adc accuracy

[renon2]

Posted on July 21, 2008 at 04:40

adc accuracy

[lanchon]

Posted on May 17, 2011 at 12:40

from what you say (I haven't looked at the specs) I wouldn't just trust the 350KOhm spec. that'd be picking and choosing, among conflicting specs, the ones that best suit your needs and wouldn't be a sound design strategy. either design to the least common denominator of specs, ask for support, or test yourself.

> the +-1uA current source I_L there. What is this source? it represents leakage current. you could double check this assumption by looking up this parameter on the specs. leakage varies greatly with process variation, so it may be difficult for you to test. leakage typically increases greatly with temperature. if you limit the upper working temperature of the die you can probably reduce the max leakage appreciably. if max leakage vs temp is not given on the specs, maybe tech support can give you typical values, which is better than nothing.

[renon2]

Posted on May 17, 2011 at 12:40

Hello,

in the STM32F103x Datasheet ( doc#13587 ) I have found ''Equation 1'' and ''Table 45'' (page 64) saying that R_aint could be even as high as 350Kohm for accurate (1/4 bit) measurement. As I intend to measure delicate, slowly changing 30V, I expect that my high imp. 2Meg/220Kohm divider will work well (e.a. with an error of a few LSB). But... I have also found the Figure 30 and the +-1uA current source “I_L� there. What is this source? How it influences the accuracy? Reading the Table 45 I have expected that the leakage has been taken into account and it was small enough to be neglected. Because if a current of 1uA would emerge via A_inX pin it would give an extra +-0.35V error on 350Kohm of R_ain :o . Yes, there is also R_adc (1Kohm) there that could almost cancel such an error (1K x 1uA = 1mV) but shouldn’t I omit this resistor at the DC analysis stage? What the STM want to suggest including this current source within the ADC model at Figure 30? Maybe it would be easier just to forget this current source and trust (give precedence) only Equation 1 and Table 45 ? Thank you very much for any help and response. [ This message was edited by: renon on 20-07-2008 11:19 ]

[obtronix]

Posted on May 17, 2011 at 12:40

Quote:

On 19-07-2008 at 23:31, Anonymous wrote: Hello, in the STM32F103x Datasheet ( doc#13587 ) I have found ''Equation 1'' and ''Table 45'' (page 64) saying that R_aint could be even as high as 350Kohm for accurate (1/4 bit) measurement. As I intend to measure delicate, slowly changing 30V, I expect that my high imp. 2Meg/220Kohm divider will work well (e.a. with an error of a few LSB). But... I have also found the Figure 30 and the +-1uA current source “I_L� there. What is this source? How it influences the accuracy? Reading the Table 45 I have expected that the leakage has been taken into account and it was small enough to be neglected. Because if a current of 1uA would emerge via A_inX pin it would give an extra +-0.35V error on 350Kohm of R_ain :o . Yes, there is also R_adc (1Kohm) there that could almost cancel such an error (1K x 1uA = 1mV) but shouldn’t I omit this resistor at the DC analysis stage? What the STM want to suggest including this current source within the ADC model at Figure 30? Maybe it would be easier just to forget this current source and trust (give precedence) only Equation 1 and Table 45 ? Thank you very much for any help and response. [ This message was edited by: renon on 20-07-2008 11:19 ] Yes there are two issues to consider for A/D accuracy, the charging of the sampling cap, which STM addressed, and the internal leakage, which STM didn't address. For a 1uA leakage the external source resistance has to be less then 366 ohms ((1/2^13)*3V/1uA) to get 1/2 LSB error out of 12 bits, 36.6 ohms at 10uA leakage. You need an external op amp, there is no other way to get around this problem. There is no magical SAR converter that can handle 350K of external resistance as the datasheet equations suggest.

[lanchon]

Posted on May 17, 2011 at 12:40

ok, I've taken a look now. table 34 says that max leakage for full temp range is +/- 1uA. but tables 47 & 48 seem to imply that max accuracy can be obtained with Rain < 10 kOhm, over full temp.

the two specs sort of conflict but in this case I'd trust the second. maybe the 1uA is overstated or does not apply in analog mode. (or maybe leakage has high frequency and its average over a sampling period is considerably smaller, but I doubt it.) if I were you, I'd aim for Rain < 5 kOhm if possible.