LD1086DT33TR and ESR

[kevin_root]

I plan to use the LD1086DT33TR LDO to output 3.3V with an input voltage of 5V. However, I am not aware of the requirements for the ESR (Equivalent Series Resistance) of the output filter capacitor for the LD1086DT33TR, and this parameter is not mentioned in the datasheet. I intend to use either the GRM31CZ71C226ME15L or the CC1206KKX7R6BB226 as the output filter capacitor, and I am considering using two such capacitors in parallel. I wonder if this configuration is appropriate.(Or a 25YXF47MFFC5X11)?

[Peter BENSCH]

Welcome @kevin_root, to the community!

The LD1086 is a quasi-LDO, which, like most older LDOs, still has the problem of requiring a device-dependent minimum ESR for stable operation, because otherwise it can tend to oscillate at high frequencies.

For the LD1086, you will find some tips for stable operation in the data sheet in figures 20 (Stability for adjustable) and 21 (Stability for 2.85 V), from which a minimum ESR of approx. 500 mohms (ADJ) and 250 mohms (2.85 V) can be read. Although the regulator seems to operate on DC, the possible rate of change of the load must still be taken into account for the frequency, so that approximately 1kHz should be assumed depending on the application.

Today's MLCCs, like the types you have selected, have a strong frequency dependence of the ESR, which for the CC1206KKX7R6BB226 (Yageo) is approximately 3 ohms at 100 Hz and has a minimum of 4 mohms at 200 kHz; at 1 kHz, the ESR is approximately 250 mohms. The GRM31CZ71C226ME15L (Murata) shows a very similar curve, but at 1kHz it is only 150mohms. Several such capacitors in parallel at the output reduce the ESR even further. For maximum stability, I recommend in any case, especially when connecting output capacitors in parallel, a series resistor to the capacitor(s) of about 250mohms as an artificial ESR.

As far as the size of the capacitor is concerned, a 10µF is perfectly adequate, but it should be located as close as possible to the LD1086. However, MLCCs also have a pronounced dependence of the capacitance value on the applied DC bias. In your case, at 3.3V, the Yageo capacitor has approx. 30% less capacitance and the Murata capacitor approx. 13% less.

Taking all other tolerances into account, a single 22µF capacitor should be perfectly adequate as an output capacitor. If you want to use a higher capacity for your application, you can move the additional capacitors closer to the load, where the ohmic resistance of the supply line will then act as ESR, preventing oscillation.

Hope that helps?

Good luck!
/Peter

[kevin_root]

Thank you very much for your reply. Then, when does the situation of "it can tend to oscillate at high frequencies" occur, and what are the consequences of such oscillation? If MLCCs are not used, then when using electrolytic capacitors, we shouldn't need to consider the issue of ESR, right? For example, the 47µF/25V electrolytic capacitor (model: 25YXF47MFFC5X11). In addition, is there any LDO that doesn't require consideration of MLCC ESR? It would be perfect if it is pin - compatible with LD1068.

[Peter BENSCH]

Well, lots of questions, but I'll try to answer them:

• When exactly oscillation occurs is hard to predict, the tendency to do so can occur due to various external conditions: sudden load jumps, pulses on the input line, etc. For this reason, the causes of this tendency should be eliminated, in the case of minimum ESR with artificial ESR or with tantalum or electrolytic capacitors. If the regulator oscillates, no DC voltage can be measured at its output, but at least a DC voltage superimposed by a more or less strong high-frequency signal. As this situation can lead to the destruction of the supplied devices, it should be avoided.
• Quasi LDOs were developed in the last millennium, when electrolytic capacitors were still prevalent and MLCCs were still rare or expensive. Nevertheless, an MLCC can also be used if it has a series resistor to simulate an ESR.
• As a rule, many newer LDOs can work stably with MLCCs, but this is also explicitly mentioned in the respective data sheet. However, they usually also have additional features, which is why they are often more expensive. Decades-old regulators such as the LD1086, however, are widely used, are produced in very large quantities and are therefore comparatively inexpensive.
• It is probably cheaper or easier to connect the series resistor mentioned for minimum ESR in series with the MLCC (or to use an electrolytic capacitor) than to look for an MLCC-capable LDO that also has to be pin-compatible.

[kevin_root]

Thank you very much for your reply, which has given me a new understanding of LDOs and ESR. For LDOs like the LD1086 and LM1117, MLCCs have been used in previous designs, and the system still operates stably (probably because there are no conditions to trigger LDO oscillation, likely due to the absence of sudden load changes). Therefore, I never paid attention to this issue before. That’s why I want to clarify it thoroughly this time.

1. I plan to use the LD1086 with a 3.3V output. Do I need to refer to the "Stability for 2.85 V" specification to select the electrolytic capacitor?
2. To keep the LDO stable, does the ESR of the capacitor need to be between 1~20 ohms?
3. At which frequency should we determine the ESR? 1kHz，100khz It seems that the electrolytic capacitor datasheet does not include a curve showing the relationship between ESR and frequency.?
4. Unfortunately, I couldn’t find the ESR parameter in the datasheets of the following electrolytic capacitors. Only high-frequency, low-impedance electrolytic capacitors provide ESR values corresponding to 100kHz. How should I proceed with the selection?
5. For the output capacitor, if I choose a 47uF/25V one, do I still need to parallel a 100nF MLCC?
6. When designing the LDO’s output capacitor, ESR is taken into account. However, in the system, this LDO supplies power to an MCU circuit, and the MCU’s 3.3V rail uses many MLCCs in parallel. In this case, the ESR is far lower than the value required for stability. How should this be explained?（Is it not possible to upload PDF attachments?）

[Peter BENSCH · Accepted solution]

1. The stability zone looks essentially the same for all voltages; with a minimum ESR of 1 ohm, you are on the safe side.
2. On the one hand, the ESR should be as small as possible for a fast response to load changes, but for stable operation it should not fall below the minimum value mentioned under point 1 above.
3. Electrolytic capacitors probably have a higher ESR anyway, so you don't need to worry about that. The frequency dependence mentioned above is more relevant for MLCCs.
4. See 3
5. If you use an electrolytic capacitor, you do not need to connect an MLCC in parallel.
6. Depending on the type of loads supplied, it is very useful or even necessary to connect an MLCC directly to that load in order to compensate for high-frequency load changes. As already mentioned, the track or supply line from the voltage regulator to the load then acts as an additional 'ESR'. If the supply line is very short and its resistance is in the order of magnitude of the total ESR of the MLCC, it may be useful to implement a low-pass filter with a small choke of e.g. 100µH...1mH between the voltage regulator and the load to mitigate the problem.

[kevin_root]

        Thank you very much for your reply. The deeper my understanding grows, the more anxious I feel. I discussed this issue with my colleagues and found that no one has paid attention to it—we have been using MLCCs all along. We tested the output of the LDO and found no abnormalities, and the system remains stable. What could be the reason for this? (How should I simulate this problem? I wonder if, since there is no oscillation occurring, it is because chips like the MCU, when running at full speed, draw high-frequency current from the capacitors on their own power supplies, thus not affecting the LDO? After carefully checking the circuit diagram, I noticed that the 3.3V output from the LDO goes through a bead and a capacitor before supplying power to the MCU.)

        I was extremely shocked when I saw point 6 in your reply: *"Whether using tantalum capacitors, electrolytic capacitors, or milliohm-level resistors in series with MLCCs, all three methods will still have the problem mentioned in point 6."* It seems that designing an LC circuit is necessary.