QucsStudio-

Tutorial
Part 2:
Harmonic Balance
Simulations

Version 1.0
Author:
Gunthard Kraus, DG8GB,
Guest Professor at the Duale Hochschule Baden-Wuerttemberg (Friedrichshafen /
Germany), assisted by the Programs Author Michael Margraf, DD6UM

Email: mail@gunthard-kraus.de
Homepage: www.gunthard-kraus.de

January 1st, 2016

1
Table of Contents Page
1. Harmonic Balance: Why? 5

2. Harmonic Balance: How? 5

3. First Example: a FET Amplifier Stage explored by a Parameter Sweep 7

3.1. Preparations 7
3.2. Simulation of the Relation between Forward Transmission S21 and
Supply Voltage 10
3.3. The„1 dB Compression Point“ 11
3.4. Linear Presentation of the Transfer Characteristic for different values of
the Power Supply Voltage Vdd 12

4. Second Example: Examining this FET Amplifier Stage for a

fixed Operating Point 12
4.1. The Simulation Schematic 12
4.2. Ideal an real Transfer Characteristic 13
4.3. Linear Presentation of S21 for different Input Voltage Amplitudes 14
4.4. Presentation of S21 in dB (including 1 dB compression point) for different
Input voltage Amplitudes 14
4.5. Rise of Harmonics for increasing Input Voltage Amplitude (linear Presentation) 15
4.6. Rise of Harmonics for increasing Input Voltage Amolitude (presentation
in dBm at 50Ώ) 16
4.7. The complete Output Spectrum 16
4.8. The IP3 Point 17
4.8.1. Fundamentals 17
4.8.2. Simulation of IP3 using the Transfer Characteristic 18
4.8.3. Simulation of IP3 using the Output Spectrum 22

5. A „Half Complex Mixer“ to generate an SSB-Signals 24

5.1. Analytic Pairs 24
5.2. Example: a Half Complex Mixer to generate an SSB Signal 24
5.3. Parameter Simulation of the „undesired Side Band's Rejection“ for different
Phase Errors 29

6. A Double Balanced Mixer (DBM) 31

6.1. The Schematic (coming from chapter 14.3. in part 1 of the tutorial) 31
6.2. Used Schematic for the Harmonic Balance Simulation 32
6.3. Harmonic Balance Parameter Sweep of the LO Level 33
6.3.1. Output Spectrum for different LO Amplitudes 33
6.3.2. Presentation of the Conversion Loss 34
6.3.3. Addition of a Diplexer to the Mixer Output 35
6.3.4. Finding the IP3 Point for a LO Peak Voltage of 2V 37

7. Summary: Harmonic Balance Simulation in Question and Answer

for an Amplifier Stage 40
7.1. Which Circuit? 40

7.2. How can I quickly find every DC Voltage of the Operating Point? 40

2
 7.3. What are the necessary Preparations for a successful HB Simulation? 40

7.4. How can I present the Input Voltage and the Output Voltage including
the gain (= linear / in dB / in dBm)? 41

7.5. I want to see the Signal to Noise Ratio for f = 1 MHz at the Output in db 42

7.6. I want to simulate the Output Voltage versus the Input Voltage. How does the
Schematic look like for this Task?s task? 43

7.7. Parameter Sweep of the Input Voltage: how do the Output Voltage and the
Transfer Function look like? 43

7.8. Parameter Sweep of the Input Voltage: at the output I would like to see
the fundamental frequency and two Harmonics 45

7.9. Parameter Sweep of the Input Voltage: how can I evalute the
Output IP3 Point? 45

7.10. Parameter Sweep of the Input Voltage: how does the noise voltage at the
output vary with an increasing Input Voltage? 46

7.11. Parameter Sweep of the Input Voltage: how can I simulate the
Signal to Noise Ratio in dB at the Output? 47

7.12. Parameter Sweep of the Input Voltage: how can I determine the
Signal to Noise Ratio in dB for a Bandwidth of 1000 Hz? 48

7.13. Parameter Sweep of the Frequency: Preparation of the HB Simulation 49

7.14. Parameter Sweep of the Frequency: curves of Output Voltage and Gain 50

7.15. Simultaneous Parameter Sweep of Frequency AND Input Voltage 51

3
4
1. Harmonic Balance: Why?

Simulations in the Time Domain (e.g. with SPICE) give a lot of information concerning the curves of currents
and voltages and an FFT provides all information to the new frequency content caused by the nonliearities of
the parts in the circuit. But this is always only a „snapshot“ because all input signal frequencies are held
constant during the simulation.

An „ACSweep“ ignores every nonlinearity and shows only the Transfer Functions for sine wave input
signals with constant amplitude.

So you find a „Simulation gap“ if you want an AC sweep and afterwards present the „nonlinearities
over frequency“ for different amplitudes or frequencies of the input signals.
This gap is now closed by the „Harmonic Balance Simulation“ which is a free part of qucsstudio.
But remember: Harmonic Balance is always a Frequency Domain Simulation!

2. Hamonic Balance: How?

This is an ingenious trick and thus patented.
If a circuit combines linear parts (like resistors, capacitors, coils..) and nonlinear parts (like diodes,
transistors, FETs) the program acts as follows:

a) Linear Parts (including connections and nodes) are collected in a „linear subcircuit“.

b) All the nonlinear parts (including connections and nodes) are collected in a „nonlinear subcircuit“.

c) Now the two subcircuits (or „blocks“) are connected together by exactly named connection lines.
Now we have again our correct circuit and the input signals can be applied.

