EMI reduction for multiplexed LED display

[Clear as mud]

In another post, I asked about multiplexing buttons and LEDs, and someone mentioned that running the multiplexed LEDs will generate EMI.  I will be running a 4-digit 7-segment display, up to 200 mA per digit, driving each segment with an output from a PIC, and the transistor for each common cathode will be driven from another output.  I guess the relatively high currents and fast edges can cause EMI.
Will bypass capacitors on the power supply and the PIC be enough to filter any radiofrequency noise?  Or is there another approach I should take?  I thought about using slow transistors, so that the LEDs will not turn on and off so fast.

Maybe I am worrying too much about this.  The circuit design is otherwise done, I just wanted to make sure the EMI will not be a problem.

[Andreas]

I thought about using slow transistors, so that the LEDs will not turn on and off so fast.

Slow edges is always a good Idea. On the other side you may get "ghost digits" on neighboured displays. So you have to introduce some waiting time between switching off and on between digits.
Decoupling capacitors are a must at each chip power supply pin pair. With 4MHz oscillator frequency on a good decoupled PIC (100nF X7R + 10uF Ta) you are well below radiation limits.

For the display the main issue is the magnetic loop area for the flowing current. So minimizing the loop area for the current flow will help to minimize the "antenna gain".  Especially if you want to build a radio controlled clock.

With best regards

Andreas

[mikeselectricstuff]

Easiest way to slow the edges would be to use FETs as the common drivers, and add series resistors to the gate, which in combination with the gate capacitance will limit the rise/fall times. With small FETs, a resistor of around 10K is a good starting point. 

[bookaboo]

Placing some 100nf and 1nf caps nice and close to the Vcc and Vdd pins of a PIC is a good idea regardless.

However if you are just driving some LED segments then EMI problems are highly unlikely, provided your MCLR reset circuit is OK.
EMI usually only rears it's ugly head when you have an inductive load such as a relay turning on and off, if that were the case then best practice is to try to kill it at source (RCs, Reverse Diodes) first before beefing up the PICs protection.

[Clear as mud]

  I laid out the board already, so hopefully I will not have to make too many changes - this is what I have now:

Input power 12V goes to a 7805 regulator.  At the regulator, I have a .33uF ceramic capacitor at Vin and a .1 uF ceramic at Vout.

I have three main indicator lamps driven directly from the 12 volt supply, of which only one will be on at a time.  They take about 210 mA at 12V, so next to the connectors and drive transistors for those lamps, I put a .33 uF ceramic and a 10 uF electrolytic capacitor.  These lamps will not be switched at high frequency (only << 1Hz).

Vout from the regulator I assigned the net name VDD, and it powers the PIC, which supplies the base drive current to all the switching transistors.  At the PIC, I have a .33 uF ceramic and a 4.7 uF tantalum capacitor between VDD and ground.  (I used 4.7 uF tantalum because I remembered I had some in stock from another project, but later I remembered the ones I have are rated for 100 V and quite expensive.   Therefore, I am going to reorder different ones anyway, so the size/type can change if needed.)

For powering the segments of the multiplexed 7-segment LED display, and my other LED indicators, I'll use a 74AC573, which can drive up to 24 mA per output (I'll have to adjust my resistors so I don't exceed the current rating).  I was going to use a darlington array as suggested, but it seemed that they invert the signal, and I couldn't figure out how to hook up the "freewheeling diodes common" terminal.  For powering the 74AC573, I created a separate power net called VCC, connected to VDD through an inductor.  I didn't know what size inductor to use.  Do I have to worry about resonant frequencies with the inductor and capacitors in the circuit?  I chose 22 uH because those seemed common on Mouser, not because I thought that value any better than anything else.  On the VCC side, I used the same sizes of decoupling capacitors at the '573 that I did at the PIC.  If resonance could be a problem, maybe using the same capacitors on both sides of the inductor was also a bad choice.

There is just one ground plane, and I made it very solid and well-interconnected.

[Andreas]

  I laid out the board already, so hopefully I will not have to make too many changes - this is what I have now:

For powering the segments of the multiplexed 7-segment LED display, and my other LED indicators, I'll use a 74AC573,

Above we spoke about slow edges for low EMI. A darlington array has around 1 us or even slower edges.
The AC-Series has up to below 1 ns rise/fall times which is more than factor 1000 faster.
So the AC573 its the best way to fall in severe EMI problems. Good luck ..... (you will need it).

With best regards

Andreas

[Clear as mud]

Well, I haven't bought the part yet.  I could change it to a Darlington array, but I thought this would be OK because the 74AC573 only drives the positive side of the LEDs.  The negative side is driven by transistors which switch the current much more slowly.  I didn't think I needed to switch both the positive and negative sides slowly to avoid fast rise times for the high-current parts of the circuit.  Also, I compared the data sheet with LS family logic, and the data sheet I looked at said the AC was slower, with about 15 ns rise times.

If I change it to a Darlington array, I will have to change the whole design, because I used a common-cathode display, so the darlington array would have to source current, not sink it.  But, I can't find any  darlington array that is currently being built that way.  There is M54562P that looks perfect, but they don't make it anymore.  So, changing to a Darlington array involves changing to a common-anode display, changing the segment drive transistors from NPN to PNP, finding a common-anode bi-color LED (and I'm not sure they exist), and changing the direction of three other LEDs and five other diodes.  Will it not suffice to keep the design I have, and count on the transistors on the LED cathodes to keep the LED current rise/fall times slower than the logic output rise/fall times?

I was hoping for some comments on whether my capacitor and inductor size choices make sense.  I guess I'm going to keep it mainly the same, but change the inductor to 2.2 uH instead of 22 mH.

-Daniel
 

[Andreas]

But, I can't find any  darlington array that is currently being built that way.   Will it not suffice to keep the design I have, and count on the transistors on the LED cathodes to keep the LED current rise/fall times slower than the logic output rise/fall times?
I was hoping for some comments on whether my capacitor and inductor size choices make sense.  I guess I'm going to keep it mainly the same, but change the inductor to 2.2 uH instead of 22 mH.

I usually use the UDN2981A as highside darlington.

At the rise/fall times of the AC-series most of the EMI current is flowing trough parasitic capacitances. So for the radiation it makes nearly no difference if your low side transistor is open or closed.
If you mention 15 ns I fear that you mix up propagation delay time with fall/rise time.

The selection of inductors is something which does not depend on the value alone. You have to regard your individual emission frequencies with respect to the resonant frequency of the inductor (parasitic capacitance). Further the dc resistance may be an important value. For the FM frequency range usually around 8-10uH are used. The effectiveness of a inductor is also determined by optimized placement on the PCB and the surrounding capacitors.

with best regards

Andreas

[Clear as mud]

You're right again, I was looking at propagation delay times, not thinking about the parasitic capacitances on the input stage of the logic chip.  I will change it to a darlington array.  Thank you for pointing out the UDN2981A. 

[TerminalJack505]

It sounds like it's too late now since you have already made the board but I thought I'd mention that a good way of introducing slew-rate limiting to a circuit with a transistor is to put a small capacitor between the base/gate and collector/drain.  This increases the transistor's Miller effect.