Vacuum Fluorescent Display Driver

[bostonman]

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.

It's a dedicated 5V without any adjustment. Maybe I'll setup a second power supply and tie the grounds together so I can keep the 6V and have the 38V for the G and SEG lines.

I hate that it costs you a chip every attempt.

I'm not in the habit of burning money, but it's in the name of engineering/science, it may benefit others who come across this post, and, most importantly, I can't let this thing get the best of me.

I'd need to review my previous posts, but believe I confirmed the 38V (in fact, I posted scope pictures showing the SEG and/or G lines - I'm calling them G lines because the schematic shows them as Gx) with the original IC. This whole issue of blowing the IC began after replacing the IC. The original issue (I know most of you probably know, but better to reiterate to be safe) was digits four and five (?) sometimes not displaying correctly, multiple periods (although maybe they were commas, so don't put much effort into thinking about that), and the lower right vertical segment on the 12th digit (most right) was always on.

The counter itself worked fine and believe I left it on overnight at least once to get it warm so I could calibrate it; or use it to calibrate a generator both referenced to a Cesium standard.

I'll plan to conduct the final tests on the VFD. The current setup is a bit wonky. I'm using (I don't recall the name of them - maybe header pin adapters) pin sockets with wires. For the most part, I have wires everywhere, a situation I hate working in because that's when accidents happen, but I keep things separated and don't rush. Ideally I'd like to finalize all the tests before dismantling the setup. Although I have no issues reassembling it, I'd rather not.

In any case, counterfeit chips is always a possibility. I placed one of the bad ICs in a socket and it sat slightly above the socket, however, the pins have solder on them, so it may not have been seated well. Regardless, I think the VFD would sit on top of the socket if I use a socket on the Display Board (this will have to be a bridge I'll cross when I get to it).

I have pictures of the new IC and the Digikey bag that I'll upload in a moment. Without deviating to dig through the notes and schematics, is the original IC on the schematic a 518? I know a 518 was purchased because it was a valid "replacement" for the part number listed either on the IC and/or the schematic (or maybe the BOM).

Maybe the issue is compatibility?

Update: also, just for reference, the IC socket I purchased is: 8444-21A1-RK-TP (these have tiny plastic alignment pins on the bottom that I didn't realize and cut off so they sit flush on the board).

 

[pqass]

Never mind what I said about the 1.5V AA cell on top of the 5V supply.  We know the filament circuit will be fine with 5V as it will draw 117mA so you should be able to just directly connect the fixed 5V supply across FIL1,FIL2 VFD pins (no 16R resistor in-series needed).  You can confirm with an ammeter but I'm sure it will be fine.  I don't think you have to worry about in-rush current.

If it came from DigiKey then I fully expect the part to be genuine but...
technically (the best kind of correct), the circuit is designed with the SN75518 in mind.  Looking at the specs for that, I see it has a max Vcc1 logic supply of 15V (not 5V) and max Vcc2 (what I've called Vpp) of 60V. There is no mention of a power up/down sequence for Vpp.  It has a PNP vs PMOS on the output.  I'm thinking it's a better part; and the HV518 doesn't quite work in this design.

So if you're not footprint constrained and go with a daughter-PCB, you can use the 40 pin DIP version of the SN75518 which are available on eBay (couldn't find the PLCC version).  Yes you risk a counterfeit, but it'll likely work and get you out of this jam. I bought one from eBay and it works fine.

[bostonman]

Believe I've completed all the tests.

When leaving all the SEG lines disconnected (i.e. open), I touched each G line individually with 37V DC and nothing turned on.

In order to tie all the SEG lines together, I used a strand of wire, however, it seems the connections weren't solid enough because I couldn't get all the segments lit. I rewound it three times and had intermittent results where segments would sometimes turn on and off. After I tried copper tape which appeared to work much better.

None of the currents went beyond around 0.5mA (without doing the math, I expected a bit higher since all twelve segments are on simultaneously).

The only anomaly I saw that was slightly concerning was the far right digit segments and part of 'u' and MHz was faintly lit.

I noticed as I ran my fingers around the G lines, I could get all twelve digits to light individually as I passed my finger across them. My guess is the G lines/segments are picking up capacitance, however, I could get the far right digit to turn off momentarily as I tinkered with the copper tape.

Maybe this is an actual problem, but due to being able to have the far right segments almost turn off, sometimes different segments in that digit would be off, then on, and sometimes fully off, I sense it's just due to the copper tape and/or weak connections.

[bostonman]

There is no mention of a power up/down sequence for Vpp.

I'd like to point out that when the new IC blew the first time, I removed it, soldered a new one, and, before I re-soldered the VFD, turned on/off the unit to measure whether the 38V was still 12V or back to 38V without the IC blowing.

