First, some background. I have two boards, Board A and Board B. Two separate, but extremely similar designs. Let's start with the similarities:

  • Based on AVRXMega
  • Exact same regulator setup. 5V in, 7833 SOT-223 3.3V regulator, and identical bypass caps
  • Both boards draw approx. 5mA unprogrammed, and 30-50mA when programmed.
  • Both are 2-layer standard 1.6mm 1oz PCB with a thorough application of ground stitching vias


  • Board A has no issues with regulation. Regardless of the load current (up to 500mA), the output is rock solid at 3.3V.
  • Board B will not regulate to 3.3V unless the input voltage is dropped below 4.5V, and even then it will only output 3.4V. Above 4.5V input, the output voltage tracks the input with an offset of 1V. So 5V in is 4V out.
  • Adding a 100Ω load resistor to Board B improves the output, it will only track up to 3.6V, and will only output 3.4V once the board has been programmed, drawing approx. 60-80mA.

I have tried half a dozen different regulators, even swapping the ones on Board A and Board B. I have used several different bare boards of A and B, even with just the regulator circuits populated.

I have even tried 100mA (TC1015-3.3), and 10mA (MIC5232-3.3) regulators, they all do the exact same thing.

  • 5
    \$\begingroup\$ You can't expect this to be answered without the actual circuit... But given one of the boards is working, even a circuit won't help as the other board is probably just faulty for one(or more) of possible hundreds of reasons. \$\endgroup\$ – Eugene Sh. Jan 19 '18 at 22:34
  • 4
    \$\begingroup\$ extremely similar designs ... what is the difference? \$\endgroup\$ – jsotola Jan 19 '18 at 22:37
  • 1
    \$\begingroup\$ sounds like Board B has a floating pin somewhere \$\endgroup\$ – jsotola Jan 19 '18 at 22:39
  • 1
    \$\begingroup\$ Test bare board impedance on input to output \$\endgroup\$ – Tony Stewart Sunnyskyguy EE75 Jan 19 '18 at 23:04
  • 3
    \$\begingroup\$ oscillation with incorrect caps? \$\endgroup\$ – Neil_UK Jan 19 '18 at 23:05

If regulator x falls on board b, but works when moved to board a, and regulator y works on board a but falls on board b, the problem is board b. If you don't post the schematic and physical board files, then we cannot answer why. The failure point is evident though. Board b has an issue.


Just a possibility:

Some suppliers of 7833 voltage regulators specify that the output needs a minimum load of 5 mA for it to regulate. If the current draw is less than 5 mA the inference is that the output voltage will rise. So, if the section of the circuit connected on the 7833 output is sub 5 mA you might be in trouble.

The TI part is good down to 100 uA but the Taitron components part is only good down to 5 mA.


Sounds like the regulator is going into bypass-mode. That normally occurs when the input voltage is too low.

However it also happens when the ground pin is floating.

Check the ground. Even if it seems ok, run a jumper wire direct to the board ground input from the common pin on the regulator and see if it comes to life.

If it does, and you are using a ground plane, you may have bad plating on the through hole.


Smaller things first.

The 7833 is not an LDO. It requires at least 5.8 volts in to work correctly.

So, when you drop the input even lower than that, the 7833 isn't working at all - it's just providing leakage/unregulated current. That's why you get poor performance. Are you sure that your input voltage is 5.8 or better on B?

Now, on to the larger issues. Your problem is simple: you've done something stupid on board B. I'm not trying to be insulting - it's the term I've applied to myself in the past when I've made the sort of mistake you have. The circuits are, as @jstola has pointed out, "Two separate, but extremely similar designs." and there's your problem. You've done something different between the two. Find out what it is. And don't confine yourself to the schematic. At a start, measure the voltage at all three pins. I'm inclined to agree with Trevor_G that the ground pin is floating. Be aware that this is the sort of thing that probing may affect, so make sure to measure the ground and outputs simultaneously.

Since you've been banging your head on this for a while, you should enlist the aid of a co-worker for 10 minutes. Walk him through the problem and your troubleshooting WHILE YOU DO IT. You'd be amazed how often a second set of eyes can supplement a blind spot.


Not the answer you're looking for? Browse other questions tagged or ask your own question.