See this nice illustration found in [1]:

The answer to the question „why all these efforts?“ is very simple:

Everything in the linear subcircuit can at once be calculated in the Frequency Domain without any
difficulties or problems.

All calculations and simulations in the nonlinear subcircuit can (after a FFT) be done in the Time
Domain. So you get as a result every „current curve distortion“ which can be FFT re-transformed into
a frequency spectrum for every connection between linear and nonlinear subcircuit. These results
are sent back to the linear subcircuit.

There is a lot to calculate....because the user must at first tell the „maximum order of the regarded
harmonics“ in the nonlinear subcircuit. Then the program starts to calculate and compares the „energy
entry from the linear subciruit into the nonlinear subcircuit (for every connection!) to the energy
„which ist given back from the nonlinear subcircuit to the linear subcircuit in form of a frequency
spectrum“. If no identy is reached then a new trial with altered values will be started.
This needs a lot of iterations up to the point of a correct „Harmonic Balance“.....and thus the user must also
enter (before starting HB simulation) the maximum number of trials and the maximum allowable error at the
end of the simulation.

Theory and mathematical background are complicated. If somebody is interested in this: please read [1]..

But the possibilities and results of a successful Harmonic Balance Simulation are fascinating. Additionally
working with a „Parameter Sweep“ you have a huge stock of data and results. You can calculate the
relationship between the input voltage and the distortions over the frequency, the 1 dB compression point,
the IP3 point, the spectrum of generated harmonics.....and so on...and so on....

Let uns start with an example which can be downloaded from the qucsstudio homepage, named

„FET_P1dB.sch“
There you find a collecton of „examples“, followed by „HB-Analysis“. Download and extract the packjage
using the path „Project / Extract package“.

6
3. First Example: a FET Amplifier Stage explored by a Parameter Sweep
3.1. Preparations
At first check the identity of your downloaded example and this schematic.

The next steps can be found in the above schematic. Please open „edit“ in the Menu taskbar and
click on

Deactivate / Activate
Now you find a small rectangle hanging on your cursor and if you now click on both „Parameter
Sweep“ and on „Harmonic Balance“, these three functions are deactivated.

(At every moment you can roll back this action by repeating the procedure)

Now impose order on your screen and shift these deactivated directives away from the schematic –
we want to examine and to explain the circuit itself!

7
 The Gate Bias Voltage is fed to
the gate by a „Bias T“ and the
voltage value is -0,43V.

The supply voltage ist named

„Vdd“ as preparation for a
Parameter Sweep

The Input RF Signal comes

from the voltage source V3. Its
frequency is f =1.9 GHz and the
amplitude is also given as a
variable named „u_RF“ (for
another Parameter Sweep).

At the Drain pin another „Bias

T“ is used to separate the
power supply voltage Vdd and
the RF output voltage at R1.

Task:
Find the DC quiescent current of the FET.

Solution:
Pick „DC Simulation“ (path: „Components / Simulations“) and simulate. The results are given in a table.

This means:
You get a Drain current of 60 mA for a Drain voltage Vdd of +4 V and a Gate voltage of
-0,43V

8
Now we continue with the Parameter Sweep

At first we activate the three

inaccessible directives (= Harmonic
Balance Simulation and two
Parameter Sweeps).

Now right click on „Harmonic

balance simulation“ to open the
properties.

On the first card you find the confirmation of chapter 2

(„How?“):
You have to enter the highest order of harmonics for the
steady state simulation, default is „8“.

The next card sets the noise simulation properties.

Please mark „no noise simulation“ for this example.

On the last card you find the simulation name „HB1“ and

a) the settings for the relative and absolute tolerance when convergence is reached

b) the maximum number of trials (= iterations) before an error message is edited.

Still a sharp look at the two Parameter Sweeps beside the schematic:

a) „SW1“ sweeps the input RF voltage „u_RF“ in 76 logarithmic steps from 10 mV up to 800 mV

b) „SW2“ sweeps the supply voltage „Vdd“ using the values „2V / 3V / 4V / 5V“.

9
With equation equ1 the start values (u_RF = 1 mV and Vdd = +4V) are given for a non-parametric
simulation.

Remember:
For a result presentation of a non-parametric simulation in the Frequency Domain you have only to
enter the name of the variable in the Graph Property equation.

For a result presentation of a Parameter Sweep the Graph Property equation must start with
„yvalue...“

But now....simulate!

3.2. Simulation of the Relation between Forward Transmission S21 and

Supply Voltage
you need a cartesian diagram for the presentation. Enter under „Graph Properties“ the following equation for
S21:

2*yvalue(output.Vb, 1.9e9) / u_RF

This can easily be understood, because this is the
well known definition of S21:

S21 = output signal at R1 divided by the

incident wave (u-RF / 2)
But have a sharp look at the output voltage
expression in the equation:

You must use „output.Vb,1e9“

which is the result of the Harmonic

Balance Simulation for the used
frequency f = 1.9 GHz.

A line thickness of „2“ is used in the diagram. The vertical axis is scaled linearly from 0.....12.
The horzontal axis is scaled linearly from 0....800 mV for the RF input voltage „u_RF“.

This is the result.

