In=20
my previous post, we talked about three =
topologies for=20
generating two high-accuracy voltage reference =
outputs. Today,=20
we=E2=80=99ll compare the performance of these solutions from three =
perspectives: total=20
error, drift tracking and matching between outputs.
Total=20
error
Equation=20
(1) converts specifications given as a percentage (%) to parts per =
million=20
(ppm).
(Equation=20
1)
The Total=20
Error performance metric of each voltage output depends on its initial =
accuracy=20
and drift over the operating temperature range, as given in Equation=20
(2).
(Equation=20
2)
In=20
Solution 1, as no typical drift for the LM4140B is given in the datasheet, we =
use the=20
maximum drift specification, over a 70=C2=BAC temperature range for =
calculation. In=20
Solution 2, the bias voltage (VBIAS) is generated from REF5030A,=20
resistor network and a buffer. Therefore, the initial accuracy and drift =
can be=20
expressed as the RSS of these three error sources, as given in Equation =
(1) in=20
Part 1. Since the REF2030 and REF5030A use the=20
box =
method to=20
determine drift, the temperature range for calculations is the entire =
operating=20
range, or 165=C2=BAC.
Table 1=20
shows that while the performance of VREF in Solution 1 is the =
same as=20
Solution 2, its VBIAS output has considerably more error. =
Note that=20
the error for VBIAS in Solution 2 includes the error from=20
VREF. With high initial accuracy and low temperature drift on =
both=20
outputs, Solution 3 has the lowest error of the three solutions. =
Table=20
1: Comparison of error contributions for each output =
voltage
Drift=20
tracking and matching
Another=20
important specification for this dual output system is drift tracking, =
which=20
describes the accuracy matching between two voltages over a particular=20
temperature range, as given by Equation (3). Figure 1 shows the typical =
drift=20
tracking performance of the REF2030.=20
(Equation=20
3)
Figure=20
1: VREF and VBIAS tracking vs. =
temperature
Since we=20
applied two independent voltage references in Solution 1, theoretically =
the two=20
references may not directly track each other, so the tracking is the RSS =
value=20
of their maximum temperature drifts (11 ppm/=C2=BAC). As the LM4140B=20
is only specified from 0=C2=B0C to 70=C2=BAC, this drift tracking only =
applies for this=20
temperature range.
In=20
Solution 2, since error in VREF is common to both outputs, =
the drift=20
tracking (=CE=B4Tracking) between VREF and =
VBIAS=20
only depends on drift from resistor network (=CE=B4RES) and =
buffer=20
(=CE=B4BUF), as given by Equation (4).
(Equation=20
4)
Given the=20
initial accuracy error, we can also calculate the matching (at =
25=C2=BAC) of the=20
outputs in terms of RSS, as shown in Equation (5).
Table 2=20
shows a comparison summary. Note that the drift tracking and output =
matching in=20
Solution 2 heavily depend on the precision of the resistors. While the =
tracking=20
of the two outputs in Solution 2 is slightly better, the matching of the =
outputs=20
is much worse than that in Solution 3. Actually, Solution 3 is about 900 =
ppm=20
better. This means, with only a 2 ppm/=C2=B0C difference in drift, it =
will take a=20
450=C2=B0C of temperature shift before Solution 2 becomes the more =
accurate solution.=20
For details on calculation, please refer to data.xlsx.
Table=20
2: Comparison of output matching and drift tracking
From our=20
calculations we know Solution 3 has the best overall performance in most =
cases.=20
However, in reality, analog engineers have to consider more than =
performance.=20
Stay tuned for next week, when we compares the three solutions with =
respect to=20
PCB space and cost.
Related=20
resources: