The Anatomy of a Radio Repair

The ask

John AB5SS asked me to power up and checkout his Collins KWM-380 transceiver after it had been in sitting unpowered for a few years at his QTH.   At some point during this effort, he suggested I write a blog about this effort.  Here it is.

The concern

John was concerned that the power supply electrolytic capacitors had degraded over time and were at risk of explosion.   Capacitors that haven’t been powered on for long periods of time have been known to blow their degraded electrolyte guts if suddenly exposed to operating voltage.  If you look closely, you’ll see small holes which are vented to normally allow gas build up to escape.  Excessive internal resistance, heat and pressure will overcome these vents.

The preventative technique of restoring the capacitors for use is to slowly ramp up the applied voltage over time, so that chemical electrolysis can reform the chemical dielectric plates.  This is called capacitor reforming or rebuilding.

Since my workbench is outfitted with an AC variac I gladly accepted the task and began by slowly ramping up the voltage to recondition the power supply caps in the radio. 

The capacitor reconditioning process

Because there’s nothing more embarrassing than breaking a good friend’s radio, I invoked an extraordinarily conservative ramp up profile over 2 days to allow the huge electrolytic cap in the power supply to sustain its full voltage.  Starting at a low voltage, I decided to incrementally increase the AC voltage to the from 60V to 120V with about 15V increments every 3 hours.  As an added safety precaution from any shorts that might occur during the power up, I added a lightbulb in line with the AC cord.  This is an old trick.  If a short circuit occurs anytime during the process, the bulb takes the brunt of the high current flow and lights up instead of damaging the test article (the radio).

As the voltage is incrementally ramped up, the in-line light bulb will glow which is normal, as more current is passed through the circuit.

Here’s a map of the ramping times, voltage increments and DC voltage on the main electrolytic cap. 

This radio used an extremely large capacitor 72000uF (not a typo), which would cause a glorious explosion of energy if indeed it had degraded and powered up untreated.  Downstream,  there’s another big cap at 23000uF.  These are huge by today’s standards.

As the radio was ramped up, the LED displays began to light up.  I noticed the first LED module had one of its legs dark.  I ignored it thinking it was an artifact of the partial power ramping.  Afterwards everything seemed normal, but now in hindsight, it could still be defective.  The final checkout was conducted at 14, 7 3 MHz so that particular segment didn’t need to light up.   I may have missed an opportunity to have fixed it- assuming it’s still no illuminating on that leg.  Darn it.

After the radio completed its reconditioning cycles on day 2, I removed the variac to begin my functional checkout of the radio.   I also restrapped the AC taps on the power transformer.  Back in the day, the residental output voltage was 110-115VAC.  Today, the power companies decided to pump up the spigot to 120-125VAC.  So running pre-1984 vintage radio plugged into today’s power, results in that extra stress on the components.  Luckily, this KWM380 has taps on the transformer which I re-configured from 110v to 120v, re-normalizing the voltage output to the power supply and precluding that extra stress with today’s higher voltage levels.

Before 

 

After

Functional Checkout

It sounded great when hooked up to my outdoor Skywire Loop antenna, but I noticed that stations were received off frequency by 30 Hz.  Not a big offset, but a calibration adjustment usually is not too difficult. Oddly, the KWM380 service manual is unobtanium, but I finally found the online maintenance notes (thank you Collins Collector’s Association) for its military counterpart, the HF380.  I quickly discovered that I had to remove the top card cage cover to gain access to the oscillator card to perform a frequency calibration adjustment.  

Military Robustness

It was immediately clear that this amateur version (KWM380) of the military HF380 cloned most of the military pedigrees as the parts and assemblies were MIL SPEC robust, high quality and expensive for its time.  As an example of robustness, the metallic cover to the PCB card cage was held in place with a ridiculous number of fasteners: Thirty-Eight  fine threaded machine screws.   Each screw was made of stainless steel with dual stainless-steel washers (a locking washer and flat washer attached them).  The components and manufacturing quality were high class as one could see by the use of Dale resistors, tiny plastic bead spacers manually attached on each through hole radial component lead, double sided PCB ground planes, and extensive labeled heat shrink tubing on all cables.   Also, there were Teflon spacers were attached to the side frames to both protect and ease the insertion and removal of the case from the chassis.  Wow, this was one very high-quality radio.

