Basic Investigations for Switching of RF Signals Werner Johler, Senior Member IEEE TYCO ELECTRONICS Logistics AG, Werk AXICOM Au Seestr. 295, 8804 Au-Waedenswil, Switzerland werner.johler@tycoelectronics.com Abstract – A major advantage of electromechanical RF relays is their capability to carry and to switch signals from DC up to the GHz range. Up to now the switching capability of miniature RF relays was limited to a few Watts, which was not really satisfactory. Basic investigations were performed to understand the mechanism when High Frequency signals are switched, as no data was previously available. When High Frequency signals are interrupted, the arc performs like a DC arc due to its thermal characteristics and burns for several periods of the signal. The arc is an adjustable resistor and just changes the amplitude of the output signal. When the contact gap increases, the arc finally extinguishes at current zero in a similar way as an AC arc. At the time the arc extinguishes the signal makes a phase shift as the open contact can be considered to be a capacitance. Practical tests were performed on a new type of high performance low-cost ultra-miniature RF relay with DC loads up to 60 W, low frequency AC loads up to 62.5 VA and RF loads with a maximum power of 37 W. The results obtained confirm the theoretical approach. Major advantages of bridge contacts providing two contact gaps in series were found, yielding excellent RF characteristics as well as outstanding switching performance in all frequency ranges from DC to RF. Even after 1 million operations no relevant changes of the RF characteristics were found, although a load of 37 W was hot switched. Keywords: RF relay, RF switch, RF switching, hot switching I. INTRODUCTION In modern telecommunications more and more information has to be transmitted faster and faster. In order to achieve higher transmission rates, more bandwidth is required. Although most information is digitalized, the transmission frequency increases proportionally. For modern radio frequency (RF) and microwave applications as well as for the related test equipment, suitable switching elements are required for cold as well as hot switching. Several competitive technologies are used to conduct and to interrupt RF signals. A. Comparison of Switching Technologies The most commonly used devices are traditional electromechanical RF relays, pin diodes and GaSFET semiconductor switches. Recently RF MEMS have also been discussed but are not commercially available yet for commercial applications. The advantages of solid-state solutions (GaAs FET and Pin diodes) are the high lifetime (in a laboratory environment), small size, high switching speed and relatively low costs. Electromechanical solutions generally provide better RF characteristics with higher isolation level and lower insertion loss within the whole frequency range. In addition, contacts are able to carry a DC signal superimposed on the RF signal and to transmit higher RF power. They yield true broadband RF characteristics from DC up to the specified frequency. Excellent isolation properties can be achieved regardless of the ambient temperature since the isolation is made of air or plastics. Metal contacts provide the lowest possible insertion loss. Furthermore electromechanical RF switches are able to transmit RF signals without any signal distortion, neither amplitude distortion nor phase distortion and without any signal modulation, neither cross modulation nor mutual modulation, regardless of the temperature. While electromechanical RF relays have traditionally been large, recently introduced designs, offering the combination of small physical size and superb RF characteristics, with isolation values up to more than 80 dB @ 1 GHz, insertion loss 0.3 dB, and VSWR values less than 1.3, are impressive. In addition to the excellent RF characteristics, RF relays yield about the same performance as standard telecom relays. Both latching and non-latching switching characteristics are available, requiring a power consumption of only 140 mW (non-latching) and 70 mW (1 coil latching). Good mechanical and electrical endurance values and high dielectric characteristics result in an extremely robust and reliable design. Standard assembly technologies can be used such as lead-free SMD soldering. B. Limiting Factors for carrying and switching RF power Miniature electromechanical RF relays with a size between 0.5 and 4 cm3 typically are able to carry 10 W and to switch a few Watt of RF