Go Down

Topic: Noise on Analog Inputs (Read 1 time) previous topic - next topic

bds

Hi everybody,
I am using the pduino firmware with Pure data, I am getting a lot of noise on the analog inputs.  
I have hooked up a bunch of different pots to the analog inputs including a couple of Spectrasymbol linear membrane pots.  
I have tried running both at different baud rates, and tried a variety of caps on the inputs from 0.001-0.1uF in an attempt to attenuate the noise.
I have also tried implementing filtering and smoothing in Pure data to no avail.
The noise is using up 8 to 13 percent of the potential signal!!!!    :o :P :P
Please tell me I can fix this....That this is not normal....AAAARGH!!!!
Can you suggest a cure?

:( :( :(
bds
Today at 10:52:02

Daniel

#1
Jun 18, 2007, 05:49 pm Last Edit: Jun 18, 2007, 08:25 pm by Daniel Reason: 1
hi

first, try a .1uf cap from the AREF pin to ground. That will filter the analog reference voltage.

Second, you should look at the circuit design, as 10% variation corresponds to .5v, which is way too high. The datasheet says something like 3 bits maximum error, which corresponds to 8/1023, which equals 40 millivolts. That's a .7% maximum error. If you are getting 5 or 10% error, there is something very wrong.  

So there is very likely something wrong with your circuitry. Is your power supply well filtered? Do you have noisy components like motors, relays or solenoids in the circuit? Do you have a filter cap near the voltage dividers for the sensors?

Are your sensors forming a proper voltage divider? With that kind of variation, it sounds like the analog input pins are floating at some point. A simple test would be to set up a fixed-resistor voltage divider on an analog pin, then measure the voltage you get repeatedly.  

D

bds

Thanks Daniel,
After following your suggestion I tested the arduino again and found it functioning well within spec. Seems that the sensors are the problem!! ::)
Your suggestion to put a cap on the AREF helped to halve the noise though....
I am now able to scale the input to a viable range to use the sensors, now it is apparent that the settling period of the sensors is not what I had expected and I will have to change my tack slightly in design.
Looks like its back to the drawing board......

Big Thanks :D
bds


Daniel

hey

no sweat... out of curiosity, what is up with the sensors? They look like nice sensors, and I'd like to use them in a project.

D

bds

I don't know if my beef is with the samples I've got or if the problem is simply the nature of the build. I will be in touch with spectrasymbol to find out more on the matter shortly.
The problem is threefold:
First off the sensor output seems to jitter/oscillate (for want of better description),  This means that the aquired data has to be scaled and/or filtered to a usable range. In this instance I have scaled to give a musically useful range of 12 intervals. Though to obtain a more discrete range one would need to attenuate the jitter somehow or simply have a much longer sensor strip. I am testing a 16mm sensor and a a gang of 3, 14mm sensors.
The second consideration is that at rest the sensors output ten percent potential and don't settle at zero resistance. Since their full resolution is still available if taken into account this 'can be' a positive advantage.
Finally and for my purposes the larger factor occurs after actuation of the sensor, the decay of the signal can take up to 5 seconds to return to a state of rest!! yet continuous control is fine.

So if I could get rid of some of the jitter and make more discrete sensor readings these small sensors would be great, though at present floating point resolution seems not so plausable.  Any suggestions that you might have would be appreciated.
bds

bds

Thought that I would share the good news that I have been getting near perfect results with my soft pots since I followed the advice on another thread that suggested pull down resistors!! ;D ;D ;D

phineus

Yes this helped me with the same problem - sorry I didn't see your post sooner.

Go Up