A whole lotsa screws (with dual washers on each) just to fasten the cover above the card cage.

 

Great workmanship to route and bundle wires.  Use of quality standoffs, and parts.  Wires are uniformly attached and with protective sleeves.  Professionally solder and secured.

This is a very very nasty place to drop screws and washers, especially if you can’t rely on your magnetic screwdriver to attract stainless steel fasteners.  

I love Dale Resistors.  Super high quality

Note the intentional labeling of the ribbon cable connectors.  Nobody does this today,

Even the single jumpers wires are heat shrinked at right angles for stress relief, and they are labeled.  This is impressive.

Here’s a master-class example of a perfect cable harness.  Note the uniformity of the bends along the well tie-wrapped multi-conductor cable assembly.

Expensive Allen Bradley trim pots.  Today’s radios use cheap plastic or weaker metal versions.

Nice plumbing on the coax lines

And way cool teflon bumpers (top and sides) to ease sliding the cabinet from the chassis.

The frequency calibration was a bit tricky, requiring the adjustment and re-adjustment of the 455KHz and 6.7MHz oscillators.  There was a bit of confusion trying to match the oscillator card and configuration to the documentation.  I soon discovered that the Miltary version used an oven to control the oscillator but the “domestic” version didn’t.   Once I was able to find the perfect balanced alignment, it seemed my job was done and complete.  … except the radio suddenly became deaf.  Ouch.

(Above) this the oscillator card elevated from it’s card cage.  The white pull lever was useless to pull it out.  I had to use 2 vice grips to work the card out of the slot.

A view of the card cage- the guts of the radio (above)

3.96MHz Crystal and Trimmer adjustment. (below)

The 4.55MHz trimmer is just below it’s crystal.

 

I was searching for this card with the crystal oven throughout the card stack until I realized it wasn’t installed.

Troubleshooting.

The first step in troubleshooting is to assess the failure and think what part of the radio might be misbehaving.  I suppose this is known as critical thinking.  This really is the fun part of radio repair, since it pits your logic alongside your observations.  But really, the true first step is to carefully review and retrace anything you may have done stupidly.  Could I have inadvertently dislodged a wire connector, left a jumper or cable unconnected? shorted out a circuit from accidental fallen debris (wire, screw, solder), etc., or perhaps accidently bumped an exposed wire or lead to ground?

Since I just finished adjusting the trimmer caps in the oscillator board during frequency adjustments, could I have induced a hidden consequence to the radio?  Could I have dislodged any of the RF connectors as I slid out the oscillator board?

No Audio

The radio seemed to tune and display, but there was no audio.  So, the thought was that the failure was in the audio amp!  I grabbed a pair of headphones and plugged it into the headphone jack hoping to hear a signal.  If audio was to be heard, that signal would be coming before the audio amp, and it would implicate the audio power amplifier.  But no joy.

Back to Basics

Always verify all the power supply voltages are proper before looking deeper.  Looking at the schematic, there are multiple voltage levels, e.g. 5V, 12V, 9V, 24V, 48V, etc. which are fed throughout the radio design. 

Checking the voltages, each was verified except for the 14V line.  The missing 14V line was traced to an open fuse.  Often a fuse simply blows as it’s weakened over the many hours of operation or possibly a manual probe accidently shorting a power line.  I assumed this was the case.   The fuse was replaced but blew out when powered up.  But I violated my typical common-sense procedure to also measure the power line to see if there was a low resistance (a short to ground) in the line which would have blown the fuse.  Well, that turns out to be what happened.  Something was shorting the 14V line.

A bad short, like almost 1 ohm which is a DEAD short.

Hunting for milli-ohms

A dead short is tough to isolate.  You must have an analog VOM with an RX1 scale.  The digital VOMS are too hard to use when trying to isolate tenths of an ohm while probing across candidate paths along the circuit board.

