Thursday, October 18, 2012

Smoke and flames from my IFR-1000S...

I'd not really intended to have two posts in a row about repairing service monitors, but fate/opportunity intervened...

Figure 1:
The gray, charred stump of the failed tantalum capacitor in the center
of the image, just above the potentiometer shaft.)  The original,
nylon extension shaft broke several years ago and here, it is
shown having been mended with two overlapping
layers of heat-shrink tubing.
Click on the image for a larger version. 
The other night I had a few minutes to spare and I decided to take a quick look at my old IFR-1000S service monitor:  I'd remembered that the last time I dragged it to a repeater site there was something about it that was flaky - but I couldn't remember what - the 'scope, I thought...

So, I plugged it in, turned it on and everything was fine until I flipped the switch to apply power to the 'scope.  At this point the front panel lights dimmed momentarily, following by a loud bang (even though it was muffled by the unit's metal case) and smoke billowed out from every gap in the front panel along with a bright, flickering yellow light from a fiercely burning flame within.

Of course, I turned it off and fortunately, the flame quickly died out!

Hmmm...

Undoing about a dozen screws I soon had the cover off and discovered the culprit:  A dipped tantalum capacitor (150uF, 15 volts) on the the high voltage power supply board for the oscilloscope had incinerated itself.  Fortunately, aside from leaving a sticky, smoky residue on the nearby components, adjacent chassis panels and the inside of the wrap-around case, there didn't appear to be any real damage.

I should say that once I saw what had "flamed out" I wasn't too surprised:  Dipped tantalum capacitors don't fail too often, but when they do, they usually fail spectacularly, often burning holes in the circuit board and destroying nearby components!

Figure 2:
Inside the wrap-around cover - evidence of smoke and flames!
Click on the image for a larger version.
Grabbing the service manual I quickly located the faulty capacitor on the schematic diagram and noted that it wasn't anything too critical - a bypass capacitor on the power supply to filter the ripple from the 'scopes high voltage power supply (essentially an oscillator) from the main 12 volt power bus - this, to keep "noise" from getting into other circuits.

Carefully unsoldering the remnants of the capacitor (now a small chunk of charred tantalum) I shook out the other pieces of the capacitor that had fallen inside the unit and the powered it up.

Everything looked good!

Now, to replace the capacitor.  The original was a dipped tantalum unit, this type chosen because of its low ESR (Equivalent Series Resistance) and its ability to effectively filter the high-frequency noise produced by the high voltage inverter.  For this task I wasn't going use an "ordinary", cheap capacitor since its filtering ability may be somewhat diminished at the frequencies involved - around 20 kHz.

Back when the unit was made the best capacitors for high-frequency filtering were tantalum units or specially-made low-ESR electrolytics, but the latter weren't extremely common.  These days, with the proliferation of switching power supplies it's quite common to find high-performance, low-ESR electrolytics designed for just this task so I rummaged around and found a 330uF, low-ESR 105C (high temperature) capacitor that appeared to be well-suited for the task.

Figure 3:
The new (blue) CDE 330uF low-ESR electrolytic.
Click on the image for a larger version.
While it would have been ideal to have completely pulled the circuit board to install the replacement capacitor, I knew this to be a chore - having done it several times before - so I was able to do a careful "top soldering" job, heating the component's through-hole vias from the component side of the board.  Not having pulled the board out of the unit also meant that some of the sticky, smoky residue remained in some of the inner recesses and on some of the adjacent components, but I was content to clean off what I could reach using denatured alcohol.

The upshot?

The unit is now working again and the new capacitor seems to be doing its job.  If I have a reason to do so in the future, I'll pull the scope module go through it to remove the last traces of the smoky residue and, perhaps, preemptively replacing the other tantalum, but for now...

I still don't remember for certain the problem for which I was checking out the service monitor!

Additional random comments:

A few months later (8/13) I noticed that sometimes the IFR-1000S would hum when it was powered up - but not always.  Clearly 120 Hz ripple, it was pervasive enough that it would register as 200-300 Hz of deviation on an otherwise unmodulated carrier, appearing as a "dirty" waveform on the output signal as well as being visible on the scope and audible via the speaker.

