AMS1117-3.3 outputs 4.4 V from a 5 V supply

Question

[disclaimer - I am new in DIY and this is the first PCB I designed myself]

I designed a PCB for a very simple ESP8266 12-F + RGB LEDs project.

I assume that I am doing something fundamentally wrong because the voltage regulator is connected to a 5.1..5.4 V, 500 mA supply but outputs 4.4 V instead of 3.3 V.

Note that the AMS1117-3.3 regulator uses the tab as another VOUT.

Note also that only the power-supply, the capacitors, and the regulator are currently mounted to the board. The ESP and LEDs are not yet on the board.

For details check the schematics screenshots below:

Any hint/help is highly appreciated!

EDIT: I am terribly sorry, just noticed that the actual design which was printed indeed was using the full-copper approach:

UPDATE: In the meanwhile my ordered 22uF smd tantalum capacitors arrived and I did some experiments:

• 22 uF capacitor without resistor -> 4.34 V
• 22 uF capacitor + 330 ohm resistor -> 4.22 V
• 22 uF capacitor + 100 ohm resistor -> 4.00-4.01 V
• 22 uF capacitor + 10 ohm resistor -> 2.68 V

So my follow-up question here would be if this is expected output? And do I need to use one of the resistors if I want to connect the ESP now?

Answer 1

The datasheet recommends a 22uF tantalum capacitor on the output. (The tantalum will have a higher ESR than ceramic and the ESR zero likely helps stabilize the loop.) The 0.1uF ceramic cap you show on the output likely isn't enough.

The layout isn't great either, the caps could be closer to the regulator with thicker traces, or even copper pours for the voltage rails.

Have you checked the output on a scope? That would show if it's unstable and oscillating due to the lack of enough output capacitance.

Answer 2

AMS1117 seems popular, probably because it sells for 2 cents on aliexpress. That's most likely counterfeit chips though, so you don't know what you're getting. Also, if a 2 cents regulator requires a special snowflake expensive capacitor, then it is no longer a 2 cents regulator.

If you like part numbers ending in 1117, I recommend LDL1117 from ST. It works with a 10µF ceramic cap, and it's also pretty cheap.

I am new in DIY and this is the first PCB I designed myself

OK, so advice:

The ESP draws quite a lot of current and I've seen a few people here that had trouble with AMS1117, it seems it is not that quick to respond to load current variation, so if you get problems like your micro randomly crashing, increase the capacitance on the output (any old electrolytic capacitor >100µF will do) and if that solves it, use a LDL1117 instead.

This SOT223 package dissipates heat through the tab, which should be connected to a copper area used as a heat sink. Since you did not, the LDO could get quite hot if the ESP uses wifi a lot. If it gets too hot and you do another board, you can either pick a regulator that has GND on the tab and connect that to the ground plane, or connect it to some copper tied to Vout on one side of the board.

Answer 3

You are perhaps not familiar with LDO’s are pullup only regulators and , pull down from the load.

Yet you have a LED load coming from 5V and driving low from 3.3V uC. since the LEDs pullup more than the uC pulls down as a lid on 3.3V the LDO voltage rises to an equilibrium where the output current is now zero.

Thus can be remedied with a a load resistor to ground on each driver pull the equilibrium down to 3.3V so the LDO is now supplying the minimum rated current, which the datasheet indicates as 5mA typ, 10mA max added to your LED load current.

I don’t see any current limiting resistors, except the uC CMOS drivers will be 25 Ohms +/-50%. if you define your LEDs, then R load can be computed.

If they are IR, Red or Yellow, don’t use 5V on the LED’s rather use 3.3. then you may avoid this issue. Normally RGB LEDs only need 2V min to 5V if using those. Show links.

Answer 4

There are most likely two factors affecting this.

First one is the output capacitor. Basically, there is much less output capacitance than suggested for stability under all conditions, and the capacitor is of wrong type. 100nF ceramic capacitors have extremely low ESR compared to the suggested tantalum capacitors. The fact that the output capacitor is further away from the regulator might even help a bit, as it is sometimes suggested when replacing tantalums with ceramics. However, 100nF can still be too low value.

Second thing is that there is no load at all. While the fixed 3.3V version should internally consume about 5mA which should make it stable, the datasheet does mention that the regulators will be stable under most conditions when it has a 10mA load.

Try to give the regulator a load, if nothing else, a 330 ohm resistor. If that does not help, add more output capacitance.