SIGMA-DELTA ANALOG-TO-DIGITAL CONVERTER CIRCUIT WITH REAL TIME CORRECTION FOR DIGITAL-TO-ANALOG CONVERTER MISMATCH ERROR

Abstract

An estimate of unit current element mismatch error in a digital to analog converter circuit is obtained through a correlation process. Unit current elements of the digital to analog converter circuit are actuated by bits of a thermometer coded signal generated in response to a quantization output signal. A correlation circuit generates the estimates of the unit current element mismatch error from a correlation of a first signal derived from the thermometer coded signal and a second signal derived from the quantization output signal.

Claims

1. A method, comprising: generating a thermometer coded signal in response to a quantization output signal; wherein bits of the thermometer coded signal control actuation of unit current elements within a digital to analog converter circuit; correlating a first signal derived from the thermometer coded signal and a second signal derived from the quantization output signal to generate estimates of unit current element mismatch error.

2. The method of claim 1, further comprising: randomly scrambling the bits of the thermometer coded signal; and applying the randomly scrambled bits of the thermometer coded signal to control actuation of unit current elements within the digital to analog converter circuit.

3. The method of claim 1, further comprising: generating from the thermometer coded signal a signal measure that is constant and independent of a thermometer code of thermometer coded signal; filtering the signal measure to generate a filtered signal measure as said first signal; and filtering the quantization output signal to generate a filtered quantization output signal as said second signal.

4. The method of claim 3, wherein filtering the signal measure comprises high pass filtering.

5. The method of claim 3, wherein the filtering the quantization output signal comprises a combination of high pass filtering and error transfer filtering.

6. The method of claim 3, wherein the filtering the quantization output signal comprises high pass filtering.

7. The method of claim 3, wherein generating comprises: summing bits of the thermometer coded signal; dividing the summed bits of the thermometer coded signal by a number of the bits in the thermometer coded signal to produce an output signal; and subtracting the output signal from the bits of the thermometer coded signal to generate the signal measure.

8. The method of claim 1, further comprising processing the quantization output signal in response to the generated estimates of unit current element mismatch error to generate a digital signal.

9. The method of claim 8, further comprising decimating the digital signal to generate an output signal.

10. A method, comprising: generating a quantization output signal; generating a thermometer coded signal in response to the quantization output signal; actuating unit current elements of a digital to analog converter (DAC) circuit in response to bits of the thermometer coded signal; and estimating unit current element mismatch error in the digital to analog converter circuit by: correlating a first signal derived from the thermometer coded signal and a second signal derived from the quantization output signal to generate estimates of unit current element mismatch error.

11. The method of claim 10, further comprising: applying a correction for unit current element mismatch error in response to the estimated unit current element mismatch error.

12. The method of claim 11, further comprising: filtering a difference signal derived from an analog signal output from the digital to analog converter circuit; and quantizing a signal generating from said filtering to generate said quantization output signal.

13. The method of claim 10, further comprising scrambling bits of the thermometer coded signal for input to the digital to analog converter.

14. The method of claim 10, further comprising: generating from the thermometer coded signal a signal measure that is constant and independent of a thermometer code of thermometer coded signal; filtering the signal measure to generate a filtered signal measure as said first signal; and filtering the quantization output signal to generate a filtered quantization output signal as said second signal.

15. The method of claim 14, wherein filtering the signal measure comprises high pass filtering.

16. The method of claim 14, wherein the filtering the quantization output signal comprises a combination of high pass filtering and error transfer filtering.

17. The method of claim 14, wherein the filtering the quantization output signal comprises high pass filtering.

18. The method of claim 14, wherein generating comprises: summing bits of the thermometer coded signal; dividing the summed bits of the thermometer coded signal by a number of the bits in the thermometer coded signal to produce an output signal; and subtracting the output signal from the bits of the thermometer coded signal to generate the signal measure

19. The method of claim 10, further comprising processing the quantization output signal in response to the generated estimates of unit current element mismatch error to generate a digital signal.

