Oversampling Delta-sigma modulation



fig. 5: noise shaping curves , noise spectrum in ΔΣ modulator



let s consider signal @ frequency





f

0





{\displaystyle \scriptstyle f_{0}}

, sampling frequency of





f


s






{\displaystyle \scriptstyle f_{\mathrm {s} }}

higher nyquist rate (see fig. 5). ΔΣ modulation based on technique of oversampling reduce noise in band of interest (green), avoids use of high-precision analog circuits anti-aliasing filter. quantization noise same both in nyquist converter (in yellow) , in oversampling converter (in blue), distributed on larger spectrum. in ΔΣ-converters, noise further reduced @ low frequencies, band signal of interest is, , increased @ higher frequencies, can filtered. technique known noise shaping.


for first order delta sigma modulator, noise shaped filter transfer function





h

n


(
z
)

=


[
1


z


1


]




{\displaystyle \scriptstyle h_{n}(z)\,=\,\left[1-z^{-1}\right]}

. assuming sampling frequency





f

s






f

0





{\displaystyle \scriptstyle f_{s}\,\gg \,f_{0}}

, quantization noise in desired signal bandwidth can approximated as:








n

0




=


e

r
m
s




π

3




(
2

f

0


τ

)


3
2





{\displaystyle \mathrm {n_{0}} \,=\,e_{rms}{\frac {\pi }{\sqrt {3}}}\,(2f_{0}\tau )^{\frac {3}{2}}}

.


similarly second order delta sigma modulator, noise shaped filter transfer function





h

n


(
z
)

=



[
1


z


1


]


2





{\displaystyle \scriptstyle h_{n}(z)\,=\,\left[1-z^{-1}\right]^{2}}

. in-band quantization noise can approximated as:








n

0




=


e

r
m
s





π

2



5




(
2

f

0


τ

)


5
2





{\displaystyle \mathrm {n_{0}} \,=\,e_{rms}{\frac {\pi ^{2}}{\sqrt {5}}}\,(2f_{0}\tau )^{\frac {5}{2}}}

.


in general,





n




{\displaystyle \scriptstyle \mathrm {n} }

-order ΔΣ-modulator, variance of in-band quantization noise:








n

0




=


e

r
m
s





π

n



2
n
+
1




(
2

f

0


τ

)



2
n
+
1

2





{\displaystyle \mathrm {n_{0}} \,=\,e_{rms}{\frac {\pi ^{n}}{\sqrt {2n+1}}}\,(2f_{0}\tau )^{\frac {2n+1}{2}}}

.


when sampling frequency doubled, signal quantization noise improved




10

log

10



(
2
)
(
2
n
+
1
)


d
b




{\displaystyle \scriptstyle 10\log _{10}(2)(2n+1)\,\mathrm {db} }







n




{\displaystyle \scriptstyle \mathrm {n} }

-order ΔΣ-modulator. higher oversampling ratio, higher signal-to-noise ratio , higher resolution in bits.


another key aspect given oversampling speed/resolution tradeoff. in fact, decimation filter put after modulator not filters whole sampled signal in band of interest (cutting noise @ higher frequencies), reduces frequency of signal increasing resolution. obtained sort of averaging of higher data rate bitstream.


example of decimation

let s have, instance, 8:1 decimation filter , 1-bit bitstream; if have input stream 10010110, counting number of ones, 4. decimation result 4/8 = 0.5. can represent 3-bits number 100 (binary), means half of largest possible number. in other words,



the sample frequency reduced factor of eight
the serial (1-bit) input bus becomes parallel (3-bits) output bus.






Comments

Popular posts from this blog

History The Vandals

Causes Portuguese conquest of the Banda Oriental

Publications Daniel Kolak