WO1991012533A1

WO1991012533A1 - A method for differential sampling

Info

Publication number: WO1991012533A1
Application number: PCT/SE1990/000079
Authority: WO
Inventors: Henrik ALMSTRÖM
Original assignee: Almstroem Henrik
Priority date: 1988-08-25
Filing date: 1990-02-07
Publication date: 1991-08-22
Also published as: SE8802974D0; SE8802974L; SE461935B

Abstract

The present invention concerns a method for differential sampling with a fixed sampling frequency fo. When sampling a signal varying between +Vmax and -Vmax, the dynamics of the sampled signal is large close to the ends of the interval but it is small close to zero. The present invention, however, gives a solution to the problem of measuring the fine structure of a signal, with constant resolution over a large amplitude interval. This is done by measuring an analogous signal and on every sampling occasion directly sampling the signal V and forming the difference, Vdiff, between the signal on this sampling occasion and the preceding. The difference is then amplified in order to increase the dynamics in the following sampling and the present signal value is reconstructed from the direct value V of the signal at a starting point, the initial value, and the sum up to the present time of the following differential values Udiff, (6), divided by their amplifications.

Description

A method for differential sampling

The present invention concerns a method for differential sampling with a fixed sampling frequency f₀.

When sampling a signal varying between +V_max and -V_max, the dynamics of the sampled signal is large close to the ends of the interval but it is small close to zero. Of course, the dynamics could be generally increased by sampling the signal with a larger number of bits, e.g. 16 bits instead of the usual 8 or 12. Close to zero the signal is still given with a small number of significant figures. Besides, a 16-bit A/D-converter is normally slower than an 8-bit or a 12-bit A/D- converter. In attempts to overcome this it has been suggested to amplify the signal so that its absolute value is only a little less than V_max. This can be done by simultaneously amplifying the signal with a number of fixed steps and sampling the most suitable. Information about the amplification together with the sample value gives all necessary information.

This method, however, does not solve the problem of measuring a small signal which is riding on a larger signal. When the absolute value of the large signal is close to V_max the signal can not be amplified and the small signal is not noticeable. In certain cases when the signal is periodic it would be possible to filter away a slowly varying signal of high amplitude and only retain the interesting fast signal of low amplitude and then amplify and sample it with good dynamics. In other cases, and especially in transient processes, such filtering can not take place with preservation of the information in the signal.

The present invention, however, gives a solution to the problem of measuring the fine structure of a signal, with constant resolution over a large amplitude interval by designing it as is evident from the following claims.

In the following, the invention will be described in more detail, with references to the enclosed drawings where fig 1 shows the invention schematically, fig 2 shows schematically a device for achieving increased dynamics, that is included in one of the embodiments of the invention,

fig 3 shows a diagram of a measuring sequence,

fig 4 shows a diagram of another measuring sequence and fig 5a-c shows the different parts of another embodiment of the device to achieve increased dynamics.

The method means that a measuring sequence is analogously measured by a sensor 1. The signal is amplified in the usual way in an amplifier 2 and is suitably then passed through a low-pass filter 3. The low-pass filter establishes an upper limiting frequency in such a suitable relation to the sampling frequency f₀ of a following sampling that the sampling theorem is met. The signal is then fed to a device 4 for achieving increased dynamics. In this there is a feed-back coupling from the A/D-converter to a delay circuit in the device which makes sure that the delay,τ, is I/f₀ all the time. In the device for achieving increased dynamics the difference V_diff between the signal at one sampling occasion and the signal at the preceding one is formed. The signal difference V_diff is then amplified to a suitable extent. This can be done in a differential amplifier 8. From the device 4 two signals emerge: the direct signal 5 as sampled at a given moment and the said amplified signal difference U_diff, 6. These two signals are fed to a computer 7 where the A/D-conversion is completed by reconstructing the momentary value from an initial value of the direct signal 5 and the sum of the thereafter up to the actual time established signal differences U_diff, 6, divided by their amplifications.

Thus the new feature of the device is that, apart from recording the direct signal, V, (n bits), successive parts of the signal V_diff, see fig. 3, are measured. In the figure we have

V_diff(t) = V(t) - V(t-τ),

τ = l/f₀ f₀ = the sampling frequency. When the signal has been recorded in the computer it can be reconstructed according to the expression V(mτ)-V(0) + (iτ), m= 1, 2...