20. The method of claim 19, further comprising decimating the digital signal to generate an output signal.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

[0026] For a better understanding of the embodiments, reference will now be made by way of example only to the accompanying figures in which:

[0027] FIG. 1 is a block diagram of a conventional sigma-delta analog-to-digital converter circuit with single bit quantization;

[0028] FIG. 2 is a block diagram of a conventional sigma-delta analog-to-digital converter circuit with multi-bit quantization;

[0029] FIG. 3 is a block diagram of a conventional sigma-delta analog-to-digital converter circuit with multi-bit quantization and scrambling of DAC codes;

[0030] FIG. 4 is a block diagram of a sigma-delta analog-to-digital converter circuit with multi-bit quantization and feedback digital-to-analog converter mismatch correction; and

[0031] FIG. 5 is a block diagram for an error estimation circuit.

DETAILED DESCRIPTION

[0032] Reference is now made to FIG. 4 which shows a block diagram of a sigma-delta analog-to-digital converter circuit 110 with multi-bit quantization and feedback digital-to-analog converter (DAC) mismatch correction. The circuit 110 includes an N-bit sigma-delta modulator circuit 112 having an input configured to receive an analog input signal A and an output configured to generate a digital output signal B comprised of a stream of N-bit code words. The sigma-delta modulator circuit 112 comprises a difference amplifier 120 (i.e., summation circuit) having a first (non-inverting) input that receives the analog input signal A and a second (inverting) input that receives an analog feedback signal D. The difference amplifier 20 outputs an analog difference signal vdif in response to a difference between the analog input signal A and the analog feedback signal D (where vdif(t)=A(t)−D(t)). The analog difference signal vdif is integrated by a K-order loop filter 116 (using, for example, K integrator circuits 122) to generate a change signal vc having a slope and magnitude that is dependent on the sign and magnitude of the analog difference signal vdif. An N-bit quantization circuit 124 samples the change signal vc at the sampling rate fs and generates for each sample a corresponding N-bit code word of the digital output signal B (also referred to as the ADC output v(k)). This signal is thermometer coded with the bits of the code word scrambled by a data weighted averaging (DWA) scrambler circuit 128 to generate a selecting signal b.sub.i(k). An N-bit digital-to-analog converter (DAC) circuit 126 in the feedback loop converts the bits of the scrambled thermometer coded selecting signal b.sub.i(k) to generate a corresponding voltage level for the analog feedback signal D.

[0033] The sigma-delta modulator circuit 112 implements a loop filter 116 with a K-order integration circuit implementation. In the illustration of FIG. 1, K=1 as there is only one integration shown (with integrator 122) in the loop filter 116, but it will be understood that this is by way of example only, and K may equal 2, 3 or more as needed by the desired noise shaping and circuit application. Those skilled in the art know how to implement a K>1 order loop filter 116 for the sigma-delta modulator circuit 112.

[0034] The stream of N-bit code words for the digital output signal B produced by the N-bit quantization circuit 124 are input to the scrambling circuit 128 to be passed through, after scrambling, to the N-bit DAC circuit 126 in the feedback path. This stream of N-bit code words for the digital output signal B is further input to a replica DAC circuit 118. In this context, the replica DAC circuit 118 is programmed by an error estimation circuit 140 to digitally model the operation of the feedback N-bit DAC circuit 126. In other words, to digitally model the ideal DAC operation plus the DAC error introduced by the non-linear operation of the unit elements. In response to this programming, the replica DAC circuit 118 processes the ADC output v(k) to remove the DAC error.

[0035] The replica DAC circuit 118 provides a digital replication of the analog N-bit DAC circuit 126, that replication specifically accounting for the non-ideal operation of the N-bit DAC circuit 126 due to element mismatch. More specifically, the replica DAC circuit 118 is programmed with a plurality of digital code words that are directly proportional to the value of the mismatched unit elements of the analog N-bit DAC circuit 126. In other words, digital codes corresponding to the analog value of the unit element error e.sub.i. The digital code words can be of any selected precision P, and are determined by the error estimation circuit 140. It will be appreciated that if the digital model provided by the replica DAC circuit 118 is substantially identical to the non-ideal actual operation of the analog N-bit DAC circuit 126, then the digital signal E output from the digital DAC copy circuit 118 will be functionally equivalent to the analog feedback signal D output from the analog N-bit DAC circuit 126. In this context, “functionally equivalent” means that an analog conversion of the digital value for the digital signal E generated in response to signal B by the digital DAC copy circuit 118 is substantially equal to the corresponding analog value for the analog feedback signal D generated in response to that same signal B. The digital signal E output by the digital DAC copy circuit 118 comprises a stream of P-bit code words (where P>N, the higher resolution provided by P bits being necessary to provide fractional components necessary to account for effects of the mismatch error). The difference in bits (P−N) defines the degree of substantial equality that is achievable.

