Vacuum Fluorescent Display Driver

[floobydust]

"Agilent 53131A Universal Counter Component Level Information" 5989-6307EN Feb. 2007 pdf does list the Supertex HV518PJ for U2, same p/n 1820-5330 as the TI SN75518FN. I say the IC is a drop-in replacement but we'd need to compare an old (75518) schematic to see if changes were made. Only the 53131/132 use that DC-DC converter.
I can only find that C129 is way too big, the hold-time of that 38V rail is large and the display likely gets blanked fast on power-down, which decreases the current drain and lengthens the ramp down.

You have be careful C129 470uF 63V (on the main board) is totally discharged before plugging in the ribbon cable or it will zap the IC. That would kill it.

The schematic shows U7 LM317 jumpered IN-OUT, front panel board +5V coming in from J1-10, J1-20. Back to J6-10, J6-20 and a ferrite bead to SW+5V. So it's the same rail as the DC-DC converter power. The 38V will lag on power-on but any possible problem is power-off, it stays up too long.
The one VFDSDIN data line from main board U19 is powered from VCC (before the SW+5V switch) but I can't see that causing 518 latchup.

[bostonman]

he schematic shows U7 LM317 jumpered IN-OUT

I completely missed the jumper. All I saw was the LM317 and +5V. Think I may have just lost some credibility on here

The schematic shows U7 LM317 jumpered IN-OUT, front panel board +5V coming in from J1-10, J1-20. Back to J6-10, J6-20 and a ferrite bead to SW+5V. So it's the same rail as the DC-DC converter power. The 38V will lag on power-on but any possible problem is power-off, it stays up too long.

Wow, you really managed to follow this well. I hate following drawings in PDF format because after jumping to another page, I forget what the other page(s) looked like. I tried printing, but for whatever reason, it printed very light; else I would have considered having them printed on C size paper at Staples.

[floobydust]

I looked again (smallest schematic dot ever) and C129 should discharge to 6V with the zener VR2 and R29 31k6 load.

[bostonman]

Earlier I removed The VFD, and, while at it, removed the IC since it's bad.

Unfortunately the IC pads don't look like they have another de-solder attempt. One pad lifted, but thankfully it's an unused pin. A few other pads appear to be on the edge of lifting or have an extremely minor bow.

Ironically I also have one new IC remaining (although I don't have any objections to ordering more).

This weekend I'll piece together momentary push-button switches (twelve - one for each G line), resistors, etc... and begin testing the VFD. Also, I'll measure the current draw on each SEG line and filament. Maybe I'll grab a second meter and also measure the current on the G lines too.

I'll update once I get more data.

[bostonman]

Yesterday I performed the bench testing.

Attached are the results. I had 64ohm 2W resistors, so I used four in parallel to make it 16ohms and 7V on the power supply gave the filament voltage 5V (approx. 132mA).

Also, I used 27V on the SEG lines rather than 30V (figured it's best to use lower and need to increase rather than go too high - but 27V seemed plenty).

From what I can tell, everything lit and nothing is shorted. SEG P and SEG Q doesn't light anything until it gets to G12. I don't see anything else that needs to be lit, so I believe SEG P and SEG Q is normal.

I'm working on a drawing to show which SEG lines are associated with lighting each segment.

Basically for the figure 8 pattern, A-F goes clockwise with A being the top horizontal segment and G and H are the middle two horizontal lines with G being the left middle and H being the right middle.

The middle stuff is:

I - Vertical top middle
J - Vertical bottom middle
K - Diagonal top right
M - Diagonal bottom right
N - Diagonal bottom left
L - Diagonal top left

The rest of the segments (Ch1, Freq, etc...) are listed in my spreadsheet.

[pqass]

132mA filament current is only just a little higher then DC-DC spec of 120mA but it'll be fine for a quick test.

Also good to see the segment currents are <<1mA using a 10K current limiting resistor.   Q1: Are the currents in the table the sum of both Gn and SEG_m currents?  Or just the segment current alone?  See attached; ie. A1+A2 or just A1?

Q2: If it's just A1, are the A2 currents similarly <<1mA using a 10K current limiting resistor? Don't bother with a detailed table of every combination. Just report back the largest draw.

