U.S. patent application number 11/901355 was filed with the patent office on 2009-03-19 for circuit and method for gain error correction in adc.
This patent application is currently assigned to Texas Instruments Incorporated. Invention is credited to Jerry L. Doorenbos, Dimitar T. Trifonov.
Application Number | 20090073011 11/901355 |
Document ID | / |
Family ID | 40364627 |
Filed Date | 2009-03-19 |
United States Patent
Application |
20090073011 |
Kind Code |
A1 |
Trifonov; Dimitar T. ; et
al. |
March 19, 2009 |
CIRCUIT AND METHOD FOR GAIN ERROR CORRECTION IN ADC
Abstract
Gain errors are corrected in an ADC chip including an integrator
(17), a comparator (30), and a digital filter (37) by storing a
gain-adjusted LSB size based on measured gain error in a memory
(44). The gain-adjusted LSB size is applied to the digital filter
to cause gain-adjusted LSB size values to be added to or subtracted
from accumulated content of the digital filter in accordance with a
first or second state, respectively, of the comparator (30) during
each cycle of the ADC. The final accumulated content after all
required cycles of the ADC is a gain-corrected digital output
signal (Dout(gain-corrected)).
Inventors: |
Trifonov; Dimitar T.; (Vail,
AZ) ; Doorenbos; Jerry L.; (Tucson, AZ) |
Correspondence
Address: |
TEXAS INSTRUMENTS INCORPORATED
P O BOX 655474, M/S 3999
DALLAS
TX
75265
US
|
Assignee: |
Texas Instruments
Incorporated
|
Family ID: |
40364627 |
Appl. No.: |
11/901355 |
Filed: |
September 17, 2007 |
Current U.S.
Class: |
341/118 |
Current CPC
Class: |
H03M 3/382 20130101;
H03M 1/1019 20130101; H03M 1/40 20130101; H03M 3/456 20130101; H03M
3/43 20130101; H03M 1/46 20130101 |
Class at
Publication: |
341/118 |
International
Class: |
H03M 1/06 20060101
H03M001/06 |
Claims
1. An analog to digital converter (ADC) comprising: (a) an
integrator for sampling an input signal, having an input coupled to
receive an input signal and also having an output; (b) a comparator
having a signal input coupled to the output of the integrator, and
an output representing a present state of the comparator depending
on a relationship of an output voltage on the output of the
integrator to a threshold voltage; (c) a digital filter having an
input coupled to the output of the comparator and also having an
output for conducting a gain-corrected digital output signal
representative of the input signal; (d) a memory for storing an
externally determined gain-adjusted LSB size number; and (e)
circuitry coupled to the memory for applying an integral number of
the gain-adjusted LSB size numbers to the input of the digital
filter to cause the integral number of gain-adjusted LSB size
numbers to be applied to content of the digital filter in
accordance with the present state of the comparator to accumulate a
value in the digital filter during successive cycles of the ADC so
as to generate the gain-corrected digital output signal.
2. The ADC of claim 1 wherein the comparator is a window comparator
having a threshold input which receives a plurality of threshold
voltages, wherein the window comparator produces a first state
which is a +1 state, a second state which is a -1 state, and
wherein the window comparator produces a third state which is a 0
state and which results in a zero input to the digital filter,
wherein the integral number of gain-adjusted LSB size numbers is
applied to an increment input of the digital filter in accordance
with the +1 state of the comparator so as to cause the integral
number of gain-adjusted LSB size numbers to be added to the
accumulated content of the digital filter, and wherein the integral
number of gain-adjusted LSB size numbers is applied to a decrement
input of the digital filter in accordance with the -1 state of the
comparator so as to cause the integral number of gain-adjusted LSB
size numbers to be subtracted from the accumulated content of the
digital filter.
3. The ADC of claim 1 wherein the integrator and the comparator
form a delta-sigma modulator.
4. The ADC of claim 3 wherein the delta sigma modulator is a first
order delta-sigma modulator.
5. The ADC of claim 1, wherein the ADC has a mixed
delta-sigma/cyclic architecture.
6. The ADC of claim 4 wherein the digital filter is configured as
an accumulator.
7. The ADC of claim 6 wherein the accumulator is configured as an
up-down counter the content of which is incremented or decremented
in the amount of the integral number of gain-adjusted LSB size
numbers.
8. The ADC of claim 1 wherein the digital filter includes an SAR
(successive approximation register) result register, an output of
which provides the gain-corrected digital output.
9. The ADC of claim 1 wherein the integrator includes an
operational amplifier having a non-inverting input coupled by a
first sampling capacitor to receive a first input signal and also
coupled by a first integrating capacitor to an inverting output of
the operational amplifier, the operational amplifier also having an
inverting input coupled by a second sampling capacitor to receive a
second input signal and also coupled by a second integrating
capacitor to a non-inverting output of the operational
amplifier.
10. The ADC of claim 9 wherein the non-inverting input of the
operational amplifier is coupled by a third sampling capacitor to a
first reference voltage, a second reference voltage, or is
unconnected to any reference voltage in accordance with an output
state of the comparator, and wherein the inverting input of the
operational amplifier is coupled by a fourth sampling capacitor to
the first reference voltage, the second reference voltage, or is
unconnected to any reference voltage in accordance with that output
state of the comparator, to prevent overflow of a delta-sigma
modulator.
11. The ADC of claim 10 wherein the input signal is equal to the
first input signal minus the second input signal.
12. The ADC of claim 1 wherein the integral number is 1.
13. The ADC of claim 1 wherein the memory is one of the group
including a fuse link memory, a one time programmable memory, and
an electrically erasable programmable memory.
14. The ADC of claim 1 wherein the memory includes means for
receiving the gain-adjusted LSB size number from an external test
and trim system.
15. The ADC of claim 1 wherein the memory and the output of the
comparator are coupled to an increment input of the digital filter
by a first ANDing circuit and are coupled to a decrement input of
the digital filter by a second ANDing circuit, wherein the first
ANDing circuit effectuates adding of the integral number of
gain-adjusted LSB size numbers from the memory to the content of
the digital filter in accordance with a first state of the
comparator, and wherein the second ANDing circuit effectuates
subtracting of the integral number of gain-adjusted LSB size
numbers from the memory to the decrement input in accordance with a
second state of the comparator.
16. A method for operating an analog to digital converter which
includes 1. an integrator having an input coupled to receive an
input signal and also having an output, 2. a comparator having a
signal input coupled to the output of the integrator, and an
output, the output of the comparator representing a first state or
a second state, depending on a relationship of an output voltage on
the output of the integrator to a threshold voltage, and 3. a
digital filter having an input coupled to the output of the
comparator, the method comprising: (a) storing an externally
determined gain-adjusted LSB size number in a memory; (b) applying
the gain-adjusted LSB size number to the input of the digital
filter to cause an integral number of gain-adjusted LSB size
numbers to be applied to content of the digital filter in
accordance with a present state of the comparator so as to
accumulate a value in the digital filter; and (c) after all
required cycles of the analog to digital converter, providing the
accumulated value at an output of the digital filter as a
gain-corrected digital output signal that represents the input
signal.
17. The method of claim 16 including adding the integral number of
gain-adjusted LSB size numbers from the memory to the content of
the digital filter in accordance with a first state of the
comparator, and subtracting the integral number of gain-adjusted
LSB size numbers from the memory in accordance with a second state
of the comparator.
18. The method of claim 16 including operating the integrator and
the comparator as a delta-sigma modulator and operating the digital
filter as an up-down counter the content of which is incremented or
decremented by the integral number of gain-adjusted LSB size
numbers.
19. The method of claim 16 wherein the ADC has a mixed
delta-sigma/cyclic architecture and the method includes providing
an output of a SAR result register as the gain-corrected digital
output.
20. The method of claim 16 including providing additional bits of
resolution in an analog to digital conversion process such that no
missing code appears during generation of a smaller number of bits
which are utilized as the gain-corrected digital output signal.
