Electronic – Connecting the linear regulator directly to AC line

It really depends on what you're going to be doing and where the device will live.

If this is a one-off board and you're going to mount it up under the dashboard, then you could most likely skip worrying about automotive/industrial temp range components and stick an RBO40 on your board, with a big electrolytic capacitor behind it (a couple hundred uF) and feed that to your normal LDO, and you'd be protected from reverse battery, load dumps, fast transients, etc.

If this is something going under the hood, read on. :)

Firstly, and most importantly, take a piece of paper and write down your goals for this device. If you're not familiar or experienced with electrical engineering, you're going to have your work set out for you well enough to properly design an automotive device, and the last thing you want is feature creep. Make a list of what you want and stick to it for your first iteration.

Now that you've listed out everything you want for the first pass of your device, here's the reality. The automotive environment is tough. Also, people are stupid. There's heat, vibration, radiated and conducted EMI just to name a few things. This means you're going to need proper temperature range parts. You'll need to consider the EMI/ESD protection on input and output lines. You'll need to consider overvoltage protection on lines that people might connect improperly. You'll have to consider vibration if you have passive filters in an analog section. Here are some generic tips I've picked up working on (not finished yet :P) two automotive devices:

Automotive-grade parts

The easiest way you'll get parts that won't poop themselves in the automotive environment is to get automotive-grade components. You'll really only be looking at this for semiconductors. They are usually specified for higher temperature ranges: -40 deg C to 125 deg C is the most common. The automotive-grade quality, which might be AEC-Q100 qualified (if so, that's a good sign) means the parts have been built under a process that ensures lower defect rates. I'm not entirely sure if I've ever seen passives that are AECQ100 (or higher) certified, but a lot of the time I'll simply make sure they are specified for the same temperature ranges or beyond. A big caveat here is that capacitors can have widely varying capacitance over their full temperature range (and voltage range, at that)... so if you have circuits that require precise capacitors, pay attention to the temperature code of the capacitors and expected operating temperature of the device.

The easiest way to spot these parts is to usually look at the part number. For TI/National parts, they indicate automotive-grade parts with a Q in the name. For example, you'll see TI parts that look like XXX1234-Q1. The Q1 is the indictator that it's automotive-grade. Other manufacturers like ON Semiconductor indicate in their parametric search if the part is qualified to AEC-Q100 so it's clear to see. It differs a little between manufacturers, but looking for a "Q" at the end of a part number or for "AEC-Q100" is a quick way to mentally filter through parts.

EMI/ESD protection

Automotive devices have inputs, there's no way around it. :) Not only do you have to deal with conducted EMI from the multiple electronic subsystems in the vehicle, but you also have to deal with potential ESD when the user installs the device or even touches it once installed. ESD can burn out microcontroller pins, ADC pins, etc, and lead to you having a very bad day. :)

The simplest way to deal with this is to use an RC filter and/or ESD diode. One of our own awesome EEs here, David Kessner, gave me his personal recommendation for ESD protection via an RC filter: 50 ohm resistor and 2.7nF capacitor, with the resistor between the connector and capacitor. This will also potentially net you some EMI protection but you need to analyze the type of signals you have and filter them accordingly. There are also ESD diodes if you're a little wary and don't mind spending a few extra pennies to protect a signal. I'd recommend doing so. It's cheap insurance.

Power supply protection (overcurrent, load dumps, reverse battery, fast transients, etc)

People will connect the power leads wrong. People make mistakes, it's inevitable. You need reverse battery protection. Some regulators give you this for free, but again.. cheap insurance in my eyes. :)

You'll want a fuse too, preferably a resettable fuse. These are commonly called PTC fuses. Overcurrent situations happen and you don't want your circuit to misbehave and toast itself. Cheap insurance.

Load dumps scared me a lot at first... but when you look around, everyone just sort of handles them the same way and quickly moves on to another problem. The easiest thing to do here is to stick a fat TVS diode on your board. Some regulators will either be able to handle the peak voltage from a load dump normally or they'll have built-in protection for load dumps. Again, to me... TVS diodes are cheap insurance. The basic idea is that when the input voltage exceeds the diode's breakdown voltage, the diode starts conducting to ground, effectively clamping that overvoltage situation. Because load dumps have a lot of energy behind them in theory, this equates to sticking a womper of a TVS on your board to handle load dumps properly. I use the SMDJ series from Littelfuse. These are SMC footprint and big. They're rated to take a lot of power, though.

