US 4,996,530 AGranted

Statistically based continuous autocalibration method and apparatus

AssigneeHewlett-Packard Company

InventorHoward E. Hilton

Life of this patent

Estimating the term…

Loading patent…

Abstract

Errors in a sampled data process are discerned statistically throughout the process, permitting their efficient removal. An exemplary application is a subranging analog-to-digital converter (ADC), in which errors associated with component digital-to-analog (DAC) current sources are discerned and corrected automatically during the circuit's normal operation. This is achieved by continually introducing a random signal into the process, statistically examining the DAC output signal to discern error terms, and correlating the occurrences of these errors with the values of the random signal applied to the DACs so as to identify the current sources to which the error terms are due. The resulting ADC output signal is compensated to remove the random signal and is further compensated to remove the DAC error terms discerned by this statistical analysis. The effect of any errors that may remain due to imperfect quantification of the DAC error terms is minimized due to their randomization, permitting the removal of these error terms by averaging techniques.

Claims26

Dependent claims

In a subranging analog-to-digital converter that includes:

first and second analog-to-digital converters (ADCs);
a digital-to-analog converter (DAC);
each of said converters having an input and an output;
an analog subtraction circuit having first and second inputs and a difference output;
an analog signal input connected to the first input of the subtraction circuit;
said first ADC having its input connected to the analog signal input and having its output connected to the input of the DAC;
said DAC having its output connected to the second input of the subtraction circuit;
said second ADC having its input connected to the difference output of the subtraction circuit;
and
means for combining the outputs of the first and second ADCs to provide a digital output signal;
an improvement wherein the DAC comprises:
first, second and third DACs, each having an input and an output;
first and second digital noise sources, each having an output;
first and second adders, each having two inputs and an output;
and
a correction circuit having a plurality of inputs and an output;
said first adder having its first input coupled to the output of the first noise source and having its second input coupled to the output of the first ADC, said adder having its output coupled to the input of the first DAC to provide the negated sum of the first noise source output and the first ADC output thereto;
said second adder having its first input coupled to the output of the first noise source and having its second input coupled to the output of the second noise source, said adder having its output coupled to the input of the second DAC to provide the sum of the first and second noise source outputs thereto;
said third DAC having its input coupled to the output of the second noise source for receiving a negated noise output therefrom;
said first, second and third DACs having their outputs coupled to the second input of the analog subtraction circuit;
said correction circuit having a first input coupled to the output of the second ADC, a second input coupled to the output of the first adder, a third input coupled to the output of the second adder, and a fourth input coupled to the output of the second noise source;
said circuit including means for compensating the digital output signal for conversion errors in the first, second and third DACs.

19 more elements

In a method of analog-to-digital conversion that includes the steps:

generating a digital approximation of an analog input signal;
converting said digital approximation into an intermediate analog signal approximately corresponding thereto;
producing an analog subrange signal by determining a difference between the analog input signal and the intermediate analog signal;
generating a digital approximation of said subrange signal;
and
combining said digital approximations to yield a final output signal;
an improvement comprising the steps:
combining the digital approximation of the analog input signal with at least one digital noise signal to yield a digital combined signal;
converting the digital combined signal into a first intermediate analog signal approximately corresponding thereto using a DAC;
processing said first intermediate analog signal to remove the noise component thereof;
and
for at least one bit of the DAC:
correlating the final output signal with a digital value applied by the digital combined signal to said bit to note any deviation from ideal DAC behavior.

10 more elements

A compensated sampled data process comprising the steps:

