linear Supply Filtering Questions (digital circuit in a car)

[tecmec]

Hello, I I'm looking for some input for designing a simple linear power supply for a microcontroller circuit I plan on powering from a vehicles power circuit.

I am using a LM7805 linear 5V regulator, and I currently have a 47uF electrolytic cap and a 220nF ceramic on the input side of the regulator, and a 100nF ceramic cap on the output side. (there are 100nF tant caps around all my ICs too, just fyi).

I haven't actually tried this in my vehicle yet, but I believe my power supply circuit is not adequately filtering my power. My reason is this: I am currently powering the circuit using a cheap 12 power supply plugged into the wall. When I plug something else into the same outlet, or switch something on or off on that outlet, some of the noise gets passed though the cheap 12 supply i'm using (not my circuit), and in tern gets passed through my 5V power supply circuit (my design). The noise that I'm seeing (on my 5v rail) is about a short 10MHz ring. This wreaks havoc on my instrumentation amplifier and in tern my a2d readings.

My load is typically 35mA, but when one of my realys is on on my board, the circuit uses about 235mA.

Could someone please provide some insight on filtering this? My cap choices are probably all wrong.

 

[betwixt]

Re: linear Supply Filtering Questions (digital circuit in a

My guess would be that the problem with the wall adapter is not directly related to the regulation. The capacitors you are using are perfectly adequate although if you made the input capacitor say 1000uF and moved the 47uF to the regulator output it might help. Connect the capacitors directly across the regulator with wiring as short as possible. You should also make sure the ground pin on the 7805 is a common 'zero' for the input and output, in other words apply the negative input and negative output together at the 7805 ground pin. Doing this will eliminate voltage drops in the wiring as the current changes from shifting the output voltage.

I would suspect though, the problem you see would not show up in a vehicle where basically, everything is shielded. Consider that the wall adapter is hooked up to possibly several Km of power wiring and probably has poor shielding between the mains side and output side. Switching any load in the vicinity could place voltage spikes of hundreds of Volts on the incoming AC and some would conduct by capacitance between the adapter transformer windings. Although you may have 12V across the adapter output, both ground and 12V could lift together to a much higher voltage as the interference strikes.

As a precautionary measure, when powering from a vehicle, drop some of the voltage in a resistor in line with the 7805 input. The reason is two fold, firstly it shares some of the heat dissipation between the regulator and the resistor and secondly in conjunction with the capacitor at the regulator input, it acts as a low pass filter to remove switching and ignition noise from the battery voltage.

Brian.

 

[tecmec]

Thanks for the informative reply. I did in fact just switch to a larger cap on the input side. The largest I could get my hands on at the moment is a 470uF, but that should be a lot better I suppose (not related to this problem, but you know...).

As for the high frequency stuff, I like your answer . I suppose 10MHz noise shouldn't be present in the car.

Thanks a bunch.

 

[Nephazz]

The board grid can get very hostile ... like 100 V DC spikes. Unfortunatly I don't know the name of this effect but you'll probably need some overvoltage protection in front of the 7805.

 

[dick_freebird]

I don't think the LM7805 has much PSRR left by the time you
get to 10MHz. There should be a curve in the back of the
datasheet, or app notes. As a rule, you want to fix the HF
PSRR problem with a LC network on the incoming supply.
More bulk capacitance does this in a sense, but only using the
"stray" inductance of the feed and if you're not choosy about
capacitor details, perhaps using a cap with too high an ESL to
be effective at the higher frequencies where you need it most.

The other thing about the automitive environment is, ground !=
ground !=ground as you wander around the chassis sheet
metal, engine block, PCM. Only clean, tight, close-in connectiions
make the grounds "close enough" - close enough for uV, I doubt.
Using sheet metal ground is a bad idea, a star ground
with independent conductor is going to be better (but you may
also want to look at fixing whatever the sensor ground is,
or at least running your IA input reference to the same point
to eliminate ground loops.

Similarly relay power and ground ought not involve the signal
conditioning power/ground at all. Coil driver ground wants a
separate conductor and no board connection between them,
cross the ground domain with "digital" signals at minimal
current where you can.