| ??? 05/19/08 18:19 Read: times |
#154915 - I don't disagree ... Responding to: ???'s previous message |
Kai Klaas said:
Richard said:
Wouldn't you conclude that any imperfection in the supply and in the input signal would be added to that? It isn't clear what's exactly going on inside a microcontroller when an analog input signal is converted in the built-in ADC. There are some certain instances when the ADC result is especially sensitive to noise, might it come form power supply or ground bounce or etc. Many micros have something like a "quiet modus", which allows you to switch off the most noisy circuitry within the micro, which is the ports driving sections, for instance, especially when driving heavy capacitive loads. I for myself use external ADCs, put the micro during the conversion into idle mode and wake him up by a /INT pulse coming from ADC. By this I can use a fast 12 bit ADC without any detoriation of performance. I have heard many times, that the internal ADC flickers, even when all sources of noise are wrestled down. So, 1...2LSB flickering noise might be the best you can get with a micro that has the ADC built-in. Kai I think that the datasheet clearly states that 2 lsb is the maximum total unadjusted error, I'd believe that that's only under absolutely ideal conditions. The up-to two lsb's of offset error are "knowable" in that once one has put the device into a nearly-ideal situation, one can adjust out the error caused by that offset. One challenge is figuring out, in firmware, what that quantity is once the system is assembled as it is to be used. As you may have guessed, I'm not a big believer in internal ADC or DAC peripherals, for the simple reason that, unless there's an essentially separate circuit path for nearly everything analog within the device, there's really no way of knowing or of controlling everything that's necessary to make intelligent choices about what's random error and what's system error and likely to become cumulative. Most internal converters of either sort are too slow for what I normally want to do with a converter, so I, too, prefer external ones. I don't know what Robert's application for this pot that he's measuring is, nor do I know how precise or accurate his readings must be. Since the ADC can be as much as 2 bits off the real value, and since the random error is apparently two bits as well, it's not likely he'll get flicker-free readings anywhere, depending, of course, on where the wiper of his pot sits with respect to full scale. It does look, however, as though as much as three bits of his readings will be partially left to chance. By their very nature, dual-slope integrating and SAR ADC's have that uncertainty in their LSB. This is the first one I've encountered that has two such bits of uncertainty. There are probably others, though. I didn't see anything that suggests that one can determine in any way what the voltage offset is, short of actually measuring it. With a two-bit uncertainty due to system noise, this might not be easy to test out. I'm thinking that it probably has some consistency within a given lot, however, so one doomed to using this part in a production situation could sample a few parts from the same lot and make an educated guess as to what the typical offset magnitude is. Perhaps, if he's just wanting to set volume or some other quantity, he can use a quadrature timing wheel rather than a pot. Then he can just count pulses. That has less uncertainty. RE |