Q3: If the Gn and SEG_m currents are <<1mA, change the grid and segment current limiting resistors from 10K to 1K and confirm that the Gn and SEG_m currents are now <<25mA.  If any Gn or SEG_m current >25mA that's going to burn-out the HV518 driver.

Q4: If the Gn and SEG_m currents are <<20mA (using 1K resistors) then switch to 100R for the grid and segment resistors and report the largest current drawn (from Gn (A2) and SEG_m (A1)).  The object here is to sneak-up on the 25mA limit. If we go over, then the display is bad because a good one will never demand more than that from a driver output pin.  Normally, there is no resistor in series with a driver output.

If the Gn and SEG_m currents are <25mA using the 100R series resistors, then finally remove the grid and segment resistors completely (each connected directly to the +27VDC supply).  Q5: What's the largest current drawn from Gn (A2) and SEG_m (A1)?

[bostonman]

I will set aside time sometime this week to perform the measurements you're asking. Unfortunately it may not be until late in the week or maybe the weekend.

The numbers I provided were the SEG current only. Gx was connected directly to 10k resistors (I'd clip lead the 27V to each resistor) without a current meter - I didn't have a spare current meter handy.

Attached is the 38V and 5V (38V was taken from a via next to E4 and 5V from pin 10 on the ribbon cable on the main PCB side).

Also, this is without the VFD installed and the VFD IC removed. Per what I remember seeing when the display was working with the new IC (until I turned off power and turned on power only to find out the IC blew), the +38V discharges slowly.

The scope capture shows after 12s the voltage is still 10V while Vcc has long since turned off.

[bostonman]

I pieced together the segment locations (see attached) that I noted while bench testing the VFD.

Maybe this is applicable to other VFDs, but thought to piece this together for anyone interested.

[bostonman]

Q2: If it's just A1, are the A2 currents similarly <<1mA using a 10K current limiting resistor? Don't bother with a detailed table of every combination. Just report back the largest draw.

Since I don't quite understand how the VFD works (although I've watched videos), I wanted some clarification: Do I need to measure all the Gn lines for every SEG combination (i.e. SEG-A with G1>G12, SEG-B with G1>G12, etc...), or can I just power a single SEG line, measure all twelve G lines, and those numbers will be the same for all SEG lines?

Not sure what I have for resistors, but maybe I'll use a decade (resistor?) box. If I remember correctly, they are 2W resistors, so that may make things easier.

[pqass]

The intent is to confirm that at NO TIME does any Gn x SEG_m combination, that EITHER current (A1 or A2) draw is >25mA. 
First using a pair of 10K, then a pair of 1K, then a pair of 100R resistors, then no series resistors. 
1/2 watts should be fine given the brief on-time to check currents.

So, I envision two ammeters. One lead from each ammeter starts from the +27V supply.  Then the other lead (with series resistor)... first on G1, second ammeter lead (with series resistor) on SEG_A, then move second to SEG_B, .... SEG_P, SEG_DP1, SEG_DP2, SEG_COM.  Each time lookup up briefly to check if the Gn or the SEG_m currents exceed 25mA.  Don't bother documenting each reading; just note if it comes close to 25mA.
Then switch first ammeter lead to G2, second on SEG_A....SEG_COM.  Then switch first ammeter lead to G3, etc.
It should just be a couple of seconds each lookup at the ammeters to see.
I don't anticipate that the grid current will change much with each different segment lit.  I just want to be sure.
If you find that as you try every combination with 1K then every combo with 100R, and you're seeing >25mA currents then stop there; don't bother to go lower since the display is FUBAR.

If you only have one ammeter, then you'll have to do every combination of Gn x SEG_m twice for every series resistor combo; once with the ammeter monitoring the A1 current then monitoring the A2 current for 10K, then again for 1K, etc.

I don't also want to rush going from 10K series resistance directly to no resistor because it might zap something inside. Although, if any A1 or A2 current >25mA that pretty much guarantees that the display is a driver chip killer so it'll be useless anyway.

Again, we're just ruling out that the display will never draw >25mA from any A1 or A2 current when no resistor is in series.

-------
VFDs are low-voltage vacuum tubes. 
You must keep the filament (heater) constantly lit by supplying the recommended voltage that does not allow it to exceed its max. current. 
Then also supply a positive voltage (with respect to the filament) for the grid AND segment (anode).  Higher grid or segment voltages will cause more electrons from the filament to strike the phosphor on the anode and cause it to get brighter.