10
Task:
Present S21 in dB and mark the vertical axis as „S21 (in dB)“. Choose a red line with thickness „2“.

Solution:
Write a new Graph property equation:

dB(2*yvalue(output.Vb, 1.9e9) / u_RF)

What a nice look....

3.3. The „1 dB Compression Point“

If the input voltage „u_RF“ is increased the output voltage will follow up to the „saturation“ and reaches a
constant value. Thus a „1 dB compression point“ is defined and this point marks a gain reduction of 1 dB.

Task::
Find the 1 dB compression point for a supply
voltage Vdd = +5V

Solution:
Use the last diagram „S21 versus u_RF“ and set
two markers:

The first marker is placed at a low value of

„u_RF“ where the gain is still constant (here:
u_RF = 20 mV). Then check S21 (here: S21=
20.7 dB).

The other marker is shifted upwards to the point

where the gain is reduced by 1 dB (now: S21 =
19.7 dB). At this point the input voltage has a
peak value of 0.6 V.

Remark:
0.6 V is the peak value of the input signal „u_RF“. The RMS value would be

u_RF = 0.6 V / sqrt2 = 0.424 V.

And if you are interested in the „level at 50 Ώ expressed in dBm“ another calculation is necessary:

„Input 1dB compression point“= 20*log10(0.424V / 0.223V) = +5,6 dBm

(0.223 V is the RMS reference voltage for P = 1mW at 50Ώ)

You get the „Output 1dB compression point“ by adding the gain (S21 = 19.7 dB) at this point. This gives a
value of
+5,6dBm + 19,7dBm = +25,3 dBm

11
Task:
Simulate the 1 dB Compression Points for Vdd = +2V / +3V / +4V
---------------------------------------------------------------------------------------------------------------------------------

3.4. Linear Presentation of the Transfer Characteristic for different

Values of the Power Supply Voltage Vdd
Open a new cartesian diagram and enter the equation:

yvalue(output.Vb,1.9e9)
This is not complicated and can be read as follows:

„Show the simulated output voltage „output.Vb“ at f = 1.9

GHz versus the input voltage „u_RF“ for the four different
power supply voltage values, using a parameter sweep“.

„--------------------------------------------------------------------------------------------------------------------------------

4. Second Example: Examining the FET Amplifier Stage for a fixed

Operating Point
4.1. Simulation Schematic

Let us use a fixed supply

voltage (Vdd = +5 V) and
deactivate the parameter
sweep SW2 used in the last
chapter.used in the last
chapter.

Please open the properties

of the „Harmonic balanced
simulation“ directive and
make the entry „noise = no“
visible.
Then simulate.

12
4.2. Ideal and real Transfer Characteristic
We use a cartesian diagram after the simulation and write the following Graph property equation for the
presentation of the real Transfer characteristic at f = 1.9 GHz:

yvalue(output.Vb,1.9e9)
(Use blue colour and line width = 2 for the curve).

To get the „ideal transfer characteristic curve“ is a little complicated:

1) Calculate the maximum value of the Input to Output ratio at f = 1.9 GHz:

plot(u_RF,(max(yvalue(output.Vb, 1.9e9) / u_RF))*u_RF)

------------------------------------------------------------------

2) Then multiply this maximum value by the input voltage „u_RF“ :

plot(u_RF,(max(yvalue(output.Vb, 1.9e9) / u_RF)) *u_RF)

-------------------------------------------------------------------

3) At last plot this result also in the same diagram. Use red colour and a line width of „2“:

plot(u_RF,(max(yvalue(output.Vb, 1.9e9) / u_RF))*u_RF)

13
4.3. Linear Presentation of S21 for different Input Voltage Amplitudes

As in the last project we use the well known formula

to calculate „S21“ at 1.9 GHz:

2*yvalue(output.Vb,1.9e9) / u_RF
Using a marker we find without any problems the
value for S21 at f = 1.9 GHz for an input voltage
u_RF = 20 mV (= peak value) to

10.9 / 123o

4.4. Presentation of S21 in dB (including 1 dB compression point) for

different Input voltage Amplitudes

Also a well known procedure:

We find (by usage of a marker) a

start value of S21 = 20,7 dB at an
input voltage u_RF = 20 mV for f
=1.9 GHz.

And shifting a second marker to

the point where S21 is now 1 dB
lower (S21 = 19.7 dB) we find that
this correlates with a peak value of
600 mV of the input voltage u_RF.

14
4.5. Rise of the Harmonics for increasing Input Voltage Amplitude (linear
Presentation)4
This is a quite simple affair because the following Graph Properties equation will do the job:

yvalue(output.Vb,[frequency])

This is the diagram showing the fundamental

frequency, the double and the triple frequency
(peak values)

4.6. Rise of the Harmonics for increasing Input Voltage Amplitude

(presentation in dBm at 50Ώ)
This diagram is needed
very often to find the level
difference between the
Fundamental and a
Harmonic Frequency.

But never forget that

qucsstudio always calculates
peak amplitude values
(...also valid for the
Harmonics).
Thus when looking for levels
scaled in „dBm“ you have at
first to change from peak
values to RMS values.

Remember:
P = 0 dBm = 1 mW at the
characteric system
resistance.