[0036] The operation of the replica DAC circuit 118 may be better understood through the use of an example. Assume that the scrambled thermometer coded selecting signal b.sub.i(k) would activate two unit current elements. If the analog N-bit DAC circuit 126 had an ideal functional operation, the analog voltage for the analog feedback signal D output from the analog N-bit DAC circuit 126 would have a value of 2*Δ. However, due to mismatch error, the voltage of the generated analog feedback signal D output from the analog N-bit DAC circuit 126 instead has a value of 2*Δ+e.sub.1Δ+e.sub.2Δ. The error estimation circuit 140 operates to process the scrambled thermometer coded selecting signal b.sub.i(k) and the ADC output v(k) to estimate the error e.sub.i for each of the unit current elements. The replica DAC circuit 118 is programmed with the estimation of the error e.sub.i. So, with consideration to the forgoing example, the digital signal E output from the replica DAC circuit 118 will be a code word with a precision of P-bits formed by summing the N-bit digital code for 2*Δ (i.e., the ideal response) plus the digital code for Δ times the sum of the programmed digital code words for the unit element errors e.sub.1 and e.sub.2 (i.e., Δ(e.sub.1+e.sub.2)) which is the introduced mismatch error.

[0037] The circuit 110 further includes a decimator circuit 114 that accumulates and averages the P-bit code words in the stream of the digital output signal E to generate a digital signal C comprised of a stream of multi-bit (M-bit) digital words at a data rate set by a decimation rate fd, where fd«fs and 1<N<P<<M. The decimator circuit 114 implements a low pass filtering to effectively remove the high-passed signal components of the quantization error and mismatch error.

[0038] As previously noted, the ADC output v(k) of the sigma-delta analog-to-digital converter circuit consists of: [0039] 1) the filtered input signal u′ (k) which is equal to u(k).Math.stf(k), where stf(k) is impulse response of the signal transfer function (STF(z)) of the system; [0040] 2) the filtered quantization noise q′ (k) which is equal to q(k).Math.ntf(k), where ntf(k) is the impulse response of the noise transfer function (NTF(z)) of the system; and [0041] 3) the filtered DAC error Σ.sub.i=1.sup.Nb′.sub.i(k).Math.e.sub.i, where b′.sub.i(k)=b.sub.i(k).Math.etf(k) and etf(k) is the impulse response of the error transfer function (ETF(z)) of the system.

[0042] The ADC output v(k) is thus given by the equation:

v(k)=u′(k)+q′(k)+Σ.sub.i=1.sup.N b′.sub.i(k).Math.e.sub.i

[0043] The error estimation circuit 140 receives the scrambled thermometer code word b.sub.i(k) output from the data weighted averaging (DWA) scrambler circuit 128 and the digital output signal B (v.sub.i(k)) produced by the N-bit quantization circuit 124. These signals are processed to estimate the unit element errors (the estimated errors referred to herein as ê.sub.i). The replica DAC circuit 118 is then programmed with these estimated errors ê.sub.i, and a DAC error correction term:

DAC err corr=Σ.sub.i=1.sup.N b′.sub.i(k).Math.{circumflex over (e)}.sub.i

is eliminated from the ADC output v(k) by the replica DAC circuit 118 to substantially cancel the DAC error Σ.sub.i=1.sup.N b′.sub.i(k).Math.e.sub.i and obtain a corrected ADC output v.sub.c(k):

v.sub.c(k)=u′(k)+q′(k)

[0044] The corrected ADC output v.sub.c(k) is then processed through the decimator 114 at the decimation rate fd to generate the digital output signal C.