Taking the cover off, the hum stopped, but I checked the filter capacitors in the power supply (some of which I'd replaced a few years ago) and found them to be good.  After having used it a few more times hum-free, the problem appeared again and this time I happened to notice, as I was picking it up while it was powered on, that the hum changed.  Pushing on the case and wiggling things I discovered that the hum changed radically when I wiggled them main AC power connector - "Jones" plug.

Upon disassembling the unit I saw that the solder joints on the connector were just fine, but that the Battery - lead from this connector (which can be used to operate the unit from DC power) shared a heavy black lead that came from the main power supply, bonding it to the chassis.

Hmmm...

Grabbing a screwdriver, I immediately noticed that this screw was a bit loose.  As it turned out it was this connection that was getting flaky, developing a slight amount of resistance.  Since it came from the power supply this caused the regulation (and consequently, ripple rejection) to suffer.  I put a drop of anti oxidant grease on the connections and properly tightened the screw, thus fixing the problem!

Friday, October 5, 2012

Resurrecting a CE-50A service monitor

Several weeks ago a friend of mine gave me his old Cushman CE-50A service monitor - a "Swiss army knife" piece of test equipment used for testing and evaluating 2-way radios, receivers and transmitters over the range 0.1 MHz to just under 1 GHz with the capability of testing receiver sensitivity, transmitter modulation and transmit power - all of this in addition to being a general-purpose audio and RF signal generator and low-end, general-purpose oscilloscope.  He'd had this unit for about a decade and had bought it in a non-working condition, having gotten it functional and had used it many times for radio maintenance.  His needs had changed and he no longer had the time and equipment to repair it so he figured that if he needed a piece of test equipment with the necessary functionality, he knew plenty of people (such as me!) from whom he could borrow the necessary gear.

Figure 1:
The now-working CE-50A with it's scope showing an "O'clock" - an
oscilloscope-based chronometer - just for fun!
Its display is slightly distorted due to my failure to compensate the
scope leads!
Click on the image for a larger version. 
Some time in the past year or so it quit working properly.  For a long time the front panel BNC connector from which the signal generator output - and the wattmeter was input - was flaky.  While a nuisance, it was still workable - at least until the built-in wattmeter quit working due to the internal RF relay not being triggered by RF any more.

Fixing the scope:

So, it fell into my hands along with the service manual.  Upon turning it on one of the first things that I noticed was that while the oscilloscope CRT was fairly dim, all displayed traces - from any source - had some "fuzz" on them at a frequency that was in the 20 kHz or so range.  My first inclination was that I should "shotgun" (e.g. replace) all of the electrolytic capacitors in the main power supply.

I should have gone with that first inclination.

I then thought that, perhaps, the scope's power supply board had lost some capacitors since not only did it supply the high voltage for the CRT, but it also generated other voltages (e.g. 90 volts for deflection and a negative voltage for other scope-related circuits) so I replaced all of the electrolytics on that board.

No change.

Poking around I then noticed that the high voltage was around -950 volts instead of, as indicated by the manual, -1400 volts.  Pulling the high voltage converter board again I realized that it didn't match the one in the book, being a different part number and further scrutiny revealed that about the only difference between the board and the one depicted in the book was that it was supposed to have a 3-stage high-voltage multiplier rather than just a 2-stage.  Rummaging around, I found some high-voltage 0.01uF disk ceramic capacitors and some 6000 volt, low-current diodes and constructed the extra stage, bringing up the high voltage to more-or-less what it should be.

This made the scope trace significantly brighter - as well as narrower and shorter.  Expectedly, the higher cathode voltage on the CRT increased the velocity of the electrons which meant that they were more difficult to deflect and this required that I re-tweak the vertical and horizontal amplifiers to get it back into calibration.  While doing this I couldn't help but notice that the service manual was obviously incomplete on some points, namely leaving out the descriptions and adjustment procedures for entire circuits within the vertical and horizontal deflection requiring a bit of on-the-spot decipherment of the apparent intent of the designers.  After a bit of tweaking, I got the scope back into calibration.  Even after all of this, the "fuzz" on the scope was still there - although slightly diminished - probably due to the increase in acceleration voltage.

