Delta modulation and demodulation

Preliminary discussion
Despite its cryptic name, delta modulation (or –modulation) is a pulse
modulation scheme. Other examples of pulse modulation schemes that you may have
learnt more about using the Emona Telecoms-Trainer 101 include: pulse amplitude
modulation (PAM), pulse code modulation (PCM) and pulse-width modulation (PWM).

Delta modulation is like PWM in that it samples the analog input signal and
converts it directly into a serial stream of 1s and 0s without the need for a discrete
analog-to-digital conversion process. As such, delta modulation is also like PWM in that
the digital signal on the delta modulator’s digital output is a complex waveform that
includes a component at the message frequency.

Despite the similarity between delta modulation and PWM, it’s actually more
useful to describe its operation by comparing it to that of PCM. PCM samples the analog
input signal at fixed intervals and the samples are converted to fixed length sequential
binary numbers. Importantly, these binary numbers are directly (or, in some cases,
mathematically) proportional to the samples’ size.

Delta modulation effectively samples the analog input signal at regular fixed
intervals also. However, it then compares the present sample to the previous one and
outputs a 1-bit data word to indicate whether the new sample is smaller or larger than
the previous one.

The block diagram of a basic delta modulation system is shown in Figure 1 below.

+ D elt a mod.
I nput L imit er S ampler
signal out

-
Clock

I nt egr at or

Figure 1
 As the delta modulation system in Figure 1 is a feedback loop, a mathematical
analysis of its operation is complicated and not attempted here. For the purpose of this
exercise, think of its operation in the following way.

The Sampler’s output (which is also the modulator’s output) is a bi-polar pulse
train. That is, instead of using 0V and 5V for the two logic states, positive and negative
voltages are used (eg ±2.5V). This is done so that the Integrator produces a ramp
voltage that changes direction in response to changes in logic state on the Sampler’s
output.

The Integrator’s output is subtracted from the message using a Summing circuit
to produce an error signal. The polarity of the error signal is a function of the relative
absolute voltages of the analog input (the message) and Integrator’s output. So, if the
Integrator’s output is greater than the message, the error signal is one polarity. If the
Integrator’s output is smaller than the message the error signal is the opposite polarity.
This allows the modulator to “know” which of the two signals is bigger (though this
information is captured only at the moment of sampling – see below) and this is essential
for achieving an Integrator output that approximates the message. Specifically, the
Integrator’s output “hunts” the analog input voltage and so its positive or negative going
ramp voltages produce a sawtooth waveform that tracks the message signal’s shape
(and this can be seen in Figure 2 on the next page).

The Limiter continuously converts the status of the relative sizes of the sawtooth
and message waveforms to a logic state. That is, one logic state indicates when the
sawtooth waveform is bigger than the message but reverses the moment that it is
smaller. The Sampler samples this information to convert it to clocked data and level
translates it to the corresponding positive or negative voltage (mentioned above).
Importantly, by definition, sampling is a periodic process and so the Sampler’s output
only reflects the relative sizes of the sawtooth and message waveforms at the sampling
moment.

Now let’s put this together. At switch-on, the Sampler’s output starts at either a
positive or negative voltage which commences a ramp on the Integrator’s output. Now,
suppose that the Integrator’s output is smaller than the message (a likely proposition)
but travelling in the same direction. This produces an error signal on the Summer’s
output with a particular polarity that is immediately converted to a particular logic state
by the Limiter. This logic state is captured on sampling and a decision is made by the
Sampler to maintain its output voltage. In other words, nothing changes.

These conditions are maintained for every sample until such time as the absolute
ramp voltage exceeds absolute message voltage (either because the ramp catches up
with the message or the message reverses direction). Once this happens, the polarity of
the error signal reverses which immediately causes the logic state on the Limiter’s
output to reverse. This new logic state is sampled and captured by the Sampler forcing to
reverse its output voltage, in turn causing the Integrator’s ramp to reverse direction.
These new conditions are now maintained for every sample until such time as the ramp
returns to an absolute voltage that is smaller than the absolute message voltage.

Experiment 16 – Delta modulation and demodulation

16-2 © Emona Instruments
 Figure 2 below shows an example of the sawtooth waveform that would be
generated on the Integrator’s output and the digital signal on the Sampler’s output for a
sinusoidal message to a delta modulator. (Note: The clock frequency of the Sampler in
this example is three times the message frequency.)

