Best design to switch a load on/off drawing ~5A

Community PCB Design
1

Hi,

I’m looking to design a simple little circuit to switch on and off a load that in the worst case will draw 5A at 3.3V.

The circuit needs to source the current, not sink it. The load is 192 LEDs and I want them all to have a common anode that I can control.

Furthermore, it needs to have a fast switching time, on the order of 1kHz at least, so I want to use something solid state.

I’m guessing some kind of power transistor will do, but I don’t really know what design to use and how to handle heat (if that will be an issue at all.) I was looking at a 2N6488.

Also, is there any special considerations I should make when designing a PCB to handle that much current? Are there any issues I should know about?

I imagine this is quite a simple problem to someone with electronics experience, but I just don’t know where to start and want to do it right.

Thanks,

Mitch

2

What is your power supply “rail”? is it +12V or +5V or what?

What are the voltage levels of your switching signal? (i.e. 0V for “off” and +5V for “on”, or what?)

So without those answers, I can’t be completely spec a design, but here are my thoughts:

(1) Reason #1 the 2N6488 might not be best: it has an hFe rating anywhere from 20-150 at the 5 Amp operating condition. This means that you may have to pump as much as a couple hundred mA into its base in order to get it to “turn on” fully at your 5 Amp operating point.

(2) Reason #2 the 2N6488 might not be best: it’s a bipolar transistor, and it has a Vce(sat) rating (collector/emitter voltage at saturation) of as much as 1.3 volts at your 5 Amp operating point. This means you’d be burning over 6 watts in the transistor, which would get pretty hot as a result. Also, if you are working from a =5V power source, this eats up over a quarter of your available voltage drop (i.e. “headroom”.)

(3) Reason #3 the 2N6488 might not be best: being an NPN bipolar transistor, and since you need to SOURCE current, you’d pretty much have to use it in an “emitter follower” mode, wiht the collector to your voltage source, and your emitter to the common anodes of your LEDs. It’s a little easier to drive a transistor to saturation (i.e. minimizing its voltage drop, power loss, and therefore HEAT) when you operate it in an open collector mode. More about that later

(4) Suggest you consider using a “P” type device in an open collector configuration instead of an “N” type device in emitter follower.

(5) suggest you consider using a MOSFET, as the gate drive requirements will be easier to do, although this depends on what your signal source is (microcontroller?) and your voltage supply rail (+12V? +5V?)

(6) Putting those last 2 together, we might end up with a P-Channel enhancement MOSFET, something like the FQP27P06, which I suggest mostly because it is available from Sparkfun. (I’m assuming I can post Sparkfun URL’s here - glad to be corrected it that’s not allowed…) http://www.sparkfun.com/products/10349 Like the 2n6488, it’s around a buck in single quantities.

Now the actual design depends on your power supply rail voltage and your switching signal characteristics. If your rail is at +5V and your signal is coming from a microprocessor, you can probably just have the micro drive the MOSFET gate directly, although note that a +5V control signal to the MOSFET gate will turn the LED drive OFF, and a GND control signal will turn the LED drive ON.

If you have a higher supply rail voltage (say, +12V) then you may need a pull-up resistor on the MOSFET gate, and another n-channel small-signal MOSFET on the output of the microprocessor, say a 2N7000 or something like that. This isolates and level-converts the microprocessor output to handle the +12V range required by the power MOSFET gate, and also has the effect of inverting your signal, so now a +5V control signal from the microwill turn the LED drive ON, and a GND control signal from the micro will turn the LED drive OFF.

If you drive this P-channel MOSFET right, it has an on resistance of .07 ohm or less. At 5 Amps of drive, this will give only 0.35 V of drop across the transistor, representing 1.75 W of heat generation. Almost 4 times less power loss (and therefore heat generation) than with the 2N6488.

As for your 1 kHz requirement, the FQP27P06 MOSFET has switching times in the hundreds of nanoseconds, so that’s not a problem there, although there’s no free lunch, and the requirement you are burdened with for all of these great advantages is that the gate has a moderate capacitance, so you need to be careful about driving it “hard” in an AC sense, to get it to turn on and off in a timely fashion. Based on the ~1400 pF max gate capacitance, and using, say a 1 k-ohm resistor for the pullup and a microprocessor output capable of 5 - 10 mA for the pull-down, you should be ableto drive that gate where it needs to go within 10 or 20 microseconds, which should be fine to switch at a 1 kHz rate. (Glad to have someone check my math here. Also considering a ~45 nC total gate charge, is another way to approach it…)

So let me know what is your switching signal source, as well as your positive voltage rail, and we can get you pointed more specifically in the right direction.

3

Hi there!

I believe a transistor would be the best choice here. there are lots of models out there, and what you need to switch is not that much. I’m currently switching +48VDC@7.6 A using RFP30N06LE MOSFET’s and it takes about 250 ns to switch them on.

Out of curiousity, why would you need to blink the lights 1000 times/second?

4

Thanks a lot rasmadrak.

So to clarify with the MOSTFET, this would be acceptable:

         Vcc
         __
          |  
          |
         /
EN----|| 
         \
          |
          |
       (load)
          |
          _
         GND

And to answer your question, the load is 192 LEDs, or more specifically 8*8 RGB LEDs. I am making a RGB LED cube where each horizontal plane draws current from a MOSFET and every LED on each vertical column is connected to a control IC that turns them on or off. I then multiplex the cube by lighting each column one at a time, hence needing the (relatively) high speed transistors.

Should I at all be concerned with heat drawing 5A from one of those FETs? Obviously I will do my own testing, but just thought I’d ask.

Mitch

5

svoynick:
What is your power supply “rail”? is it +12V or +5V or what?

What are the voltage levels of your switching signal? (i.e. 0V for “off” and +5V for “on”, or what?)

So without those answers, I can’t be completely spec a design, but here are my thoughts:

(1) Reason #1 the 2N6488 might not be best: it has an hFe rating anywhere from 20-150 at the 5 Amp operating condition. This means that you may have to pump as much as a couple hundred mA into its base in order to get it to “turn on” fully at your 5 Amp operating point.

(2) Reason #2 the 2N6488 might not be best: it’s a bipolar transistor, and it has a Vce(sat) rating (collector/emitter voltage at saturation) of as much as 1.3 volts at your 5 Amp operating point. This means you’d be burning over 6 watts in the transistor, which would get pretty hot as a result. Also, if you are working from a =5V power source, this eats up over a quarter of your available voltage drop (i.e. “headroom”.)

(3) Reason #3 the 2N6488 might not be best: being an NPN bipolar transistor, and since you need to SOURCE current, you’d pretty much have to use it in an “emitter follower” mode, wiht the collector to your voltage source, and your emitter to the common anodes of your LEDs. It’s a little easier to drive a transistor to saturation (i.e. minimizing its voltage drop, power loss, and therefore HEAT) when you operate it in an open collector mode. More about that later

(4) Suggest you consider using a “P” type device in an open collector configuration instead of an “N” type device in emitter follower.

(5) suggest you consider using a MOSFET, as the gate drive requirements will be easier to do, although this depends on what your signal source is (microcontroller?) and your voltage supply rail (+12V? +5V?)

(6) Putting those last 2 together, we might end up with a P-Channel enhancement MOSFET, something like the FQP27P06, which I suggest mostly because it is available from Sparkfun. (I’m assuming I can post Sparkfun URL’s here - glad to be corrected it that’s not allowed…) http://www.sparkfun.com/products/10349 Like the 2n6488, it’s around a buck in single quantities.

Now the actual design depends on your power supply rail voltage and your switching signal characteristics. If your rail is at +5V and your signal is coming from a microprocessor, you can probably just have the micro drive the MOSFET gate directly, although note that a +5V control signal to the MOSFET gate will turn the LED drive OFF, and a GND control signal will turn the LED drive ON.

