Analog to Digital Converters

Electronics Unit – Lecture 7

Representing a continuously varying

physical quantity by a sequence of
discrete numerical values.

03 07 10 14 09 02 00 04

LSU 10/28/2004 Electronics 7 1

 Conversion Methods
(selected types, there are others)

Ladder Comparison
Successive Approximation
Slope Integration
Flash Comparison

LSU 10/28/2004 Electronics 7 2

 Ladder Comparison

LSU 10/28/2004 Electronics 7 3

 Single slope integration
• Charge a capacitor at constant
20
current

Voltage accross the capacitor

18
16
• Vin
Count clock ticks 14
12
• Stop when the capacitor voltage 10
8
matches the input 6
4
• Cannot achieve high resolution 2
0
– Capacitor and/or comparator 0 2 4 6
Counting time
8 10 12 14 16
Tim e

Start
Conversion Start
S Q Enable
Conversion

- R

Counter
N-bit Output
C
IN
+ Oscillator Clk

LSU 10/28/2004 Electronics 7 4

 Successive Approximation

LSU 10/28/2004 Electronics 7 5

 Flash Comparison

If N is the number of bits in the

output word….
Then 2N comparators will be
required.
With modern microelectronics
this is quite possible, but will be
expensive.

LSU 10/28/2004 Electronics 7 6

 Pro and Cons
Slope Integration & Ladder Approximation
Cheap but Slow

LSU 10/28/2004 Electronics 7 7

 Pro and Cons
Flash Comparison
Fast but Expensive
Slope Integration & Ladder Approximation
Cheap but Slow

LSU 10/28/2004 Electronics 7 8

 Pro and Cons
Successive Approximation
The Happy Medium ??
Slope Integration & Ladder Approximation
Cheap but Slow
Flash Comparison
Fast but Expensive

LSU 10/28/2004 Electronics 7 9

 Resolution
Suppose a binary number with N bits is to
represent an analog value ranging from 0 to A

There are 2N possible numbers

Resolution = A / 2N

LSU 10/28/2004 Electronics 7 10

 Resolution Example
Temperature range of 0 K to 300 K to be linearly
converted to a voltage signal of 0 to 2.5 V, then
digitized with an 8-bit A/D converter

2.5 / 28 = 0.0098 V, or about 10 mV per step

300 K / 28 = 1.2 K per step

LSU 10/28/2004 Electronics 7 11

 Resolution Example
Temperature range of 0 K to 300 K to be linearly
converted to a voltage signal of 0 to 2.5 V, then
digitized with a 10-bit A/D converter

2.5 / 210 = 0.00244V, or about 2.4 mV per step

300 K / 210 = 0.29 K per step
Is the noise present in the system well below 2.4 mV ?

LSU 10/28/2004 Electronics 7 12

 Quantization Noise
Each conversion has an average uncertainty of one-
half of the step size ½(A / 2N)

This quantization error places an upper limit on the

signal to noise ratio that can be realized.

Maximum (ideal) SNR ≈ 6 N + 1.8 decibels (N = # bits)

e.g. 8 bit → 49.8 db, 10 bit → 61.8 db
LSU 10/28/2004 Electronics 7 13
 Signal to Noise Ratio
Recovering a signal masked by noise

Some audio examples

In each successive example the noise power is reduced
by a factor of two (3 db reduction), thus increasing the
signal to noise ratio by 3 db each time.

Example 1 Example 2 Example 3 Example 4

LSU 10/28/2004 Electronics 7 14

 Conversion Time
Time required to acquire a sample of the analog
signal and determine the numerical representation.

Sets the upper limit on the sampling frequency.

For the A/D on the BalloonSat board, TC ≈ 32 μs,

So the sampling rate cannot exceed about 30,000
samples per second (neglecting program overhead)

LSU 10/28/2004 Electronics 7 15

 Data Collection – Sampling Rate
The Nyquist Rate
A signal must be sampled at a rate at least twice that of the highest
frequency component that must be reproduced.

Example – Hi-Fi sound (20-20,000 Hz) is generally sampled

at about 44 kHz.

External temperature during flight need only be sampled

every few seconds at most.

LSU 10/28/2004 Electronics 7 16

 Activity E7a
®
Do the HuSAC
a party game for techies...

Human Successive Approximation Converter

LSU 10/28/2004 Electronics 7 17

 Activity E7b

Data Acquisition Using BalloonSat

LSU 10/28/2004 Electronics 7 18