How to make a simple, noise free 5v->3.3v DC linear voltage regulator circuit?

[John Celo]

I'm looking for a cheap, minimal parts count way to get 3.3V from 5V USB for my first basic microcontroller boards (can't imagine drawing more than 500mA, neither 1A), not particularly concerned about efficiency,
but I'd very much prefer it to be stable and have little to no noise.

Looking at various electronics boards I have lying around they use either LM317 or AMS1117.
LM317 isn't suitable for 5V->3.3V, so that's out.

At the first glance AMS1117-3.3 seems perfect, costs few cents on lcsc, on the second glance however, it becomes less than perfect once you read that it requires specific tantalum capacitors*, with quite specific ESR(effective series resistence) requirements, which quickly ruins the value proposition of these things.

*requirement which seems to be ignored on quite a few chinese boards, little to no tantalum orange to be seen

Searching this forum I've seen quite a few posts refering to AMS1117 as outdated and just plain out bad.

But if you want a linear voltage regulator replacement(not switching!) for the specific purpose of dropping from 5V(from usb) to 3.3v, what would it be? What are the options?

From my understanding, voltage regulators such as XC6206 with fixed output 3.3v, that work fine with 1uF ceramic capacitors on both input and output are switching regulators?

Most suggested replacements to 1117 I'm seeing online are switching regulators. (if I understand correctly, the regulators with voltage comparator that control a mosfet - would be considered switching?)

I've seen AZ1117 versions of 1117 which claim to be compatible with MLCCs (multi-layer ceramic capacitors), 10uf on input, 22uf on output in their spec page.

As well as noted that people say that large ceramic capacitors have issues with transients (voltage spikes, usually on powerup), would this affect the voltage regulator with AZ1117?

[Benta]

1 mA? 50 mA? 1 A?

[John Celo]

Arduino Nano has maximum total current of no more than 200mA (and uses AMS1117), if I understand correctly.

So 200-300mA would be the ballpark area (although i'm curious how the choices would change across the range if you need, say 500mA, or just 100mA).

[JustMeHere]

https://www.ti.com/lit/ds/symlink/lm3940.pdf

Can't get much easier than this.

[Benta]

It's all a question of power -> package -> cost.

You're looking for an LDO (low-dropout voltage regulator). TI is probably the biggest player in this market. Try lokking around here:
https://www.ti.com/power-management/linear-regulators-ldo/overview.html

[langwadt]

From my understanding, voltage regulators such as XC6206 with fixed output 3.3v, that work fine with 1uF ceramic capacitors on both input and output are switching regulators?

no, that's a regular linear regulator

[tooki]

There is a huge assortment of 1117-style (i.e. pin-compatible) regulators from numerous vendors, many of which are designed to be stable with just ceramic caps. Look at ST and Analog Devices, for example. ST’s LDL1117 is nice (I used the LDL1117S33R in a project not too long ago), as is AD’s LT3088 (requires slightly different config resistors/cap).

The schematic excerpt attached is the relevant part of that project’s schematic, showing the circuit that allows the use of fixed-voltage 1117’s, adjustable 1117’s, or the adjustable LT3088, configured for the 3.3V fixed-voltage LDL1117S33R. C21 is a 1206 ceramic, C7 and C8 are 0603 ceramic. (I didn’t go overboard on the output cap, since I have local capacitors next to the loads.) Works flawlessly, used in a piece of fairly precision measurement equipment.

[John Celo]

From my understanding, voltage regulators such as XC6206 with fixed output 3.3v, that work fine with 1uF ceramic capacitors on both input and output are switching regulators?

no, that's a regular linear regulator

If I look at the functional block diagram of XC6206 and AP2112 and similar, they generally speaking consist of
a voltage comparator connected to input and voltage reference (fixed 3.3v in this case) and output of the comparator is connected to mosfet which switches the input on and off thus charging the output capacitor to maintain the desired 3.3v.
(there's also optionally various safety features on top as far as i can tell)

Is my understanding correct?
Such a regulator is considered linear and not switching? (even though they primiarly function through switching)?

