Author Topic: VFD filament power using LM4871 replacement for obsolete LM9022 (Read 13290 times)

leonababy · « **on:** June 17, 2018, 09:52:43 pm »

Newbie here.

I am curious if anyone has used or would comment on using one of these bridge amplifiers to drive a 6-digit VFD clock's filaments. The 9022 was essentially a re-badged 4871 as I understand it, marketed at a VFD driver. The 4871 was designed to be an audio amplifier - therefore, a good match for driving filaments (low impedance, AC). Datasheet of VFD driver-marketed version: https://datasheet.octopart.com/LM9022M/NOPB-National-Semiconductor-datasheet-11896945.pdf

The 9022 datasheet includes a sample application in which the chip self-oscillates and creates a square wave from 0.15V to 4.85V with a single 5V supply, but the supply can go down to 2V. For all the VFD's I am looking at, I need about 2V rms so this looks supremely applicable - with the possible exception that for a 5V supply I may have to drop a half volt with a series resistor.

I tripped across the 9022 while researching h bridge motor controllers for the same purpose. I was concerned about using an h bridge, because I don't fully understand it and was not sure it some of the warnings about momentary simultaneous "on" states of various elements applies for driving VFD filaments. I saw one person use an h bridge made of 4 BJTs, with no protection, but I couldn't see his timing since he drove the two sides with a uC. I am not using a uC, so I am restricted to logic and transistors. Another used an LV8405V h bridge chip, but he didn't show his schematic and it is designed for a motor. I was afraid of trying to use it since I don't understand it.

I am preparing to wind some toroids as a pulse transformer with a center-tapped secondary, driven by a 555 square wave, and brought above ground with a voltage divider, and I am thinking I missed the boat. The one-chip, no toroid solution seems cleaner and simpler. Looking online, I don't see any solution that seems the most widely adapted.

I realize I'm all over the map here. Trying to evaluate options. My main question is, has anyone used the 9022/4871 or do you see a downside. If you have any successes with the other options I mentioned I'd like that too. Again, I am not using a uC. Any comments are welcome. Thanks in advance. Don

Zero999 · « **Reply #1 on:** June 17, 2018, 10:47:01 pm »

The BJT h-bridge circuit does work and it's possible to build one which self-oscillates: no MCU or 555 timer required. There will be some shoot-through but that depends on how hard the transistors are driven. It can also be limited with an inductor in series with the positive supply. Another option is to replace Q3 and Q4 with emitter followers, which would eliminate it altogether, but would increase the voltage losses in the circuit.

https://www.eevblog.com/forum/beginners/mini-h-bridge-for-driving-vfd-filament-very-warm-to-the-touch/msg567741/#msg567741

leonababy · « **Reply #2 on:** June 18, 2018, 02:24:51 am »

Thank you. I had read this thread. Yes - shoot through was the term. In an excellent introduction to h bridges I saw on the net, the author cautioned about shoot through generally speaking in h bridge design. Perhaps with a VFD filament supply it is not as critical. I mentioned that I may have to drop some voltage - back to back diodes would provide a 0.6V drop, wouldn't they? In the pulse transformer case, I can adjust the number of turns.

I appreciate this alternative. I'm hoping to hear about some others as well, and if someone has used the LM9022 or LM4871. I am planning to breadboard the pulse transformer and 4871 at this point, and will consider picking up the transistors for your solution to give that a go.

Rolo · « **Reply #3 on:** June 18, 2018, 08:05:02 am »

I use the LM4871 in my vfd psu. You can find more information here: https://www.eevblog.com/forum/projects/showing-my-vfd-psu/.

leonababy · « **Reply #4 on:** June 18, 2018, 05:52:23 pm »

Quote from: Rolo on June 18, 2018, 08:05:02 am

I use the LM4871 in my vfd psu. You can find more information here: https://www.eevblog.com/forum/projects/showing-my-vfd-psu/.

Thank you Rolo - this looks like a great example of what I am seeking. I looked only quickly, but can you please explain the reason for the 4.7V zener? I assumed that the 0.xx to 4.xxV swing of the output straight to the filament would be sufficient, centered around 2.xxV. I think I am missing something.