Example for Z = 50 Ohms:

P = 1 mW at 50 Ohms gives an RMS voltage of 0.223 V and a peak voltage of 0.223*sqrt(2) = 0.316V

So to change from „dB“ to „dBm at 50Ώ“ for this diagram the following Graph property equation was
used: (yvalue(output.Vb,[frequency]) / peak reference voltage)

15
Attention:
This equation is necessary if the charactistic system
impedance differs from the default setting of „Z = 50Ώ“
For Z = 50Ώ this calculation is much
simpler, because this case is already
prepared for you. Enter the equation

dBm(value(output.Vb,
[frequency]))

to get the same „dBm at 50Ώ“ curves

as before!

Please compare....

-------------------------------------------------------------------------------------------------------------------------

4.7. The complete Output Spectrum

If you want to see the complete
frequency spectrum at the output,
deactivate all parameter sweeps and
avoid „yvalue...“ in the graph property
equation.

Now if you are looking for all harmonics for

an input peak voltage value of u_RF = 100
mV:
Enter this value in the equation“Eqn1“ and
simulate. Then write the graph property
equation as

dBm(output.Vb)
for a cartesian diagram and use a marker
to indicate every harmonic frequency and
amplitude.

16
4.8. The IP3 Point
4.8.1. Fundamentals
In chapter 14.4. (of part 1) this text can be found:

The simulation of the „Third Order Intercept Point“ informs about the nonlinearities and distortions of a
stage when the input level increases.

When reaching the IP3 (on the “Pout / Pin curve”) the third order
distortion products would be equal to the applied RF signal at the input.
And this is not very good, because 3rd order distortion products have
nearly the same frequency as the input signal.....
But:
this IP3 point is a pure theoretical value due to the fact of compression, followed by limiting, in the stage
with increasing input levels.
That can be shown by this diagram:

But if you exactly know this IP3 point, then you can for every input or output level calculate the
level difference to the undesired 3rd order intermodulation products!
(You have simply to solve two equations for straight lines....)
And so you can for this case define an intermodulation free dynamic range with the stage's noise
as “floor”

17
4.8.2. Simulation of IP3 using the Transfer Characteristic
To determine the IP3 we feed the stage with two “in band signals” with the same
amplitudes but a very small frequency difference.
The Levels are now increased and very soon you will see the 3rd order intermodulation products at every side
of the couple of test signals. The frequency distance to the test signals is exactly the frequency
difference of the test signals. But the IP3-products can't normally be eliminated by filtering!
(...There exist also 2nd order intermodulation products with higher amplitudes. But their frequencies are far
away from the two test signals. So they don't disturb and can be eliminated by filtering....)

Example:
At the input two voltage sources (with f1 = 1.9
GHz and f2 = 1.91 GHz) are connected in
series. This gives a frequency difference of 10
MHz. In this case you'll find the IP3 products at
(1.9 GHz - 10 MHz) = 1.89 GHz
and
(1.91 GHz + 10 MHz) = 1.92 GHz

The voltage values of both sources must

always be indentical. So we use the variable
u_RF and a value range from 1 mV to 2 V in
the parameter sweep.

To present the simulation results in a cartesian

diagram use the two equations:

dBm(yvalue(output.Vb,1.91e9)
and

dBm(yvalue(output.Vb.1.92e9)
Then you'll see the output power levels in dBm
for the „fundamental frequency f2 = 1.91 GHz“
in red and the „upper IP3 product at f = 1.92 GHz“ in green

Remark:
If you want to calculate the frequencies of
the third order distortion products then use
the formula:

fIP3 = 2*f1 - f2
and you'll get f=1.89 GHz and f=1.92 GHz.
The amplitudes and relationships are the
same for both products.

But you need a logithmic horizontal x

axis to see the linear part of both curves!

Now we must approximate every curve

by a straight line, but that isn't very
difficult.

18
 Let us start with the fundamental frequency of
f = 1.91 GHz and write a simple equation for the
next graph:

dBm(u_RF)
As you can see: the pink line must be a little bit
shifted upwards and for „dBm“ scaling you have
simply to multiply u_RF by a constant factor to do
this.

The final solution is

dBm(5.5*u_RF)

For the IP3 product use the same

procedure. But remember that this is a
„third order product“ and „Third order“
means that you have to calculate

u_RF*u_RF*u_RF
by the equation. Afterwards shift the
(green) line upwards by multiplying by a
constant factor to coincide with the
simulated IP3 product.

This is the final equation form for the

green line:

dBm(1.4*(u_RF*u_RF*u_RF))

Then extrapolate the two lines (green and pink) up to the

(theroretical) break even point and set a marker to this point!

19
Now we have found what we wanted to see:

The „Output Intercept Point OIP3“ has a value of +30.8 dBm

(...calculated as peak value...) and (+30.8 dBm – 3 dB =) +27.8 dBm
calculated as RMS value.
This is achieved at an input peak voltage value of u_Rf = 2V (...see the horizontal diagram axis).

This would be an RMS value of 2V / sqrt(2) = 1.41V

And this coincides with an RMS level of 20*log10(1.41 V/ 0.224V) dBm = +16,8 dBm

Now we have found the „Input Intercept Point IIP3“ with +16.8 dBm.
-------------------------------------------------------------------------------------------------------------

Task:
An input voltage with a peak value of 0.02 V is applied to the input. Try to find the attenuation of the
third order intermodulation produkt.

20
Solution:
Use the above diagram and draw (in your brain!!) a vertical line at the point for u_RF = 0.02 V starting
at the horizontal axis.
Mark now the break even points of the vertical line with the two approximation straight lines in the
diagram using two markers.
Calculate the level difference in dB.