What am I misunderstanding?

[Manul]

If I look at the functional block diagram of XC6206 and AP2112 and similar, they generally speaking consist of
a voltage comparator connected to input and voltage reference (fixed 3.3v in this case) and output of the comparator is connected to mosfet which switches the input on and off thus charging the output capacitor to maintain the desired 3.3v.
(there's also optionally various safety features on top as far as i can tell)

Is my understanding correct?
Such a regulator is considered linear and not switching? (even though they primiarly function through switching)?

What am I misunderstanding?

No, it does not "function through switching". In such a regulator, mosfet is partialy conducting, like a controlled resistor. So it is dropping required amount of voltage. Also as a consequence it is dissipating power: voltage dropped x current. That's why it is not very efficient, especially for large voltage differencies between input and output.

[Terry Bites]

I'd plump for an AP7333. There's something to be said for stocking ajustable LDOs though.

Take you pick:

https://www.mouser.co.uk/c/semiconductors/power-management-ics/voltage-regulators-voltage-controllers/ldo-voltage-regulators/?number%20of%20outputs=1%20Output&output%20current=250%20mA~~16%20A&packaging=Cut%20Tape&polarity=Positive&instock=y&normallystocked=y&rp=semiconductors%2Fpower-management-ics%2Fvoltage-regulators-voltage-controllers%2Fldo-voltage-regulators%7C~Output%20Current%7C~Number%20of%20Outputs%7C~Polarity

[John Celo]

If I look at the functional block diagram of XC6206 and AP2112 and similar, they generally speaking consist of
a voltage comparator connected to input and voltage reference (fixed 3.3v in this case) and output of the comparator is connected to mosfet which switches the input on and off thus charging the output capacitor to maintain the desired 3.3v.
(there's also optionally various safety features on top as far as i can tell)

Is my understanding correct?
Such a regulator is considered linear and not switching? (even though they primiarly function through switching)?

What am I misunderstanding?

No, it does not "function through switching". In such a regulator, mosfet is partialy conducting, like a controlled resistor. So it is dropping required amount of voltage. Also as a consequence it is dissipating power: voltage dropped x current. That's why it is not very efficient, especially for large voltage differencies between input and output.

That clears it up! Thank you!

How do these newer linear regulators differ from AMS1117 or are they functionally roughly the same? Ie. why are many of the AMS1117 so finnicky in regards to having tantalum capacitors and don't work without specific ESR caps?

[tooki]

If I look at the functional block diagram of XC6206 and AP2112 and similar, they generally speaking consist of
a voltage comparator connected to input and voltage reference (fixed 3.3v in this case) and output of the comparator is connected to mosfet which switches the input on and off thus charging the output capacitor to maintain the desired 3.3v.
(there's also optionally various safety features on top as far as i can tell)

Is my understanding correct?

No.

Such a regulator is considered linear and not switching?

Linear…

(even though they primiarly function through switching)?

What am I misunderstanding?

…because they don’t function by switching at all.

There’s no comparator. They each use an op-amp to control a MOSFET in the linear region, meaning not fully on, not fully off. You are aware that MOSFETs aren’t only binary on-off switches, right?? In the XC6206 it’s a p-channel depletion-mode MOSFET, in the AP2112 it’s a p-channel enhancement-mode MOSFET.

The op-amp compares the feedback voltage to the reference voltage and produces an error signal which in turn varies how much the MOSFET turns off (in the ‘6206) or on (in the ‘2112).

[tooki]

How do these newer linear regulators differ from AMS1117 or are they functionally roughly the same? Ie. why are many of the AMS1117 so finnicky in regards to having tantalum capacitors and don't work without specific ESR caps?

Instability is likely a consequence of high bandwidth (which is desirable to allow high bandwidth in regulation). More stable modern variants likely use some sort of internal filtering or something to stabilize them without hurting the regulation bandwidth.

