Some microprocessors have enough jitter in their readings to use oversampling technique with the natural toggling of the least significant bit. The one-shot ADC readings of the thermistor bridge in purple are converted to ☌, as are 4096 sample readings at the default 125kHz(ps64) in grey, 250kHz(ps32) in orange and 500kHz (ps16) in green. With such a large number of samples, the averaging produces some smoothing whenever the raw ADC readings near a transition point, but if you see “rounded stair steps” like this then oversampling is not working properly – the curves shown above are all FAILURES. These are readings from a 10k NTC thermistor divider, and I’ve offset these records from each other by 0.1° for easier comparison. I usually over-sample one power more than needed for my target resolution, so I’d use for four extra bits to be sure of that order of magnitude improvement, which requires the sum of 4 4= 256 readings: Taking (4 6= 4096) samples would bump that up to ~0.0015☌ which, as the saying goes, is good enough for government work… With thermistor dividers, you typically get about 0.1☌ from the default ADC readings, and 128 samples bumps that to 0.012☌. Generally you have to add three extra bits (4 3= 128 samples) to see approximately an order of magnitude improvement in your real world resolution. And my gut feeling was that aref-based ADC bits were somehow better than any I could synthesize, but I was still curious to see if I could add over-sampled bits to the ones obtained with the bandgap trick.Īt first bounce, the method appeared to be incredibly simple, to get n extra bits of resolution, you need to read the ADC four to the power of n times. While I was figuring out how to read thermistors with our Arduino based data loggers, I came across claims that you can improve the resolution of any Analog-to-Digital converter with a technique called oversampling & decimation. I had already doubled the number of ADC bits covering my target temperature range by powering a thermistor divider from the rails and using the internal 1.1v as the analog reference. Otherwise thermal noise could make it look like your dithering is sufficient for oversampling, when it’s not. The slightest breeze makes glass bead thermistors jitterbug like crazy, so put them inside something with a decent amount of thermal inertia before you do any oversampling experiments.

Touch sensor output bouncing