TDC7200 calibration

Dear Vishy

Thank you for the reply.

Reading the calibration constants every measurement cycle takes too much time. I have to apply the calibration every measurement to correct the time reading, however I use two intermediate calculated constants, one a reciprocal of the other, in order to replace a  division with a multiplication.

  It occurs to me that I could average the calibration constants using a simple digital filter. This will reduce the variance of the calibration constants, and should also reduce the variance of the corrected measurement. This would be valid providing the calibration constants do not change too quickly. I am thinking of a filter time constant of not more than 100ms. Has this approach been considered?

regards Cosmo Little

If you have not already, please look at Data Sheet figures 6 and 7. Gives you an idea of the variation of resolution over temperature and VDD. 

thanks,

Vishy

Dear Vishy,

It seems from the data sheet that measurement variation with temperature will be negligible over a period under 1 second. As the calibration is two measurements with presumably the same random variation as a time measurement, averaging of the calibration constants may reduce the variation in the corrected measurement.

I will try it

regard

Cosmo Little

Dear Cosmo,

Could you please tell me how to "use two intermediate calculated constants, one a reciprocal of the other, in order to replace a  division with a multiplication"?

I am dealing with a similar problem in mode 1 now and I think a division with denominator = calibration2 - calibration1 is needed, but I don't know how to replace it with a multiplication. Thanks. 

Regards,

Eric

Please see the following two code fragments. The first uses the two calibration constants to correct the time value.

The second calculates the two constants from the TDC cal registers. Please note this code has only been tested on the simulator so far.

Cosmo Little

; /* read TDC7200 time to digital converter */
;
BCLR LATA, #7 ;chip select TDC A

MOV #0x1000, W0 ;SPI1BUF = 0x1000. read cycle to TDC reg TIME1
MOV W0, SPI1BUF ;start SPI
;

BCLR IFS0, #IC1IF ; clear input capture 1 interrupt flag

; Aevent_time.integerclockcycles = IC1BUF + IC2BUF * 2^16 ); get 32 bit clock counter value

MOV IC1BUF, W0
MOV IC2BUF, W1
MOV #_Aevent_data+2, W2
MOV.D W0, [W2] ;store 32 bit clock cycle count in Aevent_data.integerclockcycles
;
; wait for SPI

BTSS SPI1STAT, #SPIRBF
BRA $-2
;
; 2nd part of read cycle, first 16 bits dumped
CLR W0
MOV W0, SPI1BUF
; wait for SPI

BTSS SPI1STAT, #SPIRBF
BRA $-2
; get 16 bit uncalibrated TIME1 from TDC

MOV SPI1BUF, W0

; Correct TIME1 value using TDC calibration constants ( calculated in a different routine)

; Aevent_time.fractionalclockcycles = (TIME1 - Acalconstant)) * Arecipcalconstant >> 8

; A 16x16 multiply is performed. The 32 bit product is shifted 8 bits right ( /256)
; and the ls word is taken as the fraction cycle value. This must be complemented before storing
; as the time increment is measured to the next clock cycle, not the previous. The complemented value
; is stored in the first member of the struct Aevent_time, ie Aevent_time.fractionalclockcycles

MOV _Acalconstant, W1
SUB W0, W1, W0
BRA NN , $+4
CLR W0 ;if subtraction result is negative, zero result
MOV _Arecipcalconstant, W1
MUL.UU W0, W1, W0 ;32 bit result in W0, W1

; shift w0,w1 8 bits right, keeping lower word in w0

SL W1, #8, W2
LSR W0, #8, W0
IOR W0, W2, W0
LSR W1, #8, W1

BRA Z, $+4 ;test for multiplication overflow past 24 bits
SETM W0 ;set w0 to maximum fraction of FFFF

; fractional cycle count equals complement of W0

COM W0, W0

; store calibrated fractional clock cycles in Aevent_data.fractionalclockcycles

MOV W0, _Aevent_data

/* Function CALIBRATE_TDC */

void CALIBRATE_TDC (void)

{
static unsigned long lastPaceClock ;
unsigned long calibration1, calibration2;

if (PaceClock < ( lastPaceClock + CalibrateRate))
return ;

lastPaceClock = PaceClock;

if ( LATBbits.LATB5 == 1) /*channel A busy*/
return;
else
{

LATAbits.LATA7 = 0; /*chip select TDC A*/
SPI1BUF = 0x9C00; /* start read cycle from CALIBRATION1 register in TDC, auto increment on*/
/* do not use first 16 bit read ( 8 ms bits) */
while ( SPI1STATbits.SPIRBF == 0) /*wait for SPI*/
;
SPI1BUF = 0x0000; /* read ls 16 bits of CALIBRATION1 register */
while ( SPI1STATbits.SPIRBF == 0) /*wait for SPI*/
;
calibration1 = SPI1BUF;

SPI1BUF = 0x0000; /*read ms 8 bits of CALIBRATION2 register, but do not use */
while ( SPI1STATbits.SPIRBF == 0) /*wait for SPI*/
;
SPI1BUF = 0x0000; /* read ls 16bits from CALIBRATION2 */
while ( SPI1STATbits.SPIRBF == 0) /*wait for SPI*/
;
calibration2 = SPI1BUF;

LATAbits.LATA7 = 1; /*deselect TDC A*/

/* calculated the two calibration constants, Acalconstant and Arecipcalconstant */

Acalconstant = calibration2 - calibration1;
Arecipcalconstant = 0x1000000ul / Acalconstant;