Figure 2

There is an interesting point to observe about these signals. At first glance, it

appears as though the sampling rate is irregular. However, this is not the case. In
instances where there are wider than normal gaps between the changes in the ramp’s
direction, sampling has taken place. But, as the relative relationship between the
sawtooth waveform and the message is the same as the previous sample, there is no
change of direction in the sawtooth waveform and therefore no change in logic state on
the Sampler’s output.

Finally, Figure 1 shows a discrete amplifier section between the Sampler and the
Integrator to control the loop gain. In practice the amplification can be incorporated into
the Integrator or Summer circuits.

Experiment 16 – Delta modulation and demodulation

16-3 © Emona Instruments
The experiment
For this experiment you’ll use the Emona Telecoms-Trainer 101 to implement the
delta modulation scheme. You’ll then experiment with adjusting the Integrator’s step size
to observe the effect on the delta modulator’s slope overload and granularity
characteristics. Finally, you’ll demodulate the delta signal using low-pass filtering and
investigate the quality of the recovered message qualitatively and quantitatively using
SNDR.

It should take you about 50 minutes to complete this experiment.

Procedure

Part A – Setting up the delta modulator

Part A of this experiment gets you to implement the delta modulator using the
Emona Telecoms-Trainer 101.

1. Gather a set of the equipment listed on the previous page.

2. Locate the Adder module and set its G and g controls to the middle of their travel
(which should have the arrowheads pointing to the 12 o’clock position).

3. Connect the set-up shown in Figure 3 below.

M A S TER A D D ER U T IL ITIES M U LTIP L IER

S IGN A L S
C OM P A R ATOR
R EF

X DC

1 0 0 kH z IN OU T Y DC kX Y
S IN E R EC TIFIER S ER IA L TO
1 0 0 kH z G P A R A L L EL
C OS
1 0 0 kH z A GN D S/ P
D IGITA L D IOD E & R C L P F
8 kH z
D IGITA L
2 kH z S ER IA L X1
D IGITA L RC LP F
2 kH z g
S IN E
B GA +gB CLK X2

Figure 3

This set-up can be represented by the block diagram in Figure 4 on the next page.
The Master Signals module’s 2kHz SINE output models the message. The remaining
blocks implement the delta modulator.

Experiment 16 – Delta modulation and demodulation

16-4 © Emona Instruments
 M ast er A dder S er ial- t o- Par allel
S ignals module Compar at or Conver t er module

2 kH z A X1 D elt a mod
S/ P out put
IN

B CL K

10 0 kH z

M ast er
RC L PF
S ignals

Figure 4

Relating this functional block diagram to the generic block diagram of a delta
modulator in the preliminary discussion (Figure 1), the Utilities module’s RC LPF models
the Integrator, the Adder module implements the Summer, the Utilities module’s
Comparator implements the Limiter, and the Serial-to-Parallel Converter module models
the Sampler. Loop gain is provided and controlled by the Adder module.

A few important points should be made here before you continue with the
experiment. First, as the Telecoms-Trainer 101 doesn’t have a Sampler, the Serial-to-
Parallel Converter module is used in its place. Although this module is designed primarily
for another purpose, it has a bipolar output and provides a clocked version of the data on
its input which are the two key functions required of the Sampler in a delta modulator.
Second, the modulator’s sampling and data rate correspond with a 50kHz clock and not
the 100kHz clock connected to the Serial-to-Parallel Converter module. This is because
the module sends the data to alternate outputs on every clock pulse so the data rate of
each output corresponds with a 50kHz clock. Third, the accumulated phase shifts around
the loop return a signal to the Adder module’s B input that is inverted relative to the
message. This facilitates the delta modulator’s requirement for a subtraction of the
Integrator’s output from the message using the Adder module (in the same way that
adding a negative number to a positive number results in a subtraction of the two).

Experiment 16 – Delta modulation and demodulation

16-5 © Emona Instruments
 4. Connect the scope’s Channel 1 input to the set-up as shown in Figure 5 below.

