So say we had a 16-bit A to D converter unsigned. And we had a 0.1% tolerance Vref part. What is the accuracy of the A to D converter? So we can take, we got 16 bits unsigned and we can take 2 to the 16th times tolerance and we get 6.554. So this means that any measurement that we take is plus or minus 3.277. And really, these digits over here probably don't have any meaning. We're just going to round-up to approximately 3.3 units. The actual temperature is going to be within 3.3 units away from the actual or correct voltage. So we say well, we don't want to spend $93 on a really expensive Vref part. And we don't require that much correctness. So we're going to go with a 0.2% tolerance Vref part that we can get for $0.40. So you're going to run the calculation again. 2 to the 16th times 0.02 is 131, or we're about 65 units from the actual correct value. What we read, what we interpret, what we see in our software That's reading this. I'm assuming there was a CPU involved probably in the system is going to go out and perform this read. We know the actual temperature is somewhere plus or minus 65 units away from the actual or correct value. It's just something to be aware of. This is a picture I have in my head when I think about this. So here's a lower tolerance Vref part, and here's Vref, the A to D reading all one, so it's at Vref and down here is Vmin, and we have this tolerance range. So any given sample, I called this the region of uncertainty. We know the actual value is in this region, we just don't know where. We know it's in there someplace. This is how I visualize that, go to a higher tolerance Vref part, and that range gets tighter. And the tighter the greater and greater, the smaller and smaller the tolerance, the more and more correctness in precision we have in the measurement. Our accuracy gets better.