This slow and tedious process requires extreme diligence since there’s lots of probing and re-probing while continuously observing the slightest meter movements since it’s reading a tenth of an ohm or less.

This radio was manufactured in the 1990s. Unlike modern radios, this one did not have a silkscreened depiction of the PCB layout routes.  So, I had to painstakingly manually trace the 14V line underneath a complicated board.  This hand drawn map was used to orient my probing points and measure different resistance paths in hopes of isolating the dead short. 

The fine pincher probes are really helpful when taking measurements on component leads (above)

Use of a micro probe is super valuable.  It’s slimline yet needle pointed shape lets you reach hard-to-get-to spots without shorting against an adjacent component (below)

Using tape to de-mark the trace of interest helped me progressively isolate the path of least resistance (aka the short) (above)

I had to manually transcribe out the path route by hand by reviewing the board layout in the manual. (above)

Suspicious capacitors were circled in advance (after review the schematic) and used as targets during my probing. (above)

Here’s an example of diagramming out the jumpers, connectors, and plugs before removal.  This is essential to assure you can put everything back together. (above)

Here’s the spot of lowest resistance.. which lead to this shorted capacitor. (above)

BTW, note the fancy plastic spacers on each capacitor leg. (above)

Again, this involved measuring barely discernible movement of the VOM, since a barely moving meter movement indicated a tenth of an ohm, as I tried to find the path that distinguished itself as lower resistance than another.   Reading microfine displacements of the meter movement is a very refined technique.

The blown 2A fuse removed  (above)

RF/Exciter board fully removed.   (above)

Lots of careful skill needed to trace the slightest movement of the needle at the milli-ohm range in order to find the deflection of least resistance.  It’s harder to do this with a digital meter. 

I was almost ready to invoke a last-ditch option to isolate a fault by cutting a break in the PCB metallic trace which would vindicate one side while implicating the other (i.e. the short would be seen using a VOM only on one side of the break).  Luckily, I ultimately isolated a particular zone which exhibited a consistently lower milliohm resistance reading to ground.  This resulted in finding a shorted 47 uF tantalum capacitor!

Success, but more suspicions

With one failed cap, others are now suspect, too.  That’s a healthy NASA paradigm, too; whenever a space rated device fails a FIAR (Failure Investigative Analysis Report) is created which documents the failed device to a database used to crosschecks a failed part across the universe of other NASA flight hardware failures.  If others are found, it establishes and indicts a specific risk to existing systems that may carry a generic failure.  In this case, the failure is tracked to the level of specific vendor, value, date code and lot.

Of this vintage, cap failures are relatively common due to aging.  Tantalums have a characteristic of getting old and when they fail it is usually a short.  So, the board which had most of the electronics (Receiver/Transmit board) I counted 19 tantalum caps 1uF or larger (the smaller ones are usually very low risk since there’s less chemistry and energy density involved inside).  I reviewed the board schematic and developed a risk listing based on the stress of voltages on each cap.  I recommended to John that the large 47uF caps should be replaced.  He concurred. 

An example of going throught schematics to ID caps with potential voltage stress on it. (above)

This was a wise compromise decision since desoldering and resoldering components also introduces risk of damage to the PCB.  So, this made sense even though I was willing to change all of them, if needed.  Note:  the PCB was 2 sided with plated through hole vias and very thin metal traces- always a recipe for high risk when solder rework is performed.  I’ve ruined many PCBs due to stubborn desoldering of components.  Heat is the enemy of a delicate PCBs (and likewise susceptible surrounding components).

Soldering Rework

The topic of desoldering through hole vias needs to be discussed.  I have tried many desoldering techniques using a variety of tools including solder sucker (Unger, Paladin), electric vacuum desoldering gun (Hakko, Pace), Low temperature solder (Chipquik) and they all suck (pun intended).  the best technique is using plenty of resin (aka solder paste) and solder wick.  This gently wicks up the solder and avoids physical damage to the metal holes and the adjacent metal traces.  With generous use of resin, one can suck up the solder quickly before excessive heat damages the board and metallic vias and traces.  The choice of solder wick is super important.  Many have lousy ability to wick up quickly.  I exclusively use the Chemtronics.  A squirt of solder rosin on the joint and on the wick does the trick.  It’s slow and laborious, but it works great.