M A S TER A D D ER U T IL IT IES M U LT IP L IER

S IGN A L S
C OM P A R AT OR
R EF

X DC

1 0 0 kH z IN OU T Y DC kX Y
S IN E R EC T IFIER S ER IA L T O
1 0 0 kH z G P A R A L L EL
C OS
1 0 0 kH z A GN D S/ P
D IGITA L D IOD E & R C L P F
8 kH z
D IGITA L
2 kH z S ER IA L X1
D IGITA L RC LP F
2 kH z g
S IN E
B GA +gB CLK X2

Figure 5

5. Set up the scope per the instructions in the Appendix.

Note: Ensure that the scope’s Trigger Source control is set to the CH 1 (or INT)
position.

6. Adjust the scope’s Timebase control to view two or so cycles of the message.

7. Accurately measure the message signal’s peak-to-peak amplitude and make a

note of the value.

Experiment 16 – Delta modulation and demodulation

16-6 © Emona Instruments
 8. Modify the set-up as shown in Figure 6 below.

N OIS E
GEN ER ATOR

0 dB

-6 dB

-2 0 dB

B U F FER

GA IN

IN OU T

M A S TER A D D ER U T IL IT IES M U LT IP L IER

S IGN A L S
C OM P A R AT OR
R EF

X DC

1 0 0 kH z IN OU T Y DC kX Y
S IN E R EC T IFIER S ER IA L T O
1 0 0 kH z G P A R A L L EL
C OS
1 0 0 kH z A GN D S/ P
D IGITA L D IOD E & R C L P F
8 kH z
D IGITA L
2 kH z S ER IA L X1
D IGITA L RC LP F
2 kH z g
S IN E
B GA +gB CLK X2

Figure 6

9. Use the Buffer module’s GAIN control to set its output to the same amplitude that
you measured at Step 7.

Experiment 16 – Delta modulation and demodulation

16-7 © Emona Instruments
 10 Connect the scope’s Channel 2 input to the set-up as shown in Figure 7 below.
.

N OIS E
GEN ER ATOR

0 dB

-6 dB

-2 0 dB

B U FFER

GA IN

IN OU T

M A S TER A D D ER U TIL ITIES M U LTIP L IER

S IGN A L S
C OM P A R AT OR
R EF

X DC

1 0 0 kH z IN OU T Y DC kX Y
S IN E R EC T IFIER S ER IA L TO
1 0 0 kH z G P A R A L L EL
C OS
1 0 0 kH z A GN D S/ P
D IGITA L D IOD E & R C L P F
8 kH z
D IGITA L
2 kH z S ER IA L X1
D IGITA L RC LP F
2 kH z g
S IN E
B GA +gB CLK X2

Figure 7

The scope’s connections to the set-up can be represented by the block diagram in
Figure 8 on the next page. Notice that the Buffer module isn’t shown. This is because it’s
not contributing to the implementation of delta modulation. Instead, it’s being used to
invert the message so that it’s in phase with the Integrator’s output and the two signals
can be conveniently compared.

Figure 8
M essage
T o CH 1

A X1 D elt a mod
S/ P out put
2 kH z IN

B CL K
I nt egr at or ' s
out put
10 0 kH z
T o CH 2

11 Check that the scope’s two Vertical Attenuation controls are on the same setting
. (the 1V/div position is probably best).

Experiment 16 – Delta modulation and demodulation

16-8 © Emona Instruments
 12 Set the scope’s Mode control to the DUAL position to observe the signal on the RC
. LPF’s output as well as the message.

Note: Remember that the RC LPF is being used to model the Integrator.

13 Overlay the two signals and you should see that the RC LPF’s output tracks the
. message in much the same way as theoretically predicted in Figure 2 (in the
preliminary discussion).

Note 1: The RC LPF’s output will not be as good an approximation of the

message as the Integrator’s output in Figure 2 but it should exhibit similar
features. Reasons for the differences between them will be investigated in Part B.

Note 2: If the RC LPF’s output isn’t similar to the Integrator’s output in Figure 2,
you’ll need to double-check your wiring.

Experiment 16 – Delta modulation and demodulation

16-9 © Emona Instruments
 14 Modify the scope’s Channel 2 connection to the set-up as shown in Figure 9
. below.