Rolo · « **Reply #5 on:** June 18, 2018, 07:45:04 pm »

The zener provides a Filament Bias Voltage (Ek). The theory behind this can be found in this article : https://www.noritake-elec.com/technology/general-technical-information/vfd-operation
I can advise to read this guide, I learned a lot from it, see chapter 6.2 for the filament bias voltage. I choose 4.7V, this is about the maximum voltage swing the LM4871 can provide. If you have a lower filament voltage you can lower the zener value. You have to take this voltage into account when calculating the HV. The HV has to be "lifted" about the voltage of the zener.
You can measure the effective HV between the zener kathode and +HV.
Tip : do calculate the dissipation of the two resistors in the output of the LM4871. They can get hot if you use the wrong value.

leonababy · « **Reply #6 on:** June 18, 2018, 08:03:43 pm »

The Noritake application note is an excellent guide - I have it bookmarked and it helps to keep reviewing it. By increasing Ek I suppose you are increasing the degree to which the non-illuminated segments are off - making them more negative relative to the grid. I also appreciated in your thread the tips on selecting the output resistors.

I saw where someone had an issue (noisy sine wave instead of a square wave) breadboarding the 4871 to use in this configuration, but it worked fine once put on a PCB. These reports and tips are very helpful - I've been out of electronics for many years. Designed synchronous circuitry and classic uC hardware in the 80's. Thank you again, Rolo.

Rolo · « **Reply #7 on:** June 18, 2018, 08:17:46 pm »

I have build a circuit using the LM4871 on breadboard first, just to see the output wave, power and frequency. It worked well. Nice clean waveforms, no overshoot or ringing. Once I was sure it was going to work I designed the PCB.
I also expirimented with a 555 and a toroid, it worked but i gave nasty spikes on the Vcc line and adding a second winding on the toroid by hand was not practical. I wanted to use parts that are common and available.

leonababy · « **Reply #8 on:** June 18, 2018, 08:28:44 pm »

More very helpful information - thank you yet again. I'll be more careful when I prototype it than the person who had a problem - he had long wire lengths in several places from pics he posted. He may have prematurely abandoned it in haste.

Ian.M · « **Reply #9 on:** June 18, 2018, 11:11:39 pm »

The LM4871 will work down to 2V. If you are building a general purpose VFD PSU for experimenting with, it may be worth powering it from an adjustable LDO regulator so you can 'dial in' the desired filament voltage.

leonababy · « **Reply #10 on:** June 19, 2018, 01:39:58 am »

Excellent advice Ian. I am hoping to build with at least three different displays. DG10 on the small side, a Futaba 4-digit display in the 5LT series similar to the one in the GC-1107 on the large end, and DT1704 in between. I don't believe I will need to vary the filament voltage greatly, but making it adjustable from the start may be wise. Rolo's solution uses series resistors to do that. I built a clock using DG8 in the 70's and was dissipating almost 1 watt dropping the filament voltage. The GC-1107 as designed dropped 10.5 volts for the filament, generating about 1/2 watt in heat.

Rolo · « **Reply #11 on:** June 20, 2018, 03:35:29 pm »

That could well be my V3 design! Making the filament and HV adjustable on the board. Did some measurements on the LM4871, that gives an idea on what you can expect with different supply voltages, this is without the two series resistors.
The enable function for the filament can be tricky when using LDO before the LM4871, the input must not exceed the Vcc of the chip.
Shifting this function to a LDO with enable input or maybe use an optocoupler as level shifter here, have to give it some thoughts.
What would be a usable range for the HV Supply? 15 to 35V?

leonababy · « **Reply #12 on:** June 20, 2018, 03:55:22 pm »

I don't know if I will use a disable function for filaments, but the issue you mention sounds important to keep in mind. I suppose allowing for that option in the future makes sense.

Question - I do not understand the Vrms values. They look more like peak positive output voltages (?). Also, to clarify, is "Vin" Vcc of the 4871?

Rolo · « **Reply #13 on:** June 20, 2018, 04:51:28 pm »

Yes, Vin is the supply voltage on the LM4871. Rload stands for the filament resistance. The filament Vrms is the AC voltage on the filament outputs, measured parallel on the load resistor.

Having a switchable filament and HV makes the PSU more universal I think. I'm working on a clock that has a little PIR sensor, if nobody is arround for say one hour the filament switches of, this increases the life of the display.
Switching the HV off can be usefull when prototyping (multiplexing stops when loading new code, one grid lits up very bright) or if you have a VFD with build in controller where the HV must be turned on after Vcc. I do think the default have to be off for both in my next design.

Ian.M · « **Reply #14 on:** June 20, 2018, 05:40:46 pm »

For the shutdown pin, a Schottky diode and a pullup resistor to LDO output could do the job, or you could use a pullup resistor and a N-MOS or NPN transistor.

