Vacuum Fluorescent Display Driver

[bostonman]

Well, the sockets arrived and I wasn't comfortable with the height. They appeared borderline (after cutting the positioning posts), and, inserting an IC to test whether the IC would extend even further would result in trying to yank it out without damaging it.

Rather than risking it, I soldered in the new chip even though I really wanted a socket.

After soldering, I measured pin-to-pin to check for solder shorts (didn't find any), and checked all solder connections with the board at an angle under the microscope.

An initial power up (without the display) showed the 38V was up around 44V (probably due to being unloaded), the G lines were clocking with nice pulses, SEG lines were clocking with nice square waves pulses, and nothing was getting warm at all.

Now the question is whether installing the display will result in the chip getting damaged again.

Keep in mind, none of the display pins were shorted to (chassis) ground, however, I'm also uncertain exactly how the display works to have an idea which pins could be susceptible to shorting or cause the IC to blow.

Any suggestions on what to look for before I solder the display? The time to solder it isn't necessarily an issue, it's more than I don't want to keep applying heat to all the pads. The IC pads have already taken a beating with hot air twice and soldering twice; not including having to reflow pins that didn't flow correctly to the pads.

[pqass]

Now the question is whether installing the display will result in the chip getting damaged again.

Keep in mind, none of the display pins were shorted to (chassis) ground, however, I'm also uncertain exactly how the display works to have an idea which pins could be susceptible to shorting or cause the IC to blow.

Any suggestions on what to look for before I solder the display? The time to solder it isn't necessarily an issue, it's more than I don't want to keep applying heat to all the pads. The IC pads have already taken a beating with hot air twice and soldering twice; not including having to reflow pins that didn't flow correctly to the pads.

With the display free, I'd confirm:
a. that there is some non-zero but non-open resistance between the filament pins, and
b. no continuity exists between a filament pin and any grid pin or segment pin or between grid and segment pins. 
   ie. MM in continuity mode, black lead on pin 1, scan with red lead from pin 2 onward.  Then, black on pin 2, scan with red lead from pin 3 onward. Etc.
       There should only be a small-ish resistance between 1 and 18 (filament pins).  All other pairs should be open.

If the above passes, you can quickly confirm that everything works by...
just spreading the display pins out a bit and place them into their PCB holes; holding the display down with an elastic or just your fingers.  There should be enough springiness to maintain a connection.

Then turning on the power for a few seconds to confirm that the segments light.

[bostonman]

Thanks for the feedback, I'll make sure to take all those measurements.

Then turning on the power for a few seconds to confirm that the segments light.

Maybe I'm wrong because it was hard to see if every segment in every digit lit correctly, but after installing the first IC (the one that blew), the display seemed to work fine. It was re-powering that seemed to cause things to go south. Either way, all I can do is take initial measurements based on your feedback and keep my fingers crossed.

Most likely I'll have updates by late weekend.

[floobydust]

We don't know why the new IC failed, so I'd grasp at a few straws.

While the VFD is out, I would rule out any shorts within the VFD. I don't have direct experience with this, nothing low enough resistance to blow the driver IC, so this could all be a goose chase.
I'm not sure an ohmmeter would find them. I'd use a higher voltage test, say 25V with 10k series resistor, probing between between each of the segments, and then the digits etc. Leave the filament out/off. Just looking for a mystery low resistance occurring above a few volts.

Inside the tube, spacings are very small. I have seen some contamination/ion migration between segments cause dim or fuzzy appearance due to aging. Another person thought their short was the fine PCB-like traces running along inside the back of the tube, where the bridge occurs. You can look at the VFD's backside to see what I am talking about.
So at least look for leakage (resistance) between VFD elements.

I power a VFD up on the bench, it's not too hard just need two power supplies and a few resistors. I rejuvenate them in two ways and have good success. The contamination/ion migration between grid and segments I have no idea what it is but am able to see and reverse it. If I can find my notes maybe spell it out for people to try. If the VFD's phosphor has aged or is burned, nothing can be done there but the VFD's emission can be helped I find.