21. An analog to digital converter comprising: (a) an integrator
having an input coupled to receive an input signal and also having
an output; (b) a comparator having a signal input coupled to the
output of the integrator, and an output, the output representing a
first state or a second state, depending on a relationship of an
output voltage on the output of the integrator to a threshold
voltage; (c) a digital filter having an input coupled to the output
of the comparator; (d) means for storing an externally determined
gain-adjusted LSB size number in a memory; and (e) means for
applying the gain-adjusted LSB size number to content of the
digital filter so as to cause an integral number of gain-adjusted
LSB size numbers to be applied to content of the digital filter in
response to a present state of the comparator so as to accumulate a
gain-corrected digital value in the digital filter that, the after
all required cycles of the analog-to-digital converter, is a
gain-corrected digital value that represents the input signal.
Description
BACKGROUND OF THE INVENTION
[0001] The present invention relates generally to integrated
circuit ADCs (analog to digital converters), and particularly to
ADC circuits and methods for correcting the gain of an ADC with
minimum circuit complexity, a minimum amount of integrated circuit
chip area, and a minimum amount of time required in addition to the
basic ADC conversion time in order to accomplish the gain error
correction.
[0002] Every ADC has gain error. There are two basic ways to
correct the ADC gain error, namely analog correction and digital
correction. One analog gain error correction technique is trimming
of the reference voltage. That controls the transfer function of
the converter so as to achieve whatever correction is needed for
its gain. Another analog ADC gain correction technique is scaling
the ratios of the sampling capacitors of the integration stage of
the ADC. However, the known analog techniques for correcting ADC
gain error require increased amounts of integrated circuit die
area, and they typically require use of costly laser trimming
techniques. Furthermore, the gain correction accuracy which can be
achieved using analog ADC gain error correction techniques is less
than can be achieved using digital ADC gain error correction
techniques. Also, digital gain error correction techniques are
easier to implement during final integrated circuit testing
procedures than the analog gain error correction techniques.
[0003] Several digital methods for trimming ADC gain error have
been employed. FIG. 1 illustrates a prior art delta-sigma ADC 20A
which includes an integrator 17, the (-) output 27A of which is
coupled to the (-) input of a window comparator 30. The (+) output
27B of integrator 17 is coupled to the (+) input of window
comparator 30. Window comparator 30 may be composed of two or more
conventional comparators which receive various threshold voltages
V.sub.TH on the various conductors of window comparator threshold
bus 29, respectively. An input signal voltage
Vin=Vin.sup.+-Vin.sup.- is coupled between plates of sampling
capacitors 23A and 23B, the other plates of which are connected to
the (+) and (-) inputs, respectively, of an operational amplifier
18 of integrator 17. The (+) input conductor 26A of operational
amplifier 18 is coupled by an integrating capacitor 24A to the (-)
output conductor 27A of operational amplifier 18, and the (-) input
conductor 26B of operational amplifier 18 is coupled by an
integrating capacitor 24B to the (+) output conductor 27B of
operational amplifier 18. Input conductor 26A is coupled by a
sampling capacitor 28A to the pole of a single pole, triple throw
switch circuit S1, and input conductor 26B is coupled by a sampling
capacitor 28B to the pole of a single pole, triple throw switch
circuit S2. Switch circuit S1 couples the (+) input of operational
amplifier 18 through sampling capacitor 28A to Vref.sup.+,
Vref.sup.- or an open terminal, depending upon whether the output
CompOut of window comparator 30 is +1, -1, or 0, respectively, so
the reference Vref can be integrated in the correct direction if
necessary, depending on the decision of window comparator 30.
Similarly, switch circuit S2 couples the (-) input of operational
amplifier 18 through sampling capacitor 28B to Vref.sup.-,
Vref.sup.+, or an open terminal, depending upon whether CompOut is
+1, -1, or 0, respectively, so Vref can be integrated in the
correct direction if necessary.
[0004] The output conductors 31 of window comparator 30 are coupled
to the input of a digital filter 37, the output conductors 40 of
which produce the basic un-corrected digital representation Dout of
the input voltage Vin. The uncorrected signal Dout is multiplied by
a digital gain correction coefficient by means of a digital
multiplier 38 to produce a gain-corrected digital output signal
Dout(gain-corrected).
[0005] FIG. 2 is a flowchart which indicates the operation of ADC
20A of FIG. 1. The part of FIG. 2 within dashed line A generally
indicates how the delta-sigma modulator 15 consisting of integrator
17, window comparator 30, and switch circuits S1 and S2 operate in
conjunction with window comparator 30 so as to cause the
accumulation of new values in digital filter 37. As indicated in
block 1 of FIG. 2, integrator 17 integrates only the input voltage
Vin during the first integration cycle of delta-sigma modulator 15,
since initially window comparator output CompOut is 0. Then, as
indicated in block 2, window comparator 30 determines whether the
output of integrator 17 is greater than a threshold voltage
V.sub.TH.sup.+, less than V.sub.TH.sup.-, or between them.
Depending on the results of this comparison, the output of window
comparator 30 is "1", "-1", or "0", respectively. The threshold
voltages V.sub.TH.sup.+ and V.sub.TH.sup.- can be derived from Vref
and the scaling of various capacitors in integrator 17 so as to
provide the various desired window comparator threshold voltages
V.sub.TH (not shown) of the standard comparators of which window
comparator 30 is comprised.
[0006] In any case, window comparator output CompOut is coupled
into and added to the accumulated result in digital filter 37. If
modulator 15 is a first order delta-sigma modulator as shown in
FIG. 1, digital filter 37 can be implemented as an up-down counter,
i.e., as an accumulator. For first-order delta-sigma modulator 15
as shown in FIG. 1, the up-down counter or accumulator of digital
filter 37 is incremented by 1 if the output of window comparator 30
is +1 as indicated in decision block 3 and block 4, is decremented
by 1 if the output of window comparator 30 is -1 as indicated in
decision block 5 and block 6, or remains unchanged if the output of
window comparator 30 is "0" as indicated in block 7 of FIG. 2.
[0007] Then, as indicated in decision block 9, the process being
performed in ADC 20A determines if all of the integration cycles
required by ADC 20A have been performed. If this determination is
negative, the integrate and compare loop including blocks 1-9 is
repeated as needed to complete the basic analog-to-digital
conversion. On the second and each following required integration
cycle after the second, integrator 17 samples not only the input
voltage Vin, but also samples the reference voltage value
Vref.sup.+ or Vref.sup.- in accordance with window comparator
output CompOut. In the case wherein CompOut is too high, i.e., the
input to window comparator 30 is greater than the upper threshold
voltage Vref.sup.-, window comparator 30 makes a decision to
generate a +1 output level, and during the next integration cycle
integrator 17 samples the lower reference voltage value
V.sub.TH.sup.- in addition to sampling the input voltage Vin, and
this causes the output of integrator 17 to decrease. Similarly, in
the case wherein CompOut is too low, i.e., the input to window
comparator 30 is less than the lower threshold voltage
V.sub.TH.sup.-, window comparator 30 makes a decision to generate a
-1 output level, and during the next integration cycle integrator
17 samples the upper reference voltage value V.sub.Ref.sup.+ in
addition to sampling the input voltage Vin, and this causes the
output of integrator 17 to increase.
[0008] After the process of sampling Vin and Vref the required
numbers of times, the determination of decision block 9 eventually
is affirmative. The window comparator output CompOut then is
coupled for the last time into digital filter 37, which then
produces the uncorrected digital output Dout on bus 40, and the
basic analog-to-digital conversion is complete.
[0009] After the basic conversion of Vin to the uncorrected digital
output signal Dout on bus 40 has been completed, it is multiplied
by a gain error trim coefficient to produce the corrected final
digital output Dout(gain-corrected) of ADC 20A, as indicated in
blocks 11, 12, and 13 in the flow chart of FIG. 2.
[0010] As an example, assume that an input voltage Vin of 1.0 volt
and a reference voltage Vref are applied to ADC 20A of FIG. 1.