Another good idea is to stick a few hundred picofarads worth of capacitance near the connector on your power line to absorb fast transients. Make sure your capacitors are rated properly (200V is a smart voltage rating to go with here) and you'll be golden. This are the sort of transients you'll see when turning your windshield wipers on and off, etc. They can be in the 100 - 200V range but the series resistance and series inductance in your wiring will help tone them down and your capacitance near the connector helps as well.

Critical filters in analog sections

If you are doing any analog work and have passive filters (RC, LC, etc) then one thing you ought to consider is vibration. Your most common dielectric of ceramic capacitors, X7R, is subject to microphonics. What is microphonics, you ask? The capacitors become piezoelectric. This means they convert vibration to voltage... like a microphone. In an analog circuit, this is obviously not what you want. The solution here is to switch from X7R to C0G capacitors. They use a different dielectric and aren't subject to microphonics. Victory!

Now, this is by no means an exhaustive list, but more of a "keep these in mind while designing." As far as your personal requirements, here are my thoughts:

Your power requirements are small, but unfortunately, you're still subject to a lot of power dissipation going the normal linear regulator route. It's time to move to switching regulators.

Luckily, your current requirements are small enough where you could use a stand-in power module - like the Murata OKI series. These have a pin-compatible footprint with the classic 78XX series regulators and will give you far better efficiency and in turn you won't really have to worry about power dissipation. This particular part isn't rated to the typical automotive temperature range (only goes up to 85 deg C) but this is more of a hint than a specific answer. You could use one of these modules to get 5V power and then use a normal, cheap LDO on the board to get 3.3V. It's more current than you need, but the most important part is that it's fairly cheap and it lets you almost entirely forget about having to deal with power dissipation. Less heat generated = better for reliability.

Now, the 12V line... this is trickier. If you need a stable 12V, then you'll obviously need a regulator. Unfortunately, your current requirements again make power dissipation an issue with basic linear regulators. If you could afford to screw the regulator to the enclosure (it would have to be metal) to aid with heatsinking.. it could work. Of course, the other question is... do you need regulated 12V or are you simply powering something that could be hooked up directly to the battery of the vehicle? If it could be direct connected, you could skip having to regulate power and that problem goes away.

The most basic approach to this problem doesn't use a bridge rectifer at all but relies on a zener diode current limited by a dropper capacitor: -

The voltage across the zener can be regarded as a square wave with a positive peak at Vz and a negative peak at -0.6V. Using a conventional diode and smoothing capacitor after the zener can give a fairly smooth DC voltage of Vz - 0.7 volts: -

Smoothness is subjective of course and ultimately it depends on the smoothing capacitor value and how much current is taken by the load. You could use a 7.5 volt zener and, on the smoothed output, use an LDO regulator to give a pretty good 5V (if you really need a low ripple supply). I've used the above (without a regulator) to power a microphone, a bit of logic and a triac - it was a sound activated light built into a wall light switch. Ripple wasn't a big deal AND space was limited of course. I used a 5V6 zener as shown.

The resistor in series with the capacitor is there to prevent big surge currents should the circuit be connected to the AC at the peak of the sinewave.

This "full-wave method appears to get used quite often: -

The main idea is the same as the "bridgeless version in that the 2.2uF "drops" the AC voltage without dissipating heat. A zener diode than clamps that voltage. In this design it clamps at 6.2 volts.

It might be worth simulating it to see if the 1000 uF can be reduced in size. I would also consider putting the 100 ohm resistor on the AC side of the bridge and maybe even try this one: -

There are quite a few on-line HERE. BTW, the main problem with your design is that the 2N5550 is only rated at 140V max and it inefficiently drops all the voltage and even at 2mA collector current the average voltage across the transistor might be 200 volts and this means 0.4 watts of heat. Capacitor droppers don't have this problem.