With the shorted cap and the other high risk tantalum caps replaced, I re-installed the board back do the chassis.  Note:  there are lots of RF cables, ribbon connectors, and jumpers between boards, which must be carefully mapped.  I always take photos and/or draw diagrams of everything so I can refer to them when I might have gotten confused which connector goes where, and to assure the same routing (was cable A, installed above or below cable B; did the coax route underneath the board or alongside the low pass filter board?).  Also again, Collins took ultimate care to mechanically fasten everything with dual washers (a locking and flat).  This assures terrific fastening, but it’s a nightmare if you accidently drop one (and you will) of the washers and it lodges underneath the cavities of boards and innards.  Fat fingers, tiny screws, close spaces are a bad recipe for dropped fasteners and washers.  Note: also make sure your bench space is clean and well-lit as you will often need to go hunting for that dropped item that’s either still wedged in the chassis or possibly fell out onto the workbench or bounced onto the ground.  Yes, this happens a lot.

Success… temporary

Power up the radio, and voila, the 2A fuse didn’t blow and the radio audio came back to life.

Mission accomplished… or so I thought.  After a quick test, I noticed that the radio sounded very distorted when the front panel bandpass selectivity knob was rotated across the different filter width.  Oddly, the widest setting: 8 kHz was the only setting that sounded OK.

Yep, another problem

Back to the service manual and try to comprehend how the theory of operation works during reception.  It’s gratifying to take the time to digest the design documentation and apply logic to help isolate the problems of a failure. 

You will read the functional design, dive into the detailed design descriptions, trace the signal flow thought the schematic diagram and then find suspicious components on the board layouts to measure and test. (below)

Thinking, thinking, and more thinking

So, this iterative process to identify test points and suspicious components requires lots of the back and forth between documentation items.  I often cut and paste specific sections of schematics, test procedures, theory descriptions, and board layouts during this process; and take them up to the lab bench for diagnosis.  As I go through the isolation process, I’ll frequently go back and forth between my computer (to print out new sections to investigate), the lab bench to take measurements, and to bed where I lay down to think about next steps.  The latter is usually overnight when I can sleep on it.  For terribly tough brain stumpers, I find lots of creative thinking while long distance driving with the radio off, or just before going to bed with the issue in my head.

Let Your Brain Marinade

It’s fascinating to realize how the brain can marinade problems and sprout new ideas after a while.  I guess the subconscious gets credit for doing the background processing of complex problems.  it’s that old phase, “sleep on it”.

So, this last problem with the passband issue was a tough one.  Could I have dislodged or damaged something along the way to cause this issue?  Could the shorted 14V line and blown fuse have induced an issue to another part of the radio circuitry?  I remind myself of the KISS principle- that it’s likely something simple.  The notion that something else failed coincidently was not likely, but it was on the fault tree in my mind.

Why would the 8kHz setting of the selectivity work and the tighter filters have problems?

The filter board inputs a 455kHz, mixes this signal with another oscillator to create secondary IF for the filters, and then remixes the oscillator back to 455kHz.

Could a switch be bad.  Not likely.

Could the voltages to the board used for electronic filter switching be missing?  I rechecked and everything was properly working.

I checked the 455kHZ and the 6.7 MHz oscillator signals on a counter.  It’s spot on.

Maybe the selector ribbon cable mechanically damaged and was not properly selecting the other filters?
Maybe the other filters were not being selected fully because of the switching diode being defection.  Check all these switching diodes – all checked out…. well, I found a pair of non-conducting diodes!  Euphoria! I quickly order a batch from Digi key.  Only to fall back in disappointment when, I soon discovered that the diodes lead had a transparent non-conducting coating on the cathode.  This insulated coating tricked me into thinking the diodes were open circuited.  After I re-attached my probes by digging them deep through the coating, the diodes were measured and tested fine.  False alarm and a red herring.