[bostonman]

I have enough multimeters, so I'll monitor both, but, since I'm at it, I'll use a third to keep an eye on filament current.

Ironically, I actually performed the measurements opposite of your suggestion. I took a perf board, soldered twelve 20k resistors on it, and then soldered a 20k across each individual one to get the 10k (I had a large amount of 20k resistors so I used them). After I put a clip on a SEG line (looking at my spreadsheet, you can probably tell the order I went - left to right on both rows), and touched each G line resistor with a clip lead.

I tried thinking of the most efficient method to make it go faster and to reduce the amount of times the VFD gets touched. This way I was only touching the VFD twenty times not including connecting all the G lines initially.

On a side note, I realized I had 82ohm resistors for the filament and not 64ohm like I originally stated, and I had five in parallel making it (as it was measured) 16.1ohms. To eliminate the chances of a resistor disconnecting and causing inrush current issues, I also soldered the resistor leads.

Also, I checked earlier, the decade boxes (that's the name I'm familiar with, but they are also called substitute boxes and resistor boxes) are 1/2W resistors - thought I had the -2W version originally. Either way, 1/2W should be more than enough. At one time I had to replace some of the resistors (they were open the day I acquired them), so even if they go up in smoke, I should have some spares.

Unfortunately I didn't expect the need to perform more measurements, so I disconnected everything. Earlier I took some time to gather everything, but most likely I can't commit myself to testing until late this week or the weekend.

Edit: did you (or anyone) look at the scope capture of the +38V taking >12s to drop? Seems odd it should take that long, almost like the zener is open.

Thinking ahead (as I did previously), I'm wondering if it's worth laying out a PCB. Since the HV518 is fed by six voltage/data lines, maybe if the VFD is found to be fine (i.e. no shorts), rather than risk damaging the PCB by soldering more HV518s, a breakout board with an IC socket would be safer.

I can jumper the six lines to the breakout board and either power the filament separately, or jumper those two lines too. Also, I could use pin sockets (whatever they are called) to fit the VFD in rather than solder. The issue would be the cost for a PCB, and, each time an IC blows, it's $8 down the drain.

[pqass]

For the current tests you don't really need to solder the VFD leads; just need to touch Gn and SEG_m with resistors on crock clips. But VFD leads are robust and can take re-soldering many times.  Just make sure the leads don't move (ie mounted on a perfboard) and don't fatigue at the glass interface. 

See attached for a SPI-VFD driver I made a couple of days ago (2*74HC595+2*ULN2003+14*10K pullups, 2.8Vrms add-on secondary filament winding+16VDC "air wired" power supply). Note the VFD taped to a foam block and strip that constrains lead movement. You can do similarly using a 0.1" thru-hole 44-PLCC socket, mounting the VFD on perfboard, and temporarily soldering leads for the SPI+DCpower+FILpower from J1.  Ultimately, it would be nice if you can manage to mount machined sockets for the VFD pins and a low-profile PLCC socket on the display board.

Yeah, I saw that it took >12s to drop.  But that was without any VFD attached so no load but the zener and a 31K resistor.
As the +5V input to DC-DC dies down, the filament is still warm and the last grid and segments that where on should still draw a few mA which should discharge the 470uF cap faster than 12s. But how fast? I dunno.

[bostonman]

Yeah, I saw that it took >12s to drop.  But that was without any VFD attached so no load but the zener and a 31K resistor.

That same slow discharge rate I'm almost certain I also saw with the VFD installed and the new-new VFD Driver IC.

After installing the second new IC, I clipped onto the +38V and Vcc on two separate channels (this is why I had and still have fly leads soldered to both points). Not thinking about needing to measure discharge rate, I just looked at the voltage (probably at a fast rate) and thought the slow discharging +38V was odd after powering off the unit.

Afterwards I turned on power again and saw the +38V was now approx. +12V along with nothing on the VFD.

So don't quote me, but I'm 99% sure that slow discharge was also with the VFD installed. Also note a previous scope capture where it shows the slow discharge rate but with the blown HV518; it starts at +12V because the chip was blown, but shows the VFD didn't help the discharge rate because it was still slow.

[bostonman]

Wanted to let anyone following this that I haven't had a chance to perform follow up measurements.