power. The RF power carrying capability of relay contacts is defined mostly by the heat generated when the power passes the contacts. The heat is mainly generated by intrinsic ohmic loss. As electromechanical relays inherently have low contact resistance values, the ohmic loss is relatively small. As shown in [1], the constriction resistance of contacts depends on the frequency of the signal as skin effects result in a reduction of the spot diameters. In order to minimize this resistance, high conductive materials are used, and even coated with silver. Contacts are gold plated to keep the resistance low and stable during their entire lifetime. Depending on the ambient temperature, therefore, only a certain RF power can be carried, as the permitted tempera- Proceedings of the 53rd IEEE HOLM 2007 Conference on Electrical Contacts, September 2007 (Winner of Best Paper Award) ture level of the relay must not be exceeded. Typically this temperature is 85 °C, but for relays which are suitable for SMD soldering, a temperatures of 125 °C is possible. The switching capability of RF relays is determined by the design, the contact material used and the switching characteristics. Welded contacts or contact rivets – as used for signal or power relays – cannot be used in RF relays, as the contacts would generate discontinuities in the signal path and cause signal reflections. Whenever perfect RF performance is required welded contacts are not a viable choice. Most designs achieving low signal reflections use galvanized contacts. As there are only a few alloys suitable for switching contacts which can be deposited and the thickness of the precious metal layer (maximum 2…3 µm) which can be deposited economically is limited, the switching performance especially under load is modest. Especially when DC loads were switched, only a very limited number of switching cycles were achieved. Until now no low cost and at the same time high performing products were available. II. SWITCHING LOW LEVEL SIGNALS A. Basics If a relay is to interrupt a few amperes in the total range of frequencies from zero (DC) to 3 GHz, the mechanism for current interruption under these conditions has to be the primary consideration. At any rate, a switching arc is needed, which is not just an unwanted troublemaker as some people still believe, it is actually the ideal and unique switching element. If an arc did not appear spontaneously when a circuit is interrupted by opening the contacts, the current would be chopped discontinuously. As a consequence, the magnetic energy LI2/2 stored in the inductance of the circuit could not be dissipated in the arc but would have to be converted into capacitive energy CU2/2 in order to be stored in the stray capacitance C of the circuit. Because of the low value of C, a detrimental switching over voltage would appear even at quite low current values. Fortunately, the arc appears reliably and solves three problems at once: first, it limits the overvoltage LdI/dT, necessarily caused by any current change, to the value of its welldefined arc voltage; second, it converts the magnetic energy of the circuit LI2/2 into heat, while, third, its increasing resistance reduces the current continuously to zero until eventually the arc disappears when no longer needed - provided it is long enough not to be continuously sustained by the power source. B. DC Interruption The arc length required to interrupt current reliably depends on both the e.m.f. of the power source U0 and the current to be interrupted I0, i.e. the amperage when the contacts were still closed. By using known voltage/current arc characteristics [5], the critical arc length needed for interruption can be predicted easily. In the diagram of arc voltage vs. arc current, Fig. 1, only the points (0, U0) and (I0, 0) have to be connected by a straight line; the arc length belonging to the characteristic curve just touched by this "resistance line" gives the critical arc length required to interrupt the current in this circuit. U0 I0 Fig. 1: Voltage vs. current characteristics of arc burning on gold contacts in air [6]. C. AC interruption While DC interruption is an energy problem necessarily producing some overvoltage due to the circuit inductance when the current decreases but needed to convert the magnetic energy stored there, AC interruption is just a problem of timing because there is no energy stored in the inductance at the exact moment of current zero. Unfortunately, it is practically impossible to open the contacts