Some Russian single digit VFDs are specced at a much higher than usual Vak. How high can the boost chip you are using go?

Also, what about adjustable cathode bias? Put a 100K preset + a 100K resistor in series with its top end between Gnd and H.T, with the wiper drivng the base of a PNP Darlington, collector to Gnd in place of that Zener in the Noritake A.N.

Rolo · « **Reply #15 on:** June 20, 2018, 05:51:27 pm »

The LM2733 can boost up to 40V. Also found the TPS 73201 DBVT, a LDO with enable pin, it's high active, all above 0.7V is considered "HIGH" so it's compatible with 3.3 to 5V logic. It has very low drop voltage and no need for big caps. I like your idea about adjusing the cathode bias, can you make a drawing of this section?
Looks like a new prototype is on it's way.

Ian.M · « **Reply #16 on:** June 20, 2018, 07:25:07 pm »

In that case, 12 to 37V would be good - that leaves you with better than 30V Vak with the cathode bias at 7V, which should be enough to get good cut-off even with the max 4.5V (+/- 2.25V) filament drive. On the low end, with the cathode bias pot all the way up, Vak will only be about 5V - useful for those who want to experiment how low they can go and still get a VFD to illuminate.

I'd suggest presets on the board with unpoplulated footprints for fixed resistors in parallel + a footprint bridged by a thin track connecting the wiper, so once the required voltage asre finalised, the user can fit fixed resistors for the desired divider ratio, and either disconnect the preset, or can limit the preset's adjustment range, by putting a relatively large fixed resistor in series with its wiper.

I more or less described the whole variable cathode bias circuit - the only thing I didn't mention was the emitter goes to the C.T of the two resistors instead of the original Zener cathode. Circuit operation: the 100K preset + 100K resistor above it form a potential divider providing a bias voltage between 0V and 1/2 HT, while drawing under 0.2mA from the HT rail, and the PNP Darlington decreases the impedace of the bias voltage by its h_FE, but also adds two Vbe drops to the bias voltage. Assuming a minimum h_FE of 1000, that means the equivalent impedance of the bias circuit, as seen at the Darlington emitter will always be under 100R and typically will be half that or less.

You may also wish to consider doing a Microchip (Supertex) HV5812 breakout board. Its a dumb SPI VFD anode & grid driver with blanking and 20 outputs that can do up to 80V HT and source up to 25mA, is still in production and is reasonably affordable at around $2 each. Do four daisy-chained boards in a strip and let the user cut them up and populate the SPI headers according to the number of outputs they need - one HV5812 will handle most small VFDs, up to 7 segments, 10 digits + commas annunciators etc, two will handle virtually all starburst displays , and three will handle up to 24 characters of 5x7 dot matrix + cursors. You still need a MCU to handle the multiplexing, but as its SPI clock is up to 5MHz, there's no problem running it at fairly high refresh rates.

Rolo · « **Reply #17 on:** June 21, 2018, 05:27:39 pm »

Is this the bias circuit you had in mind?

I'm going to put that on a breadboard and do some measements, but I do not have darlingtons laying arround. I will buy some in the next parts order, and some potentiometer values. To be cont.

Als found this :

Could be usefull.

Ian.M · « **Reply #18 on:** June 21, 2018, 07:31:18 pm »

Yes, what you've drawn up as the first circuit is what I had in mind. You can prototype it with two discrete PNPs wired as a Darlington pair.

I don't like the second circuit much - its basically a boosted Vbe multiplier so has poor thermal stability, and it needs at least 10uA per volt fed to it by a pullup to the H.T. rail to bias it.

Another option would be a TL431 shunt regulator, with a JFET (+ source resistor) wired as a constant current pullup to feed it about 1.5mA from the H.T rail.

Rolo · « **Reply #19 on:** June 22, 2018, 01:03:20 pm »

I have done some drawing, this is what it could be. Note : this design is not build and tested,it only exists on paper. There are some issues with part selection. the FET has to be >40V, the choice is not great in this range, I have found the 2SK117, i's 50V, but only comes in TO-92 package. The pots wil also be trough hole, the 250K is not available in SMD. Using more trough hole parts leads to a bigger board, if this one get's it to the PCB desing. The filament enable has moved to the LDO regulator, the HV enable is the same as the fixed design. When left open the HV is ON and the Filament is OFF, maybe I wil change that later.I will buy some parts an do some prototyping on breadboard first.

