Using an H-Bridge to switch 4 LED's

16 posts / 0 new
Author
Message

I am using Cree RGBW LED's 700ma per die in a project. I have a question on the voltage necessary to fully turn on the top two mosfets in the chip.

Using the ZXMHN6A07T8TA
5VDC
See attachments and links
http://www.diodes.com/datasheets...
http://www.cree.com/products/pdf...

I have used a diode to drop the voltage feeding the LED's so that the gates G1,G2,G3,G4 can be driven by four pins on a uC. Bread boarding the design it works fine. The question is from the data sheet, how do I calculate the voltage necessary to drive gates G4 and G3. Is 1 diode drop going to be enough to fully turn on those two mosfets?

Would it be the Gate to source voltage threshold minimum of 1VDC so I should perhaps use two diode drops at .6V to make it 1.2VDC drop? The resistors are going to be under 1 ohm for three of the dies on the LED and the top two mosfets will be driving the 3.9VDC forward voltage led dies.

Note. One of the colors (RED) has a forward voltage of just 2.5 so an additional resistor will be used to connect to ground, not shown on schematic.

Attachment(s):

Originally when I bought some of these H-Bridge chips I thought I could use all four mosfets to source the led's to ground. After playing with it on the breadboard and realizing the two top mosfets would need at least the forward voltage of the LED's plus 3VDC to fully open means I have run out of voltage to trigger the gates with a 5V Uc. I guess I will just use 2 chips and G1 and G2 fets.

Any suggestions for a quad mosfet or transistor array that can take 700ma. The 4 led's will not be on at the same time so a maximum of 700ma.

The ULN2064B from ST is a quad darlington transistor array designed to work from 5v logic inputs and can sink up to 1.5A per transistor. This might meet your needs.

I wonder why a H-bridge for that.
You need four current sources, just use bipolar transistor current source (Darlington is ok, but a regular one will do as well, as AVRs can sink/source quite a significant current).

I can see you do not have the option to turn all LEDs off. Is that what you are planning?

No RSTDISBL, no fun!

Nor, is there a way to turn on two LEDs on the same side of the bridge (for mixed color). When you turn #4 off, #1 is on. As butte points out, no way to turn all off!

Jim

Jim Wagner Oregon Research Electronics, Consulting Div. Tangent, OR, USA http://www.orelectronics.net

When all transistors are off, then LEDs are connected in pairs in series so there is no way to turn all off. Current is minimal and the drop on resistor would be minimal. 0,5V on diode. Depending on colors the pair on one side will glow.

No RSTDISBL, no fun!

The ULN2064B is really what I was looking for in a mosfet and I did not realize initially it was tied together source/drain on both sides. That is the chip I will use, thanks Chris.

I did give it a go though and found you can turn all off and individual led's on as long as the forward voltages in both leds in series is the same. For example two white 20ma led's at 3V in series will not light or glow.

Color mixing can still be done alternating each color since 100% current on all 4 at one time would be brighter than any individual color mixing them as a sequence of pulses keeps the total brightness similar for mixed colors as well as single colors.

If it were not for the higher voltage necessary for the top two fets to turn on full it would actually work very well. I had forgot about that when I configured this hair brained scheme to make it work.

I will use the uln2064b though or individual mosfets.

Quote:
For example two white 20ma led's at 3V in series will not light or glow.

Your diode is RGBW so your example is rather off-topic as there are no two white LEDs in it.
Second thing is as I wrote, diode drop will be about 0,5V at ~mA range. The small value resistor will not even notice a 1mA current. Again, 5V-0,5V = 4,5V is far from your example.
Anyway, you dropped the silly ideas about H-bridge - that is ok.

You need 4 controlled current sources. The green one is going to be challenging. Bipolar voltage source + series resistor + green and a 5V supply is not possible then - you need an active feedback (op-amp).

Vce_sat of that Darling is 1.2V for 750mA (25*C, but it is going to get hot after several seconds, so it is not that bad).
Since forward voltage drop of green is 3,7V(hope it is also at 25*C, but they didn't state it in a datsheet! Crap not Cree), then a very small (low value) current sensor (resistor) is needed. If your Vcc never drops below 5V, then a ~0,1V is a max. From this you can see this is going to be about 0,15ohm one.

