Datasheet for Torex Semiconductor's XC6204/XC6205, Mark 4B2X - LDO regulator, CMOS based.

• This part shows up on a lot of ESP8266 development boards. Based on the datasheet this doesn't look like the best match.
• I also see this part on many display boards and various breakout boards.
• For me, Torex's datasheet is long which makes it very tedious to find the details, specific to a part variation.
• At least for all the variations in parameters that are listed, there is only one "Absolute Maximum Ratings" section shown for all variations.

• XC6204 vs XC6205, they support different ranges of output voltages.

  • XC6204 - 1.8V to 6.0V, with ripple rejection of 70dB@10kHz. Since we use 3.3V I will focus on this one.
  • XC6205 - 0.9V to 1.75V, with ripple rejection of 60dB@10kHz

• All variants feature a enable pin, CE, for power down.

  • One part option has an internal pull-up resistor(~300K) and CE is active low. (Semi-Custom)
  • One part option has an internal pull-down resistor(~300K) and CE is active high. (Semi-Custom)
  • Parts without an internal CE bias resistor.
    • CE is active high. (standard)
    • CE is active low. (Semi-Custom)

• A Vout tolerance of 1% and 2% are available; however, for the SOT-25 package, the part marking scheme used does not preserve this detail. I think it is safe to assume they use the one that was cheaper.

• For the SOT-25 package, there are only 4 characters to encode the part functionality.

  • The first character is either 4 (XC6204) or 5 (XC6205)
  • The second character encodes the voltage range, incrementation (100mV or 50mV), CE option, and the current rating. Look very closely at the second letter on the part and the matching letter in the table – for the presence or absence of the underscore. Those with an underscore are 300mA parts; those without are 150mA parts. For example, A mark 4A2D is a 150mA part and 4A2D 300mA part. It is easy to miss the A vs A in the datasheet's table.
  • The third character used with the second character allows you to find the output voltage in the table.
  • The fourth, last, character is the production lot number. Most of us can ignore it.

• Under PRODUCT CLASSIFICATION - Ordering Information, a lot of information is encoded in the full part number used to order. The formating of this part number is needed to find the correct relevant graphs for the part we have.

• According to datasheet section TYPICAL APPLICATION CIRCUIT, we should be good with a value of 0.1μF for CIN and 1.0μF for CL. However, the graph for CL in OPERATIONAL EXPLANATION list CIN as 1.0μF. There are also test circuits that were used for various parameter measurements that used 1.0μF for CIN. I am thinking that - if they needed 1.0μF to get the values they wanted for publishing, then maybe I should be using the same 1.0μF value for CIN in a design that does not behave properly.

• IMO, CIN and CL should be at least 1.3μF. For an X7R ceramic, this allows for a part tolerance of 10% and a Capacitance change of 15% over the range -55 to +125 degrees C. Otherwise for a generic unknown ceramic, 2.5μF. This assumes a tolerance of 20% and a Capacitance change of 50% over the range -55 to +125 degrees C. For those of us that grew up with the LM78XX line of voltage regulating ICs, these are very small values to be using. However, 1μF is the minimum value recommended for the XC6204/XC6205 CL

• Voltage spikes for high transient load changes are greatly improved with additional capacitance for CL, see middle graphics at link. IMO, the additional CL capacitance beyond what is needed to stabilize the LDO should be placed at the respective devices generating the transient load changes.

• A Schottky diode or connector/socket will introduce resistance to the LDO input supply. The low ESR CAP must be at the LDO regulator pin. Looking at the broader scope. In addition to the LDO's stabilizing CAP, any additional (must be low ESR) CAPs added to any circuit to deal with a transient load must be placed next to the load w/o any connectors or diodes in the path.

• Absolute Maximum Ratings vs Operational Maximum Ratings

  • I think of Absolute Maximum Ratings as "not to be exceeded or a catastrophic failure will occur, the magic smoke escapes."
  • For the Operational Maximum Ratings, when above and yet under the Absolute Maximum Ratings, you have no parameters that describe the behavior of the part. The part is not designed, tested, or supported for this region. It may work or may not. If it works, it may not work reliably or with the next batch of parts.

• In the Absolute Maximum Ratings section, the Power Dissipation for the SOT-25 is ~600mW. The exact value will vary with the PCB. This can easily be reached with: an input supply of 5.3V, 300mA consumed, and an LDO regulator rated for 3.3V. In the Absolute Maximum Ratings the current output is rated at 500mA across all variations. This is far outside the part's operating options of 150mA or 300mA. I am not sure how you could reach 500mA when both 150mA and 300mA parts have current foldback.

• Of concern, I do not see Thermal Shutdown as a feature for this part.

• As for the ESP8266 Development Boards that use this 150mA (sometimes 300mA) regulator, in the Espressif's Hardware Design Guideline, section 1.4.1.2. the notes specify "When using a single power supply, the recommended output current is 500 mA." I also think it is instructive to note, all the ESP8266 based IoT devices I have taken apart have a LM1117 3.3V regulator or equal, which is rated for an operational 800mA!
• While the typical power consumption of the ESP8266 is documented as 170mA/140mA/120mA for operating in 802.11b/g/n mode, at startup the ESP8266 can pull 430mA. This is shown in Ondrovo.com article "Determining the ESP8266 power consumption". That is going to trigger current foldback even with the 300mA part current limit expected at typical 380mA ("typical" implies might be better or worse). Even worse, the 150mA part current limits around 280mA - 300mA. The 430mA at startup also explains the need for a 500mA power supply as specified in the "Espressif's Hardware Design Guideline".
• This may be why, some people find, a problem is fixed by reducing the transmit power level and/or adding extra CAPs beyond what should be required. Which could lower the peak current demands within the LDO's capabilities. Even so, I would expect a notable sag in Vcc out from the LDO. A different concern, if too much capacitance is added to the power rail, power-on reset timing requirements will not be met for a reliable reset.
• Also, a problem described for sleep mode users is that the ESP8266 WiFi doesn't always work after sleep. I am thinking that maybe the WiFi Calibrate was not successful at the restart.
• Considering the above, I also find it alarming to see projects power additional devices off the 3.3V output of an ESP8266 Development Board that uses this regulator.