I think Thursday night I began setting up, but I haven't had time for anything more. Possibly I'll have time tomorrow.

[bostonman]

Earlier I performed tests with 10k resistors; this time noting the current on the G lines.

Although noting all the measurements wasn't asked, thought it would make things easier to have all of them. Entering all the G line currents did get a bit tedious and decided to take the SEG measurements from my previous test due to them being very close.

Also note that all the filament currents are the same in my spreadsheet. They varied very little and I gave up taking notes for each measurement. I noticed it would start at maybe 117mA while switching G lines (i.e. nothing lit) and then drop to about 116.5mA when the G line was connected (i.e. a segment was lit).

I know and understand the trick is to sneak up on 25mA, but is it worth trying to jump from 10k to 100ohms (or 100R as it seems to be used on here) to save steps or is it too risky?

[floobydust]

1kΩ is the lowest I would go to work with the tube. Operated as a triode, heater to screen-grid current limit is that I find.

I don't see you checking for shorts between "G-lines" (as you call the screen grid)? You've checked between segments I think.
Another check for shorted elements is to energize one segment and see if any others will light. Same for the grids, I tie a bunch of segments together high and energize each grid one at a time, looking for drama. I was staying out of the thread so you can work with pqass, haven't followed all the tests.

Lexicon reverb common problem is the DC-DC converter loses regulation and goes too high in output voltage and blows the VFD driver IC.
Here, the HV518PJ is good to 90V so not the problem but I mention it as another VFD-related issue I have seen.

[pqass]

Also note that all the filament currents are the same in my spreadsheet. They varied very little and I gave up taking notes for each measurement. I noticed it would start at maybe 117mA while switching G lines (i.e. nothing lit) and then drop to about 116.5mA when the G line was connected (i.e. a segment was lit).

I know and understand the trick is to sneak up on 25mA, but is it worth trying to jump from 10k to 100ohms (or 100R as it seems to be used on here) to save steps or is it too risky?

117mA filament current is good to see given that it's just under/at what the DC-DC filament output can provide.

Given how low the current readings are in your tables, it'll probably be fine to skip from 10K to 100R.  I was just being overly cautious by trying with 1K too.  Just only take a few seconds per measurement at most.   If you're way beyond 25mA it may be pointless to continue to record all measurements.   

If the round of measurements are below 25mA using 100R, then try the final round of measurements with NO resistor (ie. directly attaching GRIDn and SEG-m to +27V).  Again, If you're way beyond 25mA it may be pointless to continue to record all measurements.   

[pqass]

1kΩ is the lowest I would go to work with the tube. Operated as a triode, heater to screen-grid current limit is that I find.

I don't see you checking for shorts between "G-lines" (as you call the screen grid)? You've checked between segments I think.
Another check for shorted elements is to energize one segment and see if any others will light. Same for the grids, I tie a bunch of segments together high and energize each grid one at a time, looking for drama. I was staying out of the thread so you can work with pqass, haven't followed all the tests.

Eventually, bostonman will need to test all GRIDn and SEG-m pins directly attaching to the +27VDC source since that's what the driver does; no resistor in series.  Measuring the current first through a 10K, then 100R was just being overly cautious.  We talked about shorts as what to look for but really any excess current >25mA spells doom as that will destroy the driver chip. The algorithm involves connecting only one grid and one segment at a time, recording the current through each.  But, he covers all combinations so if there was some leakage current to an adjacent grid or segment then it will be reflected in the recorded current measurement.  Okay, it may still be <25mA with leakage currents, so as bostonman takes the measurements be on the lookout for multiple lit segments.

Agreed, once a round of measurements are done with NO resistor in series (with one grid and one segment on), he can then do a final multiple segment on test by shorting all segment pins together (tying them to +27VDC) and measuring the current through each GRID-n (to +27VDC); total of 12 measurements.

[bostonman]

Given how low the current readings are in your tables, it'll probably be fine to skip from 10K to 100R.  I was just being overly cautious by trying with 1K too.

I'm sure you also don't want to give advice that may blow up the property of others, so it's good to take caution. My next test will be 100ohms on both SEG and G lines. This time around maybe I'll just skim along without noting the currents and make a mental note of the highest one.