With that said, I don’t think even the ordinary ones are as fickle as you make them out to be. But moreover, it’s easy to find modern variants that aren’t fickle, and how they work is irrelevant.

[WillTurner]

Think out of the box! Raspberry Pi Pico . Output is limited to 300mA, and not a linear regulator. Parts count = 1. Clocked at 48MHz, the RP2040 overhead is 10mA. For "free" you get a LED, enable pin, and two M0+ cores.
 

[Manul]

How do these newer linear regulators differ from AMS1117 or are they functionally roughly the same? Ie. why are many of the AMS1117 so finnicky in regards to having tantalum capacitors and don't work without specific ESR caps?

In short, it's about feedback loop's phase shift. Circuit inside the regulator and the output capacitor act together to keep voltage control stable. Otherwise what happens is the following: voltage is a bit too high, make it lower, oh, now it's too low, make it higher, oh much too high, make it lower, etc. Which in technical terms called "hunting" or "oscillation".

By the way, you can emulate higher ESR capacitor by adding small series resistor, so no need for tantalum. But in modern times it is better to choose a regulator which is happy with low ESR modern capacitors out of the box.

[radiolistener]

I'm looking for a cheap, minimal parts count way to get 3.3V from 5V USB for my first basic microcontroller boards (can't imagine drawing more than 500mA, neither 1A), not particularly concerned about efficiency,
but I'd very much prefer it to be stable and have little to no noise.

Ultra low noise power supply cannot be basic and easy, designing it requires deep knowledge in RF design, EMI shielding, and professional PCB layout design for analog circuits. It requires deep electronics knowledge and skills, include measurements.

If you implement power supply without all these knowledge, you will have many issues, like noisy microcontroller ADC, faulty microcontroller PLL and sometimes even RF interference for other equipment.

Note that there are a lot of details which is very important, for example it is very important on how you designed geometry of power supply line from PSU to MCU on PCB layout, how you designed ground plane on PCB, every via and every wire geometry plays significant role here. So, it cannot be basic and easy.

At the first glance AMS1117-3.3 seems perfect, costs few cents on lcsc

I don't agree, I had very bad experience with AMS1117 and now I'm avoid using it, I prefer to use something other, for example ST LD1117.

But these are regular linear regulators. If you're looking for ultra low noise one, I can recommend LT3042..

As I remember there was a new more interesting alternative for LT3042, I found it some time ago but forgot it's name. Maybe someone will remind.

[langwadt]

I'm looking for a cheap, minimal parts count way to get 3.3V from 5V USB for my first basic microcontroller boards (can't imagine drawing more than 500mA, neither 1A), not particularly concerned about efficiency,
but I'd very much prefer it to be stable and have little to no noise.

Ultra low noise power supply cannot be basic and easy, designing it requires deep knowledge in RF design, EMI shielding, and professional PCB layout design for analog circuits. It requires deep electronics knowledge and skills, include measurements.

If you implement power supply without all these knowledge, you will have many issues, like noisy microcontroller ADC, faulty microcontroller PLL and sometimes even RF interference for other equipment.

Note that there are a lot of details which is very important, for example it is very important on how you designed geometry of power supply line from PSU to MCU on PCB layout, how you designed ground plane on PCB, every via and every wire geometry plays significant role here. So, it cannot be basic and easy.

that's exaggerated

[wasedadoc]

The OP did write "noise free". That means no noise whatsoever!

[radiolistener]

that's exaggerated

Absolutely NO. This is very serious. Designing low noise power regulator is not easy task.

I strongly recommend to strictly following PCB layout considerations from regulator datasheet in order to avoid bad noise performance of voltage regulator. Usually low noise voltage regulator has PCB layout example in datasheet and it is important to implement your PCB in the same way with all vias, ground plane, wire geometry and size. This is very important for noise performance.

[tooki]

But these are regular linear regulators. If you're looking for ultra low noise one, I can recommend LT3042..