exactly at current zero and to do it fast enough to prevent a breakdown caused by the recovery voltage. Fortunately, however, the arc which appears spontaneously carries the current during the half cycle of contact opening but ceases, also spontaneously, at the next current zero, thereby avoiding any overvoltage except the recovery voltage of the circuit - provided the recovery voltage is lower than a certain reignition value, which is not of relevance for low power signals. Certainly, the arc column does not disappear fast enough at current zero because the thermal time constant of a low current arc, e.g., 2 Arms, is of the order of 1 ms [7]. Therefore, the arc plasma does not lose its conductivity within the first millisecond after current zero until it has cooled down to about 3000 °C. But the problem is the supply of electrons from the new cathode, which was the anode just before and where no mechanism has been established, requiring very high positive ion density for field electron emission. Thus, in the very first moment after current zero, the recovery voltage appears across the arc gap, now in "reverse" direction. The very mobile free electrons of the hot arc plasma immediately move toward the new anode; but they are not replaced by other electrons leaving the cathode. Thus, "immediately" a zone of 1000 times less Proceedings of the 53rd IEEE HOLM 2007 Conference on Electrical Contacts, September 2007 (Winner of Best Paper Award) mobile and therefore practically immobile positive ions remains in front of the new cathode and the total amount of the recovery voltage drops across that positive ion zone, as shown in [8]. The density of these remaining ions, however, is far too low to cause any field electron emission from the cathode. Thus, the recovery voltage of the circuit is held by the residual ion sheet which, fortunately from our point of view, is able to withstand a certain voltage value "instantaneously". This value depends on very many parameters, but mainly on the temperature of this ion sheet, which is about the temperature of the adjacent cathode surface and therefore limited by the boiling point of the contact material. But it also depends on the way the contact surface works and the characteristics of the gas, of course. Materials such as Cu, Fe, and Ni yield various values of instantaneous reignition voltage depending mainly on their boiling point and the way they work. It is well known that this "instantaneous" reignition voltage is established within much less than 1 µs, i.e. faster than any usual recovery voltage. And, we also know that even much higher power frequencies than 60 Hz or 400 Hz can be interrupted in that way. D. RF interruption Little is known, however, about the mechanism of RF current interruption in the range of MHz and even GHz. Not even a single paper has been published on this topic so far. It has to be emphasized that there is not a quasi-steady long arc in air when such high frequencies have to be interrupted, since a current loop lasts less than a microsecond in the MHz range and less than a nanosecond in the GHz range. What is the state of matter 1 ns after contact separation? During the last millisecond before contact separation, the contact constriction area decreases, the contact resistance increases and so does the temperature of the contact site, up to melting point and beyond it. A molten metal bridge is formed by the surface tension of the melt, preventing contact separation at the very moment a gap would appear if there were no current. Instead, the bridge increases in both length and diameter while its temperature approaches boiling point. During that stage, the surface tension is decreasing while both the kinetic energy and the mean distance of the molecules are increasing until, at boiling point, the molten bridge disintegrates explosion-like. During these nanoseconds, a more or less continuous transition from liquid to gaseous (plasma) state is to be expected, followed by a rapid increase in temperature and a decrease in the density of the metal vapor. "Suddenly", the distance between the remaining solid surfaces increases while the liquid becomes vaporized or, mainly, is splashed away and the voltage across the contact gap jumps from the boiling voltage (0.75 V for Ag) to the anode arc voltage (about 8 V). "Suddenly", a so-called anode arc is established which is a rather unstable phenomenon existing inside the expanding metal vapor cloud. It is hardly known how many fractions of a nanosecond