It wasn't until I connected the VFD that the second IC blew (probably the same case as the first time).

The 38V was solid at around 40V (I think or maybe 42), the display worked, the DC/DC was cool, turned off the unit, waited about five-seconds, turned on the unit, no display, and the 38V was down to 12V again with the DC/DC getting warm.

Unless someone thinks a problem exists with the VFD due to my previous testing update, I think my next step is to measure around the DC/DC, the zener, etc...

If the conclusion is to make a daughter board to house a 40 pin DIP, I guess it will be lots of work, but the only solution. Maybe it would be a good idea to include the 44pin version too for possible future needs and/or conduct more tests that cause the IC to blow.

[pqass]

When leaving all the SEG lines disconnected (i.e. open), I touched each G line individually with 37V DC and nothing turned on.

Good. No shorts from grids to segments.

In order to tie all the SEG lines together, I used a strand of wire, however, it seems the connections weren't solid enough because I couldn't get all the segments lit. I rewound it three times and had intermittent results where segments would sometimes turn on and off. After I tried copper tape which appeared to work much better.

None of the currents went beyond around 0.5mA (without doing the math, I expected a bit higher since all twelve segments are on simultaneously).

The only anomaly I saw that was slightly concerning was the far right digit segments and part of 'u' and MHz was faintly lit.

Its okay that it was a bit flaky. This test was to compare the grid current where most segments were lit to the previous test where just one segment was lit.   So what was the current on the grid?  Was it >25mA?

[bostonman]

So what was the current on the grid?  Was it >25mA?

I only saw about 0.5mA on the SEG current and about the same on the G line current.

I was looking more at the digits to see if others lit and less on the current, but the current was very low the times I looked which didn't make sense. With a 10k series resistor the G lines were 1.5mA (doing a quick visual average) and the SEG lines were 0.15mA (also a quick visual average).

If all twelve segments were lit, and assuming 0.15mA per segment, that would be 1.8mA total; this would also be with a 10k resistor and this last test didn't have a series resistor.

Maybe since I used 5V for the filament versus 6V, this caused less current on SEG and G lines?

Update: due to lower filament voltage, the current was approx. 97mA.

[pqass]

I'd like to point out that when the new IC blew the first time, I removed it, soldered a new one, and, before I re-soldered the VFD, turned on/off the unit to measure whether the 38V was still 12V or back to 38V without the IC blowing.

It wasn't until I connected the VFD that the second IC blew (probably the same case as the first time).

The 38V was solid at around 40V (I think or maybe 42), the display worked, the DC/DC was cool, turned off the unit, waited about five-seconds, turned on the unit, no display, and the 38V was down to 12V again with the DC/DC getting warm.

I'd expect the DC-DC to be at its highest output voltage without a load attached given there's probably an inductor inside. But your experience does hint that the failure was output current related. But so far, the display tests clearly shows no grid or segment draws >25mA so the driver should not have failed for that reason alone.  Maybe the issue occurs when combined with the lingering Vpp after Vdd goes to 0V.

Unless someone thinks a problem exists with the VFD due to my previous testing update, I think my next step is to measure around the DC/DC, the zener, etc...

If either VR1 and VR2 zeners were open, you wouldn't have seen any display output at all.  If either were shorted, you would see some dim segments (that should be off), gradient, or slightly brighter segments (VR1 shorted) that would have drawn more grid/segment current (14% more).

But you can test this now while the VFD and driver have been removed.  The DC/DC output is rated for 38mA. Connect a 1K resistor (38V/0.038A=1K) between J1 pins 21 (+38V) and 23 (GND) and confirm that it can maintain the 38V.

If the conclusion is to make a daughter board to house a 40 pin DIP, I guess it will be lots of work, but the only solution. Maybe it would be a good idea to include the 44pin version too for possible future needs and/or conduct more tests that cause the IC to blow.

I don't think it would be a lot of work to create a functional solution.  I would just use a 40x7 hole perf-board, mount two 40 pin rows of machined sockets on the long edges, solder fly wires through the inside holes next to the sockets, coming out via the short edges of the perf-board, with wires sandwiched between the perf-board and the IC above.  With the perf-board centered and (double-sided) taped to the back of the display board and between the VFD pins, the 20 wires (or so) from each short side are tacked to the VFD pins and J1 pins.

For a better job, you could make a custom PCB that would be the same dimensions as the VFD pins.  You would then solder 10mm solid wire up from the VFD pins and select J1 pins into corresponding sockets from underside of the daughter PCB just like an Arduino shield.

[pqass]

I only saw about 0.5mA on the SEG current and about the same on the G line current.

