# Weird PCB stuff

1. Apr 2, 2014

### _Bd_

Hi, I recently designed a PCB in EAGLE, and I got it manufactured and I had all the pieces Sufrace Mounted on it, put it in the oven, everything seems fine and dandy, but I already killed 3 LEDs because of some weird thing going on.

I have 5 different LEDs conneced to my circuit, the leds run on 2.6V and the circuit is 3.3V so I have to reduce the voltage, therefore I use a resistor to drop it. however, even though the 5 LEDs are connected identically, only 2 drop voltage, the other 3 dont do anything =/, I used a multimeter and I checked right before each resistor (theres a hole via like a cm away before each resistor) and every Via read .9Amps and 3.3V just as I expected, however AFTER the resistors, only 2 of them dropped. here is a visual explanation:

*NOTE: ALL RESISTORS HAVE THE SAME VALUE, I ALREADY CHECKED USING MULTIMETER*

can anyone help me troubleshoot this problem?

2. Apr 2, 2014

### phinds

If you are reading the same voltage on each side of a given resistor, what does that tell you about the current through the resistor?

3. Apr 2, 2014

### meBigGuy

First off, lets mess with terminology a bit.

LED's are basically current controlled devices. At a given current they exhibit a certain voltage drop. So, you add the resistor to limit the current through the LED. For me, it is just easier to think about it that way. The brightness tracks the current, the voltage is exponential.

What current are these LED's rated at? (they nominally drop 2.6V at X ma). They won't all behave the same though. Some will drop more and some will drop less, so you need to be careful that for devices that drop less, you don't exceed the maximum current and burn them out.

Post a link to the LED spec sheet.

4. Apr 3, 2014

### _Bd_

that would tell me the current is zero, since V=iR

however, here is how it looks (like the actual EAGLE stuff):

I have some current mirrors to regulate the current on all LEDs, and I have (as mentioned) the resistors with 90 ohms each, all verified by multimeter.

I dont have the spec sheet for the LEDs, but the ones that read 2.6V do light up, the ones that read 3.3V dont light up.

according to what i know about the LEDs the max voltage @ 20mA is 2.2V

the voltage (on the anode part) reads 2.6V only on the second and last one (from left to right)

I tried the LEDs with the multimeter (checking for continuity) and they slightly light up with whatever little current the multimeter sends it, so I know the LEDs are working.

Last edited: Apr 3, 2014
5. Apr 3, 2014

### mr_pavlo

Hi,

I just wonder how you managed to measure current without breaking the trace on PCB.
And you showed us schematic with transistors and PCB with resistors. Can you please attach schematic with all parts and corresponding part of PCB.

6. Apr 3, 2014

### meBigGuy

OK, so let's assume there is something wrong on the circuit board.

Connect a 200 ohm resistor from the supply to each of the mirrors (right at the transistor) to verify they are working. If they are (I expect they are not) proceed to connect the resistor at various points until you get to the LED. You will see where it stops working.

Are these discrete transistors, or a multi-part package? Is it possible you stuffed the wrong parts.

7. Apr 3, 2014

### _Bd_

I did not measure the current, I edited the post to reflect that.

the PCB schematic is not complicated, its basically the first drawing I posted up there (just that instead of the LEDs replace with the same DF13 connector with 5 pins that connect to the 5 LEDs .

I just wanted to show where I measured the voltage across.

I've been doing some further testing and things point out to there being no current through those resistors, I am unsure how to fix this though

8. Apr 3, 2014

### phinds

Given that the resistors and diodes are in series, just take them as a unit. You have already stated correctly in post#4 that no voltage drop through the resistor means no current through either one. That means they are connected to an open circuit. WHY it is an open circuit is what you need to be looking for, as the problem obviously is not in the resistor or the diode.

9. Apr 3, 2014

### _Bd_

how can I troubleshoot it? I have little experience with this, like in theory it should work, in paper, in eagle, in LTspice

10. Apr 3, 2014

### meBigGuy

I'll say it again.

Connect a resistor to the supply. Choose a value that will drop 2V at the current your mirrors are running. (at 20ma that would be 100 ohms) Connect a wire (going nowhere yet) and a voltmeter to the other end. The meter will read 3.3V. Touch the wire to the transistor pin that is the collector of your current mirror that is driving an LED that is not working. The voltage should drop 2V. If not, the transistor is not working. If it is not working, measure the emitter voltage (sb 0) and the base voltage (~.6 or .7V) while the resistor is still connected. If it is working, mode to the next point closer to the LED and so on up to the LED. At some point it has to indicate a failure.

