Logitech Squeezebox Boom VFD Display Repair Guide Part 2: Filament Power Supply

Please refer to Part 1 for instructions on VFD common information, Boom disassembly procedures etc.!

Fixing the "Filament Starvation" Phenomenon on the Logitech Squeezebox Boom

This entry is about a problem that apparently exists mostly in Boom units. I never found it in Classics or Transporters.
So what is filament starvation anyway? I got this term from the Noritake technicians - Noritake Itron is the manufacturer of the beloved VFD displays in the older Squeezebox devices such as SLIMP, Classic, Boom, and Transporter. It means that the heater wires which you can see horizontally across the display in the frontmost position are supplied with insufficient power so the display cannot show its full brightness. Even worse, the display may show shadowy sections especially at the left and right side. No matter if you just replaced the old VFD by a brand new one, you may be greeted with a sight that looks pretty rotten.
Here is an example I took of a unit in factory button test mode. Observe the inconsistent brightness of the blocks across the display:

Or even worse: this is a brand-new display on brightness level 2 / 5 (lower brightness makes the problem much more apparent):

The same display can also look like this (same brightness level):

So how is this issue identified vs. a generally-aged display?

Idenfying the Actual Boom Display Issue

When a display begins to look shoddy, there are multiple (mostly concurrent!) indications for the respective problem behind the symptoms:

• display shows 'shadowy' sections rather in its center --> probably burnt-in pixels --> display replacement advised. This can go as far as recognizable digits from the clock screensaver which appear as a permanent dark cast across the actual display content, of course mostly in the places where the clock would have shown its digits
• display gets darker beginning at the left or right side. The center is the brightest part until it fades away completely --> power supply failure
• display is completely dark, or quickly fades from barely readable to completely dark shortly after powerup (after being separated from the power supply) --> power supply failure. This recovers when the Boom has been sitting around without power applied for some hours, but quickly comes back after powering up again

Most Booms have both simultaneously, i.e. burnt-in pixels as well as a failing power supply. When you replace the display anyway, it's worth while applying the power supply fix described here. That is, if you find that the display replacement alone did not end in the expected bright and shiny new look.

Edit 2020-06-08: there is a new blog post about this matter specifically, to be found here: https://joes-tech-blog.blogspot.com/2020/06/vacuum-fluorescent-displays-how-they.html

About VFDs

Wikipedia.com gives you extensive information about how a VFD works. Please take a look to understand the basic principles of operation.
The Noritake displays employed in Squeezebox devices need mainly four supply voltages:

a) +5V operating voltage at low current for the embedded controller chip
b) +55V grid voltage
c) +5V filament voltage #1 (left side)
d) variable voltage between +1.5 and +3.0V for filament voltage #2 (right side)

A high current flows between c) and d), I estimate it about 800mA.

EDIT 2023-01-19: well, after all these years, I have finally come to measure it and actually it is far less. It's a unit where the the diode fix is already in place. With that in the loop, the amperage is around 105mA constantly. It's a little bit more when the display brightness is higher, and is zero when the display is off.

What we are going to fix here is d). A little theory for those who did not visit the Wiki article above. A VFD is very similar to a classic vacuum tube in that it has a heated cathode that emits electrons, a grid that controls electron flow towards the anodes, and eventually anodes to "pick up" the electrons which in a VFD are the individual pixels. If a pixel is positively charged, it attracts the electrons emitted and the phosphor coating causes the pixel to glow.
While the heater wires are supposed to permanently emit electrons, the grid sections between the heater wires and the pixels control which group of pixels receives electrons at all by setting all grids that should be blocked to the a positive potential so they basically catch all free electrons. Only the grid section where electrons should pass through is switched to a neutral potential so the electrons can get past it. The neutral part in the grid sections is scanning to enable all groups of pixels in a cycle. This happens hundreds of times per second so the human eye will not notice it. A slight flicker can be observed when the eye moves quickly across a VFD though.
Filament starvation causes the heater wires to reduce or even stop electron emission. There can be any degree between hardly noticeable to completely dead. In some cases, both heater wire connections are fed with +5V which means there is no voltage drop at all. While the display may be fully operational, it will act like it's dead.
So the thermic electron emission in a directly-heated cathode (such as the heater wires in the VFD) is caused by a voltage drop across the heater wires. The higher the voltage drop, the more the heater wires will glow and thereby emit electrons. In reverse, if the voltage drop gets too low, the electron emission will slow down or stop. A considerable current is needed to heat the wires up. In dark environments, you will even see the six horizontal wires glow. This is also a reasonable limit for the driving circuit. A balance should be found between sufficient electron emission that is just enough to have a good visual impression that is not disturbed by the heater wires glowing. The corridor between both is rather narrow.

