Inductor for AP63203

[chros]

Dear Community,

I am a hobbyist electronic developer and for a project I want to use the AP63203 as Step-Down Converter to generate +3.3V from a wide range (3.8V to 32V) of input voltage.
The circuit of the Step-Down Converter is shown in the attachment.

My question is about the right inductor value:
Currently I have chosen this component: Link

The issue for me is, that in the AP63203 datasheet, an inductor with 3.9uH is stated (see Table 2). On the other hand, in the AP63203 User Guide, an inductor with 6.8uH is used (see section 9).
I got the 4.7uH from the SparkFun BabyBuck Regulator Breakout - 3.3V (AP63203) board.

I do not have a proper rationale, why to use 4.7uH instead of one of the other values. Currently I chose this value because it is in the middle of the other two values.

Has anyone experience with the AP63203 and can confirm my selected inductor or recommend a different one??

Very much appreciated!

[Phoenix]

The datasheet gives recommendations on component selection - Eq 7 will give you an inductance. You'll find a range of values will work just fine. There may be some optimization on dynamic response and ripple as the inductance varies, not a big deal.

When selecting the inductor component make sure you pay attention to the average current (thermal limit) and peak current (saturation limit).

[chros]

Thank you for the answer. In general, when I understand section 10 of the datasheet correctly, the value of the inductor depends very much on the application, the drawn current and input voltage.
In my case, I cannot determine the use case specifically, as the AP632023 is part of a breakout board of a microcontroller, which will be used in a variety of projects I do not know yet.

So, I used some worst cast values for my applications and came to an inductance of ~4.5uH. In this case, my selected inductor seems suitable. Not only in term of inductance, also in term of DCR and saturation current.

See my values in the attachment.

Big thanks!

[mariush]

The R1 and R2 are not needed.

Be careful about input capacitors, especially if you assume the input voltage will be up to 32v. 
The ceramic capacitors have to be derated with voltage - a 10uF 50v ceramic capacitor may be only 2-3 uF with 24-30v on it. I would suggest having footprints for 2 1206/1210 ceramic capacitors on input, and maybe 2 footprints on the output (up to you if you're gonna populate both)

If you think the power supply is/will be at some distance away from the switching regulator (for example you power your switching regulator through a barrel jack and power comes through a cable into the barrel jack), a solid (polymer) or even an electrolytic capacitor could help - something like a 47uF .. 100uF 35-50v rated capacitor. Long cables have resistance and some inductance and a capacitor will help with voltage spikes and other things.

On the output, as you plan to have fixed 3.3v output, you could use 16-25v rated ceramic capacitors.

Layout is very important, keep the SW-Inductor-Vout loop as small as possible, the suggested layout is quite good.
Keep the feedback trace (from V out to FB , no feedback resistors) away from the inductor, for example loop it on the other side of the input and output capacitors, then use a via to go across and connect with FB pin.

[chros]

Hey mariush,

thanks for the comment.

For the layout, I have carefully considered the datasheet and kept the layout accordingly. The Buck-Converter can be supplied by two individual sources, which are

• USB-C Connector (very close to Buck-Converter)
• External Barrel Jack (External capacitors have to be placed at carrier board, usually I use: 100nF, 1uF and 10uF)

Both input capacitors (C2 and C3) do have a 50V rating. If second option for supply is used, additional capacitors are placed on the carrier board, so this should be fine.
I planned to populate both output capacitors. Footprint is 1210 and the voltage rating is 25V.
The feedback line is taken from the output capacitors and routed on the bottom side. The PCB itself is a 4-Layer stack with two GND planes as inner layers. Top and bottom contain signal and power only. Total size of the PCB is 43mm by 30mm.

What still concerns me a bit are the differential USB Traces, which are routed right next to the inductor of the Buck-Regulator. I did this to keep the traces as short as possible and to provide the impedance matching of 90 Ohm. The traces are routed from the 27 Ohm series resistors (R6 and R7) to the IC U2, which is an USBLC6 for ESD Protection.

[tooki]

The R1 and R2 are not needed.

The EN pin is also the UVLO, so a resistive divider lets you set the minimum input voltage.

However, I wouldn’t use the circuit as shown here, with the voltage being halved, since any input voltage above around 7V would apply a potentially damaging voltage to a 3.3V microcontroller. With a maximum input voltage of 32V, half is 16V, far too much for normal MCU GPIOs. If the idea is to set an UVLO, but have the MCU be able to pull it low to disable the converter, I’d add a MOSFET to do that, so the MCU pin is not exposed to the halved input voltage. If an UVLO is not needed, omit the resistors as mariush said.

Be careful about input capacitors, especially if you assume the input voltage will be up to 32v. 
The ceramic capacitors have to be derated with voltage - a 10uF 50v ceramic capacitor may be only 2-3 uF with 24-30v on it. I would suggest having footprints for 2 1206/1210 ceramic capacitors on input, and maybe 2 footprints on the output (up to you if you're gonna populate both)

If you think the power supply is/will be at some distance away from the switching regulator (for example you power your switching regulator through a barrel jack and power comes through a cable into the barrel jack), a solid (polymer) or even an electrolytic capacitor could help - something like a 47uF .. 100uF 35-50v rated capacitor. Long cables have resistance and some inductance and a capacitor will help with voltage spikes and other things.