or even how many nanoseconds "suddenly" means. The anode arc, however, may last some time in the range between 1 ns to 10 µs depending on contact material, current and opening velocity, until at a contact distance of about 0.5...5.0 µm, it changes discontinuously into a so-called cathode arc, which is still a discharge existing in the expanding metal vapor cloud which remains from the disintegrated contact bridge but requiring some more arc voltage. Only 1...100 µs after contact separation or when about 10...1000 µm gap length is reached, the surrounding air diffusing into the arc column has replaced the expanding metal vapor and the discharge changes from the state of a metal vapor arc to a gas arc. Comparing this break arc history with the duration of RF current loops, it becomes evident that arcs at frequencies even much lower than 1 GHz attain their first current zero during the anode arc mode and the ambient of a MHz arc loop might still be metal vapor rather than air at current zero. Our limited knowledge about the short arc modes existing in an expanding high pressure metal vapor cloud is not sufficient to predict their current zero behavior. III. SWITCHING RF AND MICROWAVE SIGNALS No published information is available about make and break of RF and Microwave signals. As already described before, the most interesting question is whether the arc behaves like a DC arc due to the time constant of the arc which is longer than the period of the signal. Furthermore it is of vital interest, whether the arc is extinguished by the increase in the contact gap and/or by crossing current zero. A. Test setup The traditional way of monitoring the contact voltage and contact current is not applicable for signals in the GHz range, as the sampling rate of the fastest commercially available oscilloscope is currently 20 Gsamples/s. The test setup for measuring the break and make operation is given in Fig. 2. The signal generator and the amplifier available for these tests allowed a frequency range between 0.8 and 3 GHz to be covered, and a power range between 1 and 37 W. The following procedure was applied to investigate the make and break process. For breaking the contacts, power was applied to the closed contacts. The relay was activated and the input and output signals were compared. For making the contacts, power was applied to the closed contact, the relay activated and the signals on the closing contacts were monitored. B. Make contact Make operation is given in Fig. 3 for a RF load of 33 W @ 75 Ohm, frequency 1.4 GHz. In order to get an overview of all relevant phenomena, the overview was taken at a reduced resolution. In Fig. 3a, the input signal is given in pink, the output signal in blue. During the make operation, 3 steps can be observed in the output signal. Additionally after the first closing a bounce can be detected. The first step in the output signal appears when the contact bridge leaves the grounded position (Fig.3b). As the contact bridge approaches, a pre-strike arc is ignited (Fig.3c). Proceedings of the 53rd IEEE HOLM 2007 Conference on Electrical Contacts, September 2007 (Winner of Best Paper Award) When both contacts are closed the output signal is similar to the input signal reduced by the intrinsic loss. grounded position. Grounding the contact bridge results in a further increase of the RF isolation by about 15 dB. Fig. 3a: Overview make contact 33 W @ 50 Ohm 1.4 GHz, Amplitude 20 V/div - Time base 20 µs/Div Fig. 3b: 1st step – contact leaves grounded position. Amplitude 20 V/div - Time base 2 ns/Div Fig. 2: Test setup for the investigation of the make and break process on RF and microwave signals. C. Break contact Break operation is given in Fig. 4 for a 25 W @ 50 Ohm load, frequency 1.4 GHz. In the overview picture Fig. 4a was taken at reduced resolution, in order to get an overview of all relevant phenomena during the make operation. While the dark signal shows the input power, the light signal shows the output power. The three interesting parts for breaking the contact are shown in Fig.4a to Fig.4d. Fig. 4b represents the first step in the output signal, in which amplitude is suddenly reduced. This step is caused by the opening of the first contact of the contact bridge as an arc appears in the contact gap. Both input and output signal are still in phase, as the arc can be considered as resistor. The arc burns continuously and the signal has several current zero crossings. In Fig. 4a further reduction of the output signal can be