Dangers of Simple Usage of Microwave Software - Rohde and Hartnagel.pdf
(
1522 KB
)
Pobierz
The Dangers of simple usage of Microwave
Software
Ulrich L Rohde
*
, Hans L. Hartnagel
**
* Synergy Microwave Corp
201 McLean Blvd. Paterson, NJ, 07504, USA
e‐mail: ulr@synergymwave.com
web: www.synergymwave.com:8080/
**Fachgebiet Mikrowellenelektronik, Technische Universitäet Darmstadt,
Merckstr.25, D‐64283 Darmstadt, Germany
e‐mail: hartnagel@ieee.org
web: www.mwe.tu‐darmstadt.de
Introduction:
We have seen many dissertations about the implementation of microwave circuits, where a
student has built an oscillator or another circuit, measured it, ran a simulation, obtained
different answers and then tried to explain the reasons. Actually there are two main sources of
inaccuracy, one being the measurements and the other the simulation. In the case of an
oscillator the important parameters are output power, harmonic content and, most important,
phase noise.
These three critical parameters are determined under large‐signal conditions. Using CAD
introduces automatically two weaknesses. The device used for the application needs to be
characterized, many times by curve fitting, and needs to match a model of the simulator which
itself is mostly an analytical model rather than physics based.
In a simple oscillator case we would like to show that using a rigorous mathematical model the
educational benefit outweighs the simplicity of a CAD analysis and subsequent optimization.
The two test cases are a Driscoll oscillator with the crystal resonator in the emitter, which was
taken from the literature and the design could have never worked because of errors in the
publication data. A CAD tool would not have found the problem but an understanding of the
operation allows to find the correction. The next case is the Colpitts oscillator, which offers
many choices of design but only the large signal approach will work. This is more analytical
rather than trial and error.
This effort is based on using Bessel functions and a calculation in the time and frequency
domain. The added benefit is that all physics‐based noise models will be used and therefore the
student gains much more insight in all the concurrencies.
Once the basic set of equations is
derived, the first derivative of the feedback components vs. phase noise allows exact
optimization.
This type of circuit analysis, which can be applied to many other designs such as amplifiers and
mixers, give the best insight into the functionality of circuits in the time domain where we
discover such new things like time average loaded Q and noisy feedback or noise‐contributing
support circuits.
In this paper we will use a simple but in the end highly non‐linear circuit, where we will
demonstrate the accuracy of our approach using simulations, sets of analytical time domain
equations and of course accurate measurements using test equipment from two established
manufacturers, Agilent and R&S. Each step of this design provides much better insight in the
functionality than the standard teaching approach of this topic resorting to too much CAD. In
the following we will show three cases, which will highlight the problems.
Case study of a: Twin T‐oscillator using an Infineon BFR93aw, microwave transistor, showing
that the linear and the non‐linear analysis for the resonant frequency gives a significant
different results [1], a analysis of a Driscoll oscillator where the Cad prediction is far too
optimistic because it does not have good data for the flicker corner frequency [not provided by
the manufacturer] and flicker noise contribution of the crystal and finally the third case is the
mathematical analysis of the Colpitts oscillator using the large signal parameter and the Bessel
function to get a very close result to both the measured and the CAD simulation [2].
Case 1:
In general and until recently transistor simulations used linear analysis, which turns out to give
fairly inaccurate results. To show the base line, here is the example using a RC example. It is
based on [1] and operates at 1.6MHz. Figure 1 shows the actual circuit diagram.
Figure 1: the actual circuit diagrams for 1.6MHz
The literature is full of RC oscillators but very little information is available on its phase noise
and the difference between the linear and the non‐linear operation. So we analyzed [Figure 1]
this oscillator and scaled it to about 1MHz and using a linear simulation determined the
following resonance frequency.
Figure 2: Linear simulation result of the schematic in figure 1.
The linear case indicates the resonance frequency around 2.5GHz. The Y‐axis is the RF current in
milliamps at the junction between the two resistors, 2.1Kohms and the 150pF capacitors to the
ground. This assumes a total linear system and the purpose of this example is to show that the
linear simulator can mislead you totally. After this result we used the Ansoft serenade harmonic
balance simulator 8.7V and a validated model for the siemens transistor BFR93aw. The initial
DC analysis provides the operating point.
The results are 2.7mA for 0.86V V
ce
. The output waveform is slightly distorted and shown in
figure 3. Figure 4 shows the harmonic contents. The output frequency as seen in figure 4 is
different from the linear prediction and is 1.6MHz. The harmonic suppression is about 14dB.
The loaded output terminated into 50 ohms is –19dBm.
Figure 3: Simulated output waveform.
Figure 4: Simulated output power for the schematic in figure 1
We must keep in mind that this is a RC oscillator consisting of a notch filter and does not have a
Q in the traditional sense. These types of oscillators typically do not operate into 50 ohms but
into some CMOS gates, which are voltage and not power driven. IF we assume that the practical
load is 9Kohms then the voltage swing at the output increase to ±900mV, this is 1.8Vp‐p at the
end to drive the gate.
Figure 5: Simulated output waveform with a high impedance termination of 9Kohms.
Now to our surprise the resonant frequency is 1.679MHz a huge difference from the linear
approximation. So far we have shown output power harmonic contents and now how about the
phase noise.
This information is rarely found in the literature, but here it is shown in figure 6.
Figure 6: Predicted phase noise in dBc/Hz offset from the carrier frequency ranging from 10Hz
to 1MHz of this RC oscillator.
Plik z chomika:
umts
Inne pliki z tego folderu:
Circuit Simulation Examples - W9VES.pdf
(1529 KB)
Dangers of Simple Usage of Microwave Software - Rohde and Hartnagel.pdf
(1522 KB)
Mathematical stability problems in simulation programs - Rohde and Lakhe.pdf
(693 KB)
Using Simulation at RF - Rohde.pdf
(1325 KB)
Inne foldery tego chomika:
Chapter 10
Chapter 11
Chapter 12
Chapter 13
Chapter 14
Zgłoś jeśli
naruszono regulamin