Boom Failure Mode

While I never had any Boom with failures regarding the filament voltage #1 which is a stable +5 Volts whenever the Boom is connected to the external power supply (yes, even in the deepest standby mode), the other side of the filament is apparently the output of a power supply circuitry that has some still-unknown component that fails over time. It is meant to supply variable voltages depending on the selected brightness level, so there is a more complex circuitry behind that that is designed to handle the high current... or rather not apparently.
If you have a multimeter at hand, you will be able to measure +5V at any of the three leftmost pins of the display - positive test lead going to the display, negative test lead put to GND which can be found at any of the screws that hold the board in place while mounted. If not mounted, use the gold-plated screwholes instead. You should also measure approximately this between GND and one of the three pins on the right side:

Level 5: 1.418 Volts
Level 4: 1.414 Volts
Level 3: 1.413 Volts
Level 2: 2.513 Volts
Level 1: 2.511 Volts
Level 0 (off): circuit open

By the way, if you switch the display 'off', both filament voltages might be GND or dropping towards GND. There seem to be releases of the Boom where the display is actually turned off, in contrast to what I assumed. Just found this out. Interesting. I will probably investigate deeper here because that is what I would like to see in all Squeezebox devices when their displays are not showing anything. Just imagine how much power might be saved, and most Booms are pretty wasteful in this respect by keeping the heater wires hot all the time, no matter the operation mode.

So levels 5 and 4 mean a voltage drop of about 3.6 Volts which I think is pretty extreme. The voltage drop in levels 3 thru 1 is about 2.5 Volts.
When a Boom begins to show filament starvation, what happens is that the variable filament voltage rises too high. I measured up to 5.5 Volts in defective units which means the right side is even supplied with more volts than the left side, reversing the electron flow, but the resulting 0.5V voltage drop is not getting you anywhere.
So the failure mode regarding the right-side power supply is not that it fails and is eventually 'open' or falling to the potential of GND, but is reaching or even exceeding +5 Volts instead. So what fails is apparently some kind of pull-down circuit that is designed to provide stable voltages below 5V at a higher current.
We have +5V on the display's left side so the right side needs at least 1V more or less for a voltage drop big enough to start the magic, or better 2 Volts. The higher the voltage drop, the more the heaters will glow, so there is a voltage drop where undesired effects begin to show. Heater wires glowing red is probably not what you want. They can withstand a *lot* of power, so they will probably not take damage, but will emit a ton more electrons than actually needed. This will make the active pixels shine brighter which will more quickly consume them. Consequentially this will cause shadows in the display caused by pixels which burnt down quicker than others. Which puts you back to square one, having to replace the display again.
So let's assume that we want to define and limit the voltage drop. Otherwise, you could set the right side of the filament to GND, creating an effective voltage drop of 5V, and you'd be done. But that would be too easy, wouldn't it? A display might last some weeks or even months under this condition but it would cook itself to death.
Unfortunately, to this day, there is no technical documentation available from Logitech, and it is uncertain if it will ever surface. The majority of the Boom consists of tiny SMD parts which are hard to figure out, mounted to a multilayer board (at least 4-6 layers, I think). It's practically impossible to know what part of the circuitry is failing, and what could be replaced to fix it.
Basically, even if we knew, the design has an apparent flaw. A brand-new component will just solve the issue for a time before it starts failing all over again.

The (Pseudo-)Solution

IMPORTANT: You should not apply this fix unless you are more or less convinced that your Boom is suffering from the filament starvation phenomenon! Because there is no way of knowing if an intact power supply circuit could be damaged by the fix so things get even worse.

I have a more 'brutal' approach to this. Assuming that the right-side power supply has failed and generates +5 Volts or more continuously, why not pull it down to a more reasonable level with the help of just a few additional parts? While this will eliminate the ability to create arbitrary voltages, I found that a constant voltage drop will still allow you to fully control brightness levels. And it's really a very simple addition:

What you see here is the right side of the display. The three pins at the top are linked together (a tribute to the higher current that passes them). This is by design. Even if the three pins are not connected to anything, they share the same potential. My idea is to use three simple diodes in series from these pins towards GND, shown in yellow.
The diodes used can be run-of-the-mill 1N400x-type diodes which are available for just a few cents each. Just be careful to order ones which are capable of handling up to 1A of current. Typical diodes have a voltage drop of around 0.7 Volts each, so a chain of three creates a voltage drop of ~2.1 Volts. Assuming that the original voltage is around 5 Volts, the diodes would pull it down to about 2.9 Volts.
That is in the range of the voltage that an intact circuitry would generate; a voltage drop of 2.1 Volts is enough to give you a nice clear display in all situations, and even though this is less than the up to 3.6V voltage drop in the original circuit, giving the display a lower voltage drop is likely contributing to a longer display life. Visually it's negligible.

Current flows in the direction of the arrow towards the 'bar' in the schematic, and the bar is what you can also find on one end of the diode. That is the cathode (or minus) end, whereas the other end of it is the anode (or plus).
The nice thing about the diodes is that they cause a voltage drop but do not convert this drop into a lot of heat like a power resistor would. Furthermore, the current direction is still forced to be in sync with the original design.
This has proven to be a cheap and reliable solution in many cases (up to 60 at this point). Whenever I do a display replacement, many times I'll include this fix because the power supply decay becomes visible. Newer displays at low brightness are particularly good at helping to discover this.