A few reasons why I took extra time to note the current values was to provide a clear chart on my measurements to help dissect any possible VFD issues, and for anyone else who comes across this thread who needs help. As most may know, it's extremely time consuming to stop and write each current value and then enter them into Excel.

Also, my initial current measurements included looking at the display each time to check the associated segment was lit, and I didn't see any others that were lit simultaneously.

[floobydust]

1kΩ is the lowest I would go to work with the tube. Operated as a triode, heater to screen-grid current limit is that I find.

I don't see you checking for shorts between "G-lines" (as you call the screen grid)? You've checked between segments I think.
Another check for shorted elements is to energize one segment and see if any others will light. Same for the grids, I tie a bunch of segments together high and energize each grid one at a time, looking for drama. I was staying out of the thread so you can work with pqass, haven't followed all the tests.

Eventually, bostonman will need to test all GRIDn and SEG-m pins directly attaching to the +27VDC source since that's what the driver does; no resistor in series.  Measuring the current first through a 10K, then 100R was just being overly cautious.  We talked about shorts as what to look for but really any excess current >25mA spells doom as that will destroy the driver chip. The algorithm involves connecting only one grid and one segment at a time, recording the current through each.  But, he covers all combinations so if there was some leakage current to an adjacent grid or segment then it will be reflected in the recorded current measurement.  Okay, it may still be <25mA with leakage currents, so as bostonman takes the measurements be on the lookout for multiple lit segments.

Agreed, once a round of measurements are done with NO resistor in series (with one grid and one segment on), he can then do a final multiple segment on test by shorting all segment pins together (tying them to +27VDC) and measuring the current through each GRID-n (to +27VDC); total of 12 measurements.

Last tests I did on a HP VFD, connecting a screen-grid to +HV through a 1k yielded unexpected results.
The filament stopped glowing to the right of that grid    The display works fine. I'll see if I took pictures of that, had me stumped.
I figure cathode current was being pulled. The tube can be run as a diode- so screen current must be limited and 1kΩ was low enough for me.

[bostonman]

Earlier I tested all the lines with 100ohms and I didn't see any G lines that exceeded around 0.5mA and none of the SEG lines exceeded 8mA; so I went all in with removing the series resistors.

The final tests without resistors was basically the same (I expected slightly higher current). Also, I set the power supply current limit to 20mA (although I don't know the accuracy of the over current protection), but, obviously the power supply never went into current limit.

The only anomalies were some SEG lines had either varying current (on the order of 0.2mA), or the G12 lines on a few segments began around 5-6mA, but kept climbing. Eventually it rose to approximately 8mA (hence where I got the maximum current) and stopped.

Shall I plan to measure for shorts between G and SEG lines too?

[pqass]

Last tests I did on a HP VFD, connecting a screen-grid to +HV through a 1k yielded unexpected results.
The filament stopped glowing to the right of that grid    The display works fine. I'll see if I took pictures of that, had me stumped.
I figure cathode current was being pulled. The tube can be run as a diode- so screen current must be limited and 1kΩ was low enough for me.

First of all, VFD filaments shouldn't be seen glowing (red) at all.  It could be related to using too much voltage and/or using DC as a filament voltage vs AC which is normally used. DC produces a voltage gradient across long displays that manifests as dimmer segments on one side vs the other. AC gives a more uniform brightness.  See "5.4 DC Filament Drive" here.

I'm not a vacuum tube guy; only have a general understanding.
But "cathode current was being pulled" doesn't make sense to me.   Here's my reasoning...
the cathode heating loop doesn't change; a given floating filament voltage over a constant filament resistance equals constant current. It provides a constant source of free electrons that a second voltage loop can push out of the filament (via FIL2 lead), across the vacuum, out of the grid/segment, back into the second voltage loop source.  Any electrons from the second source replace the free electrons produced via heating.

Without a resistor in the second loop, the current is determined by the voltage applied and the physical distance between filament and grid/segment (ie. tube physical characteristics).¹  Lowering the filament voltage produces less current in the filament (unchanged resistance), less free electrons, and therefore, less light at the segment phosphor.²  Same if lowering the voltage between filament and grid/segment (second loop); results in dimmer phosphor.²

1: "Plate Voltage-Plate Current Characteristic" pages 13,14,15
2: my own experiments.

