Email: Password: Remember Me | Create Account (Free)

Back to Subject List

Old thread has been locked -- no new posts accepted in this thread
???
07/09/09 09:01
Read: times


 
#166910 - Need lots of bits
Responding to: ???'s previous message
Remember that a 10-bit ADC gives about 0.1% steps at full range. If full range is just 0..1000 (about 10 bits) then it will give 1% steps for a signal just requiring a switch to the next higher range.

In your case, you say you always need 0.1% precision. That means that you will require better than 0.01% resolution. Better since all error sources (resolution, scaling, ...) must never get above 0.1% in any part of the scale. Even if you have a full range of 4.000 (12 bits) will after range shift give 0.400 which would be 0.25% large steps. The ADC must be better than 13 bits, or you will need to use smaller range changes than a factor 10.

Optimum result is actually a range change of a factor 2, which means that your measurement is always in the upper half of the range, except for the smallest range. The bad thing is that it is easier to get good resistors with a decade distance.

An alternative is to just reqiore 0.1% at full range, and live with 1% for values just outside the previous range.

List of 85 messages in thread
TopicAuthorDate
Calibrating resistors using AT89C51RD2            01/01/70 00:00      
   The problem is not manufacturing tolerances...            01/01/70 00:00      
      those low drift 0.01% resistors are too expensive.            01/01/70 00:00      
         Now, that you enter into a discussion...            01/01/70 00:00      
            more details            01/01/70 00:00      
               Drift is well documented            01/01/70 00:00      
                  No longer published in todays datasheets...            01/01/70 00:00      
                     thanks that was very useful            01/01/70 00:00      
         You can't calibrate without a reference!            01/01/70 00:00      
            I have a Hioki DMM as a reference            01/01/70 00:00      
               But you said auto-calibration            01/01/70 00:00      
                  No            01/01/70 00:00      
                     Need lots of bits            01/01/70 00:00      
                        good analysis            01/01/70 00:00      
                           what design?            01/01/70 00:00      
               how precise is that?            01/01/70 00:00      
                  50000 counts            01/01/70 00:00      
                     you are making the usual mistake of ....            01/01/70 00:00      
         not just more precise, but lower drift as well            01/01/70 00:00      
   Calibration            01/01/70 00:00      
      can I regularly calibrate?            01/01/70 00:00      
         many issues involved            01/01/70 00:00      
            Then...            01/01/70 00:00      
         Assuming and supposing...            01/01/70 00:00      
            Ok...            01/01/70 00:00      
               How fast? How many bits? How stable input signal?            01/01/70 00:00      
                  more details            01/01/70 00:00      
               More details needed...            01/01/70 00:00      
                  spikes of source could be a problem            01/01/70 00:00      
                     Hhm...            01/01/70 00:00      
                        Hard to self-calibrate            01/01/70 00:00      
                           high sampling rate is not for calibration            01/01/70 00:00      
                        It is a constant current source            01/01/70 00:00      
                     Answers and questions...            01/01/70 00:00      
                        Answers            01/01/70 00:00      
                           Oh, I think I know what it is...            01/01/70 00:00      
                              That's true            01/01/70 00:00      
                                 Difficult!            01/01/70 00:00      
                                    since this is about calibrating known resistors ....            01/01/70 00:00      
                                       Too fast for manual change            01/01/70 00:00      
                                    My solution            01/01/70 00:00      
                                       I have seen nothing yet that gives ....            01/01/70 00:00      
                                          How?            01/01/70 00:00      
                                             two screw terminals and a precision resistor            01/01/70 00:00      
                                       What are exactly the shunts you switch?            01/01/70 00:00      
                                          currently ...            01/01/70 00:00      
                                             You should test...            01/01/70 00:00      
                                                good idea...            01/01/70 00:00      
                                                   Well, I suspect...            01/01/70 00:00      
                                                      don't expect            01/01/70 00:00      
                                                         A look into datasheet is always important...            01/01/70 00:00      
                                                      Both problems            01/01/70 00:00      
                                             Advantages of shunts in series            01/01/70 00:00      
                                                Current scheme...            01/01/70 00:00      
                                                   How make before break?            01/01/70 00:00      
                                                      It is parallel like your second scheme.            01/01/70 00:00      
                                                         Offset voltages?            01/01/70 00:00      
                                                            Yes            01/01/70 00:00      
                                                               Try the following...            01/01/70 00:00      
                                                                  Wht's the difference?            01/01/70 00:00      
                                                                     You should, to find out what's wrong!            01/01/70 00:00      
                                                                        Ok but Isn't it easier to calibrate?            01/01/70 00:00      
                                                                           Yes, but...            01/01/70 00:00      
                                                                              Purpose of auto calibration            01/01/70 00:00      
                                                                                 No!            01/01/70 00:00      
                                                                                    Did you think about the shunt error during the "make"?            01/01/70 00:00      
                                                                                       disadvantages of series scheme            01/01/70 00:00      
                                                                                          Not really...            01/01/70 00:00      
                                                                                             0.1%            01/01/70 00:00      
                                                                                                Surprise!            01/01/70 00:00      
                                                                                                   voltage drop            01/01/70 00:00      
                                                                                                      So low?            01/01/70 00:00      
                                                                                                momma! I want it all, but refuse to pay            01/01/70 00:00      
                                                                                                   Wanting to achive a precision of 0.01%...            01/01/70 00:00      
                                                                                                      next he will make a perpetuum mobile            01/01/70 00:00      
                                                                                    how true and ...            01/01/70 00:00      
                                                                                       Precise enough            01/01/70 00:00      
                                                                                          that MAY be true for VOLTS            01/01/70 00:00      
                                                                                    OK, give me a few days time....            01/01/70 00:00      
                                                                                       Try to avoid it at all!            01/01/70 00:00      
                                                                                          You need a lot of individual precision for 0.1% all-range            01/01/70 00:00      
                                                                                             Owens            01/01/70 00:00      
                                                                                                Yes            01/01/70 00:00      
                                                                                             Good discussion but...            01/01/70 00:00      
   a hypothetical example            01/01/70 00:00      

Back to Subject List