The level difference (= attenuation of third order product) is 80 dB.

Task:
Theory says that an increase of 20 dB of the input voltage u_RF causes an increase of 60 dB of the
third order intermodulation product.
Thus the level difference (= attenuation of the third order product in relation to the input level) will be
reduced by 40 dB in this case.
Please check this using the above diagram.

21
4.8.3. Simulation of OIP3 using the Output Spectrum
As already mentioned: for a spectrum presentation you have either to de-activate all parameter sweeps or to
avoid the expression „yvalue“ in a Graph property equation!

Let us use the same

schematic as in the last
chapter 4.8.2. (to simulate
the IP3 point).

But now we write

dBm(output.Vb)
as equation for the Graph
properties.

This is the result and an additional

analysis of the diagram is
necessary...because if you have a
sharp look at the lines you see them
presented as arrows....they look a
little strange...

22
Let us use the zoom function to present only the frequency range from 1.8....2.0 GHz!

This result is already well known!

We see the two amplified input signals with 1.9 GHz and 1.91 GHz at the output. Additionally we find
the IP3 products on each side of them (frequencies: 1,89 GHz und 1,92 GHz). The level difference is
exactly 80 dB – just like in the last chapter!

But now the vertical axis is scaled from -300 dB up to 0 dB. Thus we find additional products of higher order
–.. and the slope of such a signal increases with the order of the intermodulation product.

So an „overload“ of an amplifier should strictly be avoided...

23
5. A Half Complex Mixer“ to generate an SSB Signal
This is a modern version of the good old „phase method of SSB generation“. But at first we have to discuss
some new terms.

5.1. Analytic Pairs

Sounds mysterious, but isn't.
If you listen to music or a talk, a lot of signals are in action continually changing frequency and amplitude.
Thus we use the expression „frequency band“ for such kind of signals. But this has an interesting property:
you find a positive and a (mirrored!) negative band symmetrically arranged beside Zero Hertz!
If you now want to halve the bandwith: simply change to a complex signal = use either the positive OR the
negative part. Is is done as follows:

a) The signal with the double sided spectrum is the real signal (..as coming from a generator, a microphone,
a loudspeaker....) and is named „I“ Signal (= in phase signal). This signal must not be changed.

b) additionally you use a circuit named „Hilbert Transformer“. If you connect the real signal to its input, then
you get at the output a constant phase shift of exactly 90 degrees for all input frequencies! This artificial
output signal is named „Q“ signal = „Quadrature Signal“.

Using digital signal processing you have now two data streams named „I“ and „Q“ which are the two
components of a complex signal! This is the famous analyctic pair. It can be defined as

I + j*Q
and you find only
positive frequencies for „I + j*Q“ resp. only negative
frequencies for „I – j*Q“

5.2. Example: a Half Complex Mixer to generate an SSB Signal

You need the following ingredients:

a) A real information U = „Iinf“ (= speech, music..) which is converted to an Analytic Pair by the aid of a
Hilbert Transformer. Its form is now:

Uinf = Iinf + j * Qinf

b) A already complex carrier signal which can (due to Eulers law) also be written as

+jωct
Ucarrier = Ucarrier_max * e = Ucarrier_max * [cos(ωct) + j * sin(ωct)]

Attention: in practice you use the amplitude value „1“ for the carriers's amplitude and a Sine and
a Cosine voltage source with the same frequency. This represents a signal with only one positive
frequency.
.

c) Two multiplying circuits.

d) one subtracting circuit

24
Now multiply the Information's Analytic Pair by the complex carrier with an amplitude value = 1 and only
one positive frequency:
Uinf * Ucarrier = (Iinf + j * Qinf) * [cos(ωct) + j * sin(ωct)] =

[Iinf * cos(ωct) – Qinf * sin(ωct)] + j * [Qinf * cos(ωct) + Iinf * sin(ωct)]

The real part of the product is marked in red. Only this part (= USB = upper side band) can be
transmitted into the air and this part must be realized by the following circuit.

Remark:
The Hilbert Transformer with 90 degrees of constant phase shift is not complicated when realized in digital
signal processing. But an analog version with a greater bandwidth.....can be a life task.....Sorry....

This circuit generates an „Upper Side Band = USB“

If you need the „Lower Side Band = LSB“, then change the sign of the
signal in the lower part of the circuit. This can be done by replacing the
subtracting circuit by an adding circuit.

Now realize this principle in qucsstudio (see next page...).

25
This is the correct realization and by using Harmonic Balance we'll find some interesting details.
The real RF input signal is presented by two identical voltage sources. The first source is the „I“
Channel, the second source feeds the Hilbert transformer to generate the constant phase shift of 90
degrees for the Q Channel.

This is the „I“ Channel including the multiplying circuit as regarded example:
The peak internal voltage
value of the RF source is
0.316 V.
But at the input of the
multiplying circuit we
find only the half value of
0.158 V due to the
voltage divider
consisting of R7 and R6.

This input signal is named

„P_I“
The LO signal has an
amplitude peak value of
1 V and a frequency of 1
GHz.

26
That is the evidence:

at point „P_I“ we observe a peak voltage of 0.158 V and this equates to a level of -6 dBm and the RF signal
frequency is1 MHz.

