Re: aiRecord conversion

[Dirk Zimoch <dirk.zimoch@psi.ch>]

Hi Bruno,

Your device support should have a function specialLinconv() or similar, 
which is function number 6 in the device support structure:

struct {
    long      number;
    DEVSUPFUN report;
    DEVSUPFUN init;
    DEVSUPFUN init_record;
    DEVSUPFUN get_ioint_info;
    DEVSUPFUN read;
    DEVSUPFUN special_linconv;
} myAi =
{
    6,
    NULL,
    NULL,
    initRecordAi,
    getInIntInfoAi,
    readAi,
    specialLinconv
};

That function may look like this:
STATIC long specialLinconv(aiRecord *record, int after)
{
    epicsUInt32 hwSpan;
    myPrivate_t *priv = (myPrivate_t *)record->dpvt;

    if (after) {
        hwSpan = (priv->type == LCN12BMP || priv->type == LCN12BBP) ?
		4095 : 65535;
        record->eslo = (record->eguf - record->egul) / hwSpan;
        record->eoff = record->egul;
    }
    return 0;
}

Call specialLinconv() from initRecordAi(). Additionally, the record will 
call it automatically when anyone changes EGUL or EGUF.

In initRecordAi(), malloc a private data structure (I called it 
myPrivate_t) with a field "type" (and anything else you need to store 
and use later, for example in readAi). Set type and then store the 
pointer to the structure in record->dpvt before calling specialLinconv().

In the record database, set EGUF to the user value that corresponds to 
the maximum ADC value (4095 or 65535 depending on your ADC type) and 
EGUL to the value that corresponds to the minimum ACC value (0). For 
example 10 and -10 (Volts). You may change that later on the fly and 
specialLinconv will adjust ESLO and EOFF.

Dirk


Bruno Seiva Martins wrote:
> Hi Ralph,
>
> Thanks for your advice! I'm still learning EPICS, so I'm using those 
> A/D's just for experimenting. When it comes to the real deal, I'll keep 
> in mind what you said.
>
> Best regards,
>
>
> Bruno Seiva Martins
>
>
>
>
> On Tue, Aug 28, 2012 at 11:05 AM, Ralph Lange <Ralph.Lange@gmx.de 
> <mailto:Ralph.Lange@gmx.de>> wrote:
>
>     On Tue Aug 28 2012 15:28:34 GMT+0200 (CEST), Bruno Seiva Martins
>     <brunoseivam@gmail.com> <mailto:brunoseivam@gmail.com> wrote:
> >     Rectifying:
> >
> >        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 = pao->eguf; 
> >        pao->lopr = pao->egul; 
> >        pao->eslo = ((pao->eguf) - (pao->egul)) / div;
> >
> >
> >     Bruno Seiva Martins
>
>     Another comment:
>
>     The linear conversion mechanism is intended to convert the digitized
>     value into the final engineering units.
>
>     You might want to use your A/D converter for signals that correspond
>     to positions [mm], pressures [mbar], or temperatures [C].
>     In those cases, the database designer will set EGU, EGUL and EGUF to
>     e.g. "mm", -5 and 25, meaning that -10V are equivalent to -5mm, and
>     +10V are equivalent to +25mm. To limit the range that is shown in
>     the GUI, the designer might also want to set e.g. LOPR to -2 and
>     HOPR to +2, so that graphs and meters show the "interesting" range,
>     -2mm to +2mm.
>
>     To have this work correctly, setting EGUF, EGUL, HOPR and LOPR in
>     your device support may actually not be a good idea.
>     Normally, I would just set ESLO and EOFF.
>
>     Cheers,
>     ~Ralph