mini h-bridge for driving VFD filament very warm to the touch

[tiltit]

Hi all,

I have been having fun with a small calculator-like VFD display, I thought I should do things right and put together a small h-bridge the drive the filament. It works but gets in my opinion overly warm. To me transistor datasheets are still very hard to read and I am not quite sure how to choose a base resistor to simply use the transistor as a switch.
It would be greatly appreciated if someone could look and the schematic and let me know how to improve the circuit.

Note that on the schematic below +5V is in fact +3.3V, and so is the input.

Thanks all.

[Andy Watson]

You will need to drive the voltage at the base(s) to within 0.6 of the emitter(s) in order to turn the transistor off. At present both transistors in the same leg are ON simultaneously. Note that the base-emitter junction behaves like a diode - when it is biased on, it will only allow the forward voltage drop of a diode (0.6-0.7V) to occur between base and emitter. The transistors in the left-hand leg will never turn off. Only the lower, NPN transistor in the right-hand leg will turn off.

[tiltit]

Now I am really confused 
I am currently probing R6, on the right side I do see my signal but on the left side I see a constant 2.5 Volts. But the circuit as a hole seems to be working. I am lost as to what is happening.

[Andy Watson]

Consider Q2, what is the lowest voltage you expect to see on its base?
Then consider an input voltage of 0V. The resistor from the base of Q2 (560R) and the input resistor (470R) form a potential divider. It should be apparent why Q3 will never turn off.

Only Q5 can be turned off - this is probably the reason that the heater appears to be working.

[tiltit]

I think I am starting to get it.

With an input voltage of 0V the voltage between R4 and R1 should be:
(3.3 - 0.6) / 1030 * 470
Witch equals to ~1.2V

Am I correct?

So the solution would be to decrease the value of R4.

[tiltit]

Or I can I tie the bases of Q2 and Q3 together?

[amyk]

VFD filaments should be warm...

Remember that BJTs will pass current through the base, so you'll need to take into account these currents when you're designing them to switch on and off. This is very different from the behaviour of MOSFETs.

So the solution would be to decrease the value of R4.

No, that would just increase the voltage on the base of Q3. Get rid of R1 to fix that side of the bridge - that way you can pull the base of Q3 all the way to 0V.

Also this is one of those instances where running a simulation can give you a good idea of whether your circuit behaves as you'd expect it to.

[tiltit]

Ok, so I will remove R1 and R3 and replace Q1 with a pnp transistor. Maybe that will work. I will also try to find some sofware to do simulations. Hopefully something that easy to use and works on Linux.

As a side note does anyone know of a sot-23 jellybean logic level Mosfet?

[tiltit]

So I have revisited the schematic, which I am going to put together tonight. Does anyone see anything obviously wrong with it?

[tom66]

Shouldn't the emitters be connected to a common point on the H-bridge? Been a while since I've used this, but it looks like a class AB amplifier driven with a square wave.

Better to convert the PNP inverter to an NPN type (more common) or draw it so the Vcc is at the top of the device, at least...

If you have the available Vcc, it would be easier to make an ac-coupled half bridge or two-capacitor mid supply half bridge design.

[Zero999]

It still has the same problem as before. All the transistors will turn on simultaneously.

You could try swapping the NPNs and PNPs so they're emitter followers but then you lose 1.2V to 2V, depending on the current draw.

[tiltit]

So I will try and emitter follower configuration. I can see now why they would be all on. I think that I keep forgetting about the current flowing from the emitter to the base.

I have also tried before to use ac-coupling to drive the filament using two electrolytic capacitors back to back. But I got such a voltage drop that the digits on the display where barely visible. The filament should draw about 100mA, If I measure it's resistance with my multimeter I read ~50 Ohm.

[T3sl4co1l]

When you do that, you MUST have base-emitter resistors, and you must have the pull-up (R1 and R3 in the first schematic) much smaller than the series base resistors drawing from it.  Or, you need some sort of rail-to-rail buffer or logic gate to supply 0V and +5V to the driving node (the node between R4-R5, and the node between R6-R7).

Here's another way to do it,



Assuming you need full wave operation, connect your input to A and use an inverter to drive B.  Typical values might be Rb = 100 ohms, Re = 220 ohms (for 5V CMOS logic levels at A, B) and supply anywhere from 5V to 30V (one advantage of this drive method is, it is independent of supply voltage).

Tim

[Zero999]

One way of doing this is to use a complementary common base buffer to switch the transistors.

Note if you do use this for a motor, you'll need to add in some protection diodes but they're not needed for a resistive filament.

Does it actually need to be AC? It's just a heater so perhaps it'll work on DC.

[tiltit]

Thanks,  I think I will try out those schematics, they are not as simple as I would of hoped, but then I wouldn't learn anything.
The heater does work on DC but there is a noticeable gradient in the luminosity of the digits.
My end goal with this project is to have a small and cute SMD only USB powered desk clock that also has HID-Raw interface for setting the clock and displaying information like the temperature outside. I hope to learn more about circuit design with this project and that is why I do not want to take shortcuts like powering the filament with DC and call it a day.

[Zero999]

I've had another idea. Since the frequency isn't important. It's possible to use an astable multivibrator to drive it. Two additional transistors can be added to the typical astable circuit to form a self oscillating h-bridge.
http://en.wikipedia.org/wiki/Multivibrator#Operation

[tiltit]

Thanks, I just tried that last circuit on a breadboard and it works rather well.
I replaced the 1uf caps for 0.1uf ones and the frequency of the circuit is ~6KHz, I hope that the heater doesn't mind. I also don't know how well the circuit will work with the bc847 and bc857.
The good thing about this circuit is that I wont have to worry about toggling a GPIO pin anymore and that it only has 4 transistors. It just works.

