|
Analogue to
Digital Converters Explained
Bill Naylor, Electronworks Ltd
This article explains the basics of analogue (analog) to
digital converters, how they digitise an input
signal, what output formats they have and how to
write the software that drives them. You can use one
of our electronic kits to evaluate the theory.
If you have any suggestions for improving this
application note, please drop us a line at:
enquiries@electronworks.co.uk.
Introduction
This article describes how an analogue to digital
converter (ADC) works, where you use them, some of
the tricks associated with designing with them and
some fun you can have with them. As always, there is
a series of kits to go with this application note,
so you can put the theory into practice.
What is an ADC? This is a component that converts a
real world analogue signal (a voltage that can have
an infinite number of values) into a number that can
be understood by a microprocessor. Supposing you
want to design a digital thermometer…. You may have
a sensor that outputs, say 1V per °C. You may choose
to have a microprocessor that will then, say, drive
the LCD display, provide you with a number of
temperature ranges, provide alarm functions, go to
sleep if after a certain time and a whole host of
other features. But initially, you need to get the
analogue voltage into a form that the microprocessor
understands. This is where the ADC comes in.
The ADC
The part under the microscope is the TI TLC549. Why
did we pick this one? Well it is a simple low speed
ADC. It works the same as about 90% of the standard
ADCs on the market and does exactly what we need to
demonstrate how these components work. Really the
only thing that distinguishes one ADC from another
is resolution (how precise it is) and its speed (how
many samples per second it can do).
Terminology
ADCs have a language of their own. Once mastered,
you can appear like an ADC expert and confound and
confuse even the experts and even if you are not too
sure what you are talking about, you can at least
sound like you do!
Resolution
This is the first thing to get your head around.
Resolution is a measure of how many steps the ADC
can digitise the signal to. This is similar to a
ruler being able to measure in 1mm steps. Anything
finer than 1mm cannot be measured. Resolution is
specified in ‘bits’ and the TLC549 is an 8 bit ADC.
This means it can digitise to 28 (or 256)
steps. So a 5V input can be measured to an accuracy
of 5/256 steps (or 19.53mV). A 10 bit ADC can
measure to an accuracy of 210 (1024)
steps and a 12 bit ADC can measure to 4096 steps.
With a 12 bit ADC, you can measure a 5V signal to an
accuracy of 5/4096 steps or 1.22mV. Obviously if you
have a very small signal, you need an ADC with a
high resolution. 8 bit ADCs are about the lowest
resolution you can get and ADCs can go up to over 24
bits (digitising to over 16 million steps).
Speed
The ADC takes a snapshot of the input voltage and
converts it. How quickly it does this before being
able to take the next snapshot is a measure of its
speed.
Interface
This can either be serial or parallel. A parallel
interface outputs each binary bit on a pin. With an
8 bit ADC, you have 8 output pins representing
binary values from 27 (the most
significant bit, or MSB) to 20 (the
least significant bit, or LSB). These can
either be connected to a microprocessor, LEDs,
output port or another component/piece of test
equipment. If all your outputs are logic 1, the ADC
is outputting a code of 255 (1x27 + 1x26
+1x25 etc…). If all your outputs are 0V,
the ADC is outputting a code of 0 (or the part is
broken!)
Later ADCs (in the last 15 years or so) have
implemented a serial interface. Serial interfaces
mean less pins and smaller, cheaper devices. If you
are designing a small data logger, you do not want a
massive data converter taking up loads of room.
Serial devices are the way ahead and the TLC549 is a
serial output ADC.
A serial interface has 3 pins (mostly). One pin is
the chip select (CS). This turns on the device and
tells the part to get ready to digitise the input.
If you have a portable design that needs to run on
very low power, you only want your ADC turned on
when you need it. Some devices use the CS to start
the data conversion; some only use it to turn on the
device.
You then have the Data pin. This outputs the data
(duh!), normally MSB first. Then there is the clock
line. You take the clock line high (to 5V) and the
first data bit appears on the data line. You take
the clock line low then high again and the next bit
appears. Repeat this 8 times and you have clocked
out all your data and used 3 pins instead of 8. Big
deal you may say… However, imagine if you are using
a 12 bit ADC, you would use 3 pins instead of 12 and
with a 24 bit ADC you use 3 pins instead of 24. The
savings in package size start to get significant.