At the „LO_cosinus“ pin we find a peak voltage of 1 V = a level of +10 dBm (LO frequency is 1 GHz9

27
But the output is interesting:
We have to add the output signals of the two multipliers to get the LSB signal.

And a subtraction of the output signals gives the USB signal:

The levels of both sidebands equate to the RF input signal

level (= -6dBm ).

...and we get a sum frequency of 1001 MHz and a difference

frequency of 999 MHz.

28
5.3. Parameter Simulation of the „undesired Side Band's Rejection“ for
different Phase Errors
The rejection of an undesired side band depends on the exactness of the 90 degree phase shift of the Hilbert
transformer. So we start a parameter sweep to show this rejection for phase shifts between 99 and 92
degrees.

Please modify the schematic as follows:

a) In the property list of the Hilbert Transformer a variable „pdiff“ is used for the phase shift.

b) We need an equation to set the start value of the phase shift to 90 degreees.

c) The parameter sweep varies the angle „pdiff“ from 88 to 92 degrees using 101 steps. durchgespielt.

Then de-activate the parameter sweep and simulate. Compare the result to the above simulation
result – it should be identic.

29
But now.....activate the parameter sweep and simulate.

We present in a cartesian diagram the USB signal but show only the frequency range of the undesired
LSB with f = 999MHz. So we observe the level of this signal in relation of the phase error.

This is the necessary Graph property equation:

dBm(yvalue(out1_USB.Vb,999e6))

Remember that the level of the desired USB was „-6 dBm“. With this information you could now easily
calculate the „undesired side band's attenuation“ – in your brain or by the modified formula

dBm(yvalue(out1_USB.Vb,999e6))+6

This would be rthe presentation example.

30
6. A Double Balanced Mixer (DBM)
6.1. The Schematic (coming from chapter 14.3. in part 1 of the tutorial)
This circuit was investigated in the mentioned chapter and we have simulated

– the output spectrum

–
– the conversion loss
–
– the IP3 point

What can HB add to this information?

This was the

simulated output
spectrum in
chapter 14.3 of the
tutorial's part 1.

(calibration still in
„dBV“)

31
6.2. Used Schematic for the Harmonic Balance Simulation
There is only one modifaction: now Schottky diodes „1N6263“ are used because they switch faster and
show lower capacitance values. They can be found under „Schottky diodes“ in the qucsstudio diode
library. These are the properties

silicon schottky diode

60V, 50mA, 0.1ns, 2.2pF @0V
Manufacturer: ST Microelectronics

This is the simulated

output spectrum scaled in
„dBV“ .

Please compare it now to

the simulation in the Time
Domain, followed by an FFT
(found on the previous
page)

Voltage sources
properties:

a) RF input
peak voltage = 10 mV
frequency = 137 MHz

b) LO signal
peak voltage = 2 V
frequency = 37 MHz

Harmonic Balance is really a fine toy...

32
6.3. Harmonic Balance Parameter Sweep of the LO Level
6.3.1. Output Spectrum for different LO Amplitudes
For a successful simulation we need a variable for the internal voltage of the LO signal source and name it
„LO_level“. Using equation „Eqn1“ the start value is ÛLO = 2V.
The parameter sweep starts at a minimum voltage value of ÛLO = 0.1 V and rises up to 3 V in steps of
0.1 V.

„dBm(IF_out.Vb)“ for a range of -90 dBm up to

Now start the simulation and show
-40 dBm in the frequency range from 0 to 400 Mhz.
Use a marker for the LSB signal.

The spectral lines look like old christmas trees but

this is easy to explain:

The different simulation results at a regarded

frequency are written „one upon the other“. You
see that if you shift the marker and read the
entries in the marker field.
Shown example for the used marker position

LSB frequency= 100 MHz

LO level = 2 V

Level of „IF_out.Vb“ for f = 100 MHz is

-41,7 dBm

33
You can also indicate the LSB level at f = 100 MHz versus the LO level.

Therefor use the equation for the Graph properties:

dBm(yvalue(IF_out.Vb,100e6))

The marker is set again to a LO voltage value

of 2 V. Please compare to the last diagram...

6.3.2. Presentation of the Conversion Loss

The internal voltage of the RF source has a peak level of 10 mV. This can be converted to dBm by the simple
equation

dBm(RF_Source_Level)
and the result is a level of -30 dBm at f =
137 MHz(see diagram)
The Incident Wave can be calculated by the
well known relationship
„Incident Wave = Uo / 2
= RF_Source_Level/2“
Thus the Incident Wave Level (which arrives
at the mixer input pin) must be 6 dB down
and has a value of -36 dBm (= 5 mV)

Now write the equation to calculate S21 for

different LO levels:

dB(yvalue(IF_out.Vb,100e6)/5e-3)

You see:
The conversion loss is 5,71 dB for a LO
peak voltage value of 2 V
(See the marker field).

The „IF_out“ - level for f = 100 MHz must

now be

-36 dBm – 5,71 dB = -41,7 dBm

See previous page....

34
6.3.3. Addition of a Diplexer to the Mixer Output
Double Balanced Mixers equipped with diodes must have a perfect broadband 50Ώ termination at
every port to reduce distortions and to show the specifications given in the data sheet.

This can be done by a circuit named „Diplexer“ with the following properties:

a) a band pass filter characteristic for the IF-Signal output (here: f = 100 MHz), but

b) a perfect 50Ώ termination for the mixrer output – for the complete regarded frequency range

Online calculators for diplexer

can be found on the WWW.