Assume also that the analog-to-digital conversion of the applied
1.0 volt input signal is performed, and the result is not the
desired 1.0 volt digital output voltage value, but instead is a 0.9
volt digital output voltage value because of gain error of ADC 20A.
One way to correct the gain error in the digital domain is to
determine the value of a digital gain correction coefficient, which
in this simplified example is approximately 1.1, and store it in
the ADC integrated circuit die. Then the digital output conversion
value of 0.9 volt produced by digital filter 37 on bus 40 is
automatically multiplied by the stored digital gain error trim
coefficient of approximately 1.1. In this example of Vin being
equal to 1.0 volt, the multiplication result is a corrected output
value of Dout very close to 1 volt (actually, 0.99 volts). From
then on, every time an analog to digital conversion of Vin is
performed, the resulting digital output Dout on bus 40 is
multiplied by the digital gain error trim coefficient 1.1 (in this
simplified example) to thereby obtain a corrected digital output
voltage Dout(gain-corrected).
[0011] A problem with the above described prior art technique is
that implementing a digital multiplier is always expensive because
it requires complex circuitry and a large amount of integrated
circuit die area, and also because it requires a large amount of
quiescent current and hence a large amount of power dissipation.
The above described prior art technique also is very
time-consuming, because the analog-to-digital conversion must be
performed first, and then the slow digital multiplication of the
ADC conversion result must be multiplied, bit by bit, by the
digital error correction coefficient, which adds a substantial
amount to the time required for the basic ADC conversion process in
order to obtain a gain-error-corrected digital output value
Dout(gain-corrected) which accurately represents Vin. An additional
drawback of this digital multiplication process is the possibility
that the ADC transfer function may have missing codes, due to
round-off errors.
[0012] Another known digital technique for trimming ADC gain error
is by changing the number of integration cycles in a first order
delta sigma ADC. This method does not provide adequate trim
resolution if the number of integration cycles is too low. (A 0.1%
resolution change is often considered sufficient to be an
acceptable increase in resolution order to achieve an acceptable
ADC gain error correction.) In some ADC architectures, the number
of integration cycles can be very low, so the resolution of the
digital gain error correction is poor if the method of changing the
number of integration cycles is used to accomplish the gain error
correction. These architectures all have the problem that it is
quite difficult to correct the ADC gain error by changing the
number of samples.
[0013] Primary shortcomings of the prior art technique of
multiplying the conversion result by a gain error correction
coefficient are that it requires too much power dissipation and too
much circuit complexity, and also requires too much chip area, and
too much total analog-to-digital conversion time in order to obtain
the gain-error-corrected digital output value.
[0014] Thus, there is an unmet need for an ADC which corrects its
ADC transfer characteristic for gain error without increasing the
amount of time required for obtaining a gain-corrected digital
output substantially beyond the amount of time required for a basic
analog to digital conversion.
[0015] There also is an unmet need for an ADC which corrects its
ADC transfer characteristic for gain error without increasing the
amount of integrated circuit chip area for obtaining a
gain-corrected digital output substantially beyond the amount of
chip area required for a basic analog to digital conversion.
[0016] There also is an unmet need for an ADC which corrects its
ADC transfer characteristic for gain error without increasing the
amount of power consumption required in obtaining a gain-corrected
digital output substantially beyond the amount of power consumption
required for a basic analog to digital conversion.
[0017] There also is an unmet need for an ADC which corrects its
ADC transfer characteristic for gain error without the high level
of circuit complexity generally associated with using analog
circuit techniques to correct a transfer characteristic of an ADC
for gain error.
[0018] There also is an unmet need for an ADC gain error correction
technique which is particularly suitable for use in a
delta-sigma/cyclic ADC architecture, and is also useful in a SAR
ADC architecture and in a higher-order delta-sigma ADC
architecture.
SUMMARY OF THE INVENTION
[0019] It is an object of the invention to provide an ADC which
corrects its ADC transfer characteristic for gain error without
increasing the amount of time required for obtaining a
gain-corrected digital output substantially beyond the amount of
time required for a basic analog to digital conversion.
[0020] It is another object of the invention to provide an ADC
which corrects its ADC transfer characteristic for gain error
without increasing the amount of integrated circuit chip area for
obtaining a gain-corrected digital output substantially beyond the
amount of chip area required for a basic analog to digital
conversion.
[0021] It is another object of the invention to provide an ADC
which corrects its ADC transfer characteristic for gain error
without increasing the amount of power consumption required in
obtaining a gain-corrected digital output substantially beyond the
amount of power consumption required for a basic analog to digital
conversion.
[0022] It is another object of the invention to provide an ADC
which corrects its ADC transfer characteristic for gain error while
avoiding the high level of circuit complexity generally associated
with using analog circuit techniques to correct a transfer
characteristic of an ADC for gain error.
[0023] It is another object of the invention to provide an ADC gain
error correction technique which is particularly suitable for use
in a delta-sigma/cyclic ADC architecture, and is also useful in a
SAR ADC architecture and in a higher-order delta-sigma ADC
architecture.
[0024] Briefly described, and in accordance with one embodiment,
the present invention provides correction of errors in an ADC chip
including an integrator (17), a comparator (30), and a digital
filter (37) by storing a gain-adjusted LSB size based on measured
gain error in a memory (44). The gain-adjusted LSB size is applied
to the digital filter to cause gain-adjusted LSB size values to be
added to or subtracted from accumulated content of the digital
filter in accordance with a first or second state, respectively, of
the comparator (30) during each cycle of the ADC. The final
accumulated content after all required cycles of the ADC is equal
to a gain-corrected digital output signal
(Dout(gain-corrected)).
[0025] In one embodiment, the invention provides an analog to
digital converter (ADC) (20B) including an integrator (17) for
sampling an input signal (Vin), the integrator having an input
(26A,B) coupled to receive an input signal (Vin) and also having an
output (27A,B). A comparator (30) has a signal input coupled to the
output (27A,B) of the integrator (17), and an output (33A,B). The
output of the comparator represents a present state of the
comparator depending on a relationship of an output voltage on the
output (27A,B) of the integrator (17) to a threshold voltage
(V.sub.TH). A digital filter (37) has an input (48A,B) coupled to
the output (33A,B) of the comparator (30) and also has an output
(49) for conducting a gain-corrected digital output signal
(Dout(gain-corrected)) representative of the input signal (Vin). A
memory (44) stores an externally determined gain-adjusted LSB size
number. Circuitry (47A,B) coupled to the memory (44) applies an
integral number of the gain-adjusted LSB size numbers to the input
(48A,B) of the digital filter to cause the integral number of
gain-adjusted LSB size numbers to be applied to content of the
digital filter (37) in accordance with the present state of the
comparator (30) to accumulate a value in the digital filter (37)
during successive cycles of the ADC (20B) so as to generate the
gain-corrected digital output signal (Dout(gain-corrected)).
[0026] In one embodiment, the comparator is a window comparator
(30) having a threshold input (29) which receives a plurality of
threshold voltages (V.sub.TH's), wherein a first state of the
window comparator is a +1 state, a second state of the window
comparator is a -1 state, and wherein the window comparator (30)
produces a third state which is a 0 state and which results in a
zero input to the digital filter (37). The integral number of
gain-adjusted LSB size numbers is applied to an increment input
(48A) of the digital filter (37) in accordance with the +1 state of
the comparator (30) so as to cause the integral number of
gain-adjusted LSB size numbers to be added to an accumulated
content of the digital filter (37). The integral number of
gain-adjusted LSB size numbers is applied to a decrement input
(48B) of the digital filter (37) in accordance with the -1 state of
the comparator (30) so as to cause the integral number of
gain-adjusted LSB size numbers to be subtracted from the
accumulated content of the digital filter (37).
[0027] In one embodiment, integrator (17) and the comparator (30)
form a delta-sigma modulator (15). In one embodiment, the delta
sigma modulator (15) is a first order delta-sigma modulator. In one
embodiment, the ADC has a mixed delta-sigma/cyclic
architecture.
