Most modern signal-conditioning systems calibrate errors at room temperature. However, calibration of errors that result from a change in temperature is normally difficult and costly. Therefore, minimize these errors by choosing high-precision components, such as the INA819, that have improved specifications in critical areas that impact the precision of the overall system. Figure 8-10 shows an example application.

Figure 8-10 Example Application with G = 10 V/V and 1-V Output Voltage

Resistor-adjustable devices (such as the INA819) show the lowest gain error in G = 1 because of the inherently well-matched drift of the internal resistors of the differential amplifier. At gains greater than 1 (for instance, G = 10 V/V or G = 100 V/V), the gain error becomes a significant error source because of the contribution of the resistor drift of the 25-kΩ feedback resistors in conjunction with the external gain resistor. Except for very high gain applications, the gain drift is by far the largest error contributor compared to other drift errors, such as offset drift.

The INA819 offers excellent gain error over temperature for both G > 1 and G = 1 (no external gain resistor). Table 8-4 summarizes the major error sources in common INA applications and compares the three cases of G = 1 (no external resistor) and G = 10 (5.49-kΩ external resistor) and G = 100 (511-Ω external resistor). All calculations are assuming an output voltage of V_OUT = 1 V. Thus, the input signal V_DIFF (given by V_DIFF= V_OUT/G) exhibits smaller and smaller amplitudes with increasing gain G. In this example, V_DIFF = 1 mV at G = 1000. All calculations refer the error to the input for easy comparison and system evaluation. As Table 8-4 shows, errors generated by the input stage (such as input offset voltage) are more dominant at higher gain, while the effects of output stage are suppressed because they are divided by the gain when referring them back to the input. The gain error and gain drift error are much more significant for gains greater than 1 because of the contribution of the resistor drift of the 25-kΩ feedback resistors in conjunction with the external gain resistor. In most applications, static errors (absolute accuracy errors) can readily be removed during calibration in production, while the drift errors are the key factors limiting overall system performance.

Table 8-4 Error Calculation

ERROR SOURCE	ERROR CALCULATION	INA819 VALUES
		SPECIFICATION	UNIT	G = 1 ERROR (ppm)	G = 100 ERROR (ppm)	G = 1000 ERROR (ppm)
ABSOLUTE ACCURACY AT 25°C
Input offset voltage	VOSI / VDIFF	35	µV	35	350	3500
Output offset voltage	VOSO / (G × VDIFF)	300	µV	300	300	300
Input offset current	IOS × maximum (RS+, RS–) / VDIFF	0.5	nA	1	5	50
CMRR (min)	VCM / (10CMRR/20 × VDIFF)	90 (G = 1), 110 (G = 10), 130 (G = 100)	dB	316	316	316
PSRR (min)	(VCC – VS)/ (10PSRR/20 × VDIFF)	110 (G = 1), 114 (G = 10), 130 (G = 100)	dB	3	20	32
Gain error from INA (max)	GE(%) × 104	0.02 (G = 1), 0.15 (G = 10, 100)	%	200	1500	1500
Gain error from external resistor RG (max)	GE(%) × 104	0.01	%	100	100	100
Total absolute accuracy error (ppm) at 25°C, worst case	sum of all errors	—	—	955	2591	5798
Total absolute accuracy error (ppm) at 25°C, average	rms sum of all errors	—	—	491	1604	3835
DRIFT TO 105°C
Gain drift from INA (max)	GTC × (TA – 25)	5 (G = 1), 35 (G = 10, 100)	ppm/°C	400	2800	2800
Gain drift from external resistor RG (max)	GTC × (TA – 25)	10	ppm/°C	800	800	800
Input offset voltage drift (max)	(VOSI_TC / VDIFF) × (TA – 25)	0.4	µV/°C	32	320	3200
Output offset voltage drift	[VOSO_TC / ( G × VDIFF)] × (TA – 25)	5	µV/°C	400	400	400
Offset current drift	IOS_TC × maximum (RS+, RS–) × (TA – 25) / VDIFF	20	pA/°C	2	16	160
Total drift error to 105°C (ppm), worst case	sum of all errors	—	—	1634	4336	7360
Total drift error to 105°C (ppm), typical	rms sum of all errors	—	—	980	2957	4348
RESOLUTION
Gain nonlinearity		10 (G = 1, 10), 15 (G = 100)	ppm of FS	10	10	15
Voltage noise (at 1 kHz)		eNI = 8, eNO = 90	µVPP	1204	1070	3941
Current noise (at 1kHz)	IN × maximum (RS+, RS–) × √BW / VDIFF	0.13	pA/√ Hz	0.3	2	11
Total resolution error (ppm), worst case	sum of all errors	—	—	1214	1080	3956
Total resolution error (ppm), typical	rms sum of all errors	—	—	1204	1070	3941
TOTAL ERROR
Total error (ppm), worst case	sum of all errors	—	—	3802	8007	17113
Total error (ppm), typical	rms sum of all errors	—	—	1628	3530	7010