DE2511180C2 - Method and arrangement for increasing the resolution of analog-digital converters - Google Patents

Description

The invention relates to a method for increasing the resolution of analog-digital converters, which consist of a sampling device and an actual analog-to-digital converter, as well as so an arrangement for the implementation of this procedure.

A complete analog-to-digital converter is shown in block diagram form in FIG. He persists from an input 1, a sampling-hold device 2

with a sampling frequency f _a = -, where T the time-adjustable conversion accuracy is limited by the amount by which the signal still changes during the time A t _c. In this context, A t _c ia is called the aperture time of the actual analog-to-digital converter.

The smallest possible aperture time is desirable. It can be reduced by adding a scanning / holding device. The aperture time of the ideal sample and hold device is zero. It holds exactly the signal χ (/ J present at the time / ". Real sample-and-hold devices have enough aperture times, which for high-speed applications are in the order of magnitude of 1 η sec. They arise from fluctuating small deviations (jitter) from the exact one Sampling time.

The sampling jitter Δ t _a is converted into corresponding fluctuations in the output signal x (t _a ) of the sample and hold device if the slope x (t _a ) of the signal curve x (t) at the sampling time t _{a is} not equal to zero

is. For a signal limited to B <] -f _a (sampling theorem), the greatest possible slope is

if x _{max is} the maximum signal amplitude that is permitted by the modulation range of the converter. The greatest deviation of the output signal from the correct value is therefore

Ax

For a sampling frequency of f _a = 1 MHz and an aperture time of At ₀ = In see, no better resolution than 3% can be achieved. This corresponds to a quantization with 9 bits including the sign bit.

For certain applications, e.g. B. the quantization of RF signals for digital signal processing, high resolution with a high sampling frequency is required. Actual Analog-Dig '^ I converters with a resolution of 13 to 15 bits and a conversion time A t _t of 1 μsec are now possible. They require a sample-and-hold device with an aperture time of the order of magnitude f> ini <T £> r \ C \ r \ car- XkTto koraitcr ot-cuXUnt Ιΐοηοη Λ \ α. A nor-

tur times of the currently available sample-and-hold devices for high-speed applications but in the order of 1 η sec.

The invention is based on the object of providing an analog-digital converter of the type mentioned at the beginning create that allows the o. g. To achieve aperture times.

According to the invention, the object is achieved in that the on the predictable portion of the aperture to be returned to the sample-and-hold device

A Miotfakiar o1

is considered, a proper analog-to-digital converter 3 and an output 9. As will be seen from Fig. Ib, the sample and hold device holds the at sampling instant t _a present time-varying signal χ (t) for a certain period of time At <Tfest , in which the actual analog-to-digital converter carries out the implementation.

If the signal bandwidth is very small compared to the reciprocal conversion time \ IAt _{c of} the actual analog-digital converter, then a sample-and-hold device can be dispensed with. The highest signal change rate achieved is determined and stored and that the respective conversion result of the analog-digital converter is corrected by the associated stored sampling error in accordance with the associated parameters signal amplitude and signal change rate.

An expedient embodiment of the method according to the invention provides that the stored Sampling errors are supplemented by an allowance through which the stable linearity errors of the analog-to-digital converter be compensated.

In an advantageous arrangement for performing the method according to the invention is in addition to Analog-digital converter consisting of a sample-and-hold device and an actual analog-to-digital converter provided with a differentiating stage downstream auxiliary converter, which consists of a further sample-and-hold device and a further actual Analog-Digitai-Wandier consists, whereby the actual analog-digital converter and the further one the actual analog-to-digital converter is connected on the output side to the inputs of a read-only memory, the output of which is connected to a first input of a summing element, with its second input the output of the actual Aralog digital converter connected is.

The preferred read-only memory is »? Π rsur read-only memory (ROM) to use.