Now, obviously four op-amps are needed to drive transistors. One LM324 will do driving four darlings.

With these darlings - bipolar transistors (and junction diodes in general) have a nasty property their Vbe drops when heated up (negative temperature coefficient) so watch your ammeter before it is too late!! Expect darling itself to give about -4mV/*C, the green blue and white is also about -4mV/*C. The red one is only -2mV/*C.

I must warn you just in case you would like to try if you can omit op-amps and make current source with a feedback voltage divider involving Vbe - you will only see a bright green flash and a smoke when the transistor and led heat up.

No RSTDISBL, no fun!

Thank's for the detailed post Brutte. One reason I was looking for fets in the first place was because of my lack of knowledge using transistors. Thank's for reminding me about that.

What do you think about this dual mosfet. Using two of these would be a better fit for the size of the board I am looking at as well as the ability to drive with the uC pins.

I am not limited to 5V for the LED's using individual fets (or dual chips with individual fets) as the battery pack is 9VDC

I will be using an MIC input and pulsed output to control the led's. The new design you can see here in the big bass drum uses an algorithm to blink the led three times for each hit of the drum. The rate of the blink is controlled by the frequency of the beat so that a single hit will be a longer three flashes and the faster the beat the shorter the flashes are. It is working very well here. I am using a single IRL110 for this controller.

So you see, the actual on time of the led never gets above about 70% and typically 20% on time or less I suspect. The new project I am working on is lights for the snare line.

Quote:
One reason I was looking for fets in the first place was...

Do you find unipolar transistors being less challenging than bipolar?

Quote:

The technology does not matter. If you want you can use relays if the clicks do not bother you. The problem is you need some kind of current source so you need to make one (or actually four of them). And your switch from 5VDC to 9VDC changes everything..

So you have a 9VDC battery pack and you want to power 700mA LED with it.. LED use about 3,7V and with such simple solutions all the rest is dissipated as heat - the overall efficiency of this will never exceed 40%. But if that is what you need..

Quote:
So you see,

I am sorry but my laptop keeps away from youtube (64MB RAM).

You need to write the software which senses hits of the drum and enables a 700mA current sources from time to time.. I will not help you with the former, but can give you some hints on the latter.

If you have the experience with unipolar transistors then why won't you get (N-mos and a series resistor connected to 9VDC)x4? Then you can use mos as a regular switch and because of huge voltage drop (of about 5V) a resistor is not a problem. You can fit with a +-10% current accuracy even with a supply voltage and temperature variations IMHO...

With a proposed active current source (darling + op-amp) you could easily achieve 1% accuracy independent of temperature, power supply and resistor tolerances, but it seems you have neither knowledge nor experience (nor 20 leds for "learning curve" before you gain that)..

No RSTDISBL, no fun!

Quote:
With a proposed active current source (darling + op-amp) you could easily achieve 1% accuracy independent of temperature, power supply and resistor tolerances, but it seems you have neither knowledge nor experience (nor 20 leds for "learning curve" before you gain that)..

Heck I bags of thousands of led's and ready and willing to learn a new method of driving led's with an op amp/ transistor combo.

I have found though that using CL2N3 drivers for strings of low power (20mA) leds to be very nice to use.

I have included some pictures, after I designed and built my own CNC machine for routing circuit boards I also use it to cut wax molds, then I use liquid rubber to make the epoxy castings of the leds that fit into a 1/2 inch routed area under the railings on my deck. Just pointing out I don't have more than a high school education but I am able to get by.

While I was driving today it hit me from the conversation I had on this board with Jim. Since I am only lighting one led at a time I can use for example this Cat4101TV-T75 Linear led driver for my current source. That with the two mosfets i posted above should make a very nice robust system for battery packs of 9VDC or 12VDC.

Don't mind playing with the op amps and transistors if you want to post a schematic for a driver at 700ma. Actually I have a spare one on the board now Tlc272 as I am using one of the op amps for the MIC pickup connected to the ADC on a tiny13.