Have a look at the software section at the end to
see an example piece of code for clocking data out
of the TLC549.
Internal/External Reference
To get its result, the ADC needs to compare the
input voltage with a known stable voltage. This is
the reference. Some ADCs have this built in
(although they do not tend to be too accurate). Some
ADCs allow you to connect an external reference, so
you can pick what voltage you compare the input
signal to, how stable the voltage is etc. More of
this later…
Sample and Hold
The ADC needs a constant voltage to work on. If you
were half way through a mathematical calculation and
someone changed the numbers for you, you would be
pretty annoyed. The ADC is no different. It makes
sure it has a constant voltage by having a tiny
capacitor inside which stores the input voltage for
the ADC ‘engine’. Between the capacitor and the
input pin is a switch inside the chip that isolates
the ADC engine from the outside world. When you tell
the ADC to start converting, the switch opens
isolating the input voltage from the ADC engine and
the ADC works on the voltage on the capacitor.
Simple!
How does an ADC work?
As with measuring anything, you have to measure with
respect to something. Measure someone’s height and
you measure with respect to the ground they are
standing on. Measure the length of a road and you
need a start point. The same is true with an ADC.
You need an accurate, stable voltage, called a
reference voltage. The ADC looks at the reference,
then looks at the input and compares the two. Before
we explain this in more detail, we need a quick
lesson in binary. If you understand binary you can
skip the next section.
A binary number is made up of 1’s and 0’s each
carrying a different weight. As with decimal,
we go from 0 to 9 then add an extra digit to make
10. We then go from 10 to 99 then add a third digit
to make 100. With decimal, each digit we add is 100,
or 101, or 102
With binary we only have 1 and 0. Makes things
tricky, but hey, that is life!
So we have 0 then 1 then we have to add an extra
digit to get 10 (equals 2), then 11 (equals 3), then
we add a third digit to get 100 (equals 4), then
101, 110 and 111 (equals 7). With binary, each digit
we add is 20, 21, 22,
23 etc.
So to convert a number from decimal to 8 bit binary,
start with 27 (= 128) and see if this
number is larger or smaller than the number we need
to convert. If it is larger, put a 0 in that column
and move to 26. If 128 it is smaller, put
a 1 in that column and subtract 27 (128)
from your answer. Move on to 26 (= 64).
Repeat with 26 but this time subtracting
26 (64) if our number is larger than 64.
Let’s throw some examples at you to see if this
clears the water…
27 26
25 24 23
22 21 20
0 = 0 0 0
0 0 0 0
0
1 = 0 0 0
0 0 0 0 1
2 = 0 0 0
0 0 0 1 0
3 = 0 0 0
0 0 0 1 1
Now lets try 154 (a random number I picked).
Take 27 (=128). This is smaller than 154,
so we put a 1 in the 27 column and
subtract 128 from our result. My calculator tells me
this leaves 26.
Move on to 26 (=64). This is higher than
26, so we put a 0 in the 26 column. Move
on to 25 (= 32). This is higher than 26
so we put a 0 in the 25 column. Move on
to 24 (=16). Hey! This is smaller than
26, so put a 1 in the 24 column, subtract
16 (=10) and move on. Repeat this until you end up
with the following code:
27 26
25 24 23
22 21 20
1 0 0
1 1 0 1 0
So 154 decimal is equal to 10011010.
To make life easy, the Calculator function in
Windows does this for you.
If you need a number higher than 255, you have to
increase your digits (9 bits takes you up to 511, 10
bits takes you up to 1023), just like the ADC
Resolution description above.
With an ADC, we are comparing an input voltage with
a reference voltage and converting the result into a
number. With an 8 bit ADC, we are converting a
voltage to a number between 0000 0000 and 1111 1111
(in binary). So 0V input gives an output of 0000
0000 and an input voltage equal to the reference
voltage gives an output code of 1111 1111
Now here is an important bit to note: In decimal
(base 10) if you want to divide something by 10 (the
number of the ‘base’) you shift all the numbers
right by one digit. So 100 shifted right gives 10.