RF power/signal generator transfer relay:

Setting aside the fuzz on the scope for the time being I attacked the problem with the internal relay.  Normally, the "Signal Output" jack is connected to the antenna input of a receiver under test and is used to apply a variable signal level used to test the radio's performance and as a signal source to aid adjustment.  When one transmitted into this same port an internal relay was supposed to switch the signal path from the output of the signal generator and connect it to the internal power meter, allowing one to measure transmitter power from less than 1 watt to 100 watts.

Except that it didn't, and that was the problem.

Of course, the module with this relay was the most deeply-buried circuit within the entire unit.

Getting access to this module required the un-mounting and disconnecting of several modules before its nearly two-dozen screws could be accessed and the cover removed.  Having just enough wires still connected to do so, I was able to transmit some RF power into the unit and saw that an 8 volt supply that fed nothing but that power detect circuit and its relay was going from its normal 8 volts down to 3 or so under load.  Referring to the manual I then noticed - with some annoyance - that this same 8 volt supply was now buried under the modules that I had to flip open to access this circuit to take the measurement, so I had to put everything back together.

Finding the 8 volt power source - a simple 7808 3-terminal regulator bolted to the chassis near the rear of the unit - I was immediately struck by the fact that its input voltage was varying between 60 and 70 volts.  Looking at the schematic I could see that its power source was either the main 12 volt bus from the power supply, or from the battery input - the choice between the two selected automatically with a pair of diodes in "diode-OR" configuration.  Putting the voltmeter on the source voltage I determined that yes, the 12 volt supply was correct, but the battery supply - which should have been at about 14 volts for charging the not-installed battery, was in the 18-20 volt area.

This last point was definitely wrong, but where was the 60-70 volts coming from?

Grabbing an oscilloscope I noticed, in looking at the battery charge line, that it was very "dirty" with 70-80 volt spikes on it which were then being rectified and filtered by the input diode and bypass capacitor on the 7808 which was apparently shutting down under even a very light load.

It was now that I finally "shotgunned" the capacitors in the power supply and in so-doing, I found that it was in the charging circuit for the battery that a capacitor or two had failed and because because of this, the circuit had gone into oscillation and was the source of the spikes.

I really should have replaced all of the capacitors in the power supply when I started!

Finding and replacing every electrolytic capacitor on this board (nearly all of them were found to be well out of spec!) I re-installed it and observed that the input of the 7808 was now in the 13-15 volt range and that the RF relay now operated normally - and also that the "fuzz" in the scope was completely gone!  Interestingly - but not too surprisingly - the scope now appeared to be even  "brighter" than it had been before owing to the fact that without the "fuzz" to fatten all of the lines, they were now fairly narrow and crisp, giving the illusion of additional brightness.

Using a handie-talkie I fed power into the jack and observed that I was now getting a wattmeter reading - but something was still wrong:  It read very low.  Further investigation showed that the meter reading seemed to change slightly every time it was activated, indicating a mechanical problem somewhere and even more revealing, I was getting a much higher and "less incorrect" reading when transmitting at 440 MHz than I was at 145 MHz indicating an "air gap" somewhere in the signal path with capacitive coupling across it.

At this point I connected the handie-talkie to the input side of the 20 dB, 100 watt attenuator inside the unit - a point "after" the relay - and observed that  the readings were closer to being correct and consistent both from one reading to another and over frequency, being fairly consistent between VHF and UHF andpositively indicating that the problem was likely inside the module that I'd previously taken apart - and mostly likely the RF relay.

This meant tearing the unit completely apart... again... and possibly replacing the RF relay.  Fortunately, Cushman had chosen a rather common component for this - a small Switchcraft RF relay that had been used in land-mobile gear in the 60's and 70's and a type of which I had some spares that had been pulled from scrapped gear.  While the relay itself was identical, the coil was different but inspection showed that it should be possible to drill out the rivets - leaving the posts - and then epoxy the old coil onto the "new" relay were it necessary to do so.

I carefully unsoldered and removed the relay and took off its cover to inspect its insides.  This relay is fairly simple - see figure 2 - an armature inside a milled-out channel (for RF impedance matching) with a pair of contacts at the far end with a plastic button transferring the motion from the armature on the coil.  What I noticed was that the armature was out of alignment, touching the other contact with only a "glancing" blow and thus explaining why it wasn't working properly.