I was looking more at the digits to see if others lit and less on the current, but the current was very low the times I looked which didn't make sense. With a 10k series resistor the G lines were 1.5mA (doing a quick visual average) and the SEG lines were 0.15mA (also a quick visual average).

If all twelve segments were lit, and assuming 0.15mA per segment, that would be 1.8mA total; this would also be with a 10k resistor and this last test didn't have a series resistor.

Maybe since I used 5V for the filament versus 6V, this caused less current on SEG and G lines?

Update: due to lower filament voltage, the current was approx. 97mA.

Yes, less filament current means less available electrons so less current in the FIL2-grid/segment circuit (for the same voltage).
20mA less (from 117mA to 97mA) filament current isn't much of a difference.  I'm not sure which is the correct current to run the filament with but it's in the ballpark.  What's important was the 6V+16R resistor source is probably the top-end and you never came close to the 25mA current limit of the driver output.

[bostonman]

I'll double check the current values again with all the SEG pins shorted before I dismantle my setup.

Also, test the DC/DC with a resistor load. Looks like about 1.5W (38mA ^2 x 1k) and luckily (according to my list of resistors) I have should have 200ohm 0.5W (five in series), or 500ohm 200W (two in series).

For a better job, you could make a custom PCB that would be the same dimensions as the VFD pins.  You would then solder 10mm solid wire up from the VFD pins and select J1 pins into corresponding sockets from underside of the daughter PCB just like an Arduino shield.

A PCB with pin spacing to match the VFD was something I gave thought to sometime ago and agree, I think it would be much neater. Obviously I'll have to dig deeper once the decision is made to take that route.

Not sure if I mentioned this, but the irony is that I didn't have any intension of working on this counter. Don't remember what I was working on, but dragged out this counter to use. When I was reminded of the digits being off, I thought maybe it was something simple like a loose connection/connector. Before long, everything else got tossed aside and this became the main project.

[bostonman]

I performed the measurements again. This time I had all the G pins open rather than connected to resistor leads on the perf board (I was using the resistor leads to clip onto - initially I was going through the 10k resistors to limit the current) and none of the segments were ghosting; so it must been due to capacitance or whatever that I saw previously (my previous post).

The highest current draw I saw was about 0.8mA on the G lines and 0.5mA on the SEG lines. Unfortunately the copper tape wasn’t making full contact all the time, but I think it was solid enough since I saw all segments lit at short intervals.

Probably safe to say at this point the VFD is good.

Next step is to measure the DC/DC with a resistor load and then I guess plan on a daughter board.

[bostonman]

I placed two 500ohm (200W) resistors in series (overkill on the wattage, but it was the only higher wattage resistors I had) with the 38V to ground.

Attached are scope captures with the file names named accordingly.

From what I could tell, the 38V held strong under a load. The 5V into U30 is a bit noisy, but probably normal).

Also note, on some of the scope captures such as VRx, the "zero" volt line is two divisions from the bottom, but, when the power is off, the voltage sits approximately 6v above (the "zero" volt line is somewhat hard to see on the black background along with the trigger level being next to it).

[pqass]

Yup, 38V is holding up at 38mA.

And U30 pin 1 has about a 42kHz ripple of 1/2Vp-p. They could have used a 470uF for C62.

The VR1 pin 1 never goes to zero because VR2 is off below 6V.  So C129 cap just keeps it there without a load. Maybe it could get close to zero while the VFD filament cools down. It probably cools faster than the load consumes so it likely breaks the power-down rule in the HV518.

The VR2 spikes to GND are also about 42kHz.

I'm curious what the waveform looks like between FIL1 and FIL2 (U30 pin 4 and pin 6) with a 50R in between (5V @ 100mA).
Attach probe channel 1 to U30 pin 4 and channel 2 to pin 6 with ground clips to GND.
Bonus if you can generate a 3rd, math channel where it's channel 1 + channel 2.

[bostonman]

Attached are two scope captures - files named accordingly.

Not sure if I've done a 'math' function before, and certainly not with this scope (Agilent 54831M), so I'm uncertain whether it's correct.

The phase seemed to move slightly because I kept seeing it jump, but the waveforms remained solid.

Just to elaborate on the file names: the 38V was still loaded with 1k (two 500ohms in series) to ground, FL1 went into a 50ohm resistor, the other side of the resistor was connected to FL2. Technically it was two 23ohm resistor boxes in series to reduce wattage (two 25ohm settings were about 53ohms, so I used 23) and also a current meter. The filament current was about 97mA and held constant.


Maybe due to the long le

[pqass]

The waveforms aren't smooth sine waves but rather (slightly distorted) square waves; the filament won't care.
As long as the FIL1 and FIL2 outputs (relative to GND) are inverted so when added average to 0V.
The yellow math channel shows this flat avg voltage but it's raised to about 6V by the VR2 zener.