[0045] Reference is now made to FIG. 5 which shows a block diagram for the error estimation circuit 140. In order to identify and formulate a measure that remains constant and is independent of the input thermometer code from the selecting signals, the term n.sub.i(k) is defined as:

[00001]$n_i\left(k\right)=b_i\left(k\right)-\frac{1}{N}{.Math.}_{i=1}^N{b}_i\left(k\right)$

[0046] A summation circuit 160 receives the scrambled thermometer code word b.sub.i(k) and calculates the Σ.sub.i=1.sup.Nb.sub.i(k) component and a divider circuit 162 applies the 1/N fraction. The output of the divider circuit 162 is applied to one input of a subtraction circuit 164. Another input of the subtraction circuit 164 receives the scrambled thermometer code word b.sub.i(k) signal and performs the subtraction operation to generate the term n.sub.i(k).

[0047] It can be shown that for all values of b.sub.i(k), the sum of n.sub.i(k) is zero:

Σ.sub.i=1 .sup.Nn.sub.i(k)=0

[0048] A determination is now made as to whether the generated term n.sub.i(k) has any impact on the actual DAC output and the corresponding ADC output. With reference again to the non-ideal DAC output of:

a.sub.actual(k)=ΔΣ.sub.i=1.sup.N b.sub.i(k)+ΔΣ.sub.i=1.sup.N b.sub.i(k).Math.e.sub.i

and, since Σ.sub.i=1.sup.N e.sub.i=0 for a given DAC, then:

a.sub.actual(k)=ΔΣ.sub.i=1.sup.N b.sub.i(k)+ΔΣ.sub.i=1.sup.N n.sub.i(k).Math.e.sub.i

[0049] As a result, the DAC error can be represented as:

DAC error=ΔΣ.sub.i=1.sup.N n.sub.i(k).Math.e.sub.i

[0050] Substituting for the term n.sub.i(k), the ADC output can be re-written as:

v(k)=u′(k)+q′(k)+Σ.sub.i=1.sup.N n′.sub.i(k).Math.e.sub.i

where: u′(k) is the input signal A passed through the modulator STF(z) and q′(k) is the shaped quantization error. A first signal conditioning circuit 168 receives the signal n.sub.i(k) and applies a filtering 168a and a decimation 168b to generate the signal n′.sub.i(k). Filtering is needed to remove the STF passed input signal u′(k) from the modulator output. This is done to de-correlate v(k) with the input signal u(k) because any signal component present in the output will interfere with the correlation operation. In this context, n′.sub.i(k) is the STF passed version of n.sub.i(k). Recall that n.sub.i(k) is injected into the loop (via feedback) at the summation node at the input.

[0051] So, n.sub.i(k) and u(k) both pass through an identical STF and are labeled as n′.sub.i(k) and u′(k), respectively. The filtering 168a has a transfer function that is a combination of error transfer function (ETF(z)) and a high pass filtering. As a result, the overall signal conditioning operation is therefore one of a polyphase decimation filtering.

[0052] Since the scrambled thermometer code word b.sub.i(k) signal and the input signal u′(k) can be assumed to be decorrelated, the correlation between v(k) and n′.sub.i(k) can be computed to obtain an estimate of the errors. The correlation may be mathematically expressed as:

CORR[v(k),n′.sub.i(k)]=CORR[u′(k),n′.sub.i(k)]+CORR[q′(k),n′.sub.i(k)]+CORR{[Σ.sub.j=1.sup.N e.sub.j.n′.sub.j(k)], n′.sub.i(k)}

[0053] For every i there are N number of j's where N is the number of DAC elements. Intuitively, the above-expression means that N bit streams (STF passed n′.sub.i(k)) emerging from feedback DAC input (because DAC has N inputs) when correlated with the modulator output (which is a summation of applied codes to DAC+DAC error represented by [Σ.sub.j=1.sup.N e.sub.j.n′.sub.j(k)]) provides N outputs that are proportional to the DAC element error). To elaborate further, ‘i’ is the digital part of the equation, where ‘i’ data lines exist as unique physical entities. ‘j’ is the index to represent DAC passed, STF passed, quantizer passed mashed/summed data, where ‘j’ individually is no longer discernable.

[0054] The first two terms of the foregoing equation evaluate to zero under the following assumed conditions: [0055] a) e.sub.i terms follow a normal distribution and are relatively constant over a period; [0056] b) the input signal u(k) is filtered out from the ADC output; and [0057] c) the quantization error q′(k) is white, uniformly distributed and uncorrelated to n′.sub.i(k).