To avoid sampling errors during the A/D-conversion, the signal is low- pass filtered at for instance f₀ /4 Hz before the A/D-conversion. If the signal is filtered somewhat harder, the signal difference V_{dif f} will well approximate the time derivative of the signal.

((m - 1/2)τ) * V_diff(mτ)/τ , m = 1, 2...

The advantage of this expression compared with the expression above is that the absolute accuracy is constant, while it decreases with increasing m in the previous expression.

The signal difference V_diff is amplified before the A/D-conversion with an amplification factor K, where K is chosen such that

|U_diff|<V_max' where U_diff= K·V_diff, V_diff is then obtained from

V_diff=

U_diff.

Then the dynamics increases to n +

bits.

If the amplification K occasionally is too large, that is, the resulting value of U_diff is over-modulated (e.g. 12 zeros or 12 ones), the directly registered signal can be used to get a new starting point for the differential registration according to fig. 4. V(mτ) = V(0) + V_diff(iτ), m - 1, 2, ..., j

V(mτ) = V((j+1)τ) + v_diff(iτ), m = j+2, ...

The device has the property that, with increasing sampling frequency f₀, the amplification K can be increased, which gives an increased dynamics of the signal. For instance, with K=128, the resolution is given by 19 bits, using a 12-bits A/D-converter.

In one embodiment of the invention, the signal difference V_diff is amplified with a constant value K in a differential amplifier 8. When choosing the value K, one faces two conflicting interests. On the one hand, one wants the amplification to be as large as possible, on the other hand, one does not want the equipment to be over-modulated, that is to give 12 ones or 12 zeros. One has to balance the amplification K in such a way that the probability that the equipment is over- modulated is acceptably low. If the equipment is over-modulated one uses the direct signal of the next sampling occasion to get a new initial value, after which the following values are calculated from this initial value and the following amplified signal differences U_diff, 6, in the same way as in the Previous calculation.

In another embodiment of the invention, the signal difference V_diff is amplified in a number (here exemplified by 3) of parallel amplifiers 10, 11, 12, having different amplifications K₁, K₂ and K₃ where K,=l. These signals are then compared, in comparators 13, 14, 15, 16, with the given maximum value ± V_max of the equipment. With the help of the logical circuit 17, the most amplified signal 18, the absolute value of which is less than V_max, is fed into the computer 7 together with information about the used amplification 19. At the continued A/D-conversion the instantaneous value is reconstructed from an original initial value 5 and the sum of the thereafter up to the actual time measured amplified signal differences U_diff, 6, divided by their respective amplifications K₁, K₂ and K₃. Here a new initial value will normally never be used, as a value smaller than V_max is selected. In the example shown only three amplified signal differences have been used U₁= ±K₁ V_diff, U₂= ±K₂ V_diff and U₃= ±K₃ V_diff, where l = K₁< K₂< K₃. In the example, the time delay is caused by two sample-and- hold circuits 20, 21, which makes the U_i alternatingly change sign. To decrease the load of the sample-and-hold circuits 20, 21 , a i sol ation circuit 22, 23, is used before each sampl e-and-hold circuit. The signal is obtained from

V(mτ ) = V(0) ) (

where

Depending upon the circumstances it is of course in an actual case possible to use any different number of amplified, with different amplification, signals differences U_diff that is suitable.

Claims

Claims:

1. A method for differential sampling at the sampling frequency f₀ characterized in that an analogous signal is measured, that on every sampling occasion the signal V is directly sampled and the difference,V_diff' between the signal on this sampling occasion and the preceding is formed, that the difference is amplified in order to increase the dynamics in the following sampling and that the present signal value is reconstructed from the direct value V of the signal at a starting point, the initaial value, and the sum up to the present time of the following differential values U_diff, (6), divided by their amplifications.

2. A method according to claim 1, characterized in that the signal difference V_diff is amplified by a constant amplification which is chosen as large as possible taking into account that the absolute value of the amplified signal difference U_diff must not exceed the maximum signal voltage V_max of the equipment more than a limited number of times per time-unit and that every time the absolute value of the amplified signal difference U_diff has exceeded V_max the next direct signal V is used as a new initial value.

3. A method according to claim 1, characterized in that the signal difference V_diff is simultaneously amplified with a number of fixed amplifications, after which that amplified signal difference U_diff is chosen whose absolute value is less than, but closest to, the maximum signal voltage V_max of the equipment.