That is a great part. I used the LT3045 and LT3093 (the equivalent negative regulator) in a project that needed extremely low noise. But indeed PCB layout is critical; I studied the datasheets and evaluation boards in great detail, and had my boss review it critically. It worked, but took time to do. They aren’t chips I would recommend to a novice PCB designer.

[mariush]

1117 doesn't require tantalum capacitors, but some versions do require a capacitor on output that has ESR above some threshold, most common above 0.1 ohm.

For example, pretty much any electrolytic capacitor below 100uF should have an ESR value above 0.1 ohm (of course if you look hard you'll find some)

Digikey has a huge stock of NCP176 and NCP186, which are LDOs that can do up to 500mA, 1A for the NCP186, and they're both stable with ceramic capacitors.

NCP176  : https://www.digikey.com/en/products/detail/onsemi/NCP176AMX330TCG/5761739

NCP186 : https://www.digikey.com/en/products/detail/onsemi/NCP186AMX330TAG/5761743

They're quite small though, probably a bit hard to solder by a hobbyist.

AP2112K would be easier to handle, it can do 600mA with as much as 0.4v dropout voltage : https://www.digikey.com/en/products/detail/diodes-incorporated/AP2112K-3-3TRG1/4470746

AP2210K for higher input voltage (up to 13.2v) but less current, down to 300mA :  https://www.digikey.com/en/products/detail/diodes-incorporated/AP2210K-3-3TRG1/4470819


MIC5504 is also very cheap, if you're fine with maximum 300mA :  https://www.digikey.com/en/products/detail/microchip-technology/MIC5504-3-3YM5-TR/4864018

[John Celo]

1117 doesn't require tantalum capacitors

Many specsheets of 1117 specifically recommend using tantalums or have specific ESR requirements, I'm not experieced enough to ignore advice specsheets give, neither do I have an oscilloscope yet to be able measure and diagnose consequences.

I like AP2112K-3.3, is 7c @ 50units at LCSC.
Also ordered SPX3819M5-L-3-3, which looks good at 6.4c @ 50units.
AP7333 at 7c a piece (not as good value as 2112K I think).
XC6206 at 1.36c a piece, because why not.
AZ1117CH-3.3TRG1, this costs 4c @ 50u. This version of 1117 explicitly says that it is compatible with ceramics in specsheet.

Most of those have 1u and 1u on input and output (except for AZ1117, which calls for 10u on in and 22u on out)

I'm going to be using them with CH32X033, PY32F002(16cent) and similar low cost MCUs.
I hope that the selection of linear regulators I have ordered will work fine for my toy boards and price are in-line with the cost of MCUs (somewhat). Those MCUs also work from 5v supply, but...

I'm very grateful for all the advice I've gotten. Thank you!

[David Hess]

Instability is likely a consequence of high bandwidth (which is desirable to allow high bandwidth in regulation). More stable modern variants likely use some sort of internal filtering or something to stabilize them without hurting the regulation bandwidth.

Linear Technology discussed how it can be done in some of their earlier linear regulator datasheets.  They relied on a tap into the collector or drain of the pass transistor to provide AC feedback and phase lead.

The circuit model of a high ESR capacitor includes a series resistance which creates phase lead from a zero in the transfer function.  Rearrange the circuit elements to move the resistance from the ESR to be series with the output of the regulator's pass element.  Now AC feedback can come from before the series resistance, and DC feedback comes from a point across the low ESR output capacitance.

Some discrete power supplies use this configuration with the current shunt or emitter/source ballast resistance in series with the output from the pass transistor which helps prevent local oscillation.  They could tap AC feedback from before the current shunt or ballast to use a low ESR output capacitor but offhand I have not seen it done.

Ultra low noise power supply cannot be basic and easy, designing it requires deep knowledge in RF design, EMI shielding, and professional PCB layout design for analog circuits. It requires deep electronics knowledge and skills, include measurements.

A discrete low noise regulator is pretty simple.  The noise of even the best reference is several times higher than the noise of the average error amplifier so component selection is not difficult.  The trickiest part might be tuning the frequency compensation for maximum bandwidth which will produce the lowest middle frequency noise.