On the output, as you plan to have fixed 3.3v output, you could use 16-25v rated ceramic capacitors.

Layout is very important, keep the SW-Inductor-Vout loop as small as possible, the suggested layout is quite good.
Keep the feedback trace (from V out to FB , no feedback resistors) away from the inductor, for example loop it on the other side of the input and output capacitors, then use a via to go across and connect with FB pin.

Excellent advice.

[tooki]

Regarding this:

If you think the power supply is/will be at some distance away from the switching regulator (for example you power your switching regulator through a barrel jack and power comes through a cable into the barrel jack), a solid (polymer) or even an electrolytic capacitor could help - something like a 47uF .. 100uF 35-50v rated capacitor. Long cables have resistance and some inductance and a capacitor will help with voltage spikes and other things.

See this application note about it: https://www.analog.com/media/en/technical-documentation/application-notes/an88f.pdf

If power will never be hot-plugged or switched near the board, don’t worry about it. But if it will, I’d strongly suggest taking measures described here. Bear in mind that while a large input capacitor is generally desirable, if it’s connected to the USB power line, it can’t be too large. I forget what the maximum allowed by the standard is, but that’s easy enough to search for.

Before I knew about this phenomenon, I killed a few aliexpress buck converter modules by connecting them to a powered 12V source.

[chros]

Thanks, tooki, for the post and hints.

About the EN pin of the AP63203, that is correct. I removed the voltage divider and connected that pin directly to VIN. If the MCU will be able to switch that off via a transistor, I do not know yet.

About the input capacitors, it is necessary to limit the input current from the USB hot-plug. I found the AN11392 by NXP, where a maximum voltage drop of 330mV for the VBUS is stated. In section 2.5 the inrush current handling is described. There a maximum capacitance of 10uF is given.

In order to ensure that the VBUS voltage does not drop below the allowable limit, the maximum load at the end of a USB cable is defined by the USB specification to be 10 µF in parallel with 44 Ω. The 10 µF represents the capacitance directly connected to the VBUS line, in addition to any capacitance that is visible through the voltage regulator.

If I understand this correctly, the input capacitance to my Buck-Converter matches the above statement. Maybe violating it a bit, as I have 10.1uF input capacitance. Do I see this correctly?

I did not understand, what exactly killed some aliexpress buck converter modules? The inrush current was too high?

[mariush]

Both input capacitors (C2 and C3) do have a 50V rating. If second option for supply is used, additional capacitors are placed on the carrier board, so this should be fine.
I planned to populate both output capacitors. Footprint is 1210 and the voltage rating is 25V.
The feedback line is taken from the output capacitors and routed on the bottom side. The PCB itself is a 4-Layer stack with two GND planes as inner layers. Top and bottom contain signal and power only. Total size of the PCB is 43mm by 30mm.

I would be more comfortable with 2 10uF footprints and again, 50v is kind of low if you think someone's gonna use 24v or more as input voltage.
It depends on the ceramic capacitor you choose. How much the capacitor is affected by input voltage depends on model and rating.
For example, (just picking from Digikey ceramic capacitors 50v and higher, sorted by stock amount),

1210 10uF 50v X7R from Samsung  : https://product.samsungsem.com/mlcc/CL32B106KBJNNN.do   <- check dc bias graph, you can see -40% at 20v https://www.digikey.com/en/products/detail/samsung-electro-mechanics/CL32B106KBJNNNE/3889045

1210 10uF 100v X7S from Samsung : https://product.samsungsem.com/mlcc/CL32Y106KCVZNW.do  <- -35% at ~ 20v https://www.digikey.com/en/products/detail/samsung-electro-mechanics/CL32Y106KCVZNWE/15992073

1206 10uF 50v X7R from Samsung : https://product.samsungsem.com/mlcc/CL31B106KBHNNN.do < -  -70% at 20v https://www.digikey.com/en/products/detail/samsung-electro-mechanics/CL31B106KBHNNNE/5961251

1206 10uF 50v X5R from Samsung : https://product.samsungsem.com/mlcc/CL31A106MBHNNN.do  <- -80% at 20v https://www.digikey.com/en/products/detail/samsung-electro-mechanics/CL31A106MBHNNNE/5961220

[tooki]

About the EN pin of the AP63203, that is correct. I removed the voltage divider and connected that pin directly to VIN. If the MCU will be able to switch that off via a transistor, I do not know yet.

You only need the transistor if you’re using the voltage divider to set UVLO. If you’re using the EN pin as a simple on/off enable, you can connect it directly.

I did not understand, what exactly killed some aliexpress buck converter modules? The inrush current was too high?