This design comes from

http://www.changpuak.ch

At first enter the IF frequency

of 100 MHz

then the characteristic system

impedance of
Z = 50 Ώ

and at last the quality factor Q

= 10 as proposed by the
calculator for RF circuits.

Then draw a schematic to simulate the diplexer with qucsstudio.

Simulate the S parameters S11 and S21 from 0 to 200 MHz.

35
The Input Reflection S11 is below -40 dB in the complete frequency range. And if we could use lossless
parts, S21 would be 0 dB for f = 100 MHz.
Now combine the mixer and the diplexer.

The simulated spectres for the input and the output of the diplexer can be found on the next page. A
full success....

36
------------------------------------------------------------------------------------------------------------------------------------

6.3.4. Finding the IP3 Point for a LO Peak Voltage of 2V

This problem was solved for an amplifier in chapter 4.8. and we use the same procedure for the mixer circuit.

a) At the RF input two voltage sources are connected in series. The sources use the same peak voltage
value „RF_Source“ = 20 mV as starting value. The input frequencies are 137 MHz and 137.1 MHz –
thus we find a frequency difference of 0.1 MHz.

Now a HB parameter sweep for „RF_Source“ = 20 mV …...2 V is started.

This simulation is time consuming but the result is fine.

Please show the simulated spectrum around f = 100 MHz at the diplexers output by using this equation
and a frequency range from 99.5 to 100.5 MHz for the horizontal axis:

37
dBm(LSB_100MHz_out.Vb)

The LSB signals with f = 100 MHz and 100.1 MHz can easily be found in the diagram. Examining the „tops
of the arrows“ we see that the value range starts at -37 dBm and rises up to nearly 0 dBm.

The „IP3 intermodulation products“ are found at 99.9 MHz at 100.2 MHz. Their amplitude values are
nearly invisible for small values of the input voltage level „U_RF“, but thenD

To see the details we regard only the IP3 product at 99.9 MHz and the adjacent LSB signal at 100MHz for
the parameter sweep.

Then followes the game „how to find the approxmation straight lines for the simulated curves“ using
four different equations:

a) for the LSB signal at f = 100 MHz: dBm(yvalue(LSB_100MHZ_out.Vb,100e6))

b) for the associated approximation straight line for this curce: dBm(RF-Source/3.9)

c) for the IP3 product at f = 99,9 MHz: dBm(yvalue(LSB_100MHZ_out.Vb,99.9e6))

d) for the associated approximation straight line for this curve:

dBm(RF_Source*RF_Source*RF_Source)/70)

38
The result:

The break even point of the straight lines lies at the right hand side at more than „U_RF“ = 3V (peak
value). The associated output level is ca. +10 dBm at the left vertical axis.

This means that the output IP3 point has a

value of ca.
+10 dBm

39
7. Summary: Harmonic Balance Simulation in Question and Answer for
an Amplifier Stage
7.1. Which Circuit?

We use the example

„simple_amplifier.sch“ as
found in the qucsstudio
homepage (which is a good
choice).
But the amplitude value of the
input signal must be replaced
by a variable „U_in“.

-----------------------------------------------------------------------------------------------------------------------------
7.2. How can I quickly find every DC Voltage of the Operating Point?

In the task bar you find the

well known gear wheel for
the start of the simulation.
At its left hand side the
same gear wheel but over-
printed in red with „DC“.
A left hand click on it
calculates all DC voltages
(See this illustration).
A further click on to the gear
wheel lets all entries
disappear.

-----------------------------------------------------------------------------------------------------------------------------
7.3. What are the necessary Preparations for a successful HB
Simulation?
You need at first the „Harmonic balance
directive“ with the necessary entries for
a successful noise simulation (see this
illustration).

Then write an equation „Eqn1“ with the

initial value of 50 mV for the input voltage
„U_in“.

.
Now start the simulation.

40
7.4. How can I present the Input Voltage and the Output Voltage
including the gain (= linear / in dB / in dBm)?

Open a cartesian
diagram and enter
„IN.Vb“ as graph
property.

Repeat this and enter

the following equations :

OUT2.Vb

dBm(IN.Vb)

dBm(OUT2.Vb)

And now to the gain:

Use two cartesian diagrams. Write the graph equation

OUT2.Vb / IN.Vb
and
db(OUT2.Vb / IN.Vb
But the result is very irritating:

We did not expect a

maximum value of
400 dB respectively
a maximum value of
„5e19“ ...

The reason for this phenomenon is very simple:

In the „Harmonic balance property list“ we have entered a maximum harmonic order of N = 8

41
Every Harmonic frequency is marked by an arrow on the horizontal diagram axis and for every Harmonic
frequency the amplitude will be calculated. But these amplitudes decrease with the order „N“ and thus the
calculated gain for the Harmonics rises with N and rises and rises.....

The remedy is simple:

Reduce the scaling at

the vertical diagram
axis (under „Limits“) to
0...5 for linear
presentation and to
0....10 dB for dB-
presentation.

-------------------------------------------------------------------------------------------------------------------------------------------

7.5. I want to see the Signal to Noise Ratio for f = 1 MHz at the Output in
db
Use this equation:

dB(yvalue(OUT2.Vb,1e6 / OUT2.vbn)

In chapter 7.3 we have set the start frequency for

noise simulation to 1 kHz and the end to 10 kHz
using 10 steps.