Last straw to grasp, otherwise- as I mentioned a power cycle can make a '518 latch-up if the 5V rail comes up late or falls first during a power up/down per the datasheet. I think it's a separate 5V reg on the board. Maybe look at the 5V reg output cap and others, see if its value is OK.

[floobydust]

I found this pic in my collection and somebody seems to have added socket pins 
although the VFD pins are flat Kovar with some oddball plating to make solder stick.

[bostonman]

Wow, those pins look useful. Maybe I’ll try those around my fifth IC replacement

I’m about to go take some measurements and maybe solder the VFD back in afterwards.

[bostonman]

Update.

I measured all the pins on the VFD. Filament pins were about 9.3ohms, 2- to 36 were shorted (but the traces have them tied together), all other pins were high impedance.

Soldered the VFD, soldered a fly lead to the 38V, connected the scope to it, powered, and the display worked (all segments seemed to be correct). The 38V was a solid 38V.

Left it powered for about 90s, turned off power, left it unpowered for about 3-5s, turned on power, no display, and the 38V was about 12V. Several power recycles and still no display.

I’m so confused

Edit: wanted to add that the counter was free running meaning I didn't have anything connected to the BNCs, so it was displaying random numbers, however, if this matters, it didn't seem it wanted to display twelve-digits and was maybe ten instead. I noticed this with the original (bad) chip and thought maybe it auto selects the number of digits based on frequency or whatever.

[floobydust]

It seems to be eating HV518PJ's  I can hear your screams
Sounds a bit like what happened here: https://www.eevblog.com/forum/repair/hp-agilent-53181a-display-error/ shorted segments inside the VFD. I would PM the guy to confirm.
Originally I saw that you had SEG-C lit when it should have been off (when showing 2) and thought it was the '518. 

Is the VFD shorting after being warmed up a bit, or is the IC latching-up, or getting very hot (due to VFD short) and then it dies? That's why I advocated testing the VFD out of circuit. It could be a very weird Itron/Noritake failure mode.

edit: notice the tight spacings on the VFD internal bus from a similar (not the same part number) display

[pqass]

I measured all the pins on the VFD. Filament pins were about 9.3ohms,

Good, as expected.    I have a large format clock display with a 6R filament.

2- to 36 were shorted (but the traces have them tied together), all other pins were high impedance.

Shorted?  I don't understand.    Oh, you mean only the pin2-to-pin36 pair was shorted.  That may be acceptable since pin 2 and pin 17 aren't normally connected; ie. may serve an internal structural need.

Other than between pins 1 and 18, and ignoring pins 2 and 17, all other pin pairs should be open (have high resistance when the display is out-of-circuit).   If not, this is what's killing the driver chip.

You can test the display out of circuit with two DC power supplies (with common ground); +5V and +30V.   Set current limit of both supplies to 100mA max.
+5V ---47R filament resistor--- pin1,  then pin 18 to ground.
+30V ---10K grid resistor--- Gn pin  (where n=1 thru 12)
+30V ---10K segment resistor--- SEG_m pin  (where m=A thru Q, DP1, DP2, COMM)
All three must be connected for one segment of one digit to light. 
Iterate for all segments of all digits to find the short (the one that needs more current to light).
You shouldn't need less than 10K (to force higher current) for the segments to light but if you do find a short, may be intentionally driving 100mA through it (via 47R 330R vs 10K) may fix it. <== speculation on my part.
The display is likely FUBAR.

See this on how to test VFDs.

[bostonman]

Shorted?  I don't understand.    Oh, you mean only the pin2-to-pin36 pair was shorted.  That may be acceptable since pin 2 and pin 17 aren't normally connected; ie. may serve an internal structural need.

I used the word short (which they were), but it could have been a bit misleading. The follow up somewhat clarified they are suppose to be "shorted" since I followed the trace showing they are connected electrically.

I'm only thinking out loud and NOT rejecting testing the VFD and also elaborating a bit. Upon first powering after installing the new chip and the display worked, the DC/DC was not getting warm at all. It was actually cold. If the VFD had a short, rhetorically asking, why would the 518 continue working and the DC/DC staying cold? Also, this display worked fine (ignoring the bad segments) until I began installing the replacement 518's somewhat implying the VFD isn't shorted and something funky is going on with these new chips.