The issue is inductance. When you connect a fully-discharged very-low-ESR capacitor (like a ceramic cap) to a voltage, it wants infinite current initially. If your supply has a very low output impedance (like if it’s a ceramic cap), it will supply a ton of current initially. That’s fine, in and of itself. The issue is the inductance in between the two. This can be just a filter inductor in the power supply, but also the parasitic inductance of a long cable. As you know (or are about to learn), inductors resist changes in current. So if you have current flowing through an inductor, that current wants to keep flowing. If you try to stop it suddenly (by opening a switch, or in the case of the input capacitor, by becoming fully charged very quickly) the current wants to keep flowing, but can’t flow anywhere, so the voltage goes up. (Think of it as water through a pipe: the water has inertia (kinetic energy). If you then suddenly close a valve, that energy can’t continue as motion, so instead it builds pressure. This is the effect known in plumbing as “water hammer”.)

(Strictly speaking, it’s a series resonant tank circuit: https://en.wikipedia.org/wiki/LC_circuit )

By adding series resistance (limiting the inrush current), you reduce the amount of current change once the input cap is fully charged, and thus limit the voltage spike. Similarly, by adding a high-ESR capacitor in parallel to the low-ESR input cap, you get the slower charging of the high-ESR cap, but the fast transient response of the low-ESR cap for the buck converter to be happy.

And finally, as the paper also shows, you can add TVS diodes to limit the voltage.

[Gribo]

You might want to consider supporting both the 63203 and the 63200 (adjustable) variants.

[chros]

Hey everyone, thank you for the input.

I would be more comfortable with 2 10uF footprints and again, 50v is kind of low if you think someone's gonna use 24v or more as input voltage.

Makes sense to use higher voltage rating for the capacitors. Anyhow, if external supply is used (not via USB-C), then I expect additional capacitors on the carrier board. The board I am working on, is a breakout board for the RP2040 (I know there are many, but I have my reasons).
And, would I not violate the USB Spec in terms of maximum input capacitance if I use 2x10uF??
In general I fully agree, to use a higher input capacitance. My main concern is about the USB inrush current.

By adding series resistance (limiting the inrush current), you reduce the amount of current change once the input cap is fully charged, and thus limit the voltage spike. Similarly, by adding a high-ESR capacitor in parallel to the low-ESR input cap, you get the slower charging of the high-ESR cap, but the fast transient response of the low-ESR cap for the buck converter to be happy.

And finally, as the paper also shows, you can add TVS diodes to limit the voltage.

I understand. The MLCC I am using do have a low-ESR. High-ESR capacitors are Electrolytic Capacitors as far as I know. I am not so familiar with the ESR topic, but issues due to high inrush current I know.

Many boards use a Buck-Converter and allow supply via USB or external connectors. This topic I have never been seen addressed in any of these tutorials.

You might want to consider supporting both the 63203 and the 63200 (adjustable) variants.

I thought about that, but decided against that, as I will supply a specific device (RP2040) with the AP63203. So, I think no need for an adjustable output voltage.


But again, thank you all for your comments and input.

[mariush]

I understand. The MLCC I am using do have a low-ESR. High-ESR capacitors are Electrolytic Capacitors as far as I know. I am not so familiar with the ESR topic, but issues due to high inrush current I know.

Ceramic  capacitors have 1-2 or even less than 1 mOhm ESR,  solid (polymer) capacitors in the 6-20 mOhm (but you can find up to 50-100mOhm if the capacitance is big enough - 10uF polymers are in the 40 - 100mOhm, here's example: https://www.digikey.com/short/300br1n7) and electrolytic capacitors depends on series, some of the very low ESR series will  be around 100 mOhm for a 100uF  capacitor, while a 10uF capacitor could be up to 0.2-0.5 ohm ESR.

So even a solid (polymer) capacitor will be at least 10x the ESR of ceramic capacitors. 

[HwAoRrDk]

You might want to consider supporting both the 63203 and the 63200 (adjustable) variants.

Good idea. This will also allow you to substitute with pin-compatible buck converters from other manufacturers if needed. When using the fixed output chip, then just use a single zero-ohm resistor on the feedback, or the appropriate values for the adjustable.

I'd also suggest, if you have space, adding an optional footprint for a compensation/damping (I forget the exact term) capacitor alongside the divider resistors on the feedback. Some manufacturers chips (e.g. TI) require that.

[HwAoRrDk]

I thought about that, but decided against that, as I will supply a specific device (RP2040) with the AP63203. So, I think no need for an adjustable output voltage.

Gribo wasn't suggesting you consider a variable output voltage, but rather suggesting use of the adjustable output variant as a second choice of part in case you can't get hold of the AP63203. Part shortages are unfortunately still a thing.

In that case, you could get the AP63200 instead and place feedback resistors with appropriate values to output 3.3V.

[chros]

Ok, I understand, the flexibility makes sense. I will think about it.
I have to order this board fully assembled, as I do not have the capabilities to populate all components by myself.
In case I need more and a component is obsolescent or unavailable, I can change the design.
I am also happy to provide the design files to everyone who is interested in. Mo copyright or mention of my name is required.