I’m building a custom circuit for fun which I intend to use to monitor a single CT and provide temperature and humidity (and optionally pulse) readings, powered from battery, using features cherry-picked from the emonTH v2 and emonTx v3. I am going to provide all of the features I need on the board, but only populate the parts I need for a particular sensor location on each board, and upload minimal firmware, to keep battery consumption low. I also want to use a switch mode power supply as used in the emonTH so I can squeeze as much juice from the batteries as possible, but I am concerned about the ripple from the part used - the LTC3525ESC6-3.3 - which is about 25-60 mV depending on load. Does this not affect the analog readings made by the emonTH? It appears there is not much filtering of this ripple, just a bunch of reasonably large capacitors but not RC or LC or LDO stabilisation.
As I understand, the ATmega328 needs a reasonably stable reference for its ADCs, so if there is 50 mV of ripple on this line then it will affect its readings.
I could use a buck-boost to regulate the voltage to ~3.5 V and pass it through an LDO to get it down to 3.3 V with low ripple if necessary, but perhaps I don’t need to…?
I have not seen any design calculations for the emonTH, so I cannot comment about that. The only use of the ADC in the emonTH is to read the battery voltage, so any noise on that reading is less important. However, the sensors themselves might be sensitive to supply ripple, that is something that you should check if you are concerned.
However, having said that, I believe that you are correct in thinking that noise on the ADC reference can be a problem. I would suggest that you separate the reference from the supply and provide good filtering to remove as much supply ripple as possible.
If you haven’t done so, there are some Atmel application notes concerning noise, interference and grounding problems that are worth looking at.
@Robert.Wall: I’ve also got a current transducer input on the board, so there are two ADC channels in use. For this board, I ended up adding a 10 uH inductor between the AVCC and VCC pins of the ATmega328 to decouple the power supply (connected to VCC) from the analog parts. That inductor in combination with the capacitors should give good filtering at hundreds of kHz where the ripple will be large.
@Bill.Thomson: thanks for the pointer, that is a great idea! I will definitely use this technique for other designs. The voltage drop across collector and emitter of the transistor (or Darlington pair) though means this technique still doesn’t come for free for a low power design, so I’ll see how I get on without it.
I’ve ordered my boards, so I’ll report back on how the performance is once I get and stuff them (in many weeks). Thanks to both of you for your help!