EEVblog Electronics Community Forum
Electronics => Projects, Designs, and Technical Stuff => Topic started by: matseng on August 14, 2016, 05:49:02 am
-
Having packed up my lab 6 months ago for an upcoming overseas move (that drags out in time) I have to keep myself occupied with just the design and planning phases of projects. I've got one for a 6 digit hex-character display (13 segments each) using low voltage "grain of wheat" bulbs. Binary inputs for the displays with the decoders using discrete parts only.
To reduce the awfully large number of parts required for a static-driven solution (like 550 discretes) I'm looking into the possibility of muxing the the individual digits, lowering the number of decoders to 1/6th but then of course requiring a johnsson counter and a mux for the inputs.
But I can't really find any good information on the pitfalls and special requirements for muxing incandescent bulbs. If I have 2 volt bulbs and mux 1-of-6 could that be driven from 12 volt? Or would it require a higher voltage than 2*6=12 volts, or lower? With LEDs it's easy to calculate average power and look up the max peak values in the datasheets, but bulbs are quite a different beast by its non-linear nature.
Do anyone have any pointers or insights of the nature of this?
-
I think the bulb is not fast enough for POV display. You only will see dimming bulb not an functional display.
-
Right, the bulbs are slow and will not be bright if driven pwm'ed/muxed with their nominal voltage. That's why I ask about driving them at a much higher than the rated voltage. Like running 2 volt bulbs at 12 volts on a 16% duty cycle.
-
I would expect that to significantly lower their lifespan as you'll be giving higher than rated current pulses at the worst time (when they are not at operating temp), i.e. the "temperature ripple" won't be good.
-
Should work if the multiplexing frequency is high enough to avoid temperature ripple. It probably needs a 'keep warm' current through each filament to reduce the stress on it. Also each individual segment of each digit will need a series diode.
-
Even if high voltage/current pulsing used, i doubt it can work. The bulb maybe can heat fast, but the cooling time will be the same.
-
Do anyone have any pointers or insights of the nature of this?
Rather than trying to timeshare the output drivers you could multiplex the decoder and latch the segment enables for each bulb group. But multiplexing the inputs to the decoder sounds expensive in components, what decoder design do you have? DTL ROM is easy these days with dense double sided boards and diode arrays.
-
6*13 is only 78 segments. Sticking to discretes only is retro madness - not as bad as an all discrete 6502 but still fairly crazy.
A chain of ten TPIC6A595 chips gives you 80 independently controllable low side power MOSFET outputs for around $32 (from Mouser). Add an adjustable switching regulator for the common +V rail to set the bulb brightness and a MCU to convert the characters to be displayed to a SPI bit stream of segment patterns, and all-in you should be able to do a static driven solution for $50 to $60 with *FAR* less than 550 components. Lets say 10 for the regulator, 10 TPIC6A595 chips, the MCU and some decoupling and you still only come out at around 30 components + the displays.
-
Rather than trying to timeshare the output drivers you could multiplex the decoder and latch the segment enables for each bulb group. But multiplexing the inputs to the decoder sounds expensive in components, what decoder design do you have? DTL ROM is easy these days with dense double sided boards and diode arrays.
Ah yes, that is an interesting idea. I'll definitely explore that avenue for a bit and see what happens with the part count.
I've played around with Logic Friday for some different configurations of the the decoder part, going from truth table to gates, trying to minimize the gate count but even at the best of versions the total part count for doing the gates in DTL would exceed a plain simple diode rom matrix with a 4-to-16 decoder in front of it. And the routing will also be simpler with the matrix.
-
6*13 is only 78 segments.
A chain of ten TPIC6A595 chips gives you 80 independently controllable low side power MOSFET outputs for around $32 (from Mouser). Add an adjustable switching regulator for the common +V rail to set the bulb brightness and a MCU to convert the characters to be displayed to a SPI bit stream of segment patterns, and all-in you should be able to do a static driven solution for $50 to $60 with *FAR* less than 550 components. Lets say 10 for the regulator, 10 TPIC6A595 chips, the MCU and some decoupling and you still only come out at around 30 components + the displays.