Tools Needed

What you need for this repair is:

• all the tools for disassembly / reassembly (see Part 1)
  
• soldering iron with fine tip (~ 285°C, no more than 40W)
• 3 1N4001 (or 1N4002, 1N4003, whatever) silicon diodes with 1A rating
• 1 pair of pliers
• 1 small piece of heat shrink tube ~6mm diameter
• 1 small piece of heat shrink tube ~2mm diameter
• highly recommended: a "3rd / 4th" hand tool to hold the diodes in place for soldering

Creating Your Own Fix

So here is a guide on how to integrate these diodes. Please consider reading it through the end first before you start your own repair. There are some tips hidden here that could appear out of sequence.
From your collection of diodes, select three:

Keep in mind that we want a chain where one end of the overall construct is the anode and the other is the cathode. All diodes between need to be in line, the anode of one connecting to the cathode of the next. Reversing any diode in this chain will probably end in something nonfunctional that acts like an open link.

Pick two, bend the anode (unmarked) wire of one 90° to the side. Do the same with the cathode (marked) wire of the other.

Now let's join these two with a little drop of solder. First cross the bent pins, put the diodes as close together as possible, and solder:

The result is a 'Pi' pair of diodes:

Now bend one of the pins away 90° again. The third diode will be soldered to this wire. In this case, we bend the anode wire of one diode:

Also bend the cathode side of the third diode that comes in now:

Life gets easier if you clip the wires of the diode contacts that we already soldered together:

Now cross the two bent leads again and solder them together. Ensure that the anode of one diode is connected to the cathode of the next diode, or vice cersa. This needs to be consistent, otherwise the fix will not work.

Finished:

The end result is a series of three diodes in the smallest possible form factor:

The brave of heart could try a little dry test now. If you power up the Boom board, holding the diode cascade's anode to one of the three rightmost display pins and the other lead to the gold plated screw hole next to them (see schematic in "The Solution" section above), this should immediately light up the display to a much brighter level if it works. But make sure you do not touch anything else on the board with any side of the diode cascade! If nothing happens, or it gets blindingly bright and you see the heater wires getting red-hot, there is something wrong. Please stop immediately, recheck, and do not continue because this is not recovering by itself. In doubt, better send me a message and I will try to help you.

We need to isolate the diodes a little to ensure that we won't create any short circuit in the precious Boom board. I recommend heat shrink tube, one piece of ~6mm diameter (yellow in the following images) to cover the barrel of diodes, and another of ~2mm diameter (blue in the photos) which we will see shortly.

Prepare two pieces like so:

Mark one piece of the 6mm heat shrink tube so we know where the cathode of the chain is. It should correspond with a cathode marking on the last diode of the chain (it's the one closest to the camera in the following picture):

Push the 6mm heat shrink tube over the diodes, and the 2mm one over the cathode lead:

Then apply heat (150-200°C):

The anode wire (the unmarked side) will go across the three pins to the right of the display. As there is not enough space on the main board's front side, we will place this piece in the back.
First, bend the anode wire 90 degrees off:

 
Place it across the three pins as shown here and solder:

Please make sure the lead you attached to the display's pins does not touch the Wi-Fi antenna metal pad! Missing this point may damage your Wi-Fi board, or worse. The antenna pad has some degree of GND level. If both pieces make contact, you have around 5V voltage drop, so its way too high, and goes through the Wi-Fi card, too, which can't be good. Please avoid from the start. Place the lead as far right as possible. In doubt, put some isolation tape in between, like Logitech did in the original design (this strip of black plastic adhesive tape is missing in the photos). The diode contact should still touch all three solder pads of the display but this is not vital. It also works if you solder the diodes to just one of these pins as they are shorted together in multiple places on the board as well as in the display itself.

Clip the excess wire and bend the diodes so the cathode wire points to the right:

You can see in the previous picture that there is not much clearance between the diodes and the antenna. The gap should be as big as possible. Or, if you are even smarter, you first read up to here and remember to put another small piece of 2mm shrink wrap to this end of the diode cascade to isolate this properly :o)
I helped myself by bending the entire diode setup after soldering which becomes apparent in the next pictures but it's certainly not the nicest solution.

You may have to adjust the length of the 2mm shrink wrap tube. The wire needs to be soldered to the screw hole. Bend the lead to go there the shortest way. Prepare the diode's lead as well as the outmost section of the screw hole gold plating with a bit of solder:

Now join both:

Make sure the contact is stable because there will be some (mechanical) load on it once you reassemble the Boom.
You should not see any indication of the fix on the front side:

But on the back - see top left corner!

That's it folks! You're done!

I'd be overjoyed if you share your results and thoughts in the comments. Thanks for reading and for your feedback!

Final Words

Some legal stuff because you never know: please bear in mind that I am writing this as a hobbyist, not a professional. I describe personal ideas here which is only one of many ways such a repair can be achieved. I cannot guarantee that following this guide will lead to a good result, and cannot be held liable for any personal, physical, or monetary damage anybody suffers by following this guide.
I am open to advice if anything described here is wrong or can be done better. Please let me know in the comments if you find there is anything left to be desired.
Thank you!

