Adaptive Delta Modulation, a modification of Linear Delta Modulation (LDM) is a scheme that circumvents the deficiency of DM. In ADM step size Δ of the Quantizer  is not a constant but varies with time , we shall express Δ as $\Delta&space;(n)=&space;2&space;\delta&space;(n)$ .

where $\delta&space;(n)$ increases during a steep segment of input and decreases for a slowly varying segment of input.

The adaptive step size control which forms the basis of an ADM scheme can be classified in various ways such as

• Discrete or Continuous.
• instantaneous (or) syllabic (fairly gradual change).
• forward (or) backward.

we shall describe an adaptation scheme that is backward, instantaneous and discrete in practical implementation , the step size $\delta&space;(n)$ is constrained in between some pre-determined minimum and maximum values.

$\delta&space;_{min}&space;\leq&space;\delta&space;(n)\leq&space;\delta&space;_{max}$

The upper limit $\delta&space;_{max}$ controls the amount of Slope over load distortion and the lower limit $\delta&space;_{min}$ controls the granular noise (or) Idle noise.

The adaptive rule for  $\delta&space;(n)$can be expressed in the general form $\delta&space;(n)&space;=&space;g(n)&space;\delta&space;(n-1)$

where the time varying gain g(n) depends on the present binary output b(n) and M previous values b(n-1),b(n-2) ……….b(n-M). The algorithm is usually initiated with $\delta&space;_{min}$.

when M=1, b(n) and b(n-1) are compared to detect probable slope over load {b(n) = b(n-1)} (or) probable granularity  {b(n) ≠ b(n-1)} then g(n) is

• $g(n)&space;=&space;P$ if     $b(n)&space;=b(n-1)$.
• $g(n)=\frac{1}{P}$ if   $b(n)&space;\neq&space;b(n-1)$.

when $b(n)&space;=b(n-1)$ Slope overload distortion is detected and when $b(n)&space;\neq&space;b(n-1)$ Idle noise is detected.

where $P\geq&space;1$, note that $P=1$ represents LDM. $P_{optimum}=1.5$ minimizes the Quantization noise for speech signal, where $1<&space;P<&space;2$ is for broad class of signals.

(No Ratings Yet)