Analog-to-Digital Conversion

The real world is not digital. Consider temperature fluctuation as an example, it generally moves within some range of values without making large abrupt changes. We can measure aspects of our world like temperature, light intensity, forces, or whatever using analog sensors. In a digital device, the resulting signals are stored as sequential digital data. Consider the analog signal depicted below.

Image credit: Tod Kurt

In a digital device, we discretize the input signal range (i.e., -VRef to 5 +VRef) into different voltage levels or intervals called states. The number of states is the resolution of the digital representation. Common resolution values range from 256 states (i.e., stored in 8 bits) to 4,294,967,296 states (i.e., stored in 32 bits). The PIC24 offers either 10 bits (the default) or 12 bits to store the state. Assuming the 10 bit default state representation with -VRef=0 and +VRef=3.3, the smallest voltage change that can be measured is (3.3-0)/2¹⁰=3.3/1024 = 3.2mV. These calculations are expressed in your text as follows:

The translation of analog voltage values into different states is called Analog-to-Digital Conversion or ADC. For a one bit representation the process might be accomplished as follows:

A 10 bit representation can be created by extending this "flash" conversion process as depicted below.

Of course such a conversion would be quick (1 clock cycle), but very costly from a hardware perspective (e.g., consider the number of comparators needed for 10 bits---2^N-1 comparators). An alternative approach is that of "successive approximation."

At the start of the successive approximation conversion, the input signal is placed in the hold mode, and the internal DAC Voltage (VDAC) is set to mid-scale(i.e., Vref/2). A comparator determines whether the input voltage is above or below the VDAC, and the result (bit 1, the most significant bit of the conversion) is stored in the successive approximation register. The VDAC is then set either to ¼ scale or ¾ scale (depending on the value of bit 1), and compared to the input signal to set bit 2 of the conversion. The result is stored in the register, and the process continues until all of the bits of the digital representation have been determined. The full conversion takes N (i.e., the number of bits set) clock cycles. Your text represents the process as follows, where TAD refers to clock cycles:

Although slower than "flash" ADC, "successive approximation" is less expensive and provides high precision. It is the ADC process of choice for many applications including the PIC24.

The built-in ADC features of the PIC24 are as follows:

• Successive approximation
• 10-bit (default) or 12-bit resolution
• Reference voltage can be Vdd (3.3 V) or separate voltage (min AVSS + 2.7 V)
• Multiple input (more than one input channel)
• Clock source for ADC is either a divided Fosc, or and internally generated clock. The ADC clock period (TAD) cannot be less than 76 ns for 10-bit mode, or 118 ns for 12-bit mode. The internally generated clock has a period of ~ 250 ns (~ 4 MHz).