[pqass]

The final tests without resistors was basically the same (I expected slightly higher current). Also, I set the power supply current limit to 20mA (although I don't know the accuracy of the over current protection), but, obviously the power supply never went into current limit.

The only anomalies were some SEG lines had either varying current (on the order of 0.2mA), or the G12 lines on a few segments began around 5-6mA, but kept climbing. Eventually it rose to approximately 8mA (hence where I got the maximum current) and stopped.

That's good news.  I didn't know you had a lab supply otherwise we could have done this without resistors from the beginning.

Given that you tried all combinations of GRIDn and SEG-m (one grid and one segment on, without any resistors) and none of them tripped the 20mA current limit, you should raise the DC voltage from +27V to +38V (keeping the same filament arrangement of 5VDC, 117mA).  You should see the maximum current on any one grid or segment increase a bit more but should never go over 25mA.

Then one final test is to attach all (or at least most) segments to +38VDC (using a bare wire weaving around each of the top display leads).  Then measure each GRIDn current (not necessary to measure the total SEG-m current as that would be n*SEG-m or  >>25mA).  Again, I wouldn't expect a larger individual grid current even though multiple segments are lit.

If you get this far, and at no point 25mA was exceeded, and only one segment was ever lit (non-withstanding the last multi-segment test), then I'd say the display is good.  The driver failure has to be due to some other factor; not an over current on any one output.  We're back to the Vpp voltage lingering longer than Vdd.

Shall I plan to measure for shorts between G and SEG lines too?

If you run a test where the filament is energized (5VDC, 117mA) and leave all segments unconnected, then iterate connecting only a GRIDn to +38VDC, you should NOT expect to see any segments lit.   If you do, then you can conclude a segment is shorted to a grid somewhere.

[bostonman]

Yes, I have a lab power supply, actually a few. Some are linear (the one I'm using for this test) and some are switching).

The one I used is three channels with one being dedicated to 5V. I'm uncertain, but believe none of my power supplies go over 32V unless I use the series mode; which means my filament voltage that is currently 6V (approx. 5V at the pin - after the series resistor) will now be 4V at the pin (assuming a linear drop) unless I hook up a second power supply.

Just to confirm the next (three) steps (not using any series resistors):

Raise the voltage to 38V on both SEG and GRID, and then go pin-to-pin again meaning, SEGa and G1-G12, SEGb and G1-G12, etc...?

Connect all SEG lines together with 38V, connect 38V to G1, measure current through G1, then connect G2, measure current, etc...

Disconnect all SEG lines (i.e. leave open) and connect G1-G12 individually looking to see if the display has anything lit

Obviously I will conduct these tests, but, before I sacrifice cutting fly leads and/or anything else meaning messing up my setup to do these tests, do you think any other tests will be necessary to prove the VFD is good?

Also, what plan can I prepare for should the VFD show it's fine (it almost seems at this point a safe bet is we can rule out the VFD - but I will still conduct the tests)?

Soldering another chip onto the display board is basically not an option as the pads don't look like they can handle another removal. The sockets I purchased should fit, however, I haven't looked to see if the IC sits above the socket when inserted (I will install a blown IC into the socket and see if it fits flush, below, or above).

If the IC sits above the socket, then using a socket isn't an option, but, also, the socket will probably touch the back of the VFD regardless which will not provide any heat dissipation for the IC (assuming the IC sits below the socket).

The only other option I can thnk of is to make a PCB, have it built at a board house, jumper wire from the four or five leads that come in from the ribbon cable, install the IC socket, sockets for the VFD (rather than solder), etc...

It's a costly experiment, and the result will be blowing IC chips.

The only thing I haven't checked is whether maybe the zener diode is open on the DC/DC. A reason must exist for the 38V to drop so slowly, but whether that can damage the VFD driver IC is another story.

[pqass]

Yes, I have a lab power supply, actually a few. Some are linear (the one I'm using for this test) and some are switching).

The one I used is three channels with one being dedicated to 5V. I'm uncertain, but believe none of my power supplies go over 32V unless I use the series mode; which means my filament voltage that is currently 6V (approx. 5V at the pin - after the series resistor) will now be 4V at the pin (assuming a linear drop) unless I hook up a second power supply.

