I am undertaking a tilt sensing project for very small tilts.
The output is sensitive to input voltage. It is highly recommended to have a stable constant voltage.
I'm looking for guidance on very stable/constant voltage regulators. My input voltage is anywhere between 6v and 12v, I need to regulate it down to 5v.
I currently have a 5v buck regulator. My multimeter would suggest that it outputs a constant 5.0v however I suspect that it simply averages over a second or so and the noise in the micro second or millisecond scale is not captured.
Would putting a capacitor in the following position improve the stability of the voltage?
Power -> Regulator -> capacitor -> device
Or do capacitors have some sort of drift over time?
I've searched for noisy step down regulators on google to get an idea on how noisy they are (electronic noise) however I get bombarded with articles on actual noise that step down regulators can produce (i.e. longitudinal waves).
For checking the noise use oscilloscope, capacitors are often producing the noise,
even tantalum, I have changed lot of them , from 20 pcs I have find only 2 acceptable.
Depending on the current your sensor takes, there are many linear regulators which have very low noise . Add extra R-L-C smoothing if it has to be super clean.
Older regulators such as the LM317 were very good, but need the input voltage to be at least 2V higher than the output.
More modern low drop-out devices such as the LM2950 series may be suitable.
If efficiency is an issue, then consider a Switch Mode regulator [SMR] followed by a low-dropout linear regulator. Set the SMR output voltage to 5V plus the worst case dropout voltage of the linear regulator. For instance, for the LP2050-50, the worst case dropout voltage, for an output current of 100mA, is 600mV. So, set the SMR output voltage to 5 + 0.6 = 5.6V
Using the graphs in the LP2905datasheet (for instance Figure 12 and Figure 13, in the TIdatasheet), you might be able to tighten that up even more [knowing the actual specifications for your application].
That way, you get the benefit of the SMR's efficient energy translation, along with the clean output of a linear regulator. Setting the SMR as close to 5V as possible, without compromising regulation, insures minimal power loss in the linear regulator.
As allanhurst mentioned, the LP2950 is a good choice, as long as it can handle whatever current demand exists in your application (I didn't see a mention of that--but I can't imagine a tilt sensor would have much of a demand).
Any switchmode device generates loads of splat which even a good LDO can't clean up completely. The main reason is the very fast high current switching spikes coupled with the unavoidable inductances of PCB tracks and all components.
If worried about super clean dc, avoid.
Only where good efficiency and eg long battery life are important is it worth doing, and even then it may not work adequately - I tried this when designing mobile phones years ago and eventually gave up and stuck to linear.
Mind you, I needed < 1uV noise levels, so this may not be relevant in your application.
You seem to think that a stable supply voltage on your sensor will fix a problem.
"The output is sensitive to input voltage"
Maybe the (unknown) sensor has a ratiometric output, in which case it can't be detached from Aref.
If not, and you want a stable output from your device, then you also must provide a stable reference voltage for the A/D, since A/D result depends on both.
The (unknown) Arduino's default Aref could be worse than the output voltage of the (unknown) tilt sensor.
Leo..