Upon repowering, it's not even the display shows something for a brief moment, it just doesn't come on at all, the DC/DC begins getting warm, and the +38V never goes beyond approx. 12V. I did notice the +38V takes time to discharge when I turned off the unit after the first power up and wonder if this is an issue.

The failure mode in my opinion indicates the IC is latching, but, if it's latching, why does it power fine the first time (unless the +38V taking long to discharge after the first power up/down is a hint).

+5V ---47R filament resistor--- pin1,  then pin 18 to ground.
+30V ---10K grid resistor--- Gn pin  (where n=1 thru 12)
+30V ---10K segment resistor--- SEG_m pin  (where m=A thru Q, DP1, DP2, COMM)

I'm scared of blowing up the VFD, so I want to confirm.

Connect +5V to to pin 1 with a 47ohm resistor in series. Pin 18 gets connected directly to ground?

+30V to Gn pin with a 10k resistor in series - obviously to pin 1 - 12 one at a time. The ground on the +30V channel also gets connected to pin 18 (or the ground banana jack on channel 1 that's being used for +5V)?

Now I take the same 10k that's connected to (let's just say G1) and also touch it to segment A (then B, C, etc...) at the same time? I know this is the answer because segment A can't light until a G line activates the proper digit, but I want to make sure. One wrong move with the VFD and I'm really in trouble.

On a side note, I may have only one IC left. So one more failure, and I'll be ordering more and/or repairing pads that eventually lift.

[PCB.Wiz]

Left it powered for about 90s, turned off power, left it unpowered for about 3-5s, turned on power, no display, and the 38V was about 12V. Several power recycles and still no display.

Bummer !
Could be turn off, or turn on, or turn on into a partially charged load, that kills things.
It survives initial ON, so a cold start appears to be OK ?

[PCB.Wiz]

I did notice the +38V takes time to discharge when I turned off the unit after the first power up and wonder if this is an issue.

Maybe it does not like re-powering into a charged load ?
You could add some overvoltage clamps and ab active discharge, that is off with LV rail present, put pulls HV down when LV is gone ?

On a side note, I may have only one IC left. So one more failure, and I'll be ordering more and/or repairing pads that eventually lift.

Maybe time to fit the socket ?

[bostonman]

I was about to update my previous post, but saw another reply.

Bummer !
Could be turn off, or turn on, or turn on into a partially charged load, that kills things.
It survives initial ON, so a cold start appears to be OK ?

Just to clarify: install new chip, power on, everything works, power off, power on, and display no longer works.

I'm re-reading the datasheet about the power down sequence. It states Vpp should be powered down, all inputs, and then Vdd. Now if the +38V (Vpp) is slowly discharging, maybe it remains high too long and the inputs along with Vdd have already turned off.

Maybe this is the issue and I need to make the +38V turn off quicker?

[pqass]

Shorted?  I don't understand.    Oh, you mean only the pin2-to-pin36 pair was shorted.  That may be acceptable since pin 2 and pin 17 aren't normally connected; ie. may serve an internal structural need.

I used the word short (which they were), but it could have been a bit misleading. The follow up somewhat clarified they are suppose to be "shorted" since I followed the trace showing they are connected electrically.

I see. I think you should ignore pins 2 and 17 for any testing purposes since the schematic has them unconnected.

I'm only thinking out loud and NOT rejecting testing the VFD and also elaborating a bit. Upon first powering after installing the new chip and the display worked, the DC/DC was not getting warm at all. It was actually cold. If the VFD had a short, rhetorically asking, why would the 518 continue working and the DC/DC staying cold? Also, this display worked fine (ignoring the bad segments) until I began installing the replacement 518's somewhat implying the VFD isn't shorted and something funky is going on with these new chips.

