precise 5v for sensors with 3.3v arduino?

Using boards like the Uno and a MiniPro(5v), I've successfully done some analog sensor reading. Most of my sensors require an accurate 5v excitation voltage, so I just use Vcc on the 5v arduino boards. But now I'm trying to use these same sensor but on some 3.3v boards (Fio, JeeNode).

I tried using a Pololu 5v boost regulator (http://www.solarbotics.com/products/19210/) but it's too noisy and causes about 5 bits of fluctuations on the ADC sample of the sensor readings. Even with the sensors removed, I sampled the output of the boost regulator (through a resistor divider), and it still shows several bits of noise.

Looking at spec sheets, it seems like most boost regulators have a fairly large (at least ~2%) error on the output voltage, so I thought about a precision voltage reference, which can get down to 0.5% and even 0.1% error, but all seem to require more than 5v input.

What's the easiest and most accurate way to provide a precise 5v reference from a 3.3v board?

I would have thought that using 5V analog sensors on a 3.3V board is a bad idea as you will then be sending up to 5v to the analog inputs.

I use a simple voltage divider to scale the output signal of the sensor. And sometimes, some of my sensors never see the full range of it's operating specs, so it always outputs somewhere between 0 and 2.5. So either way, the ADC of the board will never see anything higher than ~3v. Like putting a water level sensor designed for 0-20 feet in a tank with a maximum capacity of 8 feet of water, the most the sensor will ever output is 2v.

I forgot to add that most of my sensors require only 5ma of excitation current, so I don't need a very beefy voltage reference.

Well I suppose the other question is how accurate is the 3.3V VCC on the arduino board? The boost converter I have tried gave me a pretty good 5V supply which was better than the 5V you normally get from the arduino regulator.

Maybe I just need to use a different boost converter. It would be great if that's all I need since it's a very simple solution. Any recommendations on a boost converter?

have you tried putting capacitors on the output of the converter?

weirdo557:
have you tried putting capacitors on the output of the converter?

Yep, and it didn't help, but maybe I didn't choose the right values or put them in the right place. Where should they go? Output to ground, and input to ground too, or just on the output?

As you are only now running the reference voltage at 3v3 any noise will appear to be bigger (more steps) that it was on 5V. You don't say what sort of sensors you have and how rapidly the signal changes but putting a capacitor across the A/D input will help. Also if your sensors have a high output impedance then taking two readings with a delay in between will also help.

Despite what you say about the sensor's range it is a bad idea and if it were me I would also add some input protection on the board, either catcher diodes or a zenner with a series resistor.

output of the converter?

You mean inputs to the converter.

i meant on the output to help smooth out the 5v rail, can you explain why they would have more benefit on the input Grumpy Mike?

I think we are at cross purposes.

By converter I meant the analogue to digital converter I think you meant the voltage converter. Adding extra capacitance on the 5V rail can help but not much in my experience. If it were that easy then the manufacturers would have done it already and got a better spec. But it is certainly something you could try.

I would be inclined to go with a switch converter up to something like 9v and then use a liner regulator to bring it down. There is much less noise like that.