For more troubleshooting and problem isolation, could I swap the placement of the working 8kHz filter with one of the other distorted filters on the board? – not a bad idea.

For each selectivity setting, a variable potentiometer is active to invoke an adjustable Passband function – i.e. it fine tunes the 6.7 MHz oscillator to shift the passband windows the center of the received signal. 

It does this by using a varactor diode, CR13,  attached in the crystal oscillator circuit.  The oscillator sightly changes in frequency as the varactor diode’s capacitance detunes the oscillator’s base frequency.  The varactor is a cool component; by changing the reverse bias of this diode, the semiconductor depletion region changes depth which effectively changes the capacitance.  Like someone moving the distance of two plates of a capacitor, which changes the capacitance in turn.

Measuring the Varactor bias

Yes, there’s a adjustable voltage bias on the varactor diode on the oscillator.  I measured a 5V swing, but is that correct?  Nothing specified in the service manual regarding the expected voltage swing.  Hmm. Usually, it’s easy to calculate but the schematic reveals a stack up of parallel resistor dividers and staged resistor paths that get turned on at different voltage levels.  Ugh, this is like those 3D resistor cube problems what are ugly to solve.  Ohm’s law shouldn’t be this tough for in this radio, but it is.  Could I more easily simulate this on a breadboard and avoid modeling this with math? 

Let’s refer to the maintenance/testing section of the manual.  The 6.7MHz oscillator is supposed to be exactly centered around the passbands of each optional filter.  Maybe that’s the problem, because I see that the center frequency is somehow mistuned!  OK the service manual spells out exactly how to re-calibrate this by making adjustments on various caps and resistor trimpots. 

Avoid the temptation to twist the knobs.

BUT WAIT.  My intuition tells me not to try this.

What is the likelihood of a failure that all of a sudden require an “out of the blue” realignment of this varactor oscillator circuit?  This sure didn’t smell right.  Recalibrating the center frequency of the Pass Band Tuning control passband clearly does not address the root cause and likely would future add new complications to a previously working circuit.  This would be akin to treating the symptom but not the cause.

Let’s look more careful about the varactor bias voltage.  Hmmm, seems that the schematic suggests there should be a wider voltage swing than what I’m measuring.

 Let’s take resistance measurements with the radio unpowered- are all the resistance values across the bias control path looking, correct?  Hmm, it measures a bit low but not enough to attenuate the voltage swing.  All the resistance paths along this control line show mostly high resistance series values, not enough to pull down a voltage level.  I decide to remove two diodes level switches which activate at different voltage levels.  This is tricky technique to add resistance to a pathway only when certain voltage thresholds are reached.  In this case it’s adding resistances to a control path to linearize the voltage swing to varactor diode.  This makes turning the variable bandpass control symmetrical and linear as one turns the know the left and right of the center frequency. 

Well Still No joy.  Next, I see that the voltage swing passes through a CMOS bi-lateral switch.  Very unlikely this switch is bad… heck these never go bad…but electrically it doesn’t make sense why the swing isn’t larger.  Something mysterious is loading down the line.   Logic says to remove this 4066 switch and measure signals on both sides of this CMOS switch to reveal which side might be dropping down the signal (voltage).

Again, unsoldering throughole PCBs is dangerous and requires care.  To make this easier, I clipped the leads off at the base of this 14pin DIP IC.  This way desoldering becomes easy; I could physically unsolder each leg independently. This precludes the ugly process of rocking the entire 14-pin device off the board as one must alternate heating all the of the pins as you incrementally raises the chip from the board.  That’s a big repetitive heat load on the board and usually leaves fry mark consequences.    Removal went like this:

•  Clip forceps to a clipped off lead on the component side.
• Head up the soldering iron to the other side.
• Gently pull out one leg at a time

After all the legs were removed.  I used solder flux and solder wick to remove the solder from each clogged hole.  Some holes were cleanly open, but many weren’t.  This is a normal headache of desoldering double sided boards.  Again, the risk due to excessive heat is not a good thing.

