Ceramic filters can be either of coaxial form or dielectrically loaded waveguide. Ceramic filters as discussed here are TEM coaxial structures using quarter wave resonators and either capacitive or magnetic coupling. The quarter wave resonator is formed from a block (usually square) of ceramic material with a hole in the middle parallel to it’s length. The outside, one end, and the hole are coated with a conductive material (usually silver) to form a shorted quarter wave length of transmission line. The length -vs- frequency of the line is determined by the dielectric constant of the ceramic. Essentially, ceramic filters are a form of distributed filters. They are much more rugged than LC filters which are susceptible to vibration and microphonics. Ceramic filters are much less expensive to produce than either LC or cavity type filters, due to the reduced labor, material, and machining costs. In addition, most modern ceramic materials are extremely temperature stable with temp coefficients of < 5ppm. Typically ceramic filters are much less temperature sensitive than either LC or Cavity filters, even after compensation.
Coaxial ceramic filters fit within the frequency range of about 500 MHz to 6000 MHz. This broad range is accomplished by using ceramic material of various dielectric constants to optimize the size of the resonators. With magnetic coupled filters, bandwidth is limited to about 8%, but magnetic coupling allows broad and deep stopbands, with 80 dB or more to 6 GHz not uncommon. Capacitively coupled filters can achieve bandwidths of 20% or more, but the stopband attenuation is limited to about 40 dB due to parasitic coupling at the front of the resonator. This problem becomes more pronounced with smaller resonators and higher frequencies.
A filter’s insertion loss is determined by the unloaded Q of the resonator, the bandwidth of the filter, and the number of resonators. Bandwidth and number of resonators are usually determined at the system level, and are given for a specific filter. This leaves unloaded Q as the primary loss determination between one filter type or another. On ceramic filters Q is comprised of a dielectric component (Qd) and a conductive component (Qc). The resultant total Q (Qt) is calculated as 1/Qt = 1/Qd + 1/Qc. Qd is the loss of the ceramic material itself and is typically in the 10’s of thousands. However, there are some materials that are lossy but provide a required dielectric constant to achieve a certain frequency range. Qc is solely determined by the coating on the outside of the ceramic. This is typically in the 100’s, therefore the process of coating the ceramic will be the primary determining factor of filter loss. At ComNav, have developed a proprietary coating process which we do in-house. This process has consistently proven to yield about 30% higher Q than any other competitive parts measured.
The rejection or attenuation a filter can achieve is determined by the number of resonators, how the resonators are connected, and the quality of the ground of the circuit in which the filter is used. As mentioned previously, magnetic coupled filters can achieve extremely high levels of attenuation if grounded properly. To assist in this grounding ComNav has developed a thru-hole mount package that provides superior and consistent performance in the end circuit. By grounding along the length of the filter, and connecting to the filter on the opposite side of the board, additional shielding and reduced leakage is achieved. This is accomplished in a open lead frame package at frequencies in the 3GHz and higher range. In addition to grounding ComNav has also developed proprietary pole/zero and cross-coupled circuit typologies which optimize the stopband by placing attenuation where it is needed. This reduces the number of resonators required and further reduces the insertion loss.
Yes. Magnetic coupled filters are not susceptible to the parasitic capacitance between the resonators that affect the performance of capacitively coupled filters. Even in leadless SMT packages with proper grounding we have consistently been able to achieve 70-80 dB rejection specs.
Ceramic filters are basically painted rocks. The size and shaped of the rocks determine the filters?performance. Due to this rugged construction, the filter is not sensitive to shock, vibration, temperature, or humidity. The only environmental effect that can distort a filter is condensing humidity on some types of designs. This is a temporary effect, after the filter dries out it will come right back into spec with no long term damage. Once a filter is tuned to spec, short of hitting it with a sledge hammer, it is virtually indestructible.
The temperature performance of ceramic filters is determined mostly by the ceramic material itself, and to a lessor extent the coupling, loading and other external components. Over the years, we have worked with our ceramic suppliers and have developed materials with the lowest possible temperature coefficient that is a compromise across our various construction techniques. The worst material we use has a temperature coefficient of 5ppm. At 5 GHz this translates into a temp drift of 1.5 MHz across our standard temp range of -40C to +85C. Most of the materials we use have temp coefficients of 1 ppm or less. This reduced temperature drift means that we can squeeze additional performance out of the filter. Since we do not have to compensate for significant temperature drift we can design closer to the spec.
Most ceramic filters are narrowband bandpass filters. But bandstop filters also lend themselves nicely to ceramic construction. Since ceramic resonators are basically narrowband devices, they do not lend themselves well to broadband lowpass or highpass filters. These devices are much better suited to LC or printed structures.
