Complex RF Switching Architectures -- Part I

In our technological world, wireless appliances are becoming extremely common, and the need for radio frequency test systems is raising at a very fast pace.

With the introduction of new parameters and new considerations, the GHz domain must be treated differently than the low frequency domain. The switching sector is not immune to this upsurge in radio frequency testing needs and in this article we treat the different solutions for this new needs.

Introduction to High Frequency

The first question that comes to mind when dealing with high frequency is a very practical one -- when is the frequency high enough to start worrying about different behaviors? The formula to compute the wavelength of a signal is



where c= speed of light (300.000.000 m/s) and is the dielectric constant of the medium where the signal is travelling. For example, for free space =1, while for a coaxial cable =2.3. A good rule of thumb is to compare the dimension of the circuit in use with the wavelength of the highest frequency component of our signal. If the circuit dimensions are more than 1/100 of the wavelength, you should think about taking high-frequency effects into account. For example, for a 300 MHz sinewave signal in free space, the lambda is equal to 1 meter (3E+8/3E+8=1), so if the circuit has dimensions bigger than 1 cm, the designer should start taking high-frequency effects into account. In cases when you are dealing with a more complex signal, such as a digital square wave, it is generally accepted to take as maximum frequency components 10 times the value of the fundamental sine wave.

Impedance Matching

One very important principle for high-frequency switching applications is the impedance matching. It is very important that the impedance of the cable matches the impedance of the source and the impedance of the load. This is the only way that we can be sure all the signals we are applying actually travel toward the destination. If the impedance of the source, cable, and load are not matched, part of the signal will not go forward but it will actually be reflected back to the source.

Crosstalk, VSWR, and Insertion Loss

Parameters that are extremely important in the radio frequency domain are Insertion Loss, Voltage Standing Wave Ratio (VSWR) and Crosstalk. They are connected to physical phenomenons that happen when the radio frequency signal travels within a system. To visualize this phenomenon, compare light travelling through air or glasses with different refraction constants.



When the light (our signal) hits the first surface, part of it is reflected and part of it continues to travel through the first layer of glass. The reflection is caused by the fact that the glass has a different refraction constant than the air. In our circuits, the characteristic impedance of the cable is equivalent to the refraction constant in the glass -- if it is different from the impedance of the source, part of the signal is reflected. Having the same impedance is equivalent to glass that behaves like air for the light; no discontinuous pattern for the signal.

The amount of signal that is reflected toward the source is quantified by a parameter called Voltage Standing Wave Ratio (VSWR). A theoretical switch should have VSWR equal to 1 (no signal reflected). In reality, the impedance will not be perfectly matched, so the VSWR has a value greater than one.

Another important parameter that the light example can help us understand is the insertion loss. The light travelling through the first glass layer 1 has some energy loss because the glass is not perfectly transparent and part of the signal is lost along the path. In the same way, a switch is not perfectly "transparent" for the signal and part of the signal energy is lost going through the switch, mainly becoming heat. A theoretical switch should have insertion loss equal to 0 (no signal loss travelling through the switch).

Signal reflection is an undesirable condition, for at least two reasons:

If only part of the signal generated can be applied to the target circuit, we are not achieving full transfer of energy and the resulting effects of the destination circuit are decreased.

The reflected part of the signal travels back toward the source, permanent damage of the signal source could occur if it is not designed to withstand signal reflection. Because an RF instrument is fairly expensive, failure to comply with the impedance matching rule could become a costly error.


In most practical systems with connectors, a perfect match will never be attained: however, having a switch with matched impedance minimizes the discontinuity to an acceptable level.

Crosstalk is a measure of how and how much a signal flowing on one channel can influence the signal on other channels. This parameter is usually expressed in dB. For reference, a crosstalk of -50 dB indicates a switch with fairly good isolation characteristics. Good crosstalk behavior guarantees that signals on channel 2 are marginally affected by signals on channel 1, for instance. As every other parameter in this scenario, crosstalk is a number that is meaningless if it is not associated with a specific value for the frequency. In general, as we expect, this parameter degrades as the frequency increases, so it is very important to have a precise idea of the frequency range of the switched signal.

For a good crosstalk parameter, it is also important to avoid negative effect because of InterModulation Distortion (IMD). Because of this physical phenomenon, if there is crosstalk interaction between channels 1 and 2 (for instance), then on channel 1 you not only see part of the signal on channel 2, but also a series of other spurious components because of the mixing of the two signals. So, while choosing a switch, it is very important to carefully consider the value of the crosstalk parameter at the frequency of interest -- it could influence the behavior of your test system dramatically.

The following table gives some guideline on the minimum level of the parameters you should look for:

Parameter	Theoretical value	Desired level at frequency of interest
Insertion loss	0 dB	Less than 1.5 dB
VSWR	1	Less than 1.5
Isolation	-	Less than -50dB

Reader Comments | Submit a comment »

incorrect formula
The formula for the frequency of a signal in a transmission line of relative dielectric constant epsilon is wrong. The formula should contain the square root of epsilon, and not epsilon to the first power. For instance, in a transmission medium of epsilon = 2.3, the signal speed is: 1/sqrt(2.3) = 66% of the speed of light.
- Carl Ellison, Creative Electronics. cellisn@gte.net - Dec 5, 2005

Incorrect Equation
The wavelength is proportional to 1 over the square root of the dielectric constant....
- Heidi Barnes, Agilent. hbarnes_2298@agilent.com - Nov 17, 2005

Errors in names, symbols and formula.
The expression for lambda has several errors. Right formula: lambda= c sub 0 over (f * sqrt(epsilon sub r)) c sub 0 is the speed of propagation of electromagnetic waves in vacuum. Right name for epsilon sub r is relative permittivity, not dielectric constant. If you write the 9 digits for the speed, better use the right value: 299 792 458 m/s (International Standard IEC 27-1 pp 43,55)
- Vicent M. Rodrigo Peñarrocha, Universidad Politécnica de Valencia. España. (Spain). vrodrig@dcom.upv.es - Jan 13, 2005

the formula for wavelength has a printing error. It must be right lambda=c0/(f*sqrt(epsilon_realtive))
- Siegfried Martius, University Erlangen-Nuremberg. leo@lhft.eei.uni-erlangen.de - Jan 9, 2005

Nice tutorial on High Frequency
Nice tutorial for basic understanding of high frequency behavior of signals. 'Looking forward for Part II Jim K
- Oct 11, 2000