I think the DC-DC (+38VDC and the +/-5V filament) outputs are fine as they can maintain their current spec. (38mA and 100mA, respectively).

[bostonman]

Time for a daughter board then for the other TH IC?

I’ll start mentally working on it and then probably have a PCB made rather than a leaf board.

[pqass]

Time for a daughter board then for the other TH IC?

You mean the SN75518?  Yeah.

[bostonman]

You mean the SN75518?  Yeah.

Yes, that one.

I will probably incorporate a socket for both the 518 and SN75518 this way I can (hopefully) insert the SN75518 and see it works fine, remove it, insert the 518, and see if it blows.

My PCB layout skills are basic, any suggestions on trace routing such as size or any "magic" designs I should incorporate to reduce noise?

[pqass]

From Reply#20: "Every ~976 us (~1024 Hz), there is a 32 bit word, with bits clocked at ~934 kHz."

I don't think the outputs changing 1024 times per second is going to generate much noise or pose a concern for board layout.
The SPI traces won't be very long; just up from J1 into the corners of the chip.

[bostonman]

So if you're not footprint constrained and go with a daughter-PCB, you can use the 40 pin DIP version of the SN75518 which are available on eBay (couldn't find the PLCC version).  Yes you risk a counterfeit, but it'll likely work and get you out of this jam. I bought one from eBay and it works fine.

SN75518FN is the part number on the schematic (I assume you just summarized the number and they are the same).

Any suggestions on which place to purchase from on eBay? I'm seeing $3.xx up to $2x.00, all from China and Hong Kong.

I assume it's a coin flip, but figured I'd ask.

Currently I have one 518 remaining, and three sockets. When I make the PCB (as mentioned), I'll incorporate a socket. I'll be able to see if the 7551s work, and, if the do, I'll perform one more test with the 518. I will blow, but it will be in the name of science.

but it'll likely work and get you out of this jam.

I could always just install the original chip and live with bad digits.

[pqass]

I bought just one "SN75518N" from a local (Canada) seller (now sold out). It looks new old stock; not a pull (splayed pins, no solder, no marks, not faded, clean). I think I paid CA$7 + CA$5 (shipping) about a year ago.  I've confirmed that I does work driving a VFD.  I don't have any experience with Chinese sellers as I don't want to deal with customs brokerage nor the wait.

Using the original flawed chip would be a bit of a let down considering your effort.

[bostonman]

Using the original flawed chip would be a bit of a let down considering your effort.

Very true, but felt maybe you and/or others felt a bit of responsibility for me going with a 518 and having these issues (assuming the 518 is the issue and not something else that developed into a failure). Obviously trying to fix this was all my doing, but, I was just stating at any point I could jump ship by returning my counter to the original (broken) state, so nothing would be lost.

I agree, waiting the several weeks for a chip from China isn't fun. Maybe tonight or tomorrow I'll do a deeper search, but my quick eBay search showed China and Hong Kong only.

A few years ago I purchased two used Agilent hybrid chips from China for my 54831 that had bad channels on 2 and 3, and so far they are working (although I fear leaving the scope on too long).

[pqass]

Your adventure has been surprising given that the HV518 works as a replacement in the HP 34401 (notice pix) lead me to believe it would work for your case. 

But just checking the 34401 schematics... and it shows that FIL1,FIL2 do not come from a DC-DC, but instead from an isolated transformer winding.  How about that!  That sure does implicate the power-down rules of the chip.

Thanks for taking one (or two) for the team.

[bostonman]

Thanks for taking one (or two) for the team.

Hahha this site and contributors have saved me more often than not. For the price of blowing up a few ICs and performing (and learning) to test the VFD, it's the least I can do.

Those scope captures I've been posting have come from my Agilent 54831M that, if not for the assistance of people on here, would have remained a two-channel out of calibrated scope. So I absolutely don't mind "taking one or two" for the team.

Technically I haven't taken anything for the team yet though. Maybe I blew up something in the unit causing the current issue and wasted the time of others. No telling unless I reinstall the original bad chip (which would be a waste) or designing the 5518 into the unit.

[bostonman]

The schematic lists an FN, but you purchased an N.

I get far less results when I search for SN75518FN than SN75518N (the one you purchased).

According to the datasheet, an 'FN' handles 1.7A and an 'N' handles 1.25W.

[pqass]

Look at the TI datasheet, first page.  N=40pin DIP package, FN=44pin PLCC package

Yes, I purchased the DIP for one of my custom projects (not as a drop in replacement) so I had flexibility.
That's all I found on eBay, anyway.