[0058] With these assumptions in place, the correlation may be simplified and mathematically expressed as:

CORR[v′(k),n′.sub.i(k)]=e.sub.i+{Σ.sub.j=1,j≠i.sup.N e.sub.j(k).CORR[n′.sub.j(k),n′.sub.i(k)]}

where: v′(k) is the ADC output filtered so that it does not contain the input signal. The error estimation circuit 140 accordingly includes a second signal conditioning circuit 172 that receives the signal v(k) as output from the quantizer and applies a high pass filtering (HPF, reference 172a) and a decimation 172b function to produce the signal v′(k). The overall signal conditioning operation is therefore one of a polyphase decimation high pass filtering.

[0059] It will be noted at this point that the signal b.sub.i(k) branches out of the scrambler 128 into the following two paths: [0060] a) for v′.sub.i(k): the path comprises a DAC, the STF, and the quantizer followed by a polyphase decimation high pass filter (using the second signal conditioning circuit 172); and [0061] b) for n′.sub.i(k): the path comprises a digital approximation of the STF and a polyphase decimation high pass filtering (using the first signal conditioning circuit 168).

[0062] A correlation circuit 180 performs the correlation CORR[v′(k),n′.sub.i(k)]. This correlation operation produces a matrix of size N×1 as follows:

[00002]$\left[\begin{array}{c}\mathrm{CORR}\left[v^{\prime }\left(k\right),n_1^{\prime }\left(k\right)\right]\\ \mathrm{CORR}\left[v^{\prime }\left(k\right),n_2^{\prime }\left(k\right)\right]\\ .Math.\\ \mathrm{CORR}\left[v^{\prime }\left(k\right),n_N^{\prime }\left(k\right)\right]\end{array}\right]=R.Math.\left[\begin{array}{c}{e}_1\\ e_2\\ .Math.\\ e_N\end{array}\right]$

where:

[00003]$R=\left[\begin{array}{ccc}1& .Math.& \mathrm{CORR}\left[n_N^{\prime }\left(k\right),n_1^{\prime }\left(k\right)\right]\\ \mathrm{CORR}\left[n_1^{\prime }\left(k\right),n_2^{\prime }\left(k\right)\right]& .Math.& \mathrm{CORR}\left[n_N^{\prime }\left(k\right),n_2^{\prime }\left(k\right)\right]\\ .Math.& \ddots & .Math.\\ \mathrm{CORR}\left[n_1^{\prime }\left(k\right),n_N^{\prime }\left(k\right)\right]& .Math.& 1\end{array}\right]$

is an N×N Matrix.

[0063] The estimated error terms can be computed as follows:

[00004]$\left[\begin{array}{c}\widehat{e_1}\\ \widehat{e_2}\\ .Math.\\ \widehat{e_N}\end{array}\right]=R^{-1}.Math.\left[\begin{array}{c}\mathrm{CORR}\left[v^{\prime }\left(k\right),n_1^{\prime }\left(k\right)\right]\\ \mathrm{CORR}\left[v^{\prime }\left(k\right),n_2^{\prime }\left(k\right)\right]\\ .Math.\\ \mathrm{CORR}\left[v^{\prime }\left(k\right),n_N^{\prime }\left(k\right)\right]\end{array}\right]$

[0064] The R matrix consists of a correlation of n′.sub.j(k) and n′.sub.i(k) which can be computed easily. The estimated errors ê.sub.l, can then be calculated as:

[00005]$\widehat{e_{\mathrm{.Math.}}}=\frac{N-1}{N}.Math.\mathrm{CORR}\left[v^{\prime }\left(k\right),n_i^{\prime }\left(k\right)\right]$

[0065] Once ê.sub.l is estimated, the replica DAC at ADC output is populated with these estimated error values. The correction block to remove the DAC error in the replica DAC is a low rate multiply and add operation which can easily be accommodated in the subsequent digital filter chain. Because the estimation and computation operations run at a much lower rate than the modulator itself, there is a significant savings in power consumption. This solution is attractive for use in wide-band, high-speed, high performance sigma-delta modulators.

[0066] While the invention has been illustrated and described in detail in the drawings and foregoing description, such illustration and description are considered illustrative or exemplary and not restrictive; the invention is not limited to the disclosed embodiments. Other variations to the disclosed embodiments can be understood and effected by those skilled in the art in practicing the claimed invention, from a study of the drawings, the disclosure, and the appended claims.