Figure 2:
The guts of the same type relay used in the CE-50 and a potentialreplacement!  The problem with the original relay was
where it emerged from the coax and into the body of the relay (on the right) and connected to the armature:  It seems that
during the original installation the polyethylene insulation melted and allowed the armature and contacts to move out of
alignment, eventually causing the relay to become unreliable.
Click on the image for a larger version.
These types of relays were originally made with short length of coaxial cable crimped onto their ends and apparently the manufacturer of the service monitor (Cushman) had simply cut off the polyethylene dielectric coax, leaving a short portion of the center conductor to be soldered into the circuit.  What had apparently happened was that upon installation, a bit too much heat had been applied while soldering and the plastic dielectric had melted, causing the armature had moved out of position.

Using a pair of needle nose and carefully applying heat to re-soften the plastic I carefully repositioned the armature into proper alignment and then, to make sure that his wouldn't happen again, encapsulated the end of the armature (the far-right end in figure 2 where it would connect to the coaxial cable) in a small amount of clear epoxy to provide a more rigid substrate than the polyethylene had provided.  After the epoxy cured I took this opportunity to inspect the relay contacts more closely - which appeared to be nearly pristine - and then burnished them with a piece of scrap paper to remove any surface oxide that might have formed.  I then put a drop of "Stabilant 22" - a synthetic contact enhancer and anti-oxidant - on the relay contacts to assure continued operation.  Finally, I very carefully bent the armature itself so that its "springiness" would be modified to more positively and forcefully make contact.  After all of this, I reassembled the relay, installed it into the circuit and tested it.

It worked!

Putting everything back together I went about recalibrating the wattmeter and found that as with the section that described the 'scope, the service manual was woefully incomplete requiring that I reverse-engineer their original intent, invent, and perform the calibration procedure - and then note it in the manual for future reference!

Why doesn't the PFM-Generate mode work?

At this point everything seemed to be working so I went about checking and recalibrating the various sub-instruments as required - until I came across the need to check and calibrate the "PFM Generate" function.  As it turns out, in addition to AM and FM, there's a "PFM" (presumably meaning "Pulse FM") mode that isn't well described in the manual.  When set to its equivalent in the "monitor" (receive) mode, this seems to insert a low-pass filter into the demodulator path to remove high-frequency noise, but when set to "generate" mode, the PFM setting seemed to do nothing at all except generate a CW (dead) carrier.

Again, I tore into the unit and with the aid of the manual I started tracing the signal path of the front panel selector switch and found that it was getting everywhere it should have.  I finally got to the audio board where these signals were routed and noticed that when in PFM, the audio path went through a separate adjustment and audio switch just for the PFM mode - and it seemed to be working.  Moving to the next circuit earlier in the audio path I found an audio gate transistor that seemed to be disabled in PFM mode with a diode.  At this point I went back and reviewed the manual's circuit description and interestingly, it described in some detail the audio paths for all modes - including PFM - but then, in a separate paragraph it mentioned in passing that this particular diode was there to disable the audio in the PFM mode!  Why, then, was there extra circuity for the PFM mode if it was ultimately disabled?  At least this answered the question and told me that "PFM-Generate" was supposed to do nothing!  While I could easily remove the diode and make this mode functional, I decided to leave it alone for now.

Comment:
Further testing and "reading between the lines" of the manual it would appear that the "PFM" mode is intended only for modulation applied through the "External Modulation" input jack - a point that the manual doesn't make clear.  Since external modulation is, in fact, possible in the other modes, it is unknown why that would have yet another switch position to accomplish this!

Hummm...