I agree that on first power up, given that the display is working, does seem to confirm that the filament nor the grid and segments aren't too much of a load on the driver chip.   However, testing the display out-of-circuit will confirm that there's no excessive draw. ie. that a 47R filament resistor, 10K segment and 10K grid resistors is enough to light a segment. And to iterate over all grids and segments to confirm there isn't a bad one in the bunch.

Upon repowering, it's not even the display shows something for a brief moment, it just doesn't come on at all, the DC/DC begins getting warm, and the +38V never goes beyond approx. 12V. I did notice the +38V takes time to discharge when I turned off the unit after the first power up and wonder if this is an issue.

The failure mode in my opinion indicates the IC is latching, but, if it's latching, why does it power fine the first time (unless the +38V taking long to discharge after the first power up/down is a hint).

All I could find in the HV518 datasheet (page 11, section 3.2) is that Vdd (+5V) should be powered before Vpp (+38V) and Vpp should be powered-down before Vdd.  Maybe by adding a 10K resistor between Vpp and ground will provide a minimal load (3.8mA) to drop that rail before the bulk caps drop the +5V rail.

+5V ---47R filament resistor--- pin1,  then pin 18 to ground.
+30V ---10K grid resistor--- Gn pin  (where n=1 thru 12)
+30V ---10K segment resistor--- SEG_m pin  (where m=A thru Q, DP1, DP2, COMM)

I'm scared of blowing up the VFD, so I want to confirm.

Connect +5V to to pin 1 with a 47ohm resistor in series. Pin 18 gets connected directly to ground?

Yes.  See attached diagram.

+30V to Gn pin with a 10k resistor in series - obviously to pin 1 - 12 one at a time. The ground on the +30V channel also gets connected to pin 18 (or the ground banana jack on channel 1 that's being used for +5V)?

Yes.  See attached diagram.

Now I take the same 10k that's connected to (let's just say G1) and also touch it to segment A (then B, C, etc...) at the same time? I know this is the answer because segment A can't light until a G line activates the proper digit, but I want to make sure. One wrong move with the VFD and I'm really in trouble.

A different 10K (from +30V) to the segment of choice.  See attached diagram.

Note, the grid and segment voltages don't have to be exact; just not more than +38V.  It should light with anything north of 20V.
Just make sure that the filament doesn't see more than 100mA (hence the 47R) and NEVER shows a red glow.  And if you do find a segment or grid that won't light with a 10K resistor in series, then lower it no less than 330R (ie. no more than allowing 100mA on any grid or segment).

The 47R and 10K resistors should limit the current fine but if you have constant current capable power supplies, set them no more than 100mA just in case of oopses.

On a side note, I may have only one IC left. So one more failure, and I'll be ordering more and/or repairing pads that eventually lift.

It sucks to find out the hard way.

[bostonman]

A different 10K (from +30V) to the segment of choice.  See attached diagram.

Rough day of yardwork and lack of sleep, I should have realized a different 10k resistor.

Thankfully I didn't use too much solder to reinstall the VFD (planned to reflow after confirming the unit was fixed), so unsoldering should be easier.

I fully agree that testing the VFD is important to ruling out at least one anomaly. If I were giving advice to someone, asking them to rule out other pieces in the puzzle would be my first approach. I'll place an amp meter in series to note the current for each segment.

As for the slow discharge of the +38V, it's not just slow, I mean (at 10V/div) the DC line crawls. I didn't measure the time, but I'd estimate a good ten-seconds and it was still >20V.

Before I remove the VFD, I'll put probe on the 5V and +38V (now approx. 12V) and check the timing during power down. I'm guessing the 5V will reach 0V long before, however, due to the chip being blown, the measurement will be skewed. It's unfortunate the datasheet doesn't specify what happens if the 'shut down' sequence isn't followed; maybe at some point tech support will help, but let me confirm the VFD first.

[bostonman]

Attached are two scope measurements - file names are named accordingly. Edit: note that Vcc was taken from pin 10 of the ribbon cable (main board end).

Remember the +38V is now being loaded by the damaged IC and measuring approx. +12V (for those who are new to this thread)

As I expected, the discharge rate on the +38V is quite fast since (I'm assuming) the chip is low impedance and sinking the voltage(s) quicker.