[0028] In one embodiment, the digital filter includes an
accumulator which is configured as an up-down counter the content
of which is incremented or decremented in the amount of the
integral number of gain-adjusted LSB size numbers.
[0029] In one embodiment, the digital filter includes an SAR
(successive approximation register) results register (450), an
output of which provides the gain-corrected digital output.
[0030] In a described embodiment, the integrator includes an
operational amplifier (18) having a non-inverting input coupled by
a first sampling capacitor (23A) to receive a first input signal
(Vin.sup.+) and also coupled by a first integrating capacitor (24A)
to an inverting output (27A) of the operational amplifier (18), the
operational amplifier (18) also having an inverting input coupled
by a second sampling capacitor (23B) to receive a second input
signal (Vin.sup.-) and also coupled by a second integrating
capacitor (24B) to a non-inverting output (27B) of the operational
amplifier (18). The non-inverting input of the operational
amplifier (18) is coupled by a third sampling capacitor (28A) to a
first reference voltage (Vref.sup.+), a second reference voltage
(Vref.sup.-), or is unconnected to any reference voltage in
accordance with an output state of the comparator (30), and wherein
the inverting input of the operational amplifier (18) is coupled by
a fourth sampling capacitor (28B) to the first reference voltage
(Vref.sup.+), the second reference voltage (Vref.sup.-), or is
unconnected to any reference voltage in accordance with that output
state of the comparator (30), to prevent overflow of the
delta-sigma modulator.
[0031] The memory (44) can be a fuse link memory, a one time
programmable memory, or an electrically erasable programmable
memory, and can be a user-accessible register which allows a user
to control the transfer characteristic of the ADC. The memory can
receive the gain-adjusted LSB size number from an external test and
trim system.
[0032] In a described embodiment, the memory (44) and the output
(33A,B) of the window comparator (30) are coupled to an increment
input (INC) of the digital filter (37) by a first ANDing circuit
(47A) and are coupled to a decrement input (DEC) of the digital
filter (37) by a second ANDing circuit (47B), wherein the first
ANDing circuit (47A) effectuates adding of the integral number of
gain-adjusted LSB size numbers from the memory (44) to the content
of the digital filter (37) in accordance with a first state of the
window comparator (30), and wherein the second ANDing circuit (47B)
effectuates subtracting of the integral number of gain-adjusted LSB
size numbers from the memory (44) to the decrement input (DEC) in
accordance with a second state of the window comparator (30).
[0033] In one embodiment, the invention provides a method for
operating an analog to digital converter which includes an
integrator (17) having an input (26A,B) coupled to receive an input
signal (Vin) and also having an output (27A,B), a comparator (30)
having a signal input coupled to the output (27A,B) of the
integrator (18), and an output (33A,B), the output of the
comparator representing a first state or a second state, depending
on a relationship of an output voltage on the output (27A,B) of the
integrator (17) to a threshold voltage (V.sub.TH), and a digital
filter (37) having an input (48A,B) coupled to the output (33A,B)
of the comparator (30), wherein the method includes storing an
externally determined gain-adjusted LSB size number in a memory
(44), applying the gain-adjusted LSB size number to the input
(48A,B) of the digital filter (37) to cause an integral number of
gain-adjusted LSB size numbers to be applied to content of the
digital filter (37) in accordance with a present state of the
comparator (30) so as to accumulate a value in the digital filter
(37), and after all required cycles of the analog to digital
converter, providing the accumulated value at an output of the
digital filter (37) as a gain-corrected digital output signal
(Dout(gain-corrected)) that represents the input signal (Vin).
[0034] In one embodiment, the method includes adding the integral
number of gain-adjusted LSB size numbers from the memory (44) to
the content of the digital filter (37) in accordance with a first
state of the comparator (30), and subtracting the integral number
of gain-adjusted LSB size numbers from the memory (44) in
accordance with a second state of the comparator (30).
[0035] In one embodiment, the method includes operating the
integrator (17) and the comparator (30) as a delta-sigma modulator
(15) and operating the digital filter (37) as an up-down counter
the content of which is incremented or decremented by the integral
number of gain-adjusted LSB size numbers. In one embodiment the ADC
has a mixed delta-sigma/cyclic architecture and the method includes
providing an output of a SAR result register as the gain-corrected
digital output.
[0036] In one embodiment, the method includes providing additional
bits of resolution in an analog to digital conversion process such
that no missing code appears during generation of a smaller number
of bits which are utilized as the gain-corrected digital output
signal.
[0037] In one embodiment, the invention provides an analog to
digital converter (20B) including an integrator (17) having an
input (26A,B) coupled to receive an input signal (Vin) and also
having an output (27A,B), a comparator (30) having a signal input
coupled to the output (27A,B) of the integrator (18), and an output
(33A,B), the output representing a first state or a second state,
depending on a relationship of an output voltage on the output
(27A,B) of the integrator (17) to a threshold voltage (V.sub.TH), a
digital filter (37) having an input (48A,B) coupled to the output
(33A,B) of the comparator (30), means (44) for storing an
externally determined gain-adjusted LSB size number in a memory
(44), and means (47A, 47B) for applying the gain-adjusted LSB size
number to content of the digital filter (37) so as to cause an
integral number of gain-adjusted LSB size numbers to be applied to
content of the digital filter (37) in response to a present state
of the comparator (30) so as to accumulate a gain-corrected digital
value (Dout(gain-corrected)) in the digital filter (37) that, the
after all required cycles of the analog-to-digital converter, is a
gain-corrected digital value (Dout(gain-corrected)) that represents
the input signal (Vin).
BRIEF DESCRIPTION OF THE DRAWINGS
[0038] FIG. 1 is a block diagram of a conventional delta-sigma ADC
including a digital multiplier for providing DC gain error
correction.
[0039] FIG. 2 is a flow chart illustrating a conventional digital
ADC gain error correction technique used in the delta-sigma ADC of
FIG. 1.
[0040] FIG. 3 is a block diagram illustrating an integrated circuit
first order delta-sigma ADC utilizing a digital gain error
correction technique according to the invention by, in effect,
digitally modifying the value of the LSB (least significant bit) of
the basic ADC conversion result.
[0041] FIG. 4 is a flow chart illustrating a digital ADC gain error
correction technique used in the delta-sigma ADC of FIG. 3.
DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0042] Referring to FIG. 3, integrated circuit delta-sigma ADC chip
20B includes integrator 17, the (-) output 27A of which is coupled
to the (-) input of a window comparator 30. The (+) output 27B of
integrator 17 is coupled to the (+) input of window comparator 30.
Window comparator 30 may be composed of two or more conventional
comparators which may receive various thresholds voltages V.sub.TH
on the conductors of window comparator threshold bus 29,
respectively. (By way of definition, it should be understood that
if a conventional comparator receives a differential input signal,
it may be considered to have a zero threshold voltage. The term
"threshold voltage" as used herein is intended to encompass a
threshold voltage applied to one input of a 2-input comparator the
other input of which receives an input signal voltage that is to be
compared with the threshold voltage, and is intended to also
encompass a zero threshold voltage of a comparator which receives a
differential input signal.)
[0043] An input signal voltage Vin=Vin.sup.+-Vin.sup.- is coupled
between plates of input sampling capacitors 23A and 23B, the other
plates of which are connected to the (+) and (-) inputs,
respectively, of operational amplifier 18 of integrator 17. The (+)
input conductor 26A of operational amplifier 18 is coupled by an
integrating capacitor 24A to the (-) output conductor 27A of
operational amplifier 18, and the (-) input conductor 26B of
operational amplifier 18 is coupled by an integrating capacitor 24B
to the (+) output conductor 27B of operational amplifier 18.