If I went for a IC solution I'd probably do something like this, os possibly with simple 595's and discrete transistors to keep the cost down (but pcb area up a bit). But in this case I want to do it as a pre-IC ~1965's era retro-style design.
-
I assume you are doing this pure bipolar, so no MOSFET tranmission gates. What logic levels are you planning to use?
Is the input in binary or 1 of N? I assume all the digits are presented in parallel.
-
I did 20x12 Volt incandescent lamps. Only 4 channels configured as a straight light chaser,
and it certainly couldn’t get to any speed for a POV display with one of four channels lit at a time.
It was a 555 with counter and then power transistors, but it was turning on & off.
I’m not sure I’d expect better modulating them with a voltage greater than zero to keep them warm or not.
-
Who said anything about a POV display apart from Han? The O.P. certainly didn't.
-
Not that long time ago incandescent filament 7-segment displays were common. Maybe at Route66 you still will find a historical gas station equipped with them :D
And sure the filaments were muxed. You need anyway a blocking diode each segment for that. For muxing you need to account for average power. If a segment/lamp is rated for say 1W and you muxe it with a duty of 1/6 you will put in 1/6 of average power also. You need to compensate for that by the voltage: Do the math so it takes 1W average again when muxed.
Further there is no objection to use incandescent filaments. These are a bit slow at turn on / turn off. But that's all. De-rating normal incandescent lamps by 10% increases lifetime dramatically.
-
I was too easily influenced :O
Who said anything about a POV display apart from Han? The O.P. certainly didn't.
-
I assume you are doing this pure bipolar, so no MOSFET tranmission gates. What logic levels are you planning to use?
Is the input in binary or 1 of N? I assume all the digits are presented in parallel.
Right.... Bipolar only, but for space savings I'll go for 60's "tech" only, not 60's encapsulation. Doing it in SMD instead of using 1/2W radial resistors and TO5-style transistors will reduce the size from a small briefcase to something more usable ;-)
I've got a discrete DTL based computer on my backburner (did some pcbs and some test circuits soon a year ago). It's using cheap BC847's running at 2.5 volt and with carefully tuned resistors I could toggle my flipflops at 10MHz. I could probably up that a bit by using something more RF'y than jellybeans for transistors.
So this display board will for sure be used for this design, but I'll make sure that it will be usable for plain 5 volt logic systems for more standard-systems.
The inputs are 4-bit binary for each digit (to be hooked up to data & address busses)
-
For muxing you need to account for average power. If a segment/lamp is rated for say 1W and you muxe it with a duty of 1/6 you will put in 1/6 of average power also. You need to compensate for that by the voltage: Do the math so it takes 1W average again when muxed.
Thanks. But how will the non-linear nature of the filament affect this - can I ignore the fact that they have significantly lower resistance when cold and only be concerned about the average power of the hot steady-state needs to be considered?
-
Thermal mass and thermal resistance of the filament act together like a RC low pass filter for the heat generated. If you multiplex fast enough, say >100Hz, the resistance will not vary much and only depends on average power. Keep in mind the lamps will do also fine at 50-60 Hz/AC. Then the power is modulated at 100-120 Hz. For the power into the filament: P = U^2 / R. So for a duty of 1/6 you need to rise the voltage by a factor sqrt(6) = 2.45
Take a safety provision the lamps are only powered when multiplexing is working, otherwise some lamps will burn out.
-
How about instead of mux of lights,
you mux a sample & hold cap that controls the light.
-
I wouldn't go as high as 16:1 mux - maybe 4 to 8.
Also remember that the filament current can be very high when cold, so your peak current at startup may be rather high - if your drivers can't handle the peak current without drooping you'll see lit lamps dim when new ones on the row are turned on.
-
Tungsten filament lamps typically have a 10:1 ratio between their hot and cold resistances. Adding 'keep warm' resistors across each segment driver transistor will not only significantly reduce this ratio, reducing the surge current through the digit driver transistor, but will also reduce the cyclic thermal stresses on the filament, improving bulb life.
-
You could always use 5 wire charlieplexing, with H-bridge buffers. Just have the 5 wires from each digit connected though low voltage triacs to one set of bridges.