detected, which is caused by the opening of the second contact of the contact bridge. As the arc does not have enough e.m.f. available it extinguishes at the current zero crossing. At the same time a phase shift can be detected, at the moment when the resistive arc extinguishes and the signal transfer is based on capacitive conduction. Fig. 4c shows another reduction of the output signal. This last step appears when the open contact bridge reaches the Fig.3c:2nd step – pre-strike when the contact approaches. Amplitude 20 V/div - Time base 1 ns/Div Fig.3d: 3rd step – contacts are closed. Amplitude 20 V/div - Time base 1 ns/Div Fig. 3: Make operation of a high frequency signal 33 W @ 75 Ohm – 1,4 GHz Proceedings of the 53rd IEEE HOLM 2007 Conference on Electrical Contacts, September 2007 (Winner of Best Paper Award) IV. Fig.4a: Overview-Break contact 25 W @ 50 Ohm load, 1.4GHz. Amplitude 20V/div-Time base 200µs/Div Fig.4b: 1st step – Contacts open and arc appears. Amplitude 20 V/div - Time base 2 ns/Div EXPERIMENTAL A. Test device All hot switching tests were performed on HF3 relays [10, 11]. The HF3 relay is given in Fig. 5. The HF3 relay is a SPDT (single pole double throw) relay designed for high frequency applications. The principle design of the carrier shown in Fig. 5. is based on a Y structure. The contact carrier is designed on the coplanar principle, the switching contact bridge is based on microstrip design. Besides the terminals for the coil connections and the three terminals for the SPDT configuration, there are eight additional terminals to provide optimum shielding. Fig. 5: Design of a low cost ultra-miniature RF relay (HF3) – dimensions 14.6 x 7.2 x 10 mm. Contacts are formed as bridge contacts, grounded in open condition (75 Ohm). HF3 Relay Coil characteristic Coil voltage Coil power consumption Contacts: Type Fig. 4c: 2nd step – arc extinguishes – phase shift. Amplitude 20 V/div - Time base 1 ns/Div Fig.4d: 3rd step – contact bridge reaches ground position. Amplitude 20 V/div - Time base 2 ns/Div Fig. 4: Make operation of a high frequency signal 25W @ 50 Ohm – 1.4 GHz Monostable and Bistable 1 or 2 coils 1.5 ... 24V 70 ... 140 mW 1 change over 1 form C / SPDT Contact material Gold over silver Maximum carrying power 50 W dc, ac and RF Maximum switching power 50 W dc, ac and RF Mechanical endurance 107 operations Electrical endurance 5 x 105 operations RF Characteristics 50 Ohm 75 Ohm Isolation - 72 dB - 72 dB @900 MHz min. - 56 dB - 56 dB @3 GHz min. Insertion loss - 0,15 dB - 0,09 dB @900 MHz max. - 0,30 dB - 0,20 dB @3 GHz max. V.S.W.R. 1,16 1,11 @900 MHz max. 1,25 1,15 @3 GHz max. Table 1: Characteristics of the newest generation of electromechanical RF relays Proceedings of the 53rd IEEE HOLM 2007 Conference on Electrical Contacts, September 2007 (Winner of Best Paper Award) C. Test evaluation The contact resistance, the number of occurrences of bridging, and the number of occurrences of sticking during the electrical endurance tests were evaluated for all AC load tests. For the evaluation of the contact resistance values, the endurance test is divided into 50 intervals. For every interval the minimum, average and maximum contact resistance values are plotted. Contact resistance limit RF load tests were performed without monitoring. After the endurance tests, the relays were opened and the contacts were analyzed using an SEM (scanning electron microscope). V. Contact circuit resistance [ohm] 10 1 0,1 0,01 20 10 1 0,1 0,01 0 20 40 60 80 Number of operations [thousands] 100 RESULTS A. AC Loads 1) 30 V and 1 A -30 VA @ 50 Hz 100,000 operations lifetime were expected at 30 V, 1 A @ 50Hz. Fig. 6 shows indeed that the contact resistance remained low and only a little erosion occurred. 0 2) 30V and 1 A – 30 VA @ 1000 Hz The same load was tested at 50 Hz and 1000 Hz, in order to investigate the impact of the load frequency. Fig. 6 and (Fig. 7) however do not show any relevant differences of contact resistance and erosion. The load was interrupted at any point on the sine wave. A short arc (Fig. 8) – duration about 1 µs – appeared and extinguished as soon as the second contact of the contact bridge opened. The duration of the arc, as already seen before, only depends on the asynchronous opening of the bridge contacts. Contact circuit resistance [ohm] B. Tests performed The aim was to compare the switching performance of the HF3 relay at: - AC loads: - 30 V and 1 A – 30 VA @ 50 Hz - 30V and 1 A – 30 VA @ 1000 Hz - RF loads: - 25 W @ 50 Ohm - 0.8, 