[0044] Just as in Prior Art FIG. 1, input conductor 26A is coupled
by a reference sampling capacitor 28A to the pole of a single pole,
triple throw switch circuit S1, and input conductor 26B is coupled
by a reference sampling capacitor 28B to the pole of a single pole,
triple throw switch circuit S2. Switch circuit S1 couples the (+)
input of operational amplifier 18 through sampling capacitor 28A to
Vref.sup.+, Vref.sup.- or an open terminal, depending upon whether
the output CompOut of window comparator 30 is +1, -1, or 0,
respectively, so the reference Vref can be integrated in the
correct direction if necessary, depending on the decision of window
comparator 30. Similarly, switch circuit S2 couples the (-) input
of operational amplifier 18 through sampling capacitor 28B to
Vref.sup.-, Vref.sup.+ or an open terminal, depending upon whether
CompOut is +1, -1, or 0, respectively, so Vref can be integrated in
the correct direction if necessary.
[0045] Note that more details of the conventional delta-sigma
modulator 15 appear in FIG. 3a of the assignee's subsequently
mentioned pending patent application Ser. No. 11/738,566, which is
incorporated herein by reference. However, those skilled in the art
can readily understand the basic operation of integrator 15 as
described and as shown in FIGS. 1 and 3 hereof.
[0046] By way of definition, the term "integrator" as used herein
is intended is intended to encompass not only a conventional
integrator such as integrator 17 shown in FIG. 3, but also an
integrator in a cyclic SAR ADC, wherein the integrator only samples
once per conversion. The present invention is as applicable to that
kind of ADC as to a delta-sigma ADC.
[0047] Output conductor 33A of window comparator 30 is connected to
an enable input of an ANDing circuit 47A which also has multiple
gain error data inputs connected to a "gain-adjusted LSB size bus"
46 which is connected to the output of a "gain-adjusted LSB size
memory" 44. Gain-adjusted LSB size memory 44 can be implemented by
means of blowable fuses, an EEPROM (electrically erasable
programmable memory), or an OTP (one time programmable) memory or
the like which is loaded from a conventional external test and trim
system (not shown) with a "gain-adjusted LSB size number",
expressed in binary format. The output of ANDing circuit 47A is
connected by conductor 48A, on which a signal INCREMENT BY
GAIN-ADJUSTED LSB SIZE is applied to an increment input (INC) of
digital filter 37, conventional internal circuitry (not shown) of
which causes a digital number equal to the gain-adjusted LSB size
to be added or "incremented" to the contents of the up-down counter
or accumulator in digital filter 37. Similarly, output conductor
33B of window comparator 30 is connected to an enable input of
another ANDing circuit 47B, which also has multiple inputs
connected to "LSB size bus" 46. The output of ANDing circuit 47B is
connected by a bus 48B, on which a signal DECREMENT BY
GAIN-ADJUSTED LSB SIZE is applied to a decrement input (DEC) of
digital filter 37, conventional internal circuitry (not shown) of
which causes a digital number equal to the gain-adjusted LSB size
to be subtracted or "decremented" from the contents of the up-down
counter or accumulator in digital filter 37. After all of the
required integration cycles have been performed, the output of
digital filter 37 is the gain-corrected digital output signal
Dout(gain-corrected).
[0048] By way of definition, the term "up-down counter" as used
herein is intended to encompass accumulating circuitry in a digital
filter the contents of which can be incremented by adding a
non-integral LSB size number thereto or decremented by subtracting
the non-integral LSB size number therefrom.
[0049] It should be understood that ANDing circuits 47A and 47B or
equivalent circuitry could be included within digital filter 37 so
as to provide adding or incrementing of integral numbers of LSB
size numbers to or subtraction of integral numbers of LSB size
numbers from the content of digital filter 37. Gain-adjusted LSB
size memory 44 could be incorporated as a register which, if
desired, could be considered to be within digital filter 37.
Gain-adjusted LSB size memory 44 also could be implemented as a
user-accessible register, which would allow the user to adjust the
transfer characteristic of the ADC.
[0050] It should be appreciated that the output bus of window
comparator 30 can have as many conductors as is needed to conduct
the number of output states that can be generated by window
comparator 30. For example, if window comparator 30 generates 3
states, a two-wire bus may be required as the output of window
comparator 30. If window comparator 30 generates 5 states, a
three-wire bus is required as the output of window comparator 30.
(However, if an ordinary non-window type of comparator is used,
then a single conductor can be used as the output of that
comparator to represent the two states, "1" and "0".)
[0051] If the window comparator 30 has more than 3 states,
appropriate integral multiples of the gain-adjusted LSB size are
added to or subtracted from the contents of digital filter 37. For
example, for a 5-level window comparator 30, its possible outputs
are +2, +1, 0, 1, and -2. In this case, the gain-adjusted
quantities available to be added to or subtracted from the contents
of digital filter 37 will include 2.times.(gain-adjusted LSB size),
which is implemented by shifting the quantity "gain-adjusted LSB
size" by one binary bit position, 1.times.(gain-adjusted LSB size),
0, -1.times.(gain-adjusted LSB size), and -2.times.(gain-adjusted
LSB size).
[0052] Those skilled in the art know that a window comparator can
be thought of as a kind of flash ADC. For example, if a window
comparator can generate 64 different states, it would be comprised
of 64 ordinary comparators coupled in parallel, the 64 comparators
having 64 different input threshold voltages. The output bus of a
window comparator, for example a window comparator that can
generate 64 different states, ordinarily is coupled to circuitry
that performs an encoding function which converts the 64 comparator
outputs into a 6-bit code. For example, window comparator 30 of
FIG. 3, with its three output states +1, 0, and 1, is composed of 2
basic comparators having 2 separate input threshold voltages (e.g.,
V.sub.TH.sup.+ and V.sub.TH.sup.-), respectively, so that one basic
comparator changes state when the window comparator input voltage
exceeds the upper threshold voltage V.sub.TH.sup.+ and the other
basic comparator changes state when the window comparator input
voltage is less than the lower threshold voltage V.sub.TH.sup.-,
and when the window comparator input voltage is between the upper
and lower threshold voltages, neither basic comparator changes
state. Two comparator output conductors 33A and 33B are required to
represent the three states.
[0053] FIG. 4 is a flowchart which indicates the process of
obtaining the gain-adjusted LSB size and also indicates the
operation of ADC 20B of FIG. 3. As indicated in block 50 of FIG. 4,
the first step in the overall procedure is to determine the gain
error trim coefficient.
[0054] The gain error coefficient value that needs to be multiplied
by the measured digital output of ADC 20B is determined on the
basis of actual measurements of Dout and the desired full scale
value thereof, to determine the percentage of gain error. During
final testing of ADC 20B, a known value of input voltage Vin is
applied to the integrated circuit chip including ADC 20B, and the
resulting analog-to-digital Dout is measured. Then the resulting
error is calculated and the error is used to calculate how much the
size of the LSB needs to be adjusted so that the error of the
measured conversion result Dout after gain correction will be zero.
The error is expressed as a percentage of the known correct value
of Dout, in binary format.
[0055] As indicated in block 51 of FIG. 4, the gain-adjusted LSB
size is determined after the gain error trim coefficient has been
determined. The relationship between the gain error trim
coefficient and the gain-adjusted LSB size can be expressed as
(gain correction coefficient).times.(uncorrected value of
Dout)=(gain-corrected value of Dout). That value of the gain
correction coefficient, which is equal to the gain-adjusted LSB
size, is entered into and permanently stored in gain-adjusted LSB
size memory 44, for example by blowing certain fuses or links or
writing it into an OTP memory inside the integrated circuit ADC.
The initial contents of gain-adjusted LSB size memory 44 can be a
default value of 1.000 . . . . Then, when the amount of ADC gain
error is determined, that default value is modified to a value
different than 1.000 . . . , so as to generate the number
GAIN-ADJUSTED LSB SIZE which is added to the accumulator or up-down
counter contents if the output of window comparator 30 is +1 or is
subtracted from the accumulator or up-down counter contents if the
output of window comparator 30 is -1.