Fixing Denon PMA-980R Input / Speaker Selector "Jumpiness" (English)

Hinweis für deutsche Leser: diesen Artikel gibt es auch auf deutsch --> Link

What to do when Denon PMA-980R "goes mad"

If it wasn't for Revox and their excellent B200 series, it is likely that a PMA-980R would still be my main amp for audio entertainment. While my first "real" amplifier was a Denon PMA-360, which was best for its price, it always lacked remote control for me, like any other component in the rack.
When PMA-980R hit the market, it was right on the spot. Denon had proven good for me for some years then, I was certain this was a good match. Besides the sound quality, a lot of other attributes were just right: it is made of solid heavy metal (near 10kg), features a quality toroid transformer, plenty of relays, a Record Selector, a Source Direct switch, and overall a very good feel about it. Thick aluminium front plate, huge heatsinks, and only the most inevitable amount of plastic.
In my opinion, the amplifier is able to compete with newer devices, thanks to the excellent Sanken transistors used in the power amplification section. No compromises were made in the volume adjustment section which consists of a motor-driven quality potentiometer for remote availability which also features a LED position indicator that can be seen easily from meters away. This was the best way to combine quality and usability. Digital attenuation was too expensive for the PMA-980R margin, and going cheaper in digital is always lossy.
Instead of mechanical switches and heavy wiring, Denon engineers employ processor-controlled relays for managing the signal path and speaker outputs. Thanks to this, everything can be controlled remotely without penalties in signal quality.
Both knobs (Speakers and Input) do not have end positions but can be turned infinitely. They have "notch" positions however which convery a very upworthy feeling. Turning a knob won't redirect any signals right away, instead, the knob action causes the processor to calculate what to do, and adjust the relays accordingly. In theory, this is much more sustainable than approaches where the signal runs through a highly complex multi-way switch, or an entire cascade of switches. Eventually, any switch in the path can be a failure point, and using sealed relays (and thereby extremely short signal paths, too) reduces these failure points to the absolute minimum. Contacts in the open air will corrode sooner or later, causing noise, interruptions and undesirable crackling and peaks, up to a complete dropout of one channel or even both.
Signal paths were also kept short for the Record Selector whose multi-way switch is positioned near the Cinch terminals and driven by a Bowden cable to make it accessible to the front. Yes, no relays here, rather classic and corrosion-prone, but another set of relays whould have made the entire amplifier considerably more costly. It's not the current subject but might be in the future, we will see...

Construction Analysis

Knobs without an end stop nowadays are a certain indication of a rotary encoder (see  my Panasonic Jog/Shuttle experiments). Well, not quite here. The PMA-980R is actually using a resistor ladder as revealed in the service manual. Basically it is a large voltage divider with 12 taps. Paths are shown here in orange for loudspeaker selector and green for input selector:

It is immediately clear that a lot more positions are possible than there are inputs or speaker groups: while there are six actual inputs, and four loudspeaker configurations, each selector has twelce positions, each resulting in a different voltage that results from the amount of resistors (10kOhms, 20kOhms etc. up to  120kOhms).
My guess is that none of these taps is actually bound to a certain input or speaker group. If they were, it would not be possible to use the amplifier via remote controller and on the device itself consistently. Probably the approach is rather to have the processor constantly measure the voltage, and a change in the voltage is interpreted as a change on the knob in clockwise or counterclockwise direction.
Inputs are TAPE2, TAPE1, PHONO, CD, TUNER und AUX. From AUX the amplifier will roll over (clockwise) to TAPE2, and from TAPE2 counterclockwise back to AUX. However, the remote controller can switch to an arbitrary input anytime. If that happened, and the knob is used, then it will still safely address the next or previous input relative to the one currently selected. So the resistances and resulting voltages must be merely an indication of the turn direction of the respective knob.
Supposing the CD input is at 20kOhm, then a resistance of 30kOhm would switch to TUNER, and 10kOhm would go to PHONO.  But actually it's not the fixed values that count, rather it's the relative change compared to the previous value. If the voltage is lower, it is typically the result of the switch being moved into one direction whereas a higher voltage means the switch moves in the other direction.
The resistor ladder is (in my opinion) a little overdimensioned. To safely recognize the turn direction, three resistors would have been sufficient. It is not clear why 12 were chosen. Anyway, the switch will sooner or later reach its last position, i.e. the 12th tap on the resistor ladder, which results in 120kOhms resistance and the respective voltage. The next position is back to the 1st tap which is just 10kOhms. So the processor needs to be able to identify this as not -110kOhms but rather see it like +10kOhms. It is the same rule for the other direction where +110kOhms actually means -10kOhms. All of this magic is probably happening inside the processor.

At powerup, the most recently used configuration concerning speaker groups and inputs is probably restored from a buffered memory, and the selection is then associated with the current resistance value of the respective knob. The other voltages are all relative then.