While chasing out the PFM-Generate mode I noticed a small amount of hum coming from the speaker.  In checking the 12 volt supply I could see that there was about 35 millivolts of ripple on it, so apart came the power supply again.  This time, I traced the 120 Hz ripple to the switcher board and then noticed that a previous modification had been done to it where an output filter capacitor had been installed with a series resistor - presumably to reduce inrush current - replacing a smaller-value capacitor that the diagram had showed as being located directly across the switcher supply's output.  Putting a scope there showed that there was a few hundred millivolts of ripple at the switching frequency (25 kHz or so) but that this was being filtered out by a later power supply stage - the one on which I'd shotgunned all the capacitors.  Using a low-ESR capacitor specifically designed for switching supplies, I reinstalled the device that had been missing and not only did the switching frequency ripple decrease significantly, but the 120 Hz ripple on the 12 volt supply went down to the 10-12 millivolt area (which was quite acceptable) - but the hum in the speaker, being much lower, was still audible.

Turning my attention to the audio amplifier I noticed that the speaker had been coupled to the output in an odd way.  This amplifier was a fairly simple, transistor-based circuit using a pair complimentary transistors in the output to the capacitively-coupled speaker.  Typically, the speaker coupling capacitor is connected between the output of the "totem-pole" transistors and the speaker, but in this case, the DC blocking/coupling caps were on the "low" side of the speaker - and there were two of them:  One between the speaker "low" side and ground and another between the +12 volt line and the speaker "low" side - and it was this latter capacitor that appeared to be coupling the power supply hum into the speaker.  I'm not sure why they did it this way, but my guess is that it prevents a loud "pop" in the speaker when the power is turned on and off, so I left it this way, deciding that the hum wasn't that bad anyway...

Front-panel connector:

The final item was to address the problem with the front-panel RF connector.  As often happens with BNC and N-type female connectors, the "leaves" on the springy center connector weaken and break off - and that had happened to the previous owner.  Unfortunately, a rather specialized chassis-mount BNC connector had been used on the end of small-diameter PTFE hardline coaxial cable.  For trouble-shooting purposes, the previous owner temporarily connected a male BNC connector via a length of cable, but now that I'd gotten the unit back together and fully functional I wanted to make a permanent fix.
Figure 3:
In Tracking Generator mode, testing a 10.7 MHz ceramic filter.  Because the
filter was fed/sourced with 50 ohms instead of 330 ohms, there's extra
passband ripple!
Click on the image for a larger version.

In rummaging around the junk box I found plenty of crimp-type BNC chassis-mount male connectors that would fit in a 1/2" diameter hole - but the one on the unit was a 3/8" hole with a flat spot and I wanted to avoid - if possible - drilling it out.  What's more, I didn't want to use a standard chassis-mount solder-on BNC connector as this RF connection had to be both RF tight and fairly impedance "flat" from essentially DC to 1 GHz - difficult to do with a solder-on connector.  Finally, I found one 3/8" O.D. chassis-mount female BNC connector with a coaxial crimp fitting it to RG-174 sized PTFE flexible coax so I carefully disassembled it and managed to successfully attach it - with a bit of soldering - to the small-diameter hardline inside the unit.  A bit of testing showed that it worked nicely over the entire frequency range, so the unit was reassembled and the project was considered to be complete!

Comment:  I later noted that the exact replacement connector was available via Pasternack Enterprises for somewhere around $50 - something to keep in mind should this repair fail at some point.

Conclusion:

Overall, the repair was a fun project, taking me several places that I didn't anticipate going and reminding me, again, of that old adage: 
"When in doubt, check the power supply!"

Additional comments:

More recently (8/13) I had a strange problem occur:  When in the spectrum analyzer mode the synthesizer lock would come and go while the frequency offset meter would slowly drift up and down.  Apparently, something was slightly unstable, preventing the main PLL from locking.

In trying to troubleshoot this I looked at some of the plug-in boards and noticed that a lot of the 100uF, 16 volt capacitors were starting to leak - and there are a couple dozen of these scattered throughout the unit on most (if not all) plug-in boards that are used as power supply bypass/filter capacitors.  While no damage seemed to have been done other than a slight amount of surface corrosion that was easily removed, I did replace pretty much all of them and at some point the problem with the synthesizer's locking went away - although I don't know for certain on which board the "fix" occurred - or if it somehow fixed itself with my re-seating the boards.

Anyway, it would be a very good idea to take a close look at all of the 100uF capacitors (and other electrolytics, as well) on the various boards and replace them as they are probably starting to leak on your CE-50, too!