[0056] The part of FIG. 4 within dashed line A generally indicates
how the first-order delta-sigma modulator 15 consisting of
integrator 17, window comparator 30, and switch circuits S1 and S2
in FIG. 3 operate in conjunction with window comparator 30 so that
during every integration cycle the input voltage Vin is integrated
from the input capacitors to the holding or feedback capacitors,
and this causes the value of the output of integrator 17 to go in
one direction. For example, if input voltage Vin is positive, the
output voltage of integrator 17 increases in response to Vin. At
some point, integrator would "overflow" and stop integrating except
for the fact that during every integration cycle, window comparator
30 makes a decision as to whether the output of integrator 17 is
beyond a particular upper threshold voltage V.sub.TH.sup.+. When
V.sub.TH.sup.+ is reached, then during the next integration cycle,
integrator 17 also samples (in addition to the input voltage Vin)
the reference voltage Vref in the appropriate direction, so that if
the output of integrator 17 is too high, then reference voltage
Vref is sampled in the negative direction (i.e., V.sub.ref.sup.- is
sampled) and as result the output voltage of integrator 17
decreases.
[0057] As indicated in block 1 of FIG. 4, integrator 17 integrates
only the input voltage Vin during the first integration cycle of
delta-sigma modulator 15, since initially window comparator output
CompOut is 0. Then, as indicated in block 2, window comparator 30
determines whether the output of integrator 17 is greater than
V.sub.TH.sup.+, less than V.sub.TH.sup.-, or between V.sub.TH.sup.+
and V.sub.TH.sup.-. Depending on the results of this comparison,
the output of window comparator 30 is "1", "-1", or "0",
respectively. (The threshold voltages V.sub.TH.sup.+ and
V.sub.TH.sup.- can be derived from Vref and the scaling of various
capacitors in integrator 17 so as to provide the various desired
window comparator threshold voltages V.sub.TH of the standard
comparators of which window comparator 30 is comprised.)
[0058] In any case, window comparator output voltage CompOut is
coupled to and causes the gain-adjusted LSB size to be added to or
subtracted from the accumulated result in digital filter 37. If
modulator 15 is a first order delta-sigma modulator as shown in
FIG. 3, digital filter 37 can be implemented as an up-down counter,
i.e., as an accumulator. For first-order delta-sigma modulator 15
as shown in FIG. 3, the up-down counter or accumulator of digital
filter 37 is incremented by the value of the digital number
GAIN-ADJUSTED LSB SIZE stored in gain-adjusted LSB size memory 44
if the output of window comparator 30 is +1, as indicated in
decision block 3 and in block 4A. Similarly, the up-down counter in
digital filter 37 is decremented by the value of the digital number
GAIN-ADJUSTED LSB SIZE stored in gain-adjusted LSB size memory 44
if the output of window comparator 30 is -1, as indicated in
decision block 5 and in block 6A, or remains unchanged if the
output of window comparator 30 is "0", as indicated in block 7 of
FIG. 4.
[0059] Then, as indicated in block 8, the resulting value of the
window comparator output voltage CompOut is used to update the
accumulated result in the up-down counter in digital filter 37
either by the amount +GAIN-ADJUSTED LSB SIZE or by the amount
-GAIN-ADJUSTED LSB SIZE.
[0060] Then, as indicated in decision block 9, the process being
performed in ADC 20B determines if all of the integration cycles
required by ADC 20B have been performed. If this determination is
negative, the integrate and compare loop including blocks 1-9 is
repeated to complete the analog-to-digital conversion. On the
second and each following required integration cycle after the
second, integrator 17 samples not only the input voltage Vin, but
also samples the reference voltage value Vref.sup.+ or Vref.sup.-
in accordance with window comparator output CompOut. In the case
wherein CompOut is too high, i.e., the input to window comparator
30 is greater than upper threshold voltage V.sub.TH.sup.+, window
comparator 30 makes the decision to generate a +1 output level, and
during the next integration cycle integrator 17 samples the lower
reference voltage value Vref in addition to sampling input voltage
Vin, and this causes the output of integrator 17 to decrease.
Similarly, in the case wherein CompOut is too low, i.e., the input
to window comparator 30 is less than lower threshold voltage
V.sub.TH.sup.-, window comparator 30 makes a decision to generate a
-1 output level, and during the next integration cycle integrator
17 samples the upper reference voltage value Vref.sup.+ in addition
to sampling input voltage Vin, and this causes the output of
integrator 17 to increase. (It should be appreciated that the input
voltage Vin, the reference voltage Vref, and the integrator output
voltage each can be either differential or single-ended-voltages,
but in most practical implementations they are differential signals
or are sampled as differential signals. For example, the input
voltage Vin and the integrator output voltage typically are
differential signals, and the reference voltage Vref is referenced
to ground but is sampled as a differential signal.)
[0061] After the process of sampling Vin and Vref the required
numbers of times for ADC 20B, the determination of decision block 9
is affirmative. The window comparator output CompOut then is
coupled for the last time into digital filter 37, which produces
the gain-error-corrected digital output Dout, and the
gain-error-adjusted analog-to-digital conversion of Vin to
Dout(gain-corrected) is complete.
[0062] This avoids the prior art process of digital multiplying of
the basic uncorrected value of Dout by a gain error trim
coefficient to produce the corrected final digital output
Dout(gain-corrected).
[0063] It should be noted that only blocks 50, 51, 4A and 6B as
shown in the flowchart of FIG. 4 are substantially different than
in the flowchart of Prior Art FIG. 2.
[0064] As an example, assume that an input voltage Vin of 1.0 volt
and a reference voltage Vref are applied to ADC 20B of FIG. 3.
Assume also that the analog-to-digital conversion of the applied
1.0 volt input signal is performed, and the result is not the
desired 1.0 volt digital output voltage value, but instead is a 0.9
volt digital output voltage value because of gain error of ADC
20B.
[0065] The value of the needed digital gain correction coefficient,
which in this simplified example is approximately 1.1, is
determined and then stored in gain-adjusted LSB size memory 44.
Then the digital output conversion value of 0.9 volt produced by
the integrator causes the gain-corrected digital output
Dout(gain-corrected) which is accumulated in digital filter 37 and
produced on bus 49 to be very close to 1 volt (actually, 0.99 volts
in this simplified example).
[0066] Thus, instead of performing a digital multiplication of the
basic uncorrected ADC conversion result Dout at the output of the
digital filter in order to correct the ADC gain error as in Prior
Art FIGS. 1 and 2, the present invention, in effect, digitally adds
an integral multiple of a gain-adjusted LSB size number to the
accumulated contents in the up-down counter of digital filter 37 in
response to each +1 level produced by window comparator 30, and
also, in effect, digitally subtracts an integral multiple of the
gain-adjusted LSB size number from the accumulated contents in the
up-down counter of digital filter 37 in response to each -1 level
produced by window comparator 30. The gain correction function is
achieved without use of a complex, slow digital multiplier as
required by the prior art, and is achieved simultaneously, on a
bit-to-bit basis as the analog-to-digital conversion progresses.
Therefore, the substantial additional amount of time for digital
multiplication by the gain correction coefficient after the basic
ADC conversion operation has completely occurred as required in the
prior art is avoided by the present invention.
[0067] The above mentioned technique of adjusting the size of the
LSB of the digital output Dout of ADC 20B can cause "missing
codes". If the LSB size is changed from 1.0 to a gain-adjusted
value and that value then is repeatedly added to and/or subtracted
from the accumulator or up-down counter of ADC 20B after each
integration cycle, the result in the accumulator will always be a
value that is equal to an integral number of gain-adjusted LSB
sizes. At the end of all of the integration cycles required by ADC
20B for a single ADC conversion, there will be a need to perform a
round-off of the value in the accumulator or op-down counter.
Assume that the LSB size is not 1.0, but is 1.1 because it is
necessary to trim the ADC gain error. Also assume that the value of
Vin, and hence the value of Dout, is very close to zero wherein
window comparator 30 goes to +1 only once during the entire
integration process. Then the final result in the accumulator of
digital filter 37 will be 1.1 because its output was incremented
only once by GAIN-ADJUSTED LSB SIZE. Also assume that the part of
the up-down counter content located after the decimal point is
discarded, so the final value of Dout(gain-corrected) is 1.0.
