Ferrari, Giorgio (Via del Faggio 3, Galliate, I-28066, IT)
Natali, Dario Andrea (Via Tasso 3, Torre Boldone, I-24020, BG)
Sampietro, Marco (Via Giuseppe Ferrari 14, Como, I-22100, IT)
Ferrari, Giorgio (Via del Faggio 3, Galliate, I-28066, IT)
Natali, Dario Andrea (Via Tasso 3, Torre Boldone, I-24020, BG)
|1.||An integrator circuit capable of integrating a current electrical signal (I), comprising an operational amplifier (7) equipped with an inverting input terminal (), a noninverting input terminal (+) and an output terminal (out), said electrical signal (I) being placed at the input to said inverting input terminal () of said operational amplifier (7), said output terminal (out) receiving feedback by means of a first feedback line (a) to said inverting input terminal (), said noninverting input terminal (+) being set at a reference voltage, characterized in that said operational amplifier (7) is further provided with feedback by means of a second feedback line (ß) to said inverting input terminal (), said first feedback line (a) being capable of producing a path for the highfrequency component (iIN) of said electrical signal (I) and said second feedback line (p) being capable of producing a path for the low frequency component (IIN) of said signal (I).|
|2.||An integrator circuit according to claim 1, characterized in that said first feedback line (a) comprises integration means (27).|
|3.||An integrator circuit according to claim 2, characterized in that said integration means (27) are a capacitor (CI).|
|4.||An integrator circuit according to claim 1, characterized in that said second feedback line (ß) comprises an electrical network (16) in series with an outflow network (y).|
|5.||An integrator circuit according to claim 4, characterized in that said electrical network (16) is a bandwidthattenuating stage with an attenuation value ("A") predefined for frequencies greater than a first frequency (46).|
|6.||An integrator circuit according to claim 5, characterized in that said bandwidthattenuating stage also has a predefined bandwidth (w), bounded at the lower end by said first frequency (46) and at the upper end by a second frequency (44).|
|7.||An integrator circuit according to claims 5 and 6, characterized in that said bandwidth attenuating stage comprises at least one operational amplifier (17,18) receiving feedback by means of an impedance (19,20), said impedance (19,20) comprising a resistance (21,23) in series with a capacitor (22, 24).|
|8.||An integrator circuit according to claim 4, characterized in that said second feedback line (p) produces a predefined gain (G) : 1 for the DC component (IIN) of said input signal (I) and a predefined attenuation (A) < 1 for the variable component (iin) of said input signal (I).|
|9.||An integrator circuit according to claim 8, characterized in that said predefined gain ("G") is comprised within the range of values 1<G<107.|
|10.||An integrator circuit according to claim 8, characterized in that said attenuation (A) has a value comprised within the range 105<G<1.|
|11.||An integrator circuit according to claim 4, characterized in that said outflow network (y) comprises an impedance having a resistance (Rf) and where appropriate a capacitor in parallel (Cp).|
|12.||An integrator circuit according to claim 11, characterized in that said outflow network (y) comprises an impedance having at least one active element.|
|13.||An integrator circuit according to claim 12, characterized in that said active element is a transistor of the MOS type, where appropriate operating in belowthreshold conditions.|
|14.||An integrator circuit according to claim 12, characterized in that said active element is a diode.|
|15.||An integrator circuit according to claim 12, characterized in that said resistive component (Rf) of said integration impedance (y) is deactivated by the electrical network (16).|
|16.||An integrator circuit according to claim 1, characterized in that said second feedback line (p) is a path for the DC component (IIN) and for the low frequency components of said signal (I) for frequencies (fmin) of less than 0.5 Hz.|
|17.||A transimpedance amplifier comprising an integrator circuit (15) to receive an electrical signal (I) from the transducer (1) so as to generate a first signal (V7B) and a differentiator circuit (29) capable of receiving said first signal (V7B) and capable of generating a second signal (VO=S), characterized in that said integrator circuit (15) is embodied according to claim 1.|
|18.||A transimpedance amplifier according to claim 17, characterized in that a first gain stage (30) is inserted between said integrator stage (15) and said differentiator stage (29).|
|19.||A transimpedance amplifier according to claim 17, characterized in that said differentiator stage comprises a second gain stage (31) placed in cascade with said first integrator stage (38) along the outgoing network.|
|20.||A transimpedance amplifier according to claim 17, characterized in that the pass band (B) of said transimpedance amplifier extends from milliHz to MegaHz.|
It is known that when there is a need to measure extremely small current electrical signals, it is necessary to have available measuring equipment capable of meeting particularly stringent requirements in terms of introduced noise.
Consequently this requirement makes it necessary to design equipment in which the introduced noise is as low as possible, which has a dynamic range as extended as possible and which has a bandwidth as extended as possible towards low frequencies so that even current signals varying slowly in time can be detected.
In fact, measuring equipment having such characteristics is used more and more widely in the most advanced technologies, such as for example in the experimental characterization of nanometric devices,
in electrical analysis systems associated with atomic- resolution vision (atomic force microscopy) and in the design of bio-electronic systems, where it is wished to analyse the electrical signal produced and transmitted by the biological material (individual proteins, networks of neurones or other cases).
The measuring equipment which can be used in the situations listed above can be reduced to a schematic diagram such as the example shown in figure 1; this consists of a transducer 1, capable of converting the value of a physical quantity which is to be measured in a given case into an electric current, a preamplification circuit 2, capable of amplifying said detected current value, and an instrument for reading the current and processing 3, capable of reading said amplified value and converting it into a useful signal, generally a voltage, for subsequent processing.