If you want to multiply something by 10 you shift
all the numbers left by 10.
In binary (base 2), it is same, but shifting the
numbers right or left by one digit divides or
multiplies by 2 (the number of the ‘base’). As an
example, 100 (=4) shifted right by one digit gives
10 (=2). 100 (=4) shifted left by one digit gives
1000 (=8). So in binary, it is very easy to multiply
and divide numbers by 2 (or multiples of 2).
Therefore with the ADC, if you input a voltage equal
to the reference voltage, the output code will be
1111 1111 (=255). If you input a voltage equal to
half the reference voltage, the output code will be
0111 1111 (=127) (the full scale code shifted right
one bit). The smarter reader will notice that half
of 255 is 127.5 and in fact if you input a voltage
equal to half the reference voltage, the ADC will
flicker between a code of 0111 1111 and 1000 0000
(=128).
OK, so back to the original question: How does an
ADC work..?
Well it does exactly what we have done above. It
looks at the input voltage and compares it to
half the reference. Thus it can establish if
the input voltage is in the upper or lower half of
the voltage range – see FIG 1. If the voltage is
higher than half the reference voltage (2.5V), it
sets the Most Significant Bit (27) high.
So in our table above the 27 column will
have a 1 in it. We then know that the input voltage
is in the top half of the input range (between 2.5
and 5V). We now need to split this range into half
again and repeat the process to see if our input
voltage is in the range (2.5 to 3.75V) or (3.75V to
5V). We repeat this until we have ‘homed in’ on our
input voltage.
Let us assume our input voltage is 3.4V and we are
using a 5V reference. The arrow represents our input
voltage. The ADC starts by setting the comparison
voltage to half the reference voltage (=2.5V) and
compared this with the input voltage to see which
half our input voltage lies in. Our input voltage is
higher than this, so a 1 is set in the Most
Significant Bit column.
We then need to determine if the input voltage is
between 2.5V and 3.75V or 3.75V and 5V.
We set the reference voltage equal to 3.75V (half
way) and do the comparison again. Our input voltage
is lower than 3.75V, so we put a 0 in the next
column (26). We know our input voltage is
somewhere between 2.5V and 3.75V. We set the
reference to 3.125V (half way) and do the comparison
again. Our input is higher than 3.125, so we put a 1
in the next column, then set the reference voltage
to the half way point and repeat the comparison…..
Actually, this varying comparison voltage is
generated using a Digital to Analogue Converter, or
DAC. This component converts a digital code into an
analogue voltage (the opposite of an ADC). See our
article on Digital to Analogue Converters
Explained for a more detailed lowdown.
As you can see, by continuously setting the
reference point to the half way point, we are making
the comparison voltage ‘home in’ on the input
voltage. Note that the above process is exactly the
same as earlier where we compare our number with 27
then subtract 27 and compare with 26,
then subtract 26…. etc
FIG 1
The ADC in FIG 1 has only 8 steps so is a 3 bit ADC
(23 = 8). Our ADC is 8 bit so has 256
steps, so as you can imagine, the steps between each
different voltage are a lot smaller (19.5mV) and FIG
1 would look a lot more crowded, so I hope you get
the idea with my 8 step example.
In fact, it takes as many steps to reach the final
output code as there are bits, so an 8 bit ADC takes
8 steps to reach the final result. This form of ADC
produces good accuracy with a fast conversion time.
Software
Here is a snippet of code that will enable you to
clock data out of the TLC549. Please note: on the
first conversion, you will end up with garbage. You
have to do 2 conversions to clock out valid data.
sample_adc
movlw 0x08
; load counter with 8
movwf count
bcf PORTA, cs
; take cs low
loop1
bcf STATUS, C ;
precondition carry bit
btfsc PORTA, din
; look at data line
bsf STATUS, C ;
set carry
rlf adc_data,
f ; rotate carry into LSB
bsf PORTA, sclk
; set clock line high
bcf PORTA, sclk
; set clock line low
decfsz count, f
goto loop1
bsf PORTA, cs
; take cs high
; we have now extracted a byte of data from the ADC
retlw 0
|