[c4757p]

Does it actually need to be AC? It's just a heater so perhaps it'll work on DC.

The filament also acts as the cathode - DC on the heater causes a voltage gradient across it, which results in a quite noticeable brightness gradient.

[Zero999]

Thanks, I just tried that last circuit on a breadboard and it works rather well.
I replaced the 1uf caps for 0.1uf ones and the frequency of the circuit is ~6KHz, I hope that the heater doesn't mind. I also don't know how well the circuit will work with the bc847 and bc857.
The good thing about this circuit is that I wont have to worry about toggling a GPIO pin anymore and that it only has 4 transistors. It just works.

I'm glad it works. R3 and R4 also control the frequency but they need to be low enough to provide sufficient base drive for Q1 and Q2. The Wikipedia article I linked to in my previous post provides more information on the calculations, if you're interested.

The heater should work at any frequency, within reason and 6kHz should be fine.

What transistors did you use for your breadboard prototype? The problem with the bc847 and bc857 is they're only rated for 100mA and it's not a good idea to run components continuously at their maximum ratings. The gain of a transistor also falls significantly as the current rises and is very low at the maximum rated current so the voltage drop could be high and it may not oscillate reliably.

How about the BC817 and BC807 which are SMT parts and electrically the same as the BC337 and BC327?

[tiltit]

I used BC327-16 and BC337-16 just like in the schematic, they are my goto transistors for breadboarding, so I had a few of them in my junk box.
I also added two 100Ohm resistors in parallel from the 3.3V rail to the emitters of the two PNP transistors. That is to drop the voltage a little since the filament requires only 2.4 volts.
I am still looking for another simple and efficient way to bring the voltage down a little but for the time being resistors seem to be the simplest solution.

[Zero999]

Oh, I didn't realise you wanted 2.4V, why not just connect a resistor in series with the heater.

[Telek]

Terribly sorry for the necropost, but just wanted to thank Zero999 for this ingenious idea. I found a collection of 15 starburst 20x2 VFDs in storage and have been trying to get them to work. I'm using HV518 for the segments (perfect as it provides 32 outputs) + HV5812 for the grid columns (perfect as it provides 20 outputs). Only thing I was struggling with was how to properly drive the filament since there is noticeable fade on DC.

Some preliminary tests seem to indicate that this works perfect! I'll add some pics in another post. I don't have 1.5kOhm resistors, so I used 1k+440 and a 1uF 400V cap (it's what I had), all other components as you described. I am getting about 600Hz under load with 4.68VDCin and 4.40VAC (cyc RMS) out and 9.72V pk-pk. Filament draw is 71mA AC. According to the datasheet, typical at 5.6Vac 60Hz RMS is 73mA, so that tells me that I have it dialled in nice.

The waveform has a little inconsistent frequency, but it seems to work! At 5.0VDC in the frequency is noticeably more stable than 4.65VDC in.

But question - the datasheet for these VFDs says that the ideal filament voltage is 5.6Vac RMS 50-60Hz, with 6.2Vac max. I know DC pretty well, and I know that overall power from AC to DC is about sqrt(2), but I don't know this situation enough to know what voltage I can apply. Since this is pretty close to a square wave, which means from a power perspective it's a lot closer to DC calculations, I've tweaked it to 6.2/sqrt(2) or 4.4VAC. With the supply at 4.68V that's within range of my 5V logic (4.5 .. 5.5) so that way I don't need another PSU, but I'd really prefer to use 5.0V for the logic. However, that brings it up to 4.77Vac - is that going to be too high? At that voltage I'm detecting about 76mA (under the maximum of 80mA that the datasheet says), so that should be ok?

Running the calculations on the mentioned wikipedia article for frequency, I get 527Hz expected, and ~600Hz observed. I assume the resistances are going to be slightly lower in-circuit due to the second set of transistors which, while on, will have significantly less resistance than the R1+R5 combo.

If I wanted to bring down the frequency a little, could I just increase the capacitors to maybe 4.7uF or even 10uF? That should bring me down closer to 60 or 100Hz without changing R3/R4.

In terms of productionizing this - what do we think of MMBT4403 - BJT 600mA 40V PNP SC-70-3? Specs look very similar (5V VEBO -0.75V CESV).

[Telek]

Here you can see the VFDs as I originally started to play, then the breadboard with the HVs and this H-bridge, with the VDCin at 4.69V and VACout at 4.45 RMS, and you can see the inconsistent frequency. That stabilizes quite a bit if VDCin is 5.0V.

[Zero999]

The RMS AC voltage and current are the same as the peak for a squarewave. The output voltage will be a little lower than the supply voltage, due to the voltage drop across the transistors.

None of the resistor values are critical. You can use a lower, or higher frequency if you like. I doubt it makes much difference for a VFD.

The circuit isn't perfect. It has one flaw: both the upper and lower transistors simultaneous turn on, when it changes from one state to the next, drawing high current spike from the power supply. The h_FE of the transistors limits them to some degree.

[Telek]

Good to know about the spikes - thank you! Should we add a capacitor or inductor to the input then to try to mitigate this?

So does this mean that if the datasheet says 5.6Vac typical, that I can do 5.6V output with this circuit without worry? Is paying attention to the current draw the more critical factor?

You mentioned that we can't really increase the resistor values that much because it needs a certain amount to drive the base, which is why I was wondering about the capacitors. I'll try it as is and see how it works out.

What do you think about surface mount 600mA BJT PNP transistors (for example, MMBT4403)? Anything special that I should look out for? I'm trying to minimize footprint on the board so would prefer to stick away from through hole transistors.

Thanks again for all your help!!