A proposed root cause for the "Withering Filters" phenomenon

Fig. 1 Photo by Jerzy,
SP6FPY.  (Click to enlarge)

After having examined the failure mechanism of more than 100 ceramic filters that have succumbed to electromigration, I think I have collected enough data to propose a root cause for the phenomenon that has caused failure of a very large numbers of filters. I should gratefully thank the many fellow radio amateurs that have sent photos of their filters to me, adding to my database on the subject. Manufacturers might want to consider changing their production methods to avoid the chain of events that I propose that leads to this phenomenon.

So, let's look at the data I have collected and a proposed theory that explains the facts:

Fig. 2. Note the water droplets.

1) In all of the ceramic filters I have examined, there was a surprising quantity of water inside the plastic case (take a look at Figs 1, 2 and observe the signs of water presence on the elements and inside the black case, also the whitish flux residue on the case of the adjacent filter in Fig. 1, which shows that a quantity of water evaporated from that surface, leaving it behind). The water inside the case causes the filter to fail soon after the transceiver circuit applies DC voltage to the input, and mainly the output pins of the filter. The quantity of water is such, that can't be attributed to a filter manufacturing flaw: it's just too much. I can't imagine that the filter manufacturers would have ignored it. Also, I haven't found water inside some new filters I have "dissected". So, there must be another reason for this quantity of water inside the filter.

2) Where does the water come from, if not from a filter manufacturing flaw? I think the answer lies with the newer technologies and materials used for the fabrication of populated printed circuit boards (PCBs). Briefly, the new methods of wave soldering with non-Pb solders extensively use organic acid (OA) fluxes. Those fluxes are removed from the PCBs after wave soldering by pressure rinsing with hot water. The rinsing process takes some time to complete.

3) I think that the water ingress can be explained thus. The rinsing process uses hot water (about 70 degrees Celsius), sometimes with ionic additives. This (as well as the previous soldering process) causes the (so far dry) air inside the plastic filter case to expand, creating positive pressure inside the filter case. This results in air escaping from inside the filter case, as the temperature softens the plastic case and compromises the (not so great to begin with) sealing at the bottom of the filter case.

When the PCB exits the rinsing process, water is held by affinity under the filter case, between the filter and the PCB surface. The filter starts cooling down, and this causes the pressure inside the filter to drop under the atmospheric pressure. This negative relative pressure tends to draw a small quantity of the water under the filter inside the filter case, through the very fine cracks at the filter case seals that the positive air pressure (when the filter was hot) created. The quantity of the water is some micrograms, but it's more than enough to cause trouble. The board is then dried, but the accident has already happened. The water is already inside the filter.

4) Once inside the filter, the water can't escape and remains trapped inside the case. Some of it condenses or moves on the filter plates, starting the electromigration process where the electric field (due to the externally applied voltage) reaches sufficient value. This most often occurs at the corners of the thin output ceramic plate (element) at the output of the filter, which is only about 0.35 mm thick. This leads to the failure of the filter after some time.

5) What might be a solution? A slight modification of the rinsing - drying process might help in avoiding this phenomenon.

If the PCB is dried IMMEDIATELY after exiting the rinsing chamber with pressurised HOT air, then the water might not get inside the filter case, as the pressure difference doesn't occur this way before the water has been removed from under the filter.

I have no further knowledge of the exact conditions used in any particular case, so the above proposal may not always apply. But each manufacturer could certainly review the method used and modify it accordingly, so as to avoid the conditions that drive water inside the filter case (and possibly inside other "sealed" components, such as miniature relays).

If someone with more knowledge of the PCB processing methods wishes to add (or correct!) something in the above, please send an e-mail to sv8ym@raag.org.

73 DE SV8YM

Addendum: The above suggest that if you repair (clean and dry thoroughly with hot air) or replace a filter that has gone bad through water ingress during the rinsing procedure with a brand new one that hasn't come in contact with water, you may not have problems in the long run, even if you don't also install DC-blocking capacitors. That is so because you need BOTH water AND a voltage gradient to start electromigration. A completely dry filter probably can withstand 7-8 V indefinitely (as experience with older rigs shows - the PCBs were washed with organic solvents a long time ago, in the paleolithic Pb era!).

 

Lift the withering spell off those ailing ceramic filters!

As promised, here is the solution to the "withering ceramic filter" problem. The mod here was performed in a TM-D710, but the concept presented holds for all similarly affected rigs.

First of all, take a look at the photo on the left (click on the pictures to enlarge). There are four ceramic filters in the '710, two (wide-narrow) for each receiver.

Our first step is to cut two PCB traces converging at each filter's input and output pins, bridge them, and then solder a size 1206 or 0603, 0.1μF / 25V or 50V SMD ceramic capacitor from each bridge to the filter's input and output pins (8 capacitors in all for the '710). This way, the DC switching voltage is blocked by the capacitor, that passes only the small AC (signal) component into and out of the filter. Removing the DC bias from the filters is all that is needed to avoid filter failure.