But because the IC body was clipped off and the legs already removed, it was a simple task to desolder using wick on the component side.  Worked great.

With the IC removed, I powered up the radio and measured and observed that the voltage swing from the Pass Band Tuning pot was WIDE.  This clearly implicated the 4066 IC as shunting the signal amplitude of the voltage needed to properly bias the 6.7MHz oscillator.

Luckily the W5OC parts bin had a 4066 and it was carefully soldered back in. 

Flux usually leaves a lot of residues after resoldering components.  I use 90% isopropyl alcohol and a Q-tip to apply on the board and a can of compressed air to speed up the drying it off the board.

Joy

With caution, I repowered up the radio and everything worked.

This is point of high gratification to have fixed two dissimilar problems.  Why did the tantalum cap shorted in the first place (to have blown the 2A fuse)?  Probably age and stress.  Why did the 4066 CMOS switch fail?  Probably stress (and age?) due to the surge of the 14V line when the 2A fuse blew?

I took the uncovered radio to the shack and listened to hams on 14 MHz and 7 mhz.  Hams nowadays operate using digital radios and usually transmit on frequency intervals of 1 kHz, so I checked for this condition after recalibrating the oscillator to the 15MHz WWV signal.  I didn’t want to reopen the radio again.  Note: it’s not easy to reinsert the case to the chassis.  It’s a very tight and tricky fit and did I forget to mention it’s a very heavy radio of 48 lbs.  Each time the case is removed and reinstalled requires tipping the radio up along the front of the radio ridge and manipulating the cover to fit correctly.  This is a recipe for scratching the radio.  I used cotton cloth underneath to protect it, but it’s not easy tipping the radio up while awkwardly trying to work the cover back on.

But before the final completion, I gave all the push button switches a shot of my favorite lube: LPS-1and worked the switches in/out multiple times to work it into the internal contacts.  These push button multi connector components *always* oxidize over time and manifest in popping and scratchy artifacts if untreated over time.  LPS1 was used at Pacific Stereo I worked as a repair technician a year out of college when I defied getting a conventional engineering job.

The more time I spent on this radio, the more I form opinions of its health and performance.

Health:  there are a bunch of additional risk tantalum caps on other boards.  This is kind of like a timebomb.  At least the big ones, all need to be replaced.

I finally noticed the power cord It’s starting to crack and exposing the inside AC wire pairs.  Needs to be taped but I’ll let John do this since it needs to be done with professional care to keep its appearance in original form.

How good it this radio.  It’s quite amazing.  Quiet and smooth. 

Today’s radios are so much more ergonomic (fast rate tuning), band segment memories, dual receivers, smaller, quieter (fans), but this one is a collector’s item- way ahead of its time.  Great job Colling Radio.

Key tools used

-  forceps

-  dual lighting lamps (luxo)

-  Simpson 260 VOM

-  microneedle probes

-  solder wick

-  solder flux

-  magnifier headset

-  clip leads

-  signal generator

-  oscilloscope

-  frequency counter

-  wattmeter

-  LPS1, canned air

-  tweezers

-  isopropyl alcohol

-  qtips

-  small tip soldering iron (metcal)

-  clip leads

-  camera (iphone or portable)

-  ribbon marker tape

-  service manual

-  Highlighters, paper, pen

Conclusion:

All repairs are almost always easy. 

Service manuals are essential.  Read the concept of operations, understand the block diagram, follow the flow of data and voltages through the schematics

Unless it’s a microprocessor fault, everything else is a simple failed device or conductor

Look for indicatosr which help isolate the problem

If stumped, let the problem marinade in your brain

Search the internet for others who may have posted a similar or same issue, along with solution.

Search the internet for vendor posted modifications on your radio

Avoid making the problem bigger by turning knobs

Take photos and make diagrams so you can retrace putting it back together

Enjoy the gratification of fixing things that are broken and brought back to life.

Don’t do this for a living- you’ll starve to death.

Mouser and Digi key are your best friends for parts

73 W5OC