Downside of "Mechanical" Components in This Design

Back to the actual issue. Corroding surfaces were reduced as much as possible, but inevitably the selector knobs will be subject to corrosion, too. No audio signals pass these selectors so the corrosion will not affect audio at all, at least not quality-wise. However, selector corrosion in the PMA-980R and similar designs may have you end up with one or more of these symptoms:

• knob not reacting at all
• the LEDs around a knob are flashing one by one or in random order, and the relays keep clicking. This may be intermittent at first and become more and more permanent
• the LEDs do not follow the turn direction
• the input or speaker selector acts randomly without any obvious trigger

Bad contact causes the selector resistance to be some random value or even a completely open connection which the processor can't deal with.
All this is usually just due to corrosion or dust accumulation or both and can be repaired with just a little time, and some tools. However, there are other factors that might cause a similar behavior. For instance, if the processor supply voltage has too much ripple (usually due to dried-out electrolytic caps), it may also start acting weird. But let's concentrate on the much more likely for now.

Repair Approach

To get down to the switches in question, we need to do this:

• remove upper housing (3 PH2 screws on each side and 2 PH2 screws at the upper back edge)
• the Speaker, Input, and Volume knobs (be careful with the latter because its illumination requires a small cable to go to the front side of the knob. Do not pull too hard or you might rip the cable)
• remove front plate (5 PH2 screws on the bottom and 2 PH2 screws on the top side)
• remove some cable connections between the front plate and the rest of the amplifier. Unfortunately, this requires to clip several cable ties that we need to replace later
• two hexagonal nuts (size 11) that connect the Speaker and Input selectors to the front plate, and thus connect the processor PCB and the front plate (the PCB can be removed easily after this step)
• two small connectors at the left and right side of the board need to be unplugged
• a metal strip across each switch needs to be desoldered, then the switches, each with 13 pins in total
• tear down the switches
• clean all contact surfaces
• add a thin layer of contact spray
• assemble the switches and solder them back in
• restore the metal strips
• test
• assemble the amplifier

What You Need

• PH2 screw driver
• a ratchet with a hexagonal bit size 11 to loosen the Speaker and Input selectors from the front plate
• side cutter to clip the cable ties
• to desolder: a vacuum desoldering station is highly recommended, otherwise a powerful soldering iron (the soldering pads are rather massive) and/or solder wick
• to solder: soldering station or a soldering iron with max. 80 Watts. Please don't use one of those soldering guns for they are not very precise and heat should be somewhat controlled. A temperature range of 250 to 300°C is usually enough to melt fresh solder, and going much higher will only weaken the PCB
• a sturdy gripper to help disassemble the switches
• optionally a flat screwdriver
• for cleaning:
• several pieces of soft cloth, as lint-free as possible
• some earbuds (Q-Tips), at least three per switch are probably needed
• Isopropylic alkohol 90% or better

Disassembly

Housing

Loosen six black PH2 screws, three on each side of the device
At the upper edge of the back side, loosen two smaller PH2 screws.
The top cover can now be removed by pulling it right upwards

Front Plate

Prior to this step it is advisable to remove the knob handles for Speaker, Input, and Volume. Keep in mind that the Volume knob is still connected with a small white cable for the illuminated LED.
All connections between the front plate and the other boards inside need to be loosened. On the way you will have to clip five or six cable ties. Be careful when pulling the connectors out of their sockets to avoid that any cable is torn. Be extra careful at the edge of the main board where the volume knob's LED connector (and others) resides. The board is flapping in the air here, and warping it too far might overstress is. Shame on you there, Denon!

The front plate is screwed down with two PH2 screws at the outer left and right edge, and another five PH2 screws on the bottom side (best accessible if you place the amplifier upside down). The bottom screws are those which are next to the front plate.
After this you should be able to pull the front plate away from the device with not much resistance. Be careful about the cables though, and remember how they were routed. A photo taken now might serve you well later.

Separate the Processor Board From the Front Plate

Mechanically, the PCB is only connected to the front plate with two size-11 hex nuts as shown here (left and right, around the axis of each switch):

Loosen and remove these to free the PCB:

There is a caveat here that is important when you put the front back together: the LEDs have no standoffs to keep them upright. All they have is their legs, cut to size, and they easily bend as shown by the warning indicator in the top right corner. You need to ensure that you un-bend all LEDs to ensure that each gets into the white guide pipe (see in the lower half). This is needed to prevent light bleeding from one LED to another. If an LED is bent, the respective light will be missing, and there is the risk of shorting something.

The processor board should already be loose because all that fixed it to the front plate were the size 11 nuts that we removed previously. Note that the board is still connected to other components of the front plate. On the left, there is a 3-wire cable between CN3C on the processor board and the IR receiver which is on its own PCB near the power knob. On the other side, there is a 4-wire cable between CN4A and the Source Direct switch and LED which are on a separate PCB as well.
Let's have a first glance at the processor PCB now:

The largest elements are the switches we need to fix. Each is secured by a metal "bridge" that goes across the entire switch and is soldered on both sides. Unfortunately, these might be a pain to remove because the metal tabs were bent on the soldering side before they were soldered down. So you will have to fight a lot of solder, and also ensure that the tabs are straightened again to get them out of the way.

Using desoldering braid is recommended to remove the majority of the solder around the bridge legs. You may free them up completely so they can be straightened and removed. Be careful not to damage the solder pads - they come loose very easily.