It's unfortunate I didn't measure the turn off rate of the +38V when the chip was working, but I remember it being quite slow. If the 5V turns off at the same rate regardless if the chip is good or bad, then a good possibility is the +38V turns off before thus going against the datasheet suggested turn-off sequence.

[bostonman]

Just wanted to clarify one thing: I still plan to test the VFD out of circuit. As I previously mentioned, before removing it, I wanted to measure the +38V discharge rate.

Hope nobody thought I was ignoring the suggestions of testing it out of circuit.

[floobydust]

What V/div? I see Ch. 2 at 1V/div for a noisy 5V, and Ch. 3 at 2V/div for 11V.

For testing a VFD, I'll use a 10Ω 5W resistor in series with a bench PSU to power the filament. This is just to limit inrush current and give coverage for an ill-behaved power supply doing some overshoot, some are terrible at power on/off or going in/out of CC mode if you use it, they surge. The tube is around 10Ω cold and 40Ω hot. To rejuvenate an old tube, I will slowly ramp up to 2x filament voltage to burn off oxides/contaminants, hold for 30 seconds and then let the VFD cool down for a few minutes. Do not bang or move the VFD at all when doing this.

The HV518PJ appears different (BIDFET) than the SN75518 (Bi-CMOS) output pin uses a BJT vs MOSFET to source current.
My money is on the tube having a short inside. Or a 5V rail cap is low value, it looks noisy.

[bostonman]

What V/div? I see Ch. 2 at 1V/div for a noisy 5V, and Ch. 3 at 2V/div for 11V.

Yes, 1V/div and 2V/div.

The HV518PJ appears different (BIDFET) than the SN75518 (Bi-CMOS) output pin uses a BJT vs MOSFET to source current.
My money is on the tube having a short inside. Or a 5V rail cap is low value, it looks noisy.

The difference in ICs wasn't something I analyzed, but my thought was the ICs are different and need a different turn-off sequence than what the unit was designed for.

Fast forwarding and assuming the VFD isn't shorted (I'll update once I get data), I've been thinking about the next step and how to somehow test the IC with the display rather than solder it in. The time and money for new ICs isn't an issue, it's more I'm worried about pads lifting due to excess heat. Also, since I don't have a clue why the ICs are failing, then it's a waste to continue soldering new ones in.

I probably know the answer to this, but is the 10ohm value essential? I have 15ohms, 50W and 20ohms, 20W I can put in parallel.

Most likely the 15ohms is fine (50W is overkill, but all I have in high wattage), but wanted to double check.

[pqass]

Given that the DC-AC (TDK CD-1848P) filament output spec is 5.2V@120mA, a 1W filament resistor would be overkill.
The higher the resistor value, the dimmer the display. If high enough, you won't see anything at all.
However, given 40R when operating (therefore, 5V/40R=125mA), adding 15R won't likely lower the current such that you won't see anything (5V/(40+15)R = 90mA).

[floobydust]

I mentioned the resistor I used 10Ω 5W because the PSU voltage has to be set higher and it's what I had it my junkbox. Just being lazy about the power rating including at rejuvenating overvoltaging it.
The 47Ω 2W you'd need almost 13V to get ~6V on the filament and a person might fret over that. 15Ω 1/2W I estimate 8V to get ~6V on the filament.
Also, I think 38V for HV is high because on the bench the VFD is not being multiplexed, so 25-30V is plenty bright for static drive. But the 10kΩ will limit current too.

I've had a couple weird effects happen with these VFD's- they don't behave like even a triode.
Even the filament glow along only half its length, ending at a digit grid with +HV on it. Segments can have an odd blue glow that goes away after a few seconds, with the anode left open. That glow never returns so I assume some contaminant between segments and something else, migrated. The segments were no longer fuzzy.

[floobydust]

On the power sequencing, where is the trace of the 38V rail?
The DC-DC converter is powered from SW+5V which is switched by Q12/C159, so it should fall fast. But C129 is huge 470uF 63V makes me wonder why.
The 5V rail for the '518, as you mentioned schematic shows U7 LM317 but it's probably a 7805? I'm not sure where the feed is from.
I'm just looking for what could scuttle the power supply sequencing.