providing an electronic system;
applying a sampled electronic data input signal to the electronic system;
processing the sampled electronic data signal in the electronic system to produce an output electronic signal corresponding thereto;
said processing step including perturbing the output signal in response to one or more process parameters, said perturbation causing the output electronic signal to deviate from an ideal value;
sensing the deviations in the output electronic signal from the ideal value;
and
compensating the electronic system in response to said sensed deviations to counteract the perturbations that cause the output signal deviations;
wherein:
the process performed by the electronic system is ideally represented as:
F(X.sub.n)=Y.sub.n
where X.sub.n is an input sequence {x.sub.1, x.sub.2, x.sub.3. . . } and Y.sub.n is a scaler output sequence corresponding thereto;
the process performed by the electronic system, including said deviations due to one or more process parameters, is represented as:
F(X.sub.n,P.sub.n)=Y.sub.n '
where X.sub.n is the input, P.sub.n is a vector representing the parameters corresponding thereto, and Y.sub.n ' is the output of the actual process corresponding to (X.sub.n,P.sub.n);
the process being such that:
F(X.sub.n)=F(X.sub.n,P)
where P is a fixed vector representing ideal process parameters;
the method further comprising the steps:
providing a random vector Z.sub.n ;
performing an operation C on (X.sub.n,Z.sub.n) to yield an output X.sub.n ';
performing the operation F on (X.sub.n ',P.sub.n) to yield an output W.sub.n ';
performing an operation H on (X.sub.n,Z.sub.n) to yield a vector output Z.sub.n ';
and
performing an operation D on (W.sub.n ',Z.sub.n ') to yield an output W.sub.n ;
calculating a parameter vector P.sub.n+1 for a next sample using the formula:
P.sub.n+1 =P.sub.n -(W.sub.n .multidot.Z.sub.n ')/K
where K is a smoothing constant;
wherein:
operations C, H and D are chosen such that:
W.sub.n =Y.sub.n =F(X.sub.n,P)=D(F(C(X.sub.n,Z.sub.n),P),H(X.sub.n, Z.sub.n))
and:
the sign of each element of the expected value of [W.sub.n Z.sub.n '] is the same as the sign of the corresponding element of the expected value of [P.sub.n -P].

29 more elements

A method for compensating a sampled data process F by removing deviations therefrom to make it more nearly ideal, said process being ideally represented as:

F(X.sub.n)=Y.sub.n
where x.sub.n is an input and Y.sub.n is a scaler output corresponding thereto;
said process actually including one or more process parameters that cause the actual process to deviate from ideal so that the actual process is represented as:
F(X.sub.n,P.sub.n)=Y.sub.n '
where X.sub.n is the input, P.sub.n is a vector representing the parameters for the corresponding sample, and Y.sub.n ' is the output of the actual process corresponding to (X.sub.n,P.sub.n);
the process being such that:
F(X.sub.n)=F(X.sub.n,P)
where P is a fixed vector representing ideal process parameters;
the method comprising the steps:
providing a random vector Z.sub.n ;
performing an operation C on (X.sub.n,Z.sub.n) to yield an output X.sub.n ';
performing the operation F on (X.sub.n ',P.sub.n) to yield an output W.sub.n ';
performing an operation H on (X.sub.n,Z.sub.n) to yield a vector output Z.sub.n ';
and
performing an operation D on (W.sub.n ',Z.sub.n ') to yield an output W.sub.n ;
calculating the parameter vector P.sub.n+1 for the next sample using the formula:
P.sub.n+1 =P.sub.n -(W.sub.n .multidot.Z.sub.n ')/K
where K is a smoothing constant;
wherein:
operations C, H and D are chosen such that:
W.sub.n =Y.sub.n =F(X.sub.n,P)=D(F(C(X.sub.n,Z.sub.n),P),H(X.sub.n,Z.sub.n))
and in which:
the sign of each element of the expected value of [W.sub.n Z.sub.n '] is the same as the sign of the corresponding element of the expected value of [P.sub.n -P];
X.sub.n comprises an analog voltage V.sub.n and three digital numbers A1.sub.n, A2.sub.n and A3.sub.n ;
the ideal process is described as:
F(X.sub.n)=V.sub.n -A1.sub.n =Y.sub.n
where Y.sub.n is a digital output with acceptable precision;
the actual process is described by:
F(X.sub.n,P.sub.n)=V.sub.n -A1.sub.n +A2.sub.n -A3.sub.n +(P.sub.n .multidot.B.sub.n)
where B.sub.n is a vector comprised of the individual bits used to encode the three digital numbers A1.sub.n, A2.sub.n and A3.sub.n, and (P.sub.n .multidot.B.sub.n) is a vector dot product yielding a scaler;
the random vector Z.sub.n consists of two random digital numbers, Z.sub.n comprises two random digital numbers Z1.sub.n and Z2.sub.n ;
the operation C generates an output X.sub.n ' comprising the analog voltage V.sub.n and three digital numbers D1.sub.n, D2.sub.n and D3.sub.n, such that:
D1.sub.n =A1.sub.n +Z1.sub.n
D2.sub.n =Z1.sub.n +Z2.sub.n
D3.sub.n =Z2.sub.n ;
the operation D generates an output W.sub.n such that:
D(W.sub.n ',Z.sub.n ')=W.sub.n '=W.sub.n ;
and
the operation H generates an output Z.sub.n ' comprising a vector of all the individual bits used to encode the three digital numbers D1.sub.n, D2.sub.n and D3.sub.n.

36 more elements

Dependencies are inferred from the claim text.