In particular, with reference to figure 2, it is known that the transducer element 1 can be represented by a generator of current IIN having in parallel a resistance RT and a capacitance CT, the latter being such as also to include stray phenomena due to the connection, while the amplification circuit 2 consists in an ideal form of an operational amplifier 7, having
an inverting input terminal"-", a non-inverting input terminal"+"and an output terminal"out". This operational unit 7 receives feedback by means of a capacitor CI and is connected to the following read circuit 3 by means of a capacitor Cd.
The read circuit 3 may be embodied according to the circuit diagram known as a differentiator, which comprises an amplifier 8 having for feedback a resistance #d and at the inverting terminal "-" the capacitor Cd.
The non-inverting terminals (+) of the amplifiers 7 and 8 are set at a fixed potential value, which is generally equal to zero.
In this way it is possible to measure the current I detected by the transducer 1 and, at least ideally, produce at the output of the amplifier circuit 3, a voltage Vour depending only on said detected current.
The diagram in figure 2 illustrates as a whole what is usually defined as a broad-bandwidth transimpedance circuit consisting of a pure integrator followed by a pure differentiator, that is the voltage VOUT produced at the output terminal of the circuit 3 is proportional to the input signal I multiplied by Cd/CI- This circuit operates in such a way that as soon
as the transducer 1 generates the current I, the latter flows entirely into the capacitor CI (this occurs only in an ideal case, that is where the operational unit 7 has infinite amplification gain and therefore the node 7A is a virtual ground point) so that it is possible for said current value I to be converted into the voltage V7B, and for this to be reconverted into current lour by means of the capacitor Cd and sent to the circuit 3 for final conversion into voltage vous, in accordance with the equation given below in terms of a Laplace transform, Voul (s) = I (s) Cd/C, Rd- Ideally, this circuit has an infinite bandwidth, limited in practice to high frequencies by the bandwidth of the operational units used. Figure 3A shows the transfer functions of the integrator alone (operational unit 7 and capacitance CI in figure 2), line 9, and of the differentiator alone (operational unit 8, resistance Rd and capacitance Cd in figure 2), line 9A, which provide the flat bandwidth"B" (figure 3B) from DC up to the limit of the operational amplifiers 7 and 8 of the transimpedance circuit illustrated in figure 2.
Referring now to figure 4, a representation is given of the current signal I at the input which may
be composed of a direct current value 10 (IIN) with a variable signal 11 (iin) superimposed, the latter being capable of representing the variations about the mean value of the physical quantity under examination.
In such circumstances, the circuit in figure 2 is ideal as regards bandwidth and noise, but nevertheless is capable of measuring the signal iin only in a limited time, defined by the saturation of the amplifier 7 because of the steady current IIN.
Where the value of the current IIN always has the same sign (i. e. is always positive or always negative), saturation of the operational amplifier 7 occurs in a"short"period of time"t".
This period of time"t"may be equal to 1 second or 1 minute or 10 minutes, for example, being deduced as a function of the values assumed by the current IIN and the capacitance CI.
This makes it impossible to carry out measurements beyond the duration of this space of time "t"without resetting the measuring equipment, that is the operation of discharging the feedback capacitance CI.
To make up for this limitation, in actual circuits a resistance Rf in parallel with the capacitance CI for feedback to the amplifier 7 is
added to the circuit diagram in figure 2, as shown in figure 5. The resistance Rf has the function of discharging from the capacitor CI the steady current integrated in it. It is chosen to have the maximum possible value so as to produce minimum noise, compatibly with the requirement of discharging the steady current IIN without saturating the operational unit. As a consequence of this necessary change, integration of the input signal Im is obtained only for frequencies greater than the frequency introduced by the pole, that is the circuit in figure 5 performs integration for frequencies higher than the frequency of the pole fp defined by the following formula: fp = 27t*l/CiRf.
This effect is shown in figures 6A and 6B. In particular in figure 6A it can be seen how the transfer function of the integrator comes to be altered because of the presence of the pole at frequency fp, line 9', while the transfer function of the differentiator circuit remains unaltered, line 9A.
The bandwidth"B'" (figure 6B) of the circuit as a whole is therefore limited by this.
Assuming that the resistance Rf is equal to 1 Gohm and that the capacitance CI is equal to 1 pF a dominant pole is obtained having a frequency equal to
approximately 160 Hz which therefore defines the pass band of the circuit and a maximum current IIN of the order of lOnA.
Use of larger resistances Rf so as to broaden the bandwidth, shifting the pole fp to a lower frequency, is difficult because they are bulky and have poor stability. Higher currents IIN and therefore smaller resistances Rf would result in still further reduced bandwidths.
In fact, referring to figure 6B, it can be seen that the minimum frequency which can be studied (that is fp) with this type of circuit is around 160 Hz.
Figure 7 on the other hand shows a graph of the equivalent noise at the input to the amplification circuit in figure 5.
In this case, the axis of abscissas indicates the frequency expressed in Hz, while the axis of ordinates gives the values of the input current noise spectral density expressed in 4/VHz ; from this graph it can be seen that at low frequencies the equivalent noise is determined by the resistance Rf alone, placed to provide feedback to the amplifier 7, line 12, while at higher frequencies the effect of the stray capacitance CT and of capacitance CI appears, line