The aperture time consists of three parts. The largest part is the "aperture delay time". This means the mean value before A t _a , i.e. the mean delay between the sampling command (time t _a ) and the actual sampling time. This component usually does not interfere, as it only corresponds to an even delay in the signal * -

The fog is orders of magnitude smaller. "Aperture jitter" (aperture delay jitter). Through it, the resolution of the analog-digital converter is explained in the above Way worsened. The aperture jitter is largely dependent on the signal. This part, which can be predicted from the signal will be called aperture modulation in the following. In the end there remains a part that cannot be predicted, which arises from a pure process of adding and here Aperture noise should be called.

In good sample and hold devices this is aperture noise Experience has shown that it is again orders of magnitude smaller than the entire aperture jitter. This offers the possibility of compensation of the predictable portion of the jitter to suppress the aperture modulation and thus the resolution-limiting Reduce apt "tur time quite considerably.

The aperture modulation is by definition lawful, because the non-social, i.e. non-determined, The stochastic portion of the aperture error was summarized as "aperture noise". It must now bioß has shown that the lawful portion s'arfc predominates, so that its compensation has a worthwhile effect on the / perture error.

To put it simply, the sampling in the sample-and-hold circuit is about the switching through of biased diodes with the aid of a switching voltage U ₀ . Since the rise time of the SchaKianke is not zero, but may have the value ι , a time A t, which is dependent on the bias voltage of the diode, elapses until the switching time. 100% of the instantaneous value of the sampled signal U i is included in the bias voltage. Consequently, the term: i "unki becomes by ¹ J be? Inf! Uß? (Anprtnr.

modulation). A rough estimate for this is

At =

The order of magnitude of the resulting aperture error is at least a few 10 ps if one assumes a technically possible minimum rise time of the switching edge of a few tenths of ns and a voltage ratio U ₀ TO ₁ = 10.

In comparison to this one has to evaluate the stochastic fluctuations. They arise from the phase noise of the oscillator controlling the sample-and-hold circuit and from thermal noise and shot noise in the sample-and-hold circuit itself, for low-noise oscillators in the frequency range 1 ... IGO MHz, which is the most important for the analog-to-digital converter according to the invention, the relative fluctuation of the period is on the order of magnitude at most 10 " ⁷ , that is to say in absolute terms far below 1 ps. The proportion of the aperture noise is thus small enough that compensation for the aperture modulation is definitely worthwhile.

The regularity of the aperture modulation, i.e. the exact functional relationship between the scanned Signal and the error, of course, is not very easy to formulate. It depends on the chosen Circuit off and becomes even more complicated compared to the derivation used above as a result of the Stray and Junction Capacities. Therefore it is useful to express the law in its general form to describe a series expansion and to assign the initially unknown coefficients by piecing determine. To do this, you need sufficiently precisely defined analog input signals with a sample-and-hold circuit are scanned. By comparing the sample value sequence obtained with the target values one obtains the errors and from them the unknown coefficients by means of a compensation calculation. the Series expansion is described below.

The aperture modulation At _mod can be derived from the signal x (t _a ) :

+ V - * ('J + V * ² (f "> + V * ³ ('") + ·· ■

+ mixed links.

This general form !, which has any dependency allowed to describe, can now in practice be reduced to a few terms, without larger ones Missing - arise, for example:

^ t _mnd = O ₁ χ + a _: ■ x ² + b, ■ χ + b ₂ ■ x ² = f (x, x).

The effect of the Aperturm.oduiaiion ajf the scanned The signal is then given by the relationship:

Ax " _md Ua) = / IxC). xUJ) - .i (t _a ) = F (x. x).

A simple way to compensate is to use the entire Fix. x) , to store it in a read-only memory (read only memory) and then to reduce each digitized sample value x (t _a ) in accordance with the associated value pair χ, χ by dXmoaUa).