[0068] Next, assume that Vin is twice as large as in the foregoing
example. Then, window comparator 30 switches to a +1 twice during
the entire integration process so at the end of the entire
integration process the final result in the accumulator or up-down
counter is 2.times.1.1, or 2.2. The part of this located after the
decimal point is rounded off, i.e. discarded, and the final result
in digital filter 37 is 2.
[0069] Next assume that Vin is 9 times greater than Vin in the
first example above. Then the result in accumulator after the
entire integration process is 9.times.1.1, i.e., 9.9. Discarding
the part located after the decimal point results in a final value
of value 9.
[0070] Next, assume Vin is 10 times greater than in the first
foregoing example. Then the result in the accumulator of digital
filter 37 is exactly 11 after all of the integration cycles
required by ADC 20B have been completed. In this case there is
nothing to round off, and the final result in the accumulator is
taken to be 11. Thus, a missing code, namely 10, has been
encountered. (Note that a missing code will occur regardless of
what rounding off procedure is used.) A solution to the above
described missing code problem is to add more bits of resolution to
the analog to digital conversion process such that no missing code
appears during the generation of the smaller number of bits which
are utilized as the output Dout. That is, adding more bits of
resolution to the conversion process means that the lower amount of
resolution actually desired for Dout is achieved before a missing
code has a chance to occur.
[0071] The technique of the present invention is equally applicable
for other methods of integration, such as higher order delta-sigma
conversion or a pure SAR (successive approximation register)
conversion. In some ADC architectures, additional bits of
resolution can be obtained at very little expense.
[0072] In one ADC architecture referred to as a "mixed
delta-sigma/cyclic architecture", conventional integration is
utilized to generate some of the output bits, after which a SAR
mode of operation is utilized and a residue on the output of the
integrator is multiplied by a factor of 2 during each cycle. The
binary content of the SAR filter is also shifted by one bit. (By
way of definition, the term "cycles" can refer to sub-cycles of the
SAR process.)
[0073] Details of such a "mixed delta-sigma/cyclic ADC
architecture", including additional details of conventional first
order delta-sigma modulator 15 shown in FIGS. 1 and 3 herein, are
shown in FIG. 3a of the assignee's pending patent application Ser.
No. 11/738,566 entitled "Hybrid Delta-Sigma/SAR Analog to Digital
Converter and Methods for Using Such", filed Apr. 23, 2007 by Jerry
Doorenbos, Marco Gardner and Dimitar Trifonov, and incorporated
herein by reference.
[0074] A cyclic SAR ADC works as follows. First, the integrator
samples the ADC input voltage. Then the ADC performs a number of
SAR successive approximation analog to digital conversion cycles.
In the case of a cyclic SAR such as one included as part of the
mixed delta-sigma/cyclic ADC architecture in the above mentioned
patent application Ser. No. 11/738,566, the residue on the output
of the integrator 310 shown in FIG. 3a after a predetermined number
of integration cycles therein is multiplied by 2 each cycle. If the
comparator output is equal to +1, then together with the
multiplication by 2, a negative reference voltage Vref is sampled.
If the comparator output is equal to -1, then together with the
multiplication by 2, a positive reference voltage Vref is sampled.
If the comparator output is equal to 0, then only the
multiplication by 2 is performed.
[0075] The procedure performed during each SAR cycle, can be
summarized by the following algorithm:
TABLE-US-00001 If (Vresidue > +Vthreshold) Then Vout =
2*(Vresidue - Vref) Else If (Vresidue < -Vthreshold) Then Vout =
2*(Vresidue + Vref) Else Vout = 2*(Vresidue) End If
The foregoing algorithm can be modified for various SAR ADC
implementations.
[0076] Next, an example is provided to show how the result is
accumulated in the SAR result register 450 in FIG. 3a of Ser. No.
11/738,566, (which can be a digital filter that in this particular
case is more like a shift register). Assume that the ADC is a 5 bit
SAR ADC and that in accordance with 5 corresponding SAR cycles,
comparator 414 generates a sequence of 5 decisions or states which
are +1, 0, +1, -1, +1. The conversion results can be produced in
two different ways that yield the same result.
[0077] The first way, which can be referred to as a "shift register
approach", is as follows. Since the comparator decision during the
first SAR cycle is +1, so the initial value (00000) plus the
comparator output +1 is loaded into the shift register. The result
is 00000+1=00001.
[0078] For the second cycle, the comparator decision is 0, so the
present contents of the shift register are shifted one position to
the left and the 00001 becomes 00010 and is added to the comparator
output 0, producing the result 00010+0=00010.
[0079] For the third cycle, the comparator decision is +1, so the
present contents of the shift register are shifted one position to
the left and the 00010 becomes 00100 and is added to the comparator
output +1, producing the result 00100+1=00101.
[0080] For the fourth cycle, the comparator decision is -1, so the
present contents of the shift register are shifted one position to
the left and the 00101 becomes 01010 and is added to the comparator
output -1, producing the result 01010-1=01001.
[0081] For the fifth cycle, the comparator decision is +1, so the
present contents of the shift register is shifted one position left
and of the 01001 becomes 10010 and is added to the comparator
output +1, producing the result 10010+1=10011. Thus, the final
conversion result is 10011.
[0082] The second way, which is referred to as the "digital filter
approach", is as follows. Since the comparator decision during the
first SAR cycle is +1, the comparator output is scaled by the
integer value 16, and the scaled comparator output becomes +10000.
The initial value 00000 plus the scaled comparator output +10000 is
loaded into the result register 450, producing the result
00000+10000=10000.
[0083] For the second SAR cycle, the comparator decision is 0, so
the comparator output is scaled by the integer value 8. Then the
previous value 10000 in the result register plus the scaled
comparator output 0 is loaded into the result register, producing
the result 10000+0=10000.
[0084] For the third cycle, the comparator decision is +1, so the
comparator output is scaled by the integer value 4, and the scaled
comparator output becomes +100. Then the previous value 10000 in
the result register plus the scaled comparator output+100 is loaded
into the result register, producing the result 10000+100=10100.
[0085] For the fourth cycle, the comparator decision is -1, so the
comparator output is scaled by the integer value 2, and the scaled
comparator output becomes -10. Then the previous value 10100 in the
result register plus the scaled comparator output (-10) is loaded
into the result register, producing the result 10100-10=10010.
[0086] For the fifth cycle, the comparator decision is +1, so the
comparator output is scaled by the integer value 1, and scaled
comparator output remains +1. Then the previous value 10010 plus
the scaled comparator output +1 is loaded into the result register,
producing the result 10010+1=10011. Thus, the final conversion
result is 10011.
[0087] Next, an example be provided to show how the basic technique
of the present invention can be applied to the mixed
delta-sigma/cyclic architecture shown in above mentioned pending
patent application Ser. No. 11/738,566. In the present invention,
instead of sending the comparator values +1, 0, -1 to the result
register 450 as described above, the properly scaled "LSB size
number" is sent to the result register. Assume that the ideal
conversion result from the ADC for the same input voltage as in the
previous example is 10100, but because of ADC gain error the actual
measured result is 10011. Then the needed LSB size number is
calculated to be 1.0001, which is a binary number representing a
6.25% gain error correction. If this gain-corrected "LSB size
number" is applied to the above mentioned algorithm above with the
same sequence of decisions from the comparator 414, the result for
each of the previously described examples will be as follows.
[0088] To apply the present invention to the previously described
the "shift register approach", for the first cycle, of the
comparator decision is +1, the initial value 00000.0000 in the SAR
result register 450 plus the "LSB size number" scaled by the
comparator output +1 is loaded into the shift register, producing
the result 00000.0000+1.0001=00001.0001.