Ian.M · « **Reply #20 on:** June 22, 2018, 01:45:06 pm »

You need to make C10 10uF as there is a risk of instability if a TL431 is shunted by 1uF.

The 2SK117 is unsuitable unless you select on test for Vgs_off due to that parameter's wide range (0.2V - 1.5V). For a one-off, breadboad the 2SK117 with a preset, adjust to 1.5mA and fit that FET + the nearest preferred value fixed resistor. For limited volume production test and bin the FETs by Vgs_off (test jig: gate grounded, drain fed by 9V battery, measure Vgs with a 10Meg input impedance DMM) so that each bin corresponds to one resistor value.

Alternatively, use a depletion mode MOSFET like BSS139 in the same circuit. Its gate threshold voltage is much better defined + it comes pre-binned with each reel having a code letter specifying a 0.2V threshold voltage range.

IMHO both enables should have the same polarity so they can be (optionally) strapped together and driven with one MCU pin. You should probably add a MOSFET to invert one of them.

Rolo · « **Reply #21 on:** June 22, 2018, 02:18:18 pm »

Thanks for the feedback. You are right on the enable issue, they should have the same behavior. Still not sure on what's best, default (pins left open) is both on or off. For experiments the first option is the best, no need for wiring the enable pins, the PSU just works when 5V is applied. When the PSU is used in a more permanent solution one can always change this behavior on the controller side.
I'm leaning too this option now. But first build a prototype. Just blew my LM4871 prototype, forgot to change the leads to another output channel of my bench PSU

So now waiting for parts....

Ian.M · « **Reply #22 on:** June 22, 2018, 02:45:31 pm »

Default (pins open) should be both off, so its off while the controlling MCU is in reset up to the point it initialises its I/O pins. However, if you make logic '0' = On, its very easy to provide two extra Gnd pins on the header next to the enable pins so a pair of jumpers can be added for always on applications.

Rolo · « **Reply #23 on:** June 22, 2018, 06:06:04 pm »

Do we need a current source in the supply line of the TL431? Would this work :

This keeps the anode voltage below the max 36V, the current wil vary with different HV values but that's no issue for the reference or the HV supply, it wil be in the 1.7 to 7 mA range, it will draw significant more current on the 5V switcher input side. It that the reason for having the current source there? Like to learn the theory behind this choice.

Ian.M · « **Reply #24 on:** June 22, 2018, 06:49:24 pm »

You don't need the Zener - The TL431 is a shunt regulator, so as long its fed through a resistor and its max current and dissipation limits are respected, it doesn't have a max input voltage, just a max output voltage.

Yes, the FET's just for efficiency. At 37V out, that 7mA is 5.5mA more than the TL431 needs to guarantee regulation. Thats over 200mW, which depending on boost converter efficiency is approximately an extra 50mA in at 5V.

Did you try the PNP Darlington circuit?

floobydust · « **Reply #25 on:** June 22, 2018, 06:58:02 pm »

What are you guys doing biasing the filament +ve?

You bias the filament at a negative voltage wrt to the anode supply, to lessen the segment driver's voltage requirement.

Rolo · « **Reply #26 on:** June 22, 2018, 07:15:01 pm »

We are biasing the (virtual) center tap of the filament. By doing this the AC filament voltage shifts up in the real GND-HV range.
This theory comes from this guide https://www.noritake-elec.com/technology/general-technical-information/vfd-operation

No I did not try the darlington circuit, can't find any PNP transistors (I know...everyone should have some). Waiting for some parts, then I will get the breadboard out.

Zero999 · « **Reply #27 on:** June 24, 2018, 01:18:00 pm »

Quote from: leonababy on June 18, 2018, 02:24:51 am

Thank you. I had read this thread. Yes - shoot through was the term. In an excellent introduction to h bridges I saw on the net, the author cautioned about shoot through generally speaking in h bridge design. Perhaps with a VFD filament supply it is not as critical. I mentioned that I may have to drop some voltage - back to back diodes would provide a 0.6V drop, wouldn't they? In the pulse transformer case, I can adjust the number of turns.

I appreciate this alternative. I'm hoping to hear about some others as well, and if someone has used the LM9022 or LM4871. I am planning to breadboard the pulse transformer and 4871 at this point, and will consider picking up the transistors for your solution to give that a go.

Shoot-through occurs when either two of the left and right transistors is on simultaneously. It doesn't happen for long and is worse at higher frequencies.

