. $(1)$.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.