A side view of the PCB:

Desoldering the Switches

Check first: if after removing the metal bridge the switch already falls apart, that's actually a good thing. It may save you the trouble of desoldering anything, presuming the green part also comes off by itself. I found it easiest to get rid of the metal strips by "flooding" the solder pads that fix them with fresh solder. It ensures that, while heating it all up, the solder liquefies everywhere.

Setting the desoldering station to 325°C, everything was heated up to that point, and vacuum was used to remove all solder. Multiple attempts are not a shame here. Once the metal tab is free enough, bend it straight with pliers and push it through to the other side.
The switches are soldered down by 12 pins in a circular arrangement, and a 13th one for the middle contact, inside this circle. They all need to be desoldered. You won't have any trouble putting them back in later because the 13th pin acts as a kind of key and ensures that only one position is actually possible.
As the pins are rather big, and solder was generously applied, one of the bigger nozzles is required on the desoldering iron. You might try with solder wick or the manual "one-shot" handheld vacuum desoldering pumps.

Disassembling the Switches

The switches after this procedure:

Bottom side. The middle contact can be seen to the right of and a little below the center axle:

The switch consists of three parts: the brown base, a green plastic piece, and the metal frame holding the axle, and clamping everything together.
The metal frame is secured to the base by two metal ends which spread apart beyond the base. They need to be pushed together far enough so the metal ends pass through the base. As the metal is quite sturdy, some force is required here. However, be careful not to push too hard or you will make your life harder when it gets to spread them back into the secured position later.
You can push them together slowly with less than maximum force and simultaneous wiggling of the plier (maybe +/- 10 degrees around the axis of the metal end).

Once the ends are close enough so they can pass through the slot in the brown base, you can try to loosen the base. I am using a flat screwdriver here as a lever:

Eventually the top part is off:

Removing the green cap requires the four "anchor" tabs at the bottom side to be pushed together, and meanwhile pushing them through the hole in the base:

You may begin with two tabs on one side, once they are loose, insert something between the base and the green plastic piece so the tabs won't snap back. Then loosen the remaining tabs and push the plastic cap out.

We can now see the slider in the green cap which in operation moves across the middle contact of the base, and makes contact with one of the twelve metal pads around it. The slider tracks are clearly visible on the metal pads as "tracks" where a silverish surface is shining through. The gold look is actually some kind of grease plus corrosion plus dust.
After a first wipe, things start to look better:

I switched to iso alcohol now, and cotton earbuds. You will be surprised how much dirt is still coming out, even if everything looks clean already. I recommend wasting earbuds until no more dirt accumulates on them.
Eventually you may treat the metal surfaces with sandpaper (800 or higher), or use a glass fiber pen if you have one. Be sure not to create scratches in the surfaces, just remove what is clearly dirt.

The counterpart should also be treated. Even if the naked eye might not see it, some dirt accumulates here as well. But be careful with these delicate metal springs, they bend easily and should not experience mechanical stress more than is absolutely unavoidable. It is enough to just clean the rounded surfaces that mate with the base contacts. Even if the rest appears dirty, better don't get there, it might ruin the spring properties.

Now for the contact spray. There are many theories about it, so I have my own as well. I think contact spray is bad if you just use it to flood devices and their switches from a distance, apparently sparing the trouble of going into disassembly or even desoldering. This might help for a time but as the dirt stays where it is, and is at best redistributed a little, the issues will keep coming back. Surfaces "treated" with the spray will accumulate dust even quicker because the oil component of the spray never evaporates. The end result is usually a mess.But a thin layer of contact spray here might help in the long run, now that everything is clean and shiny as new, the spray might help conserve this status for a longer time. Besides, we have exposed metal contact surfaces which might corrode if not treated somehow.
So my recommendation is to put contact spray on a cotton bud and wipe that across all metal surfaces. 

You can also use battery terminal grease, it will ensure good contact and smooth operation.
So here is one switch, cleaned up, treated with contact spray, prepped for reassembly:

Switch Reassembly

Basically you do everything above backwards. Clip the green plastic cap back onto the base. Don't worry about the position yet.
The axle can go in only one way because it has a flattened contour that needs to be aligned to the slot in the green cap. Push the metal frame down and route the metal tabs through the appropriate base plate slots. Once everything is pushed together fast, you can spread out the metal tabs again so they secure the entire assembly. This is needed to ensure that the gap between the base plate and the green cap is as small as possible, to avoid anything getting in there.

Soldering the Switches Back

As said above, the switches have only one position that allows them to be put into the processor board correctly. Find the middle pin and align it to the respective hole in the board. It's no matter which of the switches you put in which position as they are identical.
I recommend that you solder only one pin at first, then check whether the switch is sitting in the correct end position. It is easier to correct anything askew right now.
After all 13 pins are soldered, the brave might run a first test by connecting all the cables but not reassembling the entire devide. You do not actually need the handle knobs for the switches in order to turn them. But be careful not to cause any kind of short circuit in this test. Just do this if you really have enough experience!
Ideally you should not observe any more jumping or other random behavior now. The switches should react promptly and as expected.
The metal tabs are the final step of mounting the switches. To make your life easier next time, just don't bend the tabs on the soldering side again, like Denon production did. It is not necessary at all. Just push the metal bridge in and put solder on it. It might help meanwhile to push against the metal bridge from the other side to get it in as far as possible. This gives you the maximum surface to solder against so it is more secure later. If things become too hot for your fingers, consider using the pliers for this.

