Voltage difference with LED driver

Ok, so I hope the title was some what descriptive. I've got a project I'm working on and I was hoping to get some insight on this because I've looked through the LED drivers manual and it doesn't make too much sense and it's got to be something I don't understand.

I'm driving 16 SMD (1206) LED's (3.3V,20mA) from a TI TLC5928DBQR. I've got the driver Vcc at 3.3V (can take 3-5v) and then the LED bus, or as in the data sheet "VLED" is also 3.3V. I'm feeding both the LED and driver from a 3.3V 800mA regulator, which is fed by a 1A 5V regulator for other IC. Now, I'm getting this odd behavior that I can have the first 15 LED's on but as soon as I turn on the 16th, it goes into a funky mode which sporadic LED's going off. What I noticed when I took my DMM to it was the the power was dropping from 3.3V to about 1.9V and bouncing back up to the timing of the shifting from the driver.

So some ideas I have to correct this is to place 5V on the line and get rid of the 3.3V regulator. So finally my question. If my LED's have a Vf of 3.3V, and the driver is constant current regulating to 20mA, what is happening to the "extra" 1.7V on the bus side of the LED or as the data sheet calls it "VLED". In case it matters I've got a 1uF cap for a bypass on the driver. I hope the schematic helps understand more. Just for reference, ROUT is currently 3.3V and I wanted to bump to 5V.



In case the photo didn't work right the schematic can be view HERE

Thanks all in advance!

Hi jens, and welcome to the forum!

First, about the 'extra' 1.7V. The way that chip works is that each output has a circuit that works as follows:
A transistor is used as a variable resistor. The circuit measures the LED current and compares it to the setpoint. If the current is below the setpoint, the circuit drives the transistor harder to reduce the resistance and increase the current, and vice versa. It does this continuously, always adjusting the resistance to keep the current at the setpoint.

The extra voltage is burned off in the transistor(whose name comes from 'trans', meaning change, and 'resistor'). If you measure the output voltage to ground with the supply at 3.3V, you'd get a fraction of a volt (assuming you are using white LEDs with Vf of around 3V). That's the voltage across the transistor, and it will be equal to the supply voltage minus the Vf of the LED. So if you increase the supply voltage, the voltage across the transistor will increase, because the circuit will increase the resistance of the transistor to keep the current correct.

Since the power dissipated as heat in a resistor is equal to the voltage across it times the current through it, increasing the voltage will increase the heat generated in the resistor, er, transistor. When you add 16 circuits all together in one package, you get the chance for thermal overload pretty quickly. In your case, I think you are okay still with running the supply at 5V instead of 3.3. That would give about 5V (supply) minus 3V (LED) = 2V (transistor). The voltage times current is 2V*20mA = 40 mW. Multiply that times 16 channels and you get 640 mW. That should put the chip in a region where it's getting pretty warm, but nowhere near overheating.

By the way, assuming the regulators you are using are linear regs, they are doing exactly the same thing, except they measure the output voltage instead of current, and adjust the transistor to keep the voltage constant, instead of the current. A 3.3V regulator with a 5V input has 1.7V across it's transistor. And yes, that's the same 1.7V you were talking about earlier.

Now, as to your problems - If I understand right, the output voltage of the 3.3V regulator is what's dropping.

This almost certainly means you are overloading either the 3.3V regulator, the 5V reg that's driving it, or whatever source is providing power to the 5V reg.

It's unlikely that 15 LEDs works fine and 16 LEDs pulls the power down that far (not impossible though). My first thought is that something is wrong with that last channel, like it's miswired, or the output has failed, or both.

Can you turn only the 16th LED? If so, and it works fine the next thing I'd check is the 5V reg, both input and output. If the input droops, that's the problem. If the output does and the input doesn't, then the problem is in the reg.

If one of the regs is drooping and you are sure you aren't overloading it, then it could be oscillating. Voltage regulators can work fine at certain loads and not at others, even though both loads should be within specs. If you look at an oscillating regulator with a DMM, it looks like the output drops. That's because the oscillation is at a high frequency, which the DMM can't see. It just sees the average, which is lower because of the oscillation. Usually the cause of oscillation is improper bypassing of the output, though sometimes it can be the input or a control or noise reduction pin that's the culprit. Some regs are very sensitive to not only the value, but also the type of cap used, and the length of connection between the cap and the reg.

If you suspect a reg of oscillating, you can check with an oscilloscope. The frequency is rarely higher than a few MHz, so most any scope will do. Also, the reg's data sheet will usually mention output caps, and sometimes input caps. Pay attention to what it says, and unless you have a good PCB layout, don't push to the limits of what it suggests.

Good luck, and let us know what happens.

Thanks DIW.

I know it's been 2 weeks, but I haven't had a chance to get to this yet till this weekend. I am able to confirm that the 16th LED isn't wired wrong, it's just that I can have any combination of 15 of the LEDS. I will take a look at the reg's. Just for kicks, I took off the 3.3V reg and fed the LED's (blue) with 5V. Same results. I was hoping it was the fact I was feeding the LED's 3.3V when they have a forward of 3.3V, thinking it would be too close or would drop below the forward voltage.

I wouldn't be surprised if it was the 5V reg... it's a Radio Shack Besides that, the only thing I could think of would be communication btwn the MCU and LED Driver, but I'd tend to not think so because of being able to have any combination of any 15 LED's, just not a 16th.

I don't have an O-scope so I'll just have to wing it and see if I can check another way. I'll be reporting back once I know more.