The problem you are asking about is silver adhesion. Adhesion is a big problem with ceramic filters. The coating process is a very delicate operation. The slightest process variation, and the first thing to go is adhesion. This also affects the unloaded Q of the resonator. Most ceramic resonators, including ComNav’s, use a thick film silver sintering process to coat the resonators. We were lucky in that we had 3 years of pure R & D time to develop and refine our process and discover what can go wrong. In addition, by having the process in-house, we know immediately when something goes wrong and we fix it. As a result, over the last couple of years we have refined our process to the point where we can no longer measure adhesion, (either the ceramic or the pull tester breaks). To answer your question, no you are not doing anything wrong, and most likely your filter supplier didn’t either, as the problem started before they even received the resonators. A properly coated resonator is impervious to soldering temperatures, fluxes and normal bench handling. However an improperly coated resonator can peel, have silver leach off during soldering, or degrade over time if humidity can get under the silver and pull the silver away. At ComNav, we qualify every batch of resonators before they go into production with sampled Q and pull tests.
Over the years, we have come upon 3 distinct failure modes that can be induced by customers, one temporary, and two fatal modes. I will not go into bad coating problems since I consider that to be our problem, not yours. The temporary mode was mentioned previously, which is condensing moisture. Some construction techniques use a pin and bushing assembly to perform the impedance matching in and out of the filter. Electrically this makes up a loading capacitor. When moisture in the form of droplets get in the loading capacitor it will change the capacitance and distort the filter’s tuning. However, once it dries out it will be back to its?original tuning. The fatal modes are silver bubbling, and chipping. Silver bubbling occurs when the filter mounted on a PC board is cleaned with an ultrasonic cleaner. The silver coating is bonded to the ceramic with a glass frit. The molecular vibration of the ultrasonic cleaner more than 5 minutes is strong enough to break the glass frit under the silver. Eventually the silver will bubble, a small hole will appear, water vapor gets in, and over time shrinks and expands and pushes the silver away from the ceramic. The worst thing about this problem is that it is a time bomb. It can take years to develop and when it does, can destroy the filter, and vastly degrade the performance of the system with a problem almost impossible to troubleshoot. Our recommendation, never under any circumstances, use ultrasonic cleaners on boards with ceramic filters or any thick film component, including resistors. The third failure mode is obvious but I’ll mention it anyway, that of chipping. Ceramic along with it’s thick film coating are very brittle. If dropped on a hard surface it can chip just like a ceramic ash tray or figurine. This will distort the tuning of the filter.
In some cases, yes. Ceramic filters can achieve unloaded Q’s of 1000 or more in 12 profile resonators. So the insertion loss and stopband performance are relatively close between cavity and ceramic typologies. There have been numerous instances where our customers have replaced machined cavity filters with ceramics. In some cases not only is there a size and price advantage, but also a performance advantage. By using cross coupling and pole/zero circuits we have been able to achieve similar or even better performance with fewer resonators. There will always be a trade-off because nothing is for free. Usually it boils down to the question of whether that 0.5 dB less of insertion loss is really worth an extra $200 to $400 per filter, which is usually the price difference between a cavity and ceramic filter.
Ceramic filters lend themselves very well to diplexing. The high impedance inputs can be prototyped to provide both contiguous and non-contiguous crossovers with a full match on the opposite or output end of the filters. Bandpass/bandpass diplexers are most common, but bandpass/notch diplexers also work well. Notch/Notch diplexers, classically called duplexers, are a little more difficult and in some frequency ranges are not practical. But we have built them for several of the standard communications bands.
Since ceramic filters for the most part are relatively low impedance, arcing from high voltages is generally not a problem. The main limiting factor will be how hot can the part get before damage occurs. With this in mind, the filter’s insertion loss will be a prime determinant of power handling and will determine the temperature rise caused by the dissipated power. There can in fact be instances were a small 4 mm leadless SMT filter can handle more power than a thru-hole mount 12 mm filter. It will all depend on loss, bandwidth, heatsinking, duty cycle and will be different for each device. I use a general rule of thumb, that a 4mm filter can handle 2 watts, 6 mm 5 watts, and a 12 mm filter can handle 20 watts. But each type will probably handle more depending on the circumstances. Generally a thru-hole package, grounded to a good heat sink can double the power handling ability given in rule of thumbs.
The main limiting factor in reflow is not the ceramic, but the carrier board, and soldering of internal components. Our internal construction uses SN-95 or SN10 solder depending on the filter type. The carrier or insulating board will initially discolor, then delaminate, then boil, when exposed to excessive temperature. As limits, we recommend not exceeding the following time/temps: 215C/60 sec, 230C/30 sec, 245C/15sec, 260C/10sec.