Reassembling the Amp

Put the PCB back to the front plate. First ensure that the cables on the left and right edge are plugged back in. If you removed the 11-wire cable at the top like I did, plug it back in now because you won't be able to reach this spot once the board is back in place.

CAUTION: the LEDs around the switches are routed to the front through little pipes which have little tolerance. So each LED needs to match its respective pipe exactly. If it doesn't, the LED might bend sideways. You might notice this only some time later when you find that one or more LEDs don't seem to emit any light. They actually do, but in a direction you don't want. So keep an eye on the LEDs specifically when you put the PCB down. If they all point straight up from the board, nothing should go wrong here.
Besides, I found the IR sensor board in my PMA-980R was extremely loose. Just two plastic anchor clips would "secure" it, and they performed miserably at that, which resulted in the IR board wandering around behind the front plate, and IR commands not doing anything. So please ensure that the IR board is in its home position while you put the front plate back on. A loose IR sensor PCB could cause short circuits, and the loss of remote control is not what we want either.

Furthermore, when you reassemble the front plate, be careful to route all cables the same way they were before you started the disassembly. If a wire connection is apparently too short, that might be because the wires looped somewehere behind the front plate. Don't just pull them to their expected length. If you are in this situation, better reseat the front plate and ensure no cables are caught behind it. This is easy now, and much harder later on.

On the bottom side, push the front plate in between the chassis and the bottom metal plate, sandwich-like:

If you feel too much resistance here, or the panel cannot reach the end position, there is either a wire squished somewhere between front panel and chassis, or one of the push buttons did not find its way through the front plate. Main suspects here are Loudness, Subsonic, and Source Direct. Make sure you can see the buttons, reseat the front plate again if needed.
Eventually the front should go on without much effort. In its final position, you should be able to see the threads beneath the seven screw holes.
If you removed the plastic side covers left and right of the front, put them in now as their screws cannot be accessed once the top cover is back in place.
Reseat all cable connections, also don't forget the ground connection between the front plate and the front heatsink. Tie down the cables approximately where they were tied before.

Another test is advisable now that short circuit risks are practically out of the way (unless you drop anything into the open device under operation). Please be aware that dangerous voltages and charges are built up in the amplifier, so keep your hands away from any place but the front
 

EDIT 2025-12-22: Check for Another Design Flaw

There's another issue that many PMA-980R have: two driver transistors which miss some cooling. They get hot enough to damage their own solder points which can cause dropouts and other errors. It's a good time to check for this before putting the cover back on.

These are the transistors in question: 

Ensure the amp has cooled down, then check if these transistors can easily be moved back and forth. If they are not solid, that's a rather safe indication that they have began to desolder themselves.

Fortunately, it is not necessary to extract the PCB to fix it. The bottom cover is very service-friendly. It is in two parts, and only the one covering the PCB needs to be removed. The feet can stay mounted to the plate.

From the bottom of the PCB it might be visible that the transistor legs are moving as the transistors are wiggled at the top side. You can probably identify the soldering places of these transistors easily as the heat has caused the PCB to develop a yellow / brownish tint around the transistor pins.

Resoldering is only the second-best choice in my opinion. It will solve the problem but it's likely to come back, and the excessive amount of heat may cause more trouble in the future. Therefore, my recommendation is to move the transistors entirely from the PCB to the heatsinks nearby. Denon's engineers could have thought of that...

Sorry I have not made any pictures from the process. But here is the end result:

 

The heatsinks have some free holes that we can use to bolt the transistors on tightly. I recommend strong wire. Be aware that the diameter should be close to the diameter of the solder holes where the transistors used to be. I have used wires from a defective PC power supply, they are really good for this purpose:

 

 

Some advice for this:

• D1913 is the marking of the front-side transistor, B1274 is the marking of the one close to the rear
• both transistor markings can be read from the front. Be sure not to flip them accidentally
• I soldered the wires directly to the transistor's legs and put shrink wrap on top of that to ensure insulation. Crimping might be even better considering the heat. Any solution is better than leaving the transistors on the board
• between each transistor and the heat sink, you should apply some thermal compound, such as the stuff used in PC CPUs and their heatsinks / fans
• the transistors have plastic housings and no metal back plate, so there is no risk of shorting anything when they are mounted to the heatsink
• you can pick a screw size that fits the transistor mounting holes and the heat sinks. As there is no thread, the screws should be strong enough to make their own threads as they are turned. A rather coarse screw might serve better here than a M4 or the like
• it should be possible to bold the screws down fast enough so the transistors make full and firm contact to the heatsink
• to put the wire through the PCB for soldering, it may be required to widen the holes a little. Be careful though to leave the solder eyes on the PCB's bottom unhurt. A set of very fine drill bits is required, and this is best done manually so it won't escalate too quickly
• to prevent the wires from unwrapping while being fed through the PCB holes, their tip can be soldered. This helps keep the braid slim and still ensures it won't undo along the way
• I fixed the wires on the PCB top side with some hot glue to secure them against vibration
• before powering on, check all the transistor contacts on the PCB for shorts that may have been generated from stray wire strands or solder bridges. You should only measure rather high resistances between the contacts
• bolting the transistors down is complicated due to space restrictions by many components on the board. Only the outmost hole of the heatsink can be used, the ones further in are blocked by the large capacitors. I recommend an angled screwdriver (needs to be a good one as the force to set the screw and thread it in will be rather enormous). If you want to do this perfectly right, you will probably remove the entire PCB and free the heatsinks, then cut M4 threads, and take it from there
• an extensive test should be conducted after this. I recommend an hour of high load while watching the temperatures closely