Don't be worried by the „slope“ of the amplitudes

and have a look at the scaling of the vertical axis:

this is the fith position after decimal point....

And always remember:

Harmonic balance can only handle and calculate a collection of discrete
frequencies...but never a continuous spectrum!
-----------------------------------------------------------------------------------------------------------------------------------

42
7.6. I want to simulate with the Input Voltage as Parameter. How does
the Schematic look like for this Task?
That is it:

43
7.7. Parameter Sweep of the Input Voltage: I want to see the Output
Voltage and the Transfer Function
Write a parameter sweep declaration and sweep logarithmic from

U_in = 20 mV to U_in = 1,5 V

using 100 points at a frequency of f = 1 MHz.

Solution for the output voltage:

yvalue(OUT2.Vb,1e6) dBm(yvalue(OUT2.Vb,1e6))

And this is the transfer function in linear and double logarithmic form:
Gleichungen:

yvalue(OUT2.Vb,1e6)/U_in dB(yvalue(OUT2.Vb,1e6)/U_in)

44
7.8. Parameter Sweep of the Input Voltage: I would like to see the
Fundamental Frequency and two Harmonics in dBm at the Output
For this purpose we use the double logarithmic presentation and write three graph property equations

dBm(yvalue(OUT2.Vb,1e6))
dBm(yvalue(OUT2.Vb,2e6))
dBm(yvalue(OUT2.Vb,3e6))

------------------------------------------------------------------------------------------

7.9. Parameter Sweep of the Input Voltage: how can I evalute the
Output IP3 Point?
We use the same graph as above and extend (in our brain) the associated approximation straight lines of the
curve for 1 MHz (= fundamental frequency) and the curve for 3 MHz (= third order product). Then we
evaluate the break-even point:

45
7.10. Parameter Sweep of the Input Voltage: how does the Noise Voltage
at the Output vary with an increasing Input Voltage?
Never forget:
Noise voltages are always given as „Volts per square root of 1 Hz“ and this is the noise voltage resp.
noise power which can be measured in a bandwidth of 1 Hz. The amplitude values are extremely small and
thus they are normally given as „NanoVolts“. But 1 Volt is identical to „109 NanoVolts“ and thus the graph
equation can be written as:

1e9*yvalue(OUT2.vbn)

46
7.11. Parameter Sweep of the Input Voltage: how can I simulate the
Signal to Noise Ratio in dB at the Output?
You have two options

a) Present the output signal „OUT2.Vb“ in dBm and the output noise voltage in dBm in the same
diagram. Then determine the level difference between the curves for a desired input voltage.

Remark:
The output signal has a frequency
of f = 1 MHz.

The noise signal is calculated at a

frequency which is 10 kHz lower
(f = 990 kHz).

equations:

dBm(yvalue(OUT2.Vb)
und
dBm(yvalue(OUT2.vbn)

b) Write an equation which calculates the dBm

difference between the two curves (bandwidth = 1
Hz for the noise power):

(dBm(yvalue(OUT2.Vb,1e6)))-
(dBm(yvalue(OUT2.vbn,990e3)))

47
7.12. Parameter Sweep of the Input Voltage: how can I determine the
Signal to Noise Ratio in dB for a Bandwidth of 1000 Hz?
Very easy:

The noise power increases linearly with the bandwidth. Working with „dB“ or „dBm“, then the following
expression must be subtracted from the S / N ratio:

10*log10(regarded bandwidth / 1 Hz)

For a bandwidth of 1 kHz you get a correction factor of -30 dB for the curve of the last example (in
chapter 7.11.)
This is the new graph equation:

(dBm(yvalue(OUT2.Vb,1e6)))
-(dBm(yvalue(OUT2.vbn,990e3)))
-30

48
7.13. Parameter Sweep of the Frequency: Preparation of the HB
Simulation
This simulation needs some time for the preparation:

a) Use two variables in the property menu of the input voltage source (= „U_in“ and „freq_in“)

b) Use „equation“ to enter

fmin = 1 kHz
fmax = 400 MHz
freq_in = 1 MHz
U_in = 50 mV

c) Write the properties for the parameter sweep:

Sim=HB1
Param=freq_in
Type =log
Start=fmin
Stop=fmax
Points=100

49
7.14. Parameter Sweep of the Frequency: curves of Output Voltage and
Gain
At first write the correct equations for the result of a frequency sweep:

Linear presentation of the output voltage Output voltage in dBm

yvalue(OUT2.Vb,fmax) dBm(yvalue(OUT2.Vb,fmax)

Now gain versus frequency isn't a great problem

Gain (linear presentation): Gain (in dB)

yvalue(OUT2.Vb,fmax)/U_in) dB(yvalue(OUT2.Vb,fmax)/U_in)

In the right diagram you can see how the „upper 3 dB cutoff frequency“ was determined by two markers.

The left marker indicates a gain value of 8.2 dB at low frequencies.

The right marker has been shifted to 95.4 MHz. There the gain dropped to 5.2 dB (= 3 dB reduction).

50
7.15. Simultaneous Parameter Sweep of Frequency AND Input Voltage

Simply add a parameter sweep SW2 for the input voltage „U_in“.

The input voltage is increased from 50 mV up to 1500 mV (five steps).

The result is really impressive:

51