[bostonman]

On the power sequencing, where is the trace of the 38V rail?

I assume you're referring to my two scope captures? Channel 2 is Vcc (the green) . It's set to 1V/div and channel 3 (purple) is the +38V (but it's only measuring +13V since the IC is in "failure mode"). The time per division is 20ms.

Not sure if you already knew all this, but wasn't sure which portions of the captures you were inquirin about.

Vcc was taken from pin 10 of the ribbon cable on the main board end (I'm assuming it's the same Vcc that feeds the HV518, but the VFD is connected at the moment so I can't be 100% certain). +38V was taken directly at the ribbon cable pin on the front panel end.

As previously mentioned, upon first installing the new IC and turning off power with the VFD installed, if I remember correctly, the +38V took a very long time to discharge. Unfortunately I wasn't measuring it with intent of checking the timing and/or comparing with the Vcc discharge rate.

Also, per my statement(s) of this slow discharge being a problem, the more I thought about it, upon first installing the new IC, I powered the unit without the VFD. The chip didn't blow thus possibly eliminating the power down sequence being an issue (unless it powers down differently without the VFD connected due to a different load). Maybe the VFD truly has a short, but still remain baffled why it would work fine until power is turned off and back on.

[floobydust]

Rail sequencing is complicated. I was interested in the rail sequencing with a good IC, not with an overloaded DC-DC converter. So we don't know what the 38V rail normally does. The 38V rail cap C129 is huge.

Where does the 5V for the HV518 come from? There is a DPAK linear regulator but what provides the input?
If Q12 is shorted then SW+5V could stay up too long keeping the 38V rail up after the 5V rail is down, on power down.

[bostonman]

I was interested in the rail sequencing with a good IC, not with an overloaded DC-DC converter. So we don't know what the 38V rail normally does.

You're absolutely correct and wish I had taken such a capture. The scope captures with the "blown" IC was for general reference.

If Q12 is shorted then SW+5V could stay up too long keeping the 38V rail up after the 5V rail is down, on power down.

Another good piece of advice. Keep in mind though, this unit worked fine with the original IC (except segments were in correct) and seems to work fine with a new IC providing I don't turn off power. Not to say something didn't get damaged during my repairing.

Where does the 5V for the HV518 come from? There is a DPAK linear regulator but what provides the input?
If Q12 is shorted then SW+5V could stay up too long keeping the 38V rail up after the 5V rail is down, on power down.

I'm not quite sure the schematic makes sense to me. The front panel schematic (page 4) shows 5V into a LM317 and 5V out; that's obviously wrong. Page 53, pin 10, of the ribbon cable, shows Vcc and I believe this is the Vcc to the HV518. If I remember correctly, 5V net name gets converted to Vcc somewhere. Maybe I'm wrong and it's SW+5V goes through Q12 and is named Vcc. Edit: also I see there is a "power up" reset on page 21.

Maybe I'm wrong, and due to being uncertain about some things, I try to add as much detail in my replies to reduce any confusion.

Tomorrow, or possibly Friday, I'll remove the VFD and perform tests. I'm thinking of soldering twelve momentary push button switches with 10k resistors on a perf board to reduce constantly fiddling with clips on the leads of the VFD. Also, it will reduce time because I can power one segment, push each of the switches, and see the same segment on each digit before moving onto the next segment.

Earlier I was thinking ahead and wondering how I could perform testing on the IC without soldering it back onto the board. I thought about soldering the IC socket to a perf board, running jumper wires from the ribbon cable for the data line and other essential signals, but then I'm stuck wiring all the segments and G lines manually.

The advantage is I can blow IC chips without the need to solder, still a $8 loss each time, but better than soldering.

The only other thought was to layout a PCB with sockets for the VFD pins (or hope the pins make a connection on unsoldered holes). I'd still need to layout a PCB, cost, and wait for it to arrive.