Attachment(s):

Last Edited: Sun. Jul 24, 2011 - 10:49 PM

Marching band LED unit

The internal resistance of the batteries combined with the resistors was calculated to be 1000 mA maximum pulse, as the batteries lose voltage the lights get dimmer but with 6 AA batteries they last about 2 hours of marching without any sign of dimming.

These are thr Red/orange 3watt Luxeon emitters, they can take a maximum 2000ma surge and run at 1500ma.

Attachment(s):

Why are these photos 0,5MB each!!

Anyway, CL2N3 is a constant current source. But not very precise (+-10%). With leds in series you will not notice difference in light intensity, but with several chips driving groups of leds placed close together 10% is much too much.

But that is not the subject. A 700mA constant current source is.
All the solutions we are discussing in here are linear. I have never used a switched current sources. But since the weigh of the battery pack is not an issue as the drum itself weighs much more..

There are other solutions which are suitable for your design, like LM317T (also has thermal shutdown) which can be easily converted to a 0,1-1,5A current source, but the voltage drop there must be about 1.25V+2.25V=3.5V minimum (which is ok for a 9V source, but unacceptable for a 5V source you proposed).

This CAT4101 seems to be suitable - it is integrated and has all you need. This way you can avoid building a current source out of discrete components. Chip can dissipate about 3W but that is still not enough. Fortunately it has a low dropout below 0,5V in your case, so instead of connecting your leds between Vcc and CAT, you should put additional resistor in series with it so that CAT didn't fight with 6V voltage drop at full 700mA current (that would be 4,2W which is too much for this chip).

Additionally the chip has internal circuits which prevent damaging it in case of overheat - that is an advantage. However it does not protect the LED itself.

When you make such a device, you must always remember LEDs do fail sometimes. Those can fail to open circuit, and that is ok for linear regulators (no fire). They can fail to short circuit and it is also fine as long as you design a proper power dissipation in external components (especially in the CAT current source) because the current is going to be exactly the same after such failure (no fire).
The problem is when the LED fails somewhere in between (it draws nominal current and the fuse is of no use, but the voltage across LED is higher than operating voltage, thus power dissipated is higher).
Make your calculations so that at any condition and with any failure mode an external circuit (driver) didn't provide more power (P=U*I) than led case can dissipate.

No RSTDISBL, no fun!

Hey Brutte, thanks for all the help. I was worried a bit about the cat taking the whole 6V drop so initially I thought I might use an ultra low dropout regulator. I selected this one http://www.diodes.com/datasheets...

Since it is adjustable I thought I would tweek the voltage to match the led/cat requirements. From the datasheet:

Quote:
The AP1184 can easily be programmed with the addition of two external resistors to any
voltages within the range of 1.25V to 15.5V. Another major requirement of these graphic chips is the need to switch the load current from
zero to several amps in tens of nanoseconds at the processor pins, which translates to an approximately 300 to 500ns of current step at
the regulator. In addition, the output voltage tolerances are also extremely tight and they include the transient response as part of the
specification.

I am also using a mic2920a 400 ma regulator for the Uc. This new project I will be using a tiny84 as there will be various different modes of operation and an interface for the user.

You have a bit worried on that other projects light balance on the individual strings. I am going to make up two strings and a few of the cl2n3 regulators to see if that 10% will make a visual difference. Thank's for the tip.

Quote:
so initially I thought I might use an ultra low dropout regulator.

What for? You do not need any regulator. Just get four CATs and four 5W resistors and you are done. Project finished.

Does your battery pack voltage drop below 8,5V at 700mA? CAT needs only 0,5V for that 0,7A current. Green led takes 3,7V so all the rest must be dissipated in the resistor. That gives 8,5V-0,5V-3,7V=4,3V. Then you need a 4,3V/0,7A=6,1ohm resistor for a green one. Slightly higher for red one.
With 6ohm resistor you will get a 0,7*0,7*6=3W heat and a 4,2V voltage drop on a resistor.
When LED shorts, you will get 8,5V*0,7A-3W=2,95W energy dissipation on CAT. Will get hot but still I do not see a smoke.

No RSTDISBL, no fun!