If you have a higher supply rail voltage (say, +12V) then you may need a pull-up resistor on the MOSFET gate, and another n-channel small-signal MOSFET on the output of the microprocessor, say a 2N7000 or something like that. This isolates and level-converts the microprocessor output to handle the +12V range required by the power MOSFET gate, and also has the effect of inverting your signal, so now a +5V control signal from the microwill turn the LED drive ON, and a GND control signal from the micro will turn the LED drive OFF.

If you drive this P-channel MOSFET right, it has an on resistance of .07 ohm or less. At 5 Amps of drive, this will give only 0.35 V of drop across the transistor, representing 1.75 W of heat generation. Almost 4 times less power loss (and therefore heat generation) than with the 2N6488.

As for your 1 kHz requirement, the FQP27P06 MOSFET has switching times in the hundreds of nanoseconds, so that’s not a problem there, although there’s no free lunch, and the requirement you are burdened with for all of these great advantages is that the gate has a moderate capacitance, so you need to be careful about driving it “hard” in an AC sense, to get it to turn on and off in a timely fashion. Based on the ~1400 pF max gate capacitance, and using, say a 1 k-ohm resistor for the pullup and a microprocessor output capable of 5 - 10 mA for the pull-down, you should be ableto drive that gate where it needs to go within 10 or 20 microseconds, which should be fine to switch at a 1 kHz rate. (Glad to have someone check my math here. Also considering a ~45 nC total gate charge, is another way to approach it…)

So let me know what is your switching signal source, as well as your positive voltage rail, and we can get you pointed more specifically in the right direction.

Thanks a lot for your VERY hepful reply! 3.3v supply and logic levels. If you felt like going into more detail of the design it would be much appreciated.

Do you think I need to be concerned with heat?

I plan to have 8 of these FETs connected to a decoder and power them in a one-hot manner.

Cheers,

Mitch

6

m1tch37:
Thanks a lot rasmadrak.

So to clarify with the MOSTFET, this would be acceptable:

         Vcc

     __
      |  
      |
     /

EN----||

|
|
(load)
|
_
GND




And to answer your question, the load is 192 LEDs, or more specifically 8*8 RGB LEDs. I am making a RGB LED cube where each horizontal plane draws current from a MOSFET and every LED on each vertical column is connected to a control IC that turns them on or off. I then multiplex the cube by lighting each column one at a time, hence needing the (relatively) high speed transistors.



Should I at all be concerned with heat drawing 5A from one of those FETs? Obviously I will do my own testing, but just thought I'd ask.



Mitch

Yes, that’s correct, although usually the load is placed on the vcc side of things.

A resistor between gate and gnd to force it shut unless opened should also be used.

7

Hi rasmadrak.

This is why I wanted to clarify. I need the circuit to be Vcc → Transistor → Load → Gnd. This is because the load is common annode tri-colour LEDs, and I must use the common annode to switch them all on and off. Because of this, putting the load before the transistor simply wont work for my design.

Most simple circuit I find online are as you describe. How can I make one to my requirements?

Mitch

Edit: I am now looking at: http://www.vishay.com/docs/71937/71937.pdf

8

Let’s be clear here… On any given LED, are you switching all 3 colors always at the same time? Or are you intending to turn on and off different colors at different times? Because the way you have described it, with a switching transistor driving the common anode, I have been assuming you’d have all the cathodes to ground.

This means that if you drive the common anode with your switch, then all 3 colors will come on at the same time, which is no different than just connecting all the cathodes together, and driving the LED by putting your V+ on the common anode, and switching the connected cathodes to ground with an “N-type” transistor. (i.e. Vcc - Anode - LED - 3 cathodes together - transistor - ground.)

Is that what you are doing - turning them all on and off at once? - or are you intending to turn on the different colors at different times?

