Hi Dirk,

I wrote a custom Device Support for this record, so DTYP is set to my custom device. Indeed, it is the device support code that sets those parameters, but only in the init_record routine:

   epicsFloat64 div;

   pao->eguf = 10.0;

   pao->egul = (type == LCN12BMP || type == LCN16BMP) ? 0.0 : -10.0;

   div = (type == LCN12BMP || type == LCN12BBP) ? 4095.0 : 65535.0;

   pao->hopr = *eguf; 

   pao->lopr = *egul; 

   pao->eslo = ((pao->eguf) - (pao->egul)) / div;

After the initialization, those parameters are not changed. The read_ai routine only reads into the RVAL field.

   struct loconState *priv = pao->dpvt;

   epicsUInt16 value = 0;

   LOCON_read_analog(priv->address, priv->type, &value);

   pao->rval = (epicsInt32) value;

   return 0;

Bruno Seiva Martins

On Tue, Aug 28, 2012 at 10:12 AM, Dirk Zimoch <[email protected]> wrote:

Hi Bruno,

Many drivers overwrite EOFF and ESLO with their own calculations (from EGUL and EGUF and their own assumptions about the raw value range).

So the important missing information is: What driver are you using (DTYP)?

Dirk


Bruno Seiva Martins wrote:

Hi all,

I am having trouble trying to adjust the conversion parameters of an analog input record for a A/D converter.
My AD accepts voltages ranging from -10 to 10 volts and has a resolution of 12 bits.

For -10 volts, it produces a digital output of 0
For 10 volts, 4095

I configured the conversion fields as follows:

EGUF: 10
EGUL: -10
AOFF: 0
ASLO: 1
EOFF: 0
ESLO 0.004884   // which is (EGUF - EGUL) / 4095
ROFF: 0
LINR: LINEAR

Those values seem right to me but, for example, RVAL=1005 gets converted to VAL=4.9084, when it should be VAL=-5.0916. The problem seems to be that the conversion formula isn't taking into account the EGUL field.

According to the EPICS record reference for aiRecord:

    The record processing routine performs the following algorithm for
    all records except those that use the |Soft Channel| device support
    routine. Be aware that step 3 is performed only when the record
    specifies LINEAR:
    1. Val= RVAL + ROFF
    2. Val = Val * ASLO + AOFF

    If the conversion algorithm is |LINEAR|, the raw value is converted

    via the equation:
    3. val = val * ESLO + EGUL


Seems like the step 3 is not being completely executed, or considering EGUL = 0...


Does anybody know what is wrong? Probably I'm missing something really obvious, but I couldn't figure it out...


Thanks in advance,


Bruno Seiva Martins