1.5 and 3 GHz- 37 W @ 50 Ohm – 1.5 GHz - 37 W @ 75 Ohm – 1.5 GHz Number of operations: 1 x 106 Contact material • Fixed contacts: 3 µm Ag + 0.5 µm AuCo 0.5 • Moveable contacts: 3 µm Ag + 0.5 µm Au Ambient temperature: 85°C for AC tests, 23°C for RF tests 40 60 80 100 Number of operations [thousands] Fig. 7: Resistive load 30 V and 1 A @1000 Hz. Minimum, average and maximum contact resistance values (top figure) and SEM photos of the bridge contacts. Power [W] Impedance=50 Ohm Impedance=75 Ohm [W] U [Vp] I [mA] U [Vp] I [mA] 1 10 141 12 115 5 22 316 27 258 6 24 346 30 283 9 30 424 37 346 10 32 447 39 365 20 45 632 55 516 25 50 707 61 577 37 61 860 74 702 50 71 1000 87 816 Table 2: Correlation between RF power and impedance on the one hand and voltage and current on the other hand. Fig. 6: Resistive load 30 V and 1 A @50 Hz. Minimum, average and maximum contact resistance values (top figure) and SEM photos of the bridge contacts. B. RF Loads While all AC electrical endurance tests were fully monitored (measurement of contact and isolation resistance after every single operation) this was not possible for hot switching RF loads. On these relays contact resistance as well as insertion loss was measured before and after the switching tests. In addition the contacts were analyzed using a SEM Proceedings of the 53rd IEEE HOLM 2007 Conference on Electrical Contacts, September 2007 (Winner of Best Paper Award) (Scanning Electron Microscope). In RF engineering, the load is not given in voltage and current values, but it is usually defined by the power and by the system impedance. A correlation between power and voltage and current respectively is given in . Voltage 20V/DIV 2) RF Load 25 W @ 50 Ohm – 3GHz In order to judge the impact of the load frequency the same test was performed also at a frequency of 3GHz. As the pictures in Fig. 10 indicate, the erosion is approximately the same. The precious metal layer is still present in the contact area. Arcing during the hot switching tests showed some impact on the contacts but still not critical. Detail Current 0.5A/DIV 500µs/DIV 10µs/DIV Fig. 8: Voltage and current during break contact. Fig. 10: RF Load 25 W @ 50 Ohm – 3 GHz - SEM photos of the contacts after 1 million operations. Fig. 9: RF Load 25 W @ 50 Ohm - 0.8 GHz - SEM photos of the contacts after 1 million operations. 1) RF Load 25 W @ 50 Ohm - 0.8 GHz The HF3 relay was switched for 1 million operations (make and break). The power was monitored during this time. After the tests the insertion loss was measured once more and than opened for SEM analysis. The analysis of the contacts showed some minor erosion of the contacts after the hot switching tests (Fig. 9). Considering the load of 25 W and the 1 million operations the results are remarkable. Only the gold layer is removed in the contact area, but the thin silver layer was able to withstand the arcing. 3) RF Load 37 W @ 50 Ohm – 1.5 GHz Although the power which has to be carried and switched in systems with a 50 and 75 Ohm impedance is the same, there are significant differences in the voltage and current. While the 37 W in 50Ohm systems result in a contact load of 61 Vp / 860 mA, in the 75 Ohm system the contact load is 74 Vp / 702 mA. Apart from the different contact load the design of the 50 and 75 Ohm type is only slightly different, but might have a significant impact for power switching. While the relays suitable for 50 Ohm impedance have a bifurcated contact bridge, the 75 Ohm version has only a single contact bridge. In order to compare the hot switching performance of both versions, tests at 37 Watt were performed. The results achieved for hot switching of 37 W @ 50 Ohm - 1.5 GHz (Fig. 11) show remarkably low contact erosion. After 1 million operations no significant contact erosion can be detected. Measurements of the contact resistance values before and after the hot switching tests also did not show any differences. 4) RF Load 37 W @ 75 Ohm – 1.5 GHz The contact erosion for the 37 W load @ 75 Ohm impedance shows very similar results to those achieved at 50 Ohm. The erosion is visible but still in the precious metal layer (Fig. 12), and therefore absolutely not critical. Just the opposite in fact, as the erosion is remarkably low for Proceedings of the 53rd IEEE HOLM 2007 Conference on Electrical Contacts, September 2007 (Winner of Best Paper Award) this very high load of 37W which was hot switched for 1 million operations. 