N OIS E
GEN ER ATOR

0 dB

-6 dB

-2 0 dB

B U F FER

GA IN

IN OU T

M A S TER A D D ER U T IL IT IES M U LT IP L IER

S IGN A L S
C OM P A R AT OR
R EF

X DC

1 0 0 kH z IN OU T Y DC kX Y
S IN E R EC T IFIER S ER IA L T O
1 0 0 kH z G P A R A L L EL
C OS
1 0 0 kH z A GN D S/ P
D IGITA L D IOD E & R C L P F
8 kH z
D IGITA L
2 kH z S ER IA L X1
D IGITA L RC LP F
2 kH z g
S IN E
B GA +gB CLK X2

Figure 9

The new scope connection to the set-up can be represented by the block diagram in
Figure 10 on the next page.

Experiment 16 – Delta modulation and demodulation

16-10 © Emona Instruments
 M essage
T o CH 1

A X1 D elt a mod
S/ P out put
2 kH z IN
T o CH 2
B CL K

10 0 kH z

Figure 10

Question 1
The digital signal on the delta modulator’s output is a serial stream of 1-bit
numbers. What do these numbers represent?

Question 2
The digital signal on the delta modulator’s output is a complex waveform (that is,
it’s made up of many sinewaves). State the frequency of one of these sinewaves.
Tip: No calculations are needed to answer this question.

Question 3
What’s the significance of this sinewave?

Part B – Slope overload and granularity

You will have noticed at Step 13 that the RC LPF’s sawtooth output is not as good
an approximation of the message as the Integrator’s output in Figure 2. This is an
important issue with delta modulation because, the more closely the Integrator’s output
approximates the message, the easier it is to recover the message from the digital data
without distortion and noise.

There are several factors that the affect closeness of the Integrator’s output to
the message and these include: the integration rate, the loop gain and the sampling

Experiment 16 – Delta modulation and demodulation

16-11 © Emona Instruments
frequency. Part B gets you to investigate one of these and introduces you to two delta
modulation performance specifications called slope overload and granularity.

15 Return the scope’s Channel 2 input to the RC LPF’s output as shown in Figure 11
. below.

N OIS E
GEN ER ATOR

0 dB

-6 dB

-2 0 dB

B U F FER

GA IN

IN OU T

M A S TER A D D ER U T IL IT IES M U LT IP L IER

S IGN A L S
C OM P A R AT OR
R EF

X DC

1 0 0 kH z IN OU T Y DC kX Y
S IN E R EC T IFIER S ER IA L T O
1 0 0 kH z G P A R A L L EL
C OS
1 0 0 kH z A GN D S/ P
D IGITA L D IOD E & R C L P F
8 kH z
D IGITA L
2 kH z S ER IA L X1
D IGITA L RC LP F
2 kH z g
S IN E
B GA +gB CLK X2

Figure 11

16 Set the scope’s two Input Coupling controls to the GND position.
.

17 Align the traces then return the Input Coupling controls to the AC position.
.
Note: This overlays the RC LPF’s output with the message so that their centre
lines are aligned.

18 Look closely at the two signals and note when the sawtooth waveform changes
. back and forth across the message and when it doesn’t.

Experiment 16 – Delta modulation and demodulation

16-12 © Emona Instruments
 You’ll notice that the sawtooth waveform changes back and forth across the
message when the message’s rate of change is relatively slow (at the peaks in the case
of a sinewave). But, you’ll also notice that the sawtooth waveform doesn’t cross the
message when the message’s rate of change is relatively fast (at the sinewave’s zero
crossover points). This inability of the Integrator’s output to keep up with quick changes
in the message is called slope overload and is also shown in Figure 12 below.

S lope over load

Figure 12

Slope overload produces spectral components in the digital data signal at

frequencies near the copy of the message and so can distort it at the demodulator’s
output. To minimise slope overload, the Integrator’s “step size” must be increased. That
is, the amount that the ramp can change in a given sampling period must be increased.
This can be achieved in one of two ways. Either increase the integration rate (which we
cannot do here because this is a fixed property of the RC LPF) or increase the loop gain.
This second option is available to us.

19 Slowly increase the voltage gain of the Adder module’s B input by turning its g
. control clockwise and observe the effect on the RC LPF’s output.

Note: This increases the loop gain, in turn increasing the Integrator’s step size.