11. Apr 3, 2014

### Baluncore

First the obvious, what colour are the LEDs?
And are they all installed the right way round?
When using discrete transistors in a mirror to regulate currents you should use some resistance in the emitter lead. That prevents the one carrying most current getting hotter and taking more current due to the negative temperature coefficient of Vbe.

12. Apr 3, 2014

### _Bd_

they are red, and yes they are hooked up correctly, in fact if I swap LEDs from one wire to another they light up normally, but on those 3 particular channels (wires) since there is no voltage drop, the LEDs either dont light up or burn out.

*EDIT* reading previous posts, i didnt see some of them*

The transistors are in an array (chip), and yes it's the correct one.

13. Apr 3, 2014

### Integral

Staff Emeritus
The 3 LEDs that are reading 3.3V, or source voltage. There is an open between the 2 points you measured. First thing is to reflow all of the concerned solder joints. You could try to localize the open by measuring between different points. Voltage readings across the open will always read source voltage.

14. Apr 3, 2014

### davenn

there's another obvious fault in the circuit

the left hand transistor has it's base and collector legs shorted out
.... both legs going to the bottom of R12

Dave

15. Apr 3, 2014

### analogdesign

That transistor is connected fine. It's a diode connected device set to control a current mirror.

16. Apr 3, 2014

### analogdesign

Use your multimeter to ohm out the board.

First check that there is close to zero ohms between the LED cathode and the transistor collector for the LEDs that don't light up. Then check the resistance between the LED anode and the resistor. Finally check the resistance between the resistor and VCC. I think you'll find your problem that way.

17. Apr 3, 2014

### davenn

how do you figure that ?

so the way its wired the transistor would ALWAYS be on in full saturation so what's the point of having it there ?

because the junction of R12, C6 is going directly to the other transistors
that transistor is being bypassed anyway
but because the transistor is fully on, the C6/R12 junction is effectively always grounded

just wanna know

Dave

Last edited: Apr 3, 2014
18. Apr 3, 2014

### Baluncore

R11 and R12 set the collector current of the first transistor, right at the defined edge of saturation.
The Vbe generated is then paralleled with all others so they have the same collector currents.
It is a ground referenced current mirror with multiple parallel output sinks.

19. Apr 3, 2014

### AlephZero

I wonder if the problem is component tolerances. Looking at the MMBT3904 data sheet, it gives a 0.2V spread (0.65 to 0.85V) of base-emitter voltage at 10mA, which is roughly the collector current in U1.

With a 3.3V supply and 2.6V across the LEDs, there is no supply voltage headroom above the 0.7V needed to turn the transistors on. Maybe some of them are stuck at "off".

I expect this would work fine in a simulator where all the transistors are mathematically identical, but real world electronics isn't like that.

If you can isolate this circuit (so you don't fry anything else) and slowly increase the supply voltage up to say 4V, that would check if this is the problem.

Last edited: Apr 3, 2014
20. Apr 3, 2014

### analogdesign

The C6/R12 junction is not grounded at DC. It is ground from the point of view of noise which is what you want. It is just the input device of a simple current mirror.

21. Apr 3, 2014

### mjhilger

What is the exact part number of the transistor chip you are using? The one on the schematic is a chip but an individual transistor (1). Unless the batch is the same, there is a high likelihood that the Hfe is not the same among the transistors and some are not getting enough current to conduct as much as the others, though I would expect some current to flow. If your part is a chip with multi transistors in one package, then you need to make sure all emitters are connected to ground or common(some chips will show GND but require all GND pins be connected to GND.

Last edited: Apr 3, 2014
22. Apr 3, 2014

### Baluncore

I would check the pinout on the mirror chip very carefully. Maybe the numbering convention is different from that expected, the chip may be pictured inverted.

AlephZero's suggestion of increasing the supply voltage is worth trying.

Another possibility would be to momentarily short two transistor collectors.
If one LED was off and the other on then; if both come on at half brightness, the mirror transistor is the problem. If both turn off, one LED is the problem.

23. Apr 3, 2014

### AlephZero

There is a part number on the circuit diagram. The data sheet I found http://www.nxp.com/documents/data_sheet/MMBT3904.pdf implies they are separate transistors, not an array of matched transistors on one chip (which would be a better design option IMO).

24. Apr 3, 2014

### meBigGuy

I certainly would probe as I described to prove it is the mirror. You will have to do that eventually. It's ridiculously simple, and isolates the problem.