0.0 -0.1 [dB] -0.2 -0.3 -0.4 -0.5 0 500 1000 1500 2000 2500 3000 Frequency [Mhz] NO-Init NC-Init NO-Finale NC-Finale Fig. 13: Change of insertion loss caused by hot switching - Load 37 W @ 50 Ohm – 1.5 GHz VI. Fig. 11: RF Load 37 W @ 50 Ohm – 1.5 GHz - SEM photos of the contacts after 1 million operations. 5) RF Load 37 W @ 75 Ohm – 1.5 GHz The contact erosion for the 37 W load @ 75 Ohm impedance shows very similar results to those achieved at 50 Ohm. The erosion is visible but still in the precious metal layer (Fig. 12), and therefore absolutely not critical. Just the opposite in fact, as the erosion is remarkably low for this very high load of 37W which was hot switched for 1 million operations. DISCUSSION AND PROSPECTS The HF3 relay has been designed to optimize RF characteristics. For good RF isolation in the open position, bridge contacts were used and grounded in the open state. The electrical endurance tests performed with AC and RF loads showed that bridge contacts improve the switching performance even in load areas where bridge contacts are not necessarily required in order to interrupt the load circuit, e.g., at higher DC voltages. A. AC loads According to published data, current interruption of the loads was expected at power frequency current zero. For the tested loads, this breaking mechanism was not confirmed. This can be explained for the 30 V / 1 A load. When the second contact of the contact bridge opened, the minimum voltage to maintain an arc was determined to be about 35 V. As this voltage was not available, the arc extinguished immediately. For the 30 V / 1 A load, the load was interrupted at any point of the sine wave. The typical arcing time was always less than a few µs. Current 0.5A/DIV Fig. 12: RF Load 37 W @ 75 Ohm – 1.5 GHz - SEM photos of the contacts after 1 million operations. Contact resistance values and contact erosion are good indicators for the RF characteristics. Hot switching does not have an impact on isolation but on the insertion loss values, due to potential changes of the contact surface. The insertion loss before and after a hot switching test of 37 W @ 50 Ohm for 1 million operations was measured. The results presented in Fig. 13 show no change in the insertion loss, caused by the switching operations. Voltage 20V/DIV 5µs/DIV Fig. 14: Voltage and current during break contact. Proceedings of the 53rd IEEE HOLM 2007 Conference on Electrical Contacts, September 2007 (Winner of Best Paper Award) A. RF loads For the first time ever, the switching of RF signals was investigated in detail. The analysis of the break and make operation showed, that arcing appears in the same voltage/current range as for DC loads. As long as the voltage is less than the minimum arc voltage no arcing occurs. When the load is increased at about the same load levels as for DC loads an arc appears during make or break contacts. During break operation of the contacts, an arc is ignited when the contact opens and burns for the duration of several periods of the signal – similar to a dc arc. This can be explained with the thermal time constant of the arc, which is longer than the duration of a period. As the contact gap increases, the minimum arc voltage increases and the voltage required for the arc cannot be maintained anymore. Finally the arc extinguishes in a similar way to an AC arc, during system frequency current zero. At the same time when the circuit is interrupted the signal makes a phase shift from a resistive to a capacitive circuit. Although loads up to 37 W at both 50 and 75 Ohm impedance were tested, the arcing during make and break did not cause significant damage to the contacts. This is a remarkable result, as 1 million operations were made under load. The excellent performance was also confirmed by the unchanged contact resistance and insertion loss values of contacts which were subjected to the hot switching tests. VII. CONCLUSIONS Electromechanical RF switches provide significant advantages compared to solid-state RF switches. As the isolation is made from air or plastics, excellent isolation can be achieved regardless of the ambient temperature. Metal contacts provide the lowest possible insertion loss. Furthermore, electromechanical RF switches are able to transmit RF signals without signal distortion and signal modulation. Electromechanical RF relays provide real broadband characteristics over the whole frequency range from DC to GHz. For the first time ever, detailed investigations into the switching phenomena of RF signals were performed. It was found that making or breaking RF loads