This one doesn't suffer from shoot-through and only uses NPN transistors. The voltage loss is high, with the output being 3.7V, at 5V in. the Schottky diodes can be replaced with ordinary silicon diodes, but the output voltage will drop further.

Rolo · « **Reply #28 on:** June 24, 2018, 03:19:04 pm »

I did some experiments with the special H bridge timer functions on a STM32 series controller, you can specify the "dead time", this is the time the two sections (A and B) are both off. It worked, choosing a longer dead time's even controlled the total output power of the filament. So you have a software controlled filament driver. I have added the AC coupler and virtual center tap in this schematic, but you get the idea. By using fets's the efficiency of this circuit is high, but the down side is you have to have the dead time. Use a scope to observe the behavior of the brigde, not having the timing right can lead to strange effect and high power consumption.
The selection of the fet's are just what I had laying arround in my parts drawer. Other types could do the job also.

Zero999 · « **Reply #29 on:** June 24, 2018, 07:16:30 pm »

Quote from: Rolo on June 24, 2018, 03:19:04 pm

I did some experiments with the special H bridge timer functions on a STM32 series controller, you can specify the "dead time", this is the time the two sections (A and B) are both off. It worked, choosing a longer dead time's even controlled the total output power of the filament. So you have a software controlled filament driver. I have added the AC coupler and virtual center tap in this schematic, but you get the idea. By using fets's the efficiency of this circuit is high, but the down side is you have to have the dead time. Use a scope to observe the behavior of the brigde, not having the timing right can lead to strange effect and high power consumption.
The selection of the fet's are just what I had laying arround in my parts drawer. Other types could do the job also.

The downside to that circuit is it has a a lot of shoot-through and large current spikes when both transistors turn on. Fortunately the BSS138 and BSS84 have a fairly high on resistance: 6Ω and 8Ω respectively, at 5V, which will limit the spike current to just under 360mA.

What does D1, C2, R2 and R3 do except waste power?

Rolo · « **Reply #30 on:** June 24, 2018, 07:33:47 pm »

Quote

What does D1, C2, R2 and R3 do except waste power?

They provide the Filament Bias Voltage (Ek), discussed earlier in this topic. You are right about the shoot trough and current spikes. That's why I choose the LM4781, this gives a much smoother output and is self oscillating, this gives a wider choice of controllers.

Ian.M · « **Reply #31 on:** June 24, 2018, 09:05:46 pm »

Of course as 5V of cathode bias is suitable for many VFDs, one can often get away without D1 and C2 and simply return the cathode center tap resistors to the +5V rail, saving the power wasted in D1.

Zero999 · « **Reply #32 on:** June 25, 2018, 09:28:17 am »

How about a transformer and suitable driver IC, such as the SN6501?
http://www.ti.com/lit/ds/symlink/sn6501.pdf

The cathode can be biased with a single resistor, saving the power wasted in a potential divider. It's also possible to use multiple windings to get other voltages.

Rolo · « **Reply #33 on:** July 04, 2018, 07:22:42 pm »

Update on this project, I stopped working on the "universal" vfd psu for a while, want to do more programming now using my original vdf psu.
https://www.eevblog.com/forum/projects/showing-my-vfd-psu/

leonababy · « **Reply #34 on:** July 08, 2018, 02:44:00 pm »

Quote from: Rolo on June 20, 2018, 04:51:28 pm

Yes, Vin is the supply voltage on the LM4871. Rload stands for the filament resistance. The filament Vrms is the AC voltage on the filament outputs, measured parallel on the load resistor.

{snip}

I apologize for not being back earlier. I tested my circuit around this time, and discovered while making measurements that your readings are twice what I expected because I didn't realize a bridged output effectively doubles the output. Because I only was expecting about 2.3Vrms, I simply drive from one (the positive) output to ground, with a series resistance. The power wasted in the series resistance is less than 1/2 of what the elements use (I am running about 1.3 to 1.4Vrms in the VFDs I have tested so far). The LM4871 itself uses about 450mW to drive 6 elements. I am extremely happy with the result, and the overall power consumption of the design is great for my purposes. I see this topic took legs and am interested in the other replies and designs. You guys are on a whole other level. Thank you for the bandwidth - much appreciated.


EEVblog Main Site	EEVblog on Youtube	EEVblog on Twitter	EEVblog on Facebook	EEVblog on Odysee

Author Topic: VFD filament power using LM4871 replacement for obsolete LM9022 (Read 13290 times)

Share me