[0089] For the second cycle, the comparator decision is 0 so for
present contents of the shift register is shifted one position to
the left and the 00001.0001 becomes 00010.0010 and is added to the
"LSB size number" scaled by the comparator output 0, producing the
result 00010.0010+0=00010.0010.
[0090] For the third cycle, the comparator decision is +1, so the
present contents of the shift register are shifted one position to
the left and the 00010.0010 becomes 00100.0100 and is added to the
"LSB size number" scaled by the comparator output +1, producing a
result 00100.0100+1.0001=00101.0101.
[0091] For the fourth cycle, the comparator decision is -1, so the
present contents of the shift register are shifted one position to
the left and the 00101.0101 becomes 01010.1010 and is added to the
"LSB size number" scaled by the comparator output -1, producing the
result 01010.1010-1.0001=01001.1001.
[0092] For the fifth cycle, the comparator decision is +1, so the
present contents of the contents of the shift register are shifted
one position to the left and the 01001.1001 becomes 10011.0010 and
is added to the "LSB size number" scaled by the comparator output
(+1), producing the result 10011.0010+1.0001=10100.0011.
[0093] After rounding off (discarding) the numbers after the
decimal point, the final gain-adjusted ADC conversion result is
10100.
[0094] To apply the present invention to the previously described
"digital filter approach", for the first SAR cycle, the comparator
decision is +1, so the "LSB size number" is scaled by the
comparator output and by the integer value 16 so the scaled value
becomes +10001.0000. The initial value 00000.0000 plus the scaled
comparator output +10001.0000 is loaded into the digital filter,
producing the result 00000.0000+10001.0000=10001.0000.
[0095] For the second SAR cycle, the comparator decision is 0, so
the "LSB size number" is scaled by the comparator output and by the
integer value 8 so the scaled value becomes 0.0000 since the
comparator output states is 0. Then the previous value 10001.0000
plus the scaled "LSB size number" 0.0000 is loaded into the digital
filter, producing the result 10001.0000+0=10001.0000.
[0096] For the third cycle, the comparator decision is +1, so the
"LSB size number" is scaled by the comparator output and by the
integer value 4 so the scaled value becomes +100.0100. Then the
previous value 10001.0000 plus the scaled "LSB size
number"+100.0100 is loaded into the digital filter, producing the
result 10001.0000+100.0100=10101.0100.
[0097] For the fourth cycle, the comparator decision is -1, so the
"LSB size number" is scaled by the comparator output and by the
integer value 2 so the scaled value becomes -10.0010. Then the
previous value 10101.0100 plus the scaled "LSB size number"-10.0010
is loaded into the digital filter, producing the result
10101.0100-10.0010=10011.0010.
[0098] For the fifth cycle, the comparator decision is +1, so the
"LSB size number" is scaled by the comparator output and by the
integer value 1 so the scaled value remains +1.0001. Then the
previous value (10011.0010) plus the scaled "LSB size
number"+1.0001 is loaded into the digital filter, producing the
result 10011.0010+1.0001=10100.0011.
[0099] After rounding off or discarding the numbers after the
decimal point, the final gain-corrected ADC conversion result is
10100.
[0100] There are various implementations of SAR ADC but all of them
accumulate the desired digital result in the ways exemplified
above, and all of them can be corrected for gain error by means of
the above described algorithm.
[0101] For application to the architecture described in above
mentioned Ser. No. 11/738,566, the present invention works in the
same way. First, in the integration mode the "LSB size number" is
applied during every integration cycle to the digital filter
contents as scaled by the comparator output (+1, 0 or -1). No
shifting to the left is performed. When the mixed integrating/SAR
circuit switches from integrating mode to SAR mode, the accumulated
number in the register (i.e., the DS Result Counter 460 in FIG. 3a
of Ser. No. 11/738,566) is taken as "initial value" and the
gain-corrected conversion proceeds in the above described SAR
mode.
[0102] As an example, assume that there are 8 integration cycles
followed by 5 SAR cycles. Also assume that during the integration
cycles the comparator switched states six times to +1, then once to
0, and then to -1. Then the integration result will be 6-1=5 (i.e.,
binary 101). In SAR mode, the initial value of the result register
will be 00000101 using the "shift register" algorithm or 10100000
using the "digital filter" approach.
[0103] According to the present invention, and with the same "LSB
size number"+1.0001 after the integration mode operation and before
the SAR mode operation, the initial value of the result register
will be 00000101.0101 using the above described "shift register"
algorithm or 10101010.0000 using the "digital filter" approach. As
in the "shift register" approach, the initial value is shifted one
position left every SAR cycle so the final effect of the
integration cycles (which determine the MSBs of the result) on the
result is the same.
[0104] Fundamentally, the invention provides an LSB size number and
then scales it by an integer and then adds or subtracts the scaled
amount to the contents of the digital filter, according to the
state of the comparator. For example, in the case of using the
invention in a successive approximation ADC, the LSB size number
might be successively scaled by numbers such as 16,384, 8192, 4096,
2048, 1024, 512, 256 etc., before the LSB size number is added to
or subtracted from the content of the SAR register according to the
status of the comparator.
[0105] The invention thus provides a method of correction of the
gain error of an ADC by changing the value of its LSB which is
added to or subtracted from the digital filter contents in response
to the output states generated by the window comparator. For
example, if the nominal value of the LSB of ADC 20B in FIG. 3 is
"1", its value may be adjusted by an amount in the range from
"0.111110000000" to "1.000001111111". With 8 bits available to trim
the value of the LSB, this provides a +/-3% trim range and 0.024%
trim resolution.
[0106] An advantage of the above described embodiment of the
invention is that it does not require an additional digital
multiplier as required by the prior art, and therefore does not
require additional time and additional integrated circuit die area
for the multiplication required using the prior art techniques, and
therefore also reduces the amount of power consumption required for
a DC gain correction. The technique of the present invention can be
easily implemented as a so-called final test trimming operation
which takes into account all shifts of the ADC performance during
the usual wafer back grinding process and integrated circuit
packaging operations. The only penalty is the increased size of the
accumulation register of the digital filter. The described
embodiments of the invention also provide a convenient way of
dealing with missing output codes which occur if the LSB size is
trimmed to a value greater than 1, by simply increasing the
resolution. This is especially easily accomplished in a mixed delta
sigma/cyclic architecture and in a SAR ADC architecture, because
each additional bit of resolution requires only one additional
integration cycle.
[0107] Other advantages of the above described embodiment of the
present invention and adjusting LSB size of the ADC to trim gain
error include the fact that it is in an entirely digital correction
and does not require any additional analog circuitry. It is capable
of providing any desirable trim range and trim resolution, does not
substantially increase the overall gain-adjusted ADC conversion
time, does not require additional mathematical operations, and
requires less integrated circuit chip area and less power
consumption than the prior art techniques for correcting gain error
of an ADC. The technique of the present invention also improves DNL
(differential non-linearity) and INL (integral non-linearity) of
the ADC by adding more bits of resolution to the conversion result,
and can be applied to various ADC architectures, such as SAR ADCs
and high order delta sigma ADCs.
[0108] While the invention has been described with reference to
several particular embodiments thereof, those skilled in the art
will be able to make various modifications to the described
embodiments of the invention without departing from its true spirit
and scope. It is intended that all elements or steps which are
insubstantially different from those recited in the claims but
perform substantially the same functions, respectively, in
substantially the same way to achieve the same result as what is
claimed are within the scope of the invention. For example, the
invention applies not only to first order delta-sigma ADC
architecture as described above, but also to a mixed
delta-sigma/cyclic architecture, or to a complete SAR ADC, or a
higher order delta-sigma ADC. It applies to all ADC architectures
that require a low number of integration cycles and to accomplish
an analog to digital conversion. It should be appreciated that if
the modulator 15 is a SAR ADC, then the digital filter 37 is more
like a shift register than an up-down counter. If modulator 15 is a
higher order delta-sigma modulator, then the digital filter
typically is a much more complex digital filter that depends on the
architecture of the ADC.
* * * * *