9

Hi svoynick,

Yes, what I originally said is what I need. I want to be able to switch on/off all three colours in the tri-color LEDS. I want to be able to control the common anode.

To elaborate, my plan is to have 8 MOSFETs sourcing current and 8 TLC5947s sinking current. By multiplexing the MOSFETs I now have a 8196 LED matrix. Also, before you suggest I split it into eight 824 matrices, this will not work because it is a 3D matrix (888 [*3 for RGB]) and doing anything else would greatly complicate the lattice.

My plan should work, I just seeking some assistance in choosing the right transistors. In hindsight I should have elaborated in my OP, but my question was fairly specific and I’m not really looking for alternative solutions to my problem.

Thanks!

Mitch

10

OK, that puts it all in context. And I realize I missed the 3.3V power supply that you did put in your OP, so if it seemed weird why I was asking you that, it was weirdness at my end…

Now that I see what you are doing, it makes sense, but the 3.3V supply creates some fairly narrow parameters you are having to work within. Make sure you look at how your 3.3V headroom is “spent”… From Fig. 12 in the TLC5947 datasheet, we see that when the device is running 10-20 mA of current, it will have an output voltage somewhere in the 0.3 - 0.5 V range. And from our earlier calculation, the MOSFET may drop as much as 0.35 V, so you have 3.3V - 0.35 - 0.5 or only about 2.5 V left across your LED. Make sure all your LED’s can run at the desired current with this voltage or less…

So with a P-Channel MOSFET, your circuit would be VCC - MOSFET Source - (MOSFET) - MOSFET Drain - LED Common Anode group - (LED) - LED Cathode - TLC5947 channel connection. Yes, it’s the MOSFET Source that goes to the supply rail - this is a P-channel MOSFET, so you kinda use it “upside down”.

Note that as long as your microprocesor output is actively driven both high and low, this should work fine. If your micro output is an “open drain”, though (i.e. pulls down only) you might need to include a resistor (say 1k - 4.7k) from the GATE of the power MOSFET to the positive rail. This will pull the positive rail back UP to turn the MOSFET off, in case your micro output doesn’t have an active drive (but it most likely does…)

Finally, now that I have it clear that you are working from a 3.3V rail, in looknig back at that Sparkfun MOSFET, I see that it’s Vgs gate threshold is listed in a range from 2.0 to 4.0 volts. You want to drive a MOSFET gate beyond its threshold in a switching application like this, so you may want to pick a power MOSFET with a LOWER Vgs gate threshold, since the most you have to work with is 3.3V. You’ll want to find something with a low Rds(on) (to keep your voltage drop as small as possible) and a low Vgs (don’t we all!)

When I look for stuff like this, I go to a distributor website, and search on “MOSFET” then start narrowing the parameters… P-Channel, greater than 6A continuous Drain Current, < 2.2V Vgs, “In-Stock” parts only, and so on, until I get down to about 50 parts and then start looking through them. One that pops up is a Fairchild FDS4465. THis has a nice, low Rds(on) around 10 milli-ohm, so at 5A, it will only be dropping about .05 V, and dissapating about 0.25W. These are both definitely better than the example I gave earlier, and are both important to your application. It also has a Vgs of 1.5V (or better, on the “typical” spec) so that will be better in your 3.3V application. Note, it’s an SMD device, and it runs about 1.75 ea in small quantities.

So that’s an example of the kinds of tradeoffs you make in designing something like this, and notice that it was imporant to know what came “downstream” of the LED’s, to factor in that effect…

11

Thanks for your comments svoynick. It is very helpful.

I have a 3.3v logic level, but I suppose there is nothing stopping me using a higher voltage for the LEDs. The TLC5947s could handle it, and the power supply for my little project is TBD, so I am somewhat flexible.

Will going to a higher voltage make my life easier?

Sorry for asking such basic questions, I intend to research further what you have been saying, but I have an exam in a few hours :smiley:

Mitch