20. Stop increasing the voltage gain of the Adder module’s B input when its g control
is at the 3 o’clock position.

Experiment 16 – Delta modulation and demodulation

16-13 © Emona Instruments
To maintain the comparison between the Integrator’s output and the message, the Buffer
module’s gain must be reduced.

21 Reduce the Buffer module’s gain but stop the moment the Integrator’s sawtooth
. waveform crosses the message every time it changes direction.

Note: When the Buffer’s gain is the correct level, the two signals should look
similar to Figure 13 below.

Figure 13

Notice that the Integrator’s output now has several more “teeth” during the time
that the message signal changes quickly (at the sinewave’s zero crossing points) than it
did when the Integrator’s step size was smaller. With the bigger step size the Integrator’s
output can better track the message and so the slope overload is reduced.

However, a price has been paid. Notice now that the teeth in the Integrator’s
output at the time that the message signal changes slowly (at the sinewave’s peaks)
deviate further from the message on one side than they did previously. This is a
quantisation error problem which contributes to quantisation noise or granularity on the
demodulator’s output. Thus, obtaining an appropriate step size for the Integrator is a
compromise between overload distortion and granularity.

Question 4
Which is worst when the Integrator’s step size is relatively small?

 Slope overload
distortion
 Granularity

Question 5
Which is worst when the Integrator’s step size is relatively large?

 Slope overload
distortion
 Granularit
y
An alternative to managing the compromise between slope overload distortion
and granularity involves increasing the sampling rate. This allows the step size to be

Experiment 16 – Delta modulation and demodulation

16-14 © Emona Instruments
greater while preventing the steps from overshooting the message by so much when the
message changes slowly. From a spectrum perspective, increasing the sampling rate
forces the other sinewaves that make up the delta signal higher up the spectrum and so
easier to remove at demodulation. As such, oversampling is widely used with delta
modulation.

The sampling rate can be increased here by replacing the Master Signals
module’s 100kHz DIGITAL output with the VCO module’s DIGITAL output and setting the
VCO module’s output frequency to maximum. However, as the VCO module’s output and
Master Signals module’s 2kHz SINE output are not synchronised, the Integrator’s output
signal will be unstable on the scope’s display rendering the improvement difficult to
observe. This can be overcome by capturing one sweep of the display using a storage
scope if one is available.

Part C – Delta demodulation

Recall that delta modulation produces a pulse train with a spectral composition
that includes a copy of the message signal. It’s possible to recover the message using
low-pass filtering and Part C gets you to do this.

22. Locate the Tuneable Low-pass Filter module and set its Gain control to the middle
of its travel.

23. Turn the Tuneable Low-pass Filter module’s Cut-off Frequency Adjust control fully
clockwise.

24. Modify the set-up as shown in Figure 14 below.

M A S TER A D D ER U T IL IT IES M U LT IP L IER T U N EA B L E

S IGN A L S LP F
C OM P A R AT OR
R EF

X DC f C x1 0 0

1 0 0 kH z IN OU T Y DC kX Y
S IN E R EC T IFIER S ER IA L T O
1 0 0 kH z G P A R A L L EL
C OS fC
1 0 0 kH z A GN D S/ P
D IGITA L D IOD E & R C L P F
8 kH z
D IGITA L
2 kH z S ER IA L X1
D IGITA L GA IN
RC LP F
2 kH z g
S IN E
B GA +gB CLK X2 IN OU T

Figure 14
The entire set-up including the modification in Figure 14 can be represented by
the block diagram in Figure 15 below. The Tuneable Low-pass Filter module is used to
demodulate the delta signal and recover the 2kHz message.

Experiment 16 – Delta modulation and demodulation

16-15 © Emona Instruments
 M essage
T o CH 1
T uneable
L ow- pass F ilt er

A X1 Recover ed
S/ P message
2 kH z IN
T o CH 2
B CL K

10 0 kH z

M essage M odulat ion D emodulat ion

Figure 15

25. Slowly reduce the Tuneable Low-pass Filter module’s cut-off frequency until the
message has been recovered (ignoring phase shift).

Question 6
What change to the delta modulator can be made to allow simpler filters to be
used by the demodulator without introducing more noise and distortion?

Experiment 16 – Delta modulation and demodulation

16-16 © Emona Instruments
Experiment 16 – Delta modulation and demodulation
16-17 © Emona Instruments