Open ended LM2937-5 outputs 12V

Question

The LDO, LM2937-5, with 12V input voltage and nothing connected to the output terminal, outputs 12 V (measuring using voltmeter). As soon as I connected 100 uF capacitor, it shows 5V.

Datasheet clearly states that output capacitor is required, but it does not say about overvoltage, rather than about resulting oscillation. In addition, all the tables list minimal current of 5 mA.

Now imagine circuit with failed capacitors (all of them), and microcontroller consuming less than 1mA. Microcontroller will fry?

I have another device on hand TS2937, and when its output is open (no cap) it shows more or less 5V, and surely not 12V.

What do I miss here?

Update: 10 uF does the job too.

Update: the practical question - is it safe using this LDO at all? I have never experienced this behavior from LDO. And as I said in the condition above - will circuit fry?

Update: thanks all for the responses. This is a repair, therefore using 7805 is not an option (need SOT-223 package), and the original part was TS2937. There're two 10 uF caps on the rail (however for 6.3V rating, and I personally do not like this fact), and LED with 220 Ohm series resistor, therefore the circuit meets the requirements :)

Answer 1

I wouldn't use an LDO regulator at all if the nominal input voltage is 12V. Something like the good old 78M05 has no special requirements for output capacitor capacitance and ESR to ensure stability and has a reasonable drop-out voltage in relation to the input-output differential. Also much cheaper, multiple-sourced and more available.

Both regulators are guaranteed with load > 5mA only, but I would not expect any issues with zero load current except possibly at elevated temperatures. There is a nominal 1mA load built-in, in the form of the output divider network on the 78M05, and no doubt something similar in the LM2937 (though the values are not published that I can see). Since it is guaranteed to regulate with a 5mA load at Tj = 125° if we assume that the leakage doubles every 10°C then no-load should be okay up to ~100°C Tj.

The LM2937 should have a similar relationship, though without the divider chain current it's not possible to calculate it. That part is also a nasty bipolar design using a low-hFE lateral PNP pass transistor so the Iq shoots up significantly near dropout (unlike the almost constant Iq current of the 78M05).

In any case, you are violating the requirements for stability by omitting the output capacitor (and note that a 10uF ceramic capacitor is almost certainly inadequate in that the capacitance will be less than nominal and the ESR will be too low). A 22uF to 100uF aluminum electrolytic capacitor would be okay.

As to your reading- I suspect it's oscillating at hundreds of kHz or more and your meter does not deal with that by showing the average but rather by reading something weird.

Answer 2

The datasheet mentions oscillation as a condition that can occur due to missing capacitance, but it does not need to give an exhaustive list. Simply put, the regulator is not operating in its rated/designed conditions, and its behavior is undefined -- a capacitor must be added.

The actual structure of the regulator is some control plus a PNP pass element:

If the PNP element is even slightly too-strongly-on without a significant load, then the output will get driven to 12 V (since the load of the feedback resistors is likely negligible). It is very plausible that there may be internal oscillations or dysregulation which manifests itself as the output being stuck at 12 V. It's also plausible that with a real load (but insufficient capacitance), the internal oscillations might actually appear as external oscillations of the output voltage.

Furthermore, it's plausible that different design or device parameters in TS2937 may make it somewhat stable without a capacitor, although it may still fail to meet its line/load regulation specification.

To answer the practical question - yes. Tons of circuits will do weird things when one deliberately ignores their datasheet requirements for input/output capacitance or their other requirements. If you have an issue with capacitors regularly failing open on your boards, you have bigger issues than the choice of regulator and attempt to prevent frying MCUs.