Our full technical support staff does not monitor this forum. If you need assistance from a member of our staff, please submit your question from the Ask a Question page.


Log in or register to post/reply in the forum.

FLOAT4 resolution for BrHalf4W PT100 measurements


LGS May 6, 2022 08:17 AM

Hi,

We have a CR1000 measuring PT100 using 4 wires half bridge and saving the data in FLOAT format. We are using a CS multiplexer.

The code is (relevant parts):

  Const Slow_Record_Interval = 5 'in sec


Public TEMP

...

DataTable(Slow,True,-1)
DataInterval(0,Slow_Record_Interval,Sec,Slow_Record_Interval)

CardOut(0,-1)

...

Sample(1, TEMP, Float)

...

PulsePort(6,5000)
BrHalf4W(TEMP, 1, mV25C, mV25C, 3, Vx1, 1, 2200, True, False, 0, _50Hz, 1, 0)

If a create an histogram of the absolute difference between two adjacent value in time, I got that (one day of data, air temperature so range is close to 1 explicity [1.004756,1.030377]):

  dif    | count  
----------+-------
0.000000 | 11519
0.000001 |    86
0.000002 |    73
0.000003 |    26
0.000004 |    44
0.000005 |    11
0.000149 |     4
0.000150 |     3
0.000151 |     3
0.000152 |     4
0.000153 |  1216
0.000154 |  1200
0.000155 |  1418
0.000156 |   483
0.000157 |   403
0.000158 |   254
0.000160 |     1
0.000162 |     1
0.000306 |    28
0.000307 |    90
0.000308 |   115
0.000309 |   143
0.000310 |    43
0.000311 |    68
0.000312 |     5
0.000313 |     1
0.000314 |     3
0.000315 |    10
0.000316 |     3
0.000462 |     9
0.000463 |     3
0.000464 |     6
0.000465 |     3

It looks like that values around 0.000155 (and multiples of it) are by far the most occuring. But this is not expected for 32 bits floating point values. ~7 digits resolution was expected and some of them are present (but always around a multiple of 0.000155).

From a precision point of view, this is not an issue but if the noise of the measurement is monitored, the resolution should be known. Could you confirm that the real resolution is ~5 digits and not ~7 digits?

Regards


nsw May 9, 2022 10:51 AM

The CR1000 A/D resolution is 13 bits. The internal core processing is 32 bit.

As you are storing the value as measured (i.e. un-processed data), you are seeing the 13 bit resolution, i think. 

If you were processing the values, say storing averages over a minute, you should then see the 32 bit resolution working as you expect.

Log in or register to post/reply in the forum.