Fin vnrtpilhaffpc Aucfiihrnnochjsicnii »! t * ir \ pr Annrrf-

tion to carry out the procedure according to Invention is illustrated in Fig. 2 and will be described in the following describe in more detail.

The time-variable signal x (t) to be converted is fed via an input ί on the one hand to a sampling device 2 with a downstream actual analog-to-digital converter 3, with its aid

the digitized sampled values still afflicted with an error caused by the aperture modulation .v (/ ") are obtained, and on the other hand one Differentiator 4 with downstream auxiliary converter

for the purpose of sampling and digitizing the time derivative χ (t). The auxiliary converter consists of a further sample-holding device 5 and a further actual analog-digital converter 6 and is designed for a few bits in accordance with the desired improvement in resolution of the arrangement. Associated value pairs χ (/ "), χ (/") are given to the inputs of a read-only memory 7, which contains, for example, 1024 correction values, whereby the whole number space for the roughly quantized quantities χ and χ is covered. A correction value Δ * "*" / ('") = F (x ('J> * ('J)) corresponding to the value pair χ (t _a ), χ (t _a ) on the input side is transferred from the read-only memory to a first Input of a summing element 8. Since the sample x (t _a ) digitized by the analog-to-digital converter 3 is simultaneously present at the second input of the summing element, a digitized sampling value is obtained at one output 9 in which the error caused by the aperture modulation is compensated for .

The limit of the resolution improvement is determined by the aperture noise, which cannot be determined from the signal. In the case of the best abbot holding devices available today with Λ hoBch ^ -S P see, it is about two orders of magnitude below

the aperture modulation (At _milJ ^ 250ρ see). With the measures described, a 500 kHz wide band with a resolution of 14 to 15 bits can be digitized. The correction function F (x _f x) is expediently supplemented by a supplement that compensates for the stable linearity errors of the converter.

The individual correction function of an analog-digital converter can be determined by means of a signal analysis with two measuring transmitters. The signals of the two measuring transmitters of unequal frequency ^ <&</ _a / 2 are added in an analog summing network and thus applied to the analog-digital converter. The digital samples on the output side are converted into the spectral range in blocks with the aid of the discrete Fourier transformation. In addition to the two measuring transmitter frequencies and their harmonics, the spectrum shows a number of combination signals at the frequencies jf + a, f _t -j5, If _x + A, 2Ji -Λ, lh "Ti, .. Distortions of the sample-and-hold device-analog-digital converter system, caused in particular: by the aperture modulation, arise The coefficients of the correction equation F (x, x) can be calculated from the distribution of the combination signals.

1 sheet of drawings

Claims (4)

Patent claims: 1. Method for increasing the resolution of analog-digital converters that consist of a Sampling-Haite device and an actual analog-to-digital converter exist, characterized in that that the on the predictable portion of the aperture of the Abtasi-holding device Sampling errors to be traced back as a function of the signal amplitude and the rate of signal change is determined and stored and that the respective conversion result of the analog-digital converter according to the associated parameters signal amplitude and signal rate of change is corrected by the associated stored scanning error. 2. The method according to claim 1, characterized in that the stored scanning errors by a guarantee can be added through which the sta- 2Q bile linearity error of the analog-digital converter compens. be rt. 3. Arrangement for carrying out the method ^ according to claim 1 or 2, characterized in that that in addition to the analog-to-digital converter consisting of a sampling Haite device (2) and an actual analog-to-digital converter (3) a differentiating stage (4) is provided with a downstream Auxiliary converter consisting of a further scanning-Haite device (5) and a further 3C actual analog-to-digital converter (6) consists that the actual analog-to-digital converter (3) and the other actual analog-digital converter (6) on the output side with inputs of a read-only memory (7) are connected, the output of which is connected to a first input of a Sumrni, r link (8) whose second input is the output of the actual analog-to-digital converter (3) is connected (Fig. 2). 4. Arrangement according to claim 3, characterized in that that the read only memory is a read-only memory (ROM).