Microcontroller LED multiplexing / General Science and Electronics / Forums

Forums

4hv.org :: Forums :: General Science and Electronics

« Previous topic | Next topic »

Microcontroller LED multiplexing

Move Thread

LAN_403

GeordieBoy

Tue Dec 06 2011, 10:09PM

Registered Member #1232 Joined: Wed Jan 16 2008, 10:53PM
Location: Doon tha Toon!
Posts: 881

Hi and seasonal greetings to all,

Here's a quick question for the electronics designers here, and those who frequently drive lots of LEDs from microcontrollers etc...

Almost all of the examples of LED multiplexing from microcontrollers that I can find online and in service manuals of commercial equipment show the common-anodes (matrix column connections) of the LEDs being driven from the positive supply rail via PNP transistors operating in common-emitter mode. These boost the available current from the microcontroller I/O line to support the sum of all of the possible LED currents on that column. For example:

My question is why are PNPs used in common-emitter mode when NPNs would seem to be a better choice operating in emitter-follower mode?

Eg:

I know that the saturation voltage of a PNP in common-emitter is lower than the 0.7v drop of an NPN operating as an emitter-follower. However in LED driver applications operating from 5v or more you still have to drop a few volts across the LED's current limiting resistor anyway, so you might as well drop 0.7v of it across the transistor and adjust the resistor calculation accordingly.

For all commerical projects i've done to date I have just copied the norm and used PNP transistors to drive the common-anode columns with the appropriate base resistors. This has generally worked well. However, on a few occasions I have had to add "dead-time" delays into the multiplexing software to allow time for one column transistor to turn off before changing the row data in order to prevent ghosting.

It seems to me that NPN transistors operating in emitter-follower mode would switch faster, are often cheaper and also don't need resistors to limit their base currents. So I am left wondering why this arrangement isn't the norm for multiplexed driving of panel LEDs in typical embedded control applications.

Any thoughts are much appreciated,

-Richie,

klugesmith

Wed Dec 07 2011, 12:11AM

Registered Member #2099 Joined: Wed Apr 29 2009, 12:22AM
Location: Los Altos, California
Posts: 1716

Hi. Has it been a while since you've posted here?

The NPN emitter follower has no current gain to speak of, so its digital input needs to source the entire anode current for as many segments as are lit. It (the uC output port) probably has non-negligible output impedance, especially when sourcing rather than sinking. So the voltage would drop & brightness would be change according to the number of segments.
[edit] As 2Spoons pointed out, that's just plain wrong. My bad; must remember to engage brain before shifting fingers into gear.

The PNP saturated switch requires much less base current, and its V_CE varies little with load current.

2Spoons

Wed Dec 07 2011, 01:43AM

Registered Member #2939 Joined: Fri Jun 25 2010, 04:25AM
Location:
Posts: 615

Rubbish. An emitter follower has current gain set by the Hfe of the transistor. It's voltage gain is unity.

klugesmith

Wed Dec 07 2011, 01:56AM

Registered Member #2099 Joined: Wed Apr 29 2009, 12:22AM
Location: Los Altos, California
Posts: 1716

2Spoons wrote ...
Rubbish. An emitter follower has current gain set by the Hfe of the transistor. It's voltage gain is unity.

Right you are, 2Spoons. I spoke in haste, and was just plain wrong, and regret it.
On a non-technical note, you mistakenly put an apostrophe in its.

>>In the NPN case its across emitter-base, which can usually only handle ~5V of reverse bias.

Hey wait a minute, how does the emitter follower config ever reverse bias the EB junction?
Are you talking about the brief period at turn-off, when the load (selected LEDs and series resistors) doesn't pull the NPN emitter down far enough and fast enough to keep up with the base?

I think the OP did mention that the NPN emitter-follower wastes some supply voltage.
Weren't the original multiplexed LED displays found in early-1970's pocket calculators, with NMOS LSI chip sets and perhaps with 2.4-volt batteries for power?