Final Assembly

The cover slides back home just being put down straight from above with the sides bent slightly outwards. Reset the two PH2 backside screws first, then the six side screws.

That's it! Should be like new, enjoy it!

Some Design Criticism About the PMA-980R

Denon did almost everything right with the PMA-980R. However there are some things left to be desired. What I am missing might be due to the price margin though.
The jumpiness issue might have been avoided with a) less than 12 positions for the selector switches, or b) with optical encoders. As discussed above, the different resistance values just serve the purpose of determining the turn direction. For this, three different resistance values would be sufficient. I cannot think of why you need twelve, it's unnecessarily complex.
A purely optical encoder might suffer from dust accumulation but as there is no mechanical-electrical contact, it won't suffer nearly as bad. An open optical encoder system (like in early computer "ball" mice) could also be cleaned easily without a full disassembly of the component.
Either Denon's engineers didn't think this far, or it was too costly at the time. It's unlikely that planned obsolescence played any role back in the day but in 1992 engineers knew corrosion problems well as practically all hi-fi devices had open potentiometers and switches forever.
I don't want to think about the amount of amplifiers and similar devices that ended up in the landfill just because of such a minuscule cause, and owners could not figure out that it might be fixed rather simply.

Another issue may surface during rough handling of the amplifier. Certainly a little out of spec, but I had one victim of an accidental drop that was beyond fixing. The reason is that the mainboard in the PMA-980R is badly lacking mechanical support. As you can see from the next picture, the mainboard is pretty large, filling up the center and right space:

However, the board is bolted down in just four places, near the heatsinks! The only things that secure the right side of the board are the cinch connector terminal (which is not designed for this job) as well as the motor-driven potentiometer which is surrounded by some metal. It's likewise not designed for holding the board in place because actually the solder joints of the potentiometer are the place where this happens.
The right front edge of the board is not supported at all, i.e. if you pull a connector or push it in, the board flexes considerably. It could go as far as the board getting a crack in the most stressed place.
This is particularly sad because the PCB designer provided some holes in that corner, but the case designer did not put anything near there for a screw to hold.
The board offers enough space for additional screws but there is nothing beneath the board to screw against.
I think that at least three or four additional screws would have been really nice.
All is well until a certain degree of acceleration is exceeded, especially when the device is dropped flat from some height. The drop damage may face in a split through the entire board.
Because an even greater design mistake is this: the buffer capacitors (the large brown / shiny cans in the middle of the board) are neither protected from vibration (other manufacturers would surround them with a rubber or foam rubber support), nor are they mechanically decoupled by any means from the board. Their weight becomes a considerable problem combined with acceleration. On a drop, they will put all their weight down on the rather thin board, and a crack rips right through the middle of the board (horizontally, in the picture above). It might not be a visible crack but just a hair line, anyway it's not economically fixable.
If the place of the PCB where these caps are positioned would have even the slightest support, e.g. from a plastic frame beneath the board, or a metal frame above the caps that clamps them in, this could not happen, or would at least not end so badly. A PCB of better quality (epoxy instead of the "hard paper" type they appear to have used) would also have been a good choice.
The following picture illustrates that there is nothing but air beneath the board 😟:

Further Advice:

A camera is expressly recommended for documentation of cable routing, screw positions, erc. Better make one more photo than missing one later.
As a matter of course, be aware that working on the device in its powered-up state is dangerous and should be avoided. It is not only dangerous for yourself but also for the device under repair. So if you work on its internals, be sure to power it off and even better, pull the plug. Components such as capacitors can store a lot of energy for a long time even after the plug was pulled so if you are working on these components or places where their connections are exposed, be sure to discharge them first. The easiest and most effective way to discharge a high-voltage capacitor is to connect it directly to a 230V light bulb. This will quickly eat all energy stored. If there is too much energy, the light bulb may be blown out, but that's still better than you getting this kind of charge.
Some legal stuff because you never know: please bear in mind that I am writing this as a hobbyist, not a professional. I describe personal ideas here which is only one of many ways such a repair can be achieved. I cannot guarantee that following this guide will lead to a good result, and cannot be held liable for any personal, physical, or monetary damage anybody suffers by following this guide.
I am open to advice if anything described here is wrong or can be done better. Please let me know in the comments if you find there is anything left to be desired.
Thank you!