Well, you could still use the 5V supply for the filament since we know the characteristics of that circuit; no chance of it going into current-limiting mode.  Just put a 1.5V AA cell in series with 5V supply to raise its max output (before the 16R series resistor) then adjust the voltage knob so that you maintain 5V (@117mA) between the FIL1, FIL2 pins as before.

Then you can use the other two supplies in series to dial in +38V (with 25mA current-limit) for the FIL2-grid/segment circuit.

Just to confirm the next (three) steps (not using any series resistors):

Raise the voltage to 38V on both SEG and GRID, and then go pin-to-pin again meaning, SEGa and G1-G12, SEGb and G1-G12, etc...?

Yes.
Confirming that no combination consumes more than 25mA in a grid or segment.

Connect all SEG lines together with 38V, connect 38V to G1, measure current through G1, then connect G2, measure current, etc...

Yes.
I figure you could use a bare stranded wire weaving around the top leads of the display (all segments but two), their elasticity keeping each segment touching the bare wire.  Again, confirming that no grid current exceeds the 25mA limit.

Disconnect all SEG lines (i.e. leave open) and connect G1-G12 individually looking to see if the display has anything lit

Yes.
Which confirms that there isn't a segment-grid short if nothing is lit.

Obviously I will conduct these tests, but, before I sacrifice cutting fly leads and/or anything else meaning messing up my setup to do these tests, do you think any other tests will be necessary to prove the VFD is good?

I'm not sure of your current setup but can you leave the other ends of the fly leads (if they're soldered to the display) open before you attempt the multi-segment lit test?

The filament circuit was tested at its rated spec. You've confirmed the current in each individual grid and segment. And you will confirm that the grid current doesn't exceed 25mA when most/all the segments are lit.  I'm not sure what else we can test the display for. 

Also, what plan can I prepare for should the VFD show it's fine (it almost seems at this point a safe bet is we can rule out the VFD - but I will still conduct the tests)?

So far the VFD is looking good. I'm not sure what comes next.
The DC-DC, the zeners and caps in the Vpp supply were chosen with the HV518 in mind.  The designers would have known the power up/down requirement of the driver chip.  I'd be second-guessing the designers if we changed the large 470uF C129 cap to something like 47uF or add a larger 10K resistor (vs relying on the existing 31K+zener) across that C129 cap to force a faster power down of Vpp.  It's worth trying though.  I hate that it costs you a chip every attempt.     A smaller C129 may result in a slight flicker in the display and a 10K (or lower value) across C129 will add to the max 38mA provided by the DC-DC which may starve the segment/grid needs.

Soldering another chip onto the display board is basically not an option as the pads don't look like they can handle another removal. The sockets I purchased should fit, however, I haven't looked to see if the IC sits above the socket when inserted (I will install a blown IC into the socket and see if it fits flush, below, or above).

If the IC sits above the socket, then using a socket isn't an option, but, also, the socket will probably touch the back of the VFD regardless which will not provide any heat dissipation for the IC (assuming the IC sits below the socket).

The only other option I can thnk of is to make a PCB, have it built at a board house, jumper wire from the four or five leads that come in from the ribbon cable, install the IC socket, sockets for the VFD (rather than solder), etc...

It's a costly experiment, and the result will be blowing IC chips.

I figured the PLCC socket plus VFD machined sockets would maintain the same distance between top of the driver chip and the underneath the VFD; just bringing the front of the VFD 1mm closer to the enclosure.  But if the traces are in that bad a shape, then yeah, make a small board to hold the PLCC socket then tack short wires to the VFD pins and J1 pins on the backside of the display PCB.

The only thing I haven't checked is whether maybe the zener diode is open on the DC/DC. A reason must exist for the 38V to drop so slowly, but whether that can damage the VFD driver IC is another story.

I thought you'd confirmed that the DC/DC was outputting the correct voltage earlier; that it produced +38VDC between J1 pin 21 and pin 23 (DCOM, GND)?  However, I don't recall anything about VR2 cathode; the center-tap of the filament winding.  It should be +6.2V above GND.  If VR2 was open then nothing would be lit since the center-tap of the filament would be at +38V. If VR2 was shorted there'd be a gradient or poorly lit/off segments.  You can confirm this now with the driver IC and display currently removed. I wouldn't try to fix the DC/DC and just replace it.


One more thing...   wondering if the HV518 could be counterfeit.  Are you buying them from a reputable source?