2Spoons

Wed Dec 07 2011, 01:58AM

Registered Member #2939 Joined: Fri Jun 25 2010, 04:25AM
Location:
Posts: 615

Fair enough, probably the lunch time sake doing the spell checking.

Dr. Slack

Wed Dec 07 2011, 08:05AM

Registered Member #72 Joined: Thu Feb 09 2006, 08:29AM
Location: UK St. Albans
Posts: 1659

It might be a hang-over from when the world was TTL, when, to within an order of magnitude, logic only sank current and never sourced it, and a logic '1' was anything above 2.4v on a 5v rail. In these CMOS days, outputs are more symmetrical, but even so, NPN structures are still better than PNPs, so rail to rail outputs still tend to sink a little better than they source.

This means is that it was easier to sink current out of a PNP hanging on the +ve rail, than it was to source current into an NPN follower and get a decent voltage out to drive the LEDs. The readily available rows decoders, see 74LS138 for instance, always had active low outputs, which reinforced this preference.

Anybody making new stuff, even to take advantage of new technology which overcomes former disadvantages, will have a tendency to follow in the wheel ruts of the old for compatibility, or familiarity of the users.

Steve Conner

Wed Dec 07 2011, 10:33AM

Registered Member #30 Joined: Fri Feb 03 2006, 10:52AM
Location: Glasgow, Scotland
Posts: 6706

This rang a bell. It turned out that I have done it with the emitter followers before!

Nothing bad happened.

Note the multiplexer would have been a 74HC238.

Electra

Thu Dec 08 2011, 12:25AM

Registered Member #816 Joined: Sun Jun 03 2007, 07:29PM
Location:
Posts: 156

I did think about the reverse bias, but this will be shared over the led and the transistor b-e junction when the row is high and column select low, unless the driver/microcontroller is using open collector outputs.
Probably isnâ€™t going to be a big problem at just 5v though.

Pnp drivers would really be an advantage if you were using a higher supply voltage or a different supply rail for the for larger size displays, as driven from an open collector would do the level shifting aswell.

Voltage drop is probably the most significant reason though, particularly as Dr. Slack mentioned if your logic high is not quite equal to +5v in the first place, and those green / yellow leds have a higher forward voltage drop again.

GeordieBoy

Thu Dec 08 2011, 10:11PM

Registered Member #1232 Joined: Wed Jan 16 2008, 10:53PM
Location: Doon tha Toon!
Posts: 881

Thanks to everyone for their thoughts.

Neil, I suspect you've hit the nail on the head with the reasons for most equipment being designed the traditional way with common-emitter stages. I hadn't thought about TTL outputs that struggled to get close to the +ve rail when outputting a logic '1'. In this case PNP common-emitter stages would work better. They would also be a good choice if using white LEDs where a much larger voltage needs to be dropped across the LED.

Steve, thanks for the vote of confidence...

"It turned out that I have done it with the emitter followers before! Nothing bad happened."

That's what I was hoping someone was going to say!

My proposed "LED and switch matrix" looks very much like yours except that:

1. I'm going to use one common set of eight "row lines" from the micro to support both driving a matrix of LEDs and also reading a matrix of switches.

2. There will be three column lines for driving LEDs through NPN emitter-followers, and three un-buffered column lines for driving the diode/switch matrix.

The only added complication is flipping the eight port lines used for the rows between inputs and outputs during the multiplexing to support reading the switches and driving the LEDs. I'll spend as little time as possible reading the columns of switches in order to keep the duty ratio for the LEDs sensible, and keep the LEDs as bright as possible for a given peak current stress.

Quickly sketched up in Eagle:

-Richie,

Moderator(s): Chris Russell, Noelle, Alex, Tesladownunder, Dave Marshall, Dave Billington, Bjørn, Steve Conner, Wolfram, Kizmo, Mads Barnkob