causes arcs which burn over several periods of the signal, as the time constant of the arc is longer than the cycle time of the signal. The arc extinguishes as the contact gap increases by frequency current zero. Excellent switching characteristics were found for • DC loads up to 60 W • AC loads up to 30 VA at 50 Hz and 1kHz • RF loads up to 37 W @ 50 and 75Ohm impedance and frequencies up to 3 GHz Due to the bridge contacts, which were initially designed in order to optimize RF isolation values, excellent switching performance was also achieved. The two contact gaps in series resulted in a minimum arcing voltage of a total of more than 30 V. All loads with an open circuit voltage less than approximately 35 V can be interrupted with no or only a very short arc. The arcing time is defined by the synchro- nism of the contact bridge. Due to manufacturing tolerances, the two contact spots of the bridge contacts do not open at exactly the same time. Then the arc burns on the contact pair which opens first and until the second contact pair open too. The switching mechanisms can be characterized as follows: • On AC loads, breaking at any point on the sine wave was detected, even at 360 V peak voltages. On all AC loads tested, the arc extinguished before the power frequency current zero. • On RF loads basically a mixture between DC and AC was found. The arc burns over several periods and finally extinguishes at power frequency current zero. Regardless of the mechanism involved, the switching characteristic is remarkable for the present design. Before the tests were performed, this level of switching capability for DC, AC and RF loads, with such a thin precious metal layer, would not have been expected. VIII. ACKNOWLEDGMENT The author thanks Dr. Klaus Poefel, Mike Ludwig and Bert Krapfenbauer for conducting the tests and preparing the analyses. WERNER JOHLER Werner Johler received his Ph.D. degree in electrical engineering from the Technical University of Vienna, Austria in 1988 and his MBA in 2003. From 1984 to 1988 he was a scientific staff member at the Institute of Switchgear at the Technical University of Vienna. Since 1988 he has been with Tyco Electronics in Au, Switzerland. He is Director for AXICOM relays. He has been Chairman of the Technical Committee TC94 “All or nothing relays” within CENELEC since 1999 and member of the Board of Directors of IRSTC (International Relay and Switch Technology Conference) since 2002 (former NARM). He received the Scientific Award from the state of Vorarlberg, Austria in 2004, the Albert Keil Award from VDE in 2005 and the IEC 1906 award in 2007. He has published more than 60 papers on relay technology, miniaturization of electromechanical devices, contact physics and reliability of electromechanical relays REFERENCES [1] [2] Robert D. Malucci, “The Impact of Contact Resistance on High Speed Digital Signal Transmission” 48th IEEE Holm Conference on Electrical Contacts 2001 C. Wheeler: “ The Superior RF Switch Technology- Part I – The Advantages of MEMS“, MagLatch Proceedings of the 53rd IEEE HOLM 2007 Conference on Electrical Contacts, September 2007 (Winner of Best Paper Award) [3] [4] [5] [6] [7] Design Guide, SectionII-Types of Microwave Switches, Dow-Key Microwave Corporation, Ventura, CA P.D. Grant, R.R. Mansour, M.W. Denhoff: “ A Comparison Between RF MEMS Switches and Semiconductor Switches”, Can.J.Elect.Comput.Eng., Vol.27, No.1, Jan.2002 W.Rieder: “Low Current Arc Modes of Short Length and Time: A Review, IEEE Transactions CPT, Vol.23, No.2, June 2000; pp 286 –292 R. Holm: Electrical Contacts, Springer-Verlag, 1967 A. v. Engel, M. Steebeck, "Messung des zeitlichen Verlaufes der Gastemperatur in der Säule eines Wechselstrom-Luftlichtbogens", Wiss.Veröff. Siemens, Vol. 12, pp. 74-89, 1933. [8] 9] [10] [11] [12] J. Slepian, "Extinction of an AC arc", Trans. Am. Inst. El. Eng., Vol. 47, pp. 1398-1408, 1928. W. Johler, W.F. Rieder, "Switching RF Signals”, 50th IEEE Holm Conference on Electrical Contacts, 2004, pp. 168 ... 175. W. Johler: „High Performance Ultra-miniature High Frequency Relays“, Proceedings 51st International Relay Conference, 2003, pp. 5-1 ... 5-11. W. Johler: “RF Performance of Ultra-miniature High Frequency Relays“, Proceedings 49th IEEE Holm Conference on Electrical Contacts, 2003, pp. 179 ... 189. P.G. Slade: Electrical Contacts: principles and applications; Marcel Dekker, 1999. Proceedings of the 53rd IEEE HOLM 2007 Conference on Electrical Contacts, September 2007 (Winner of Best Paper Award)