The first photo shows (with the red arrows) the eight traces to be cut with a very fine-tip grinding tool. Notice that there are 16 cuts for the 710, eight are shown completed, the other eight are in the red circle on the left (the circuit topology is identical).

Take a look at the second photo (click to enlarge), showing  the completed job for a filter set. As you see, fine wire bridges have been soldered, bypassing the cut-off portion of the circuit that was going to the filter pins, and the ceramic capacitors are also in place, soldered between the wire bridges and the filter's input / output pins.

The third photo (click to enlarge) shows the completed job for all four filters.

I must say, that although very simple, this mod is a bit difficult because of the very small dimensions of the traces and SMD components. Extreme care is required, along with the proper tools and ability to work with them. It's easy to make a disaster out of the job, so please be careful!! If you don't feel up to the job, have somebody qualified do it!

The same concept holds for any transceiver that has DC bias voltage applied to ceramic filters. The objective is to block the DC bias with the capacitor, but otherwise maintain intact the circuit's filter switching functionality (if of course there is one), which in most rigs is accomplished by properly biasing switching diodes.

Please note: If your filters have already deteriorated, they must be replaced or repaired before the mod. The symptoms of deteriorated or failed filters are: "deaf" receiver, crackling noises heard during reception of otherwise full-quieting level signals in FM receivers, crackling noises with no antenna connected in SSB receivers.

Good luck!

It's getting hotter...

In a previous post I discussed the problem of imperfect contact between the power amplifier modules and the heat sink in an TM-D710. Last weekend I worked on three more D-710s, which had developed the infamous "withering filter" illness, and discovered that the imperfect contact matter may reach extreme proportions!

Take a look at the photos (click on them to enlarge). The power modules barely made contact with the heat sink, only near the affixing screws. Less than 20% of the surface was in contact! Not a very good scenario for the longevity of the modules!

The other two rigs showed about 25% and 10% loss of contact, and only at one of their two modules, respectively.  The photos are of the "worst case" rig. As an afterthought, I should have tried to see which is at fault, the modules or the heat sink, but I foolishly didn't do that. I will in the next transceiver that comes along, and let you know. My wild guess is that the die-cast heat sink machine finishing (leveling) at that spot is at fault; but  I am certainly not sure.

What to do if your rig suffers this way? A couple of drops of gear oil will probably slide your problems away and cool things off, see the older post.

Not Flat Enough - or, Oiling a Transceiver

During an operation to install a new ceramic IF filter in a Kenwood TM-D710, I had to remove the rig's PCB from the diecast casing (heat sink), to gain access to the filter's pins (more on the filter issue to come soon in the "Mysterious Case of the Withering Filters"). The rig hadn't ever been serviced before. After unscrewing all of the screws, including those of the hybrid power modules, I lifted the PCB. I immediately noticed that the heat transfer compound spread indicated that the contact area of the heat sink (or that of the hybrid?) wasn't exactly flat, and a comparatively large spot hadn't been making contact at all (clicking on a photo enlarges it). The corresponding surface of the hybrid had been covered with the right thickness of thermal transfer compound, but it never made contact with the heat sink at all. This can be a serious situation, endangering the expensive hybrid amplifier module because of reduced heat transfer and the unequal thermal and mechanical stresses that develop as a consequence.
Using more heat transfer compound in such cases doesn't help much, because the compound itself is not a spectacular heat conductor, it just helps by filling up microscopic surface irregularities at the contact interface and replacing air, that would be an even worse heat conductor. The contact surfaces must be clean and flat and the heat transfer compound layer thickness must be exceedingly small for the compound to serve its purpose effectively, and this fact is emphasized in every power semiconductor manufacturer's application notes.
I thought of another heat transfer agent that is frequently used in other cases (for example, in oil-filled dummy loads). The right viscosity (heavy) oil would fill the gap nicely, stay there due to the forces of affinity and help transfer the heat to the heat sink at the problem spot. During reassembly, I used a couple of drops of SAE 80W-90 gear oil on the trouble area and gently pushed down the hybrids with a sliding motion, so all of the air trapped between the contact surfaces came out. The screws affixing the hybrids must be tight enough but not overtightened, as this could cause the ceramic substrate inside the module to crack. Just apply enough torque so that the split-ring on the screws closes, and then about an eighth of a turn more - but not much more!
This isn't a new trick for this dog - during my misspent youth, I used vaseline jelly in place of the (then) expensive and hard-to-obtain silicon grease. A couple of linear power supplies I built then (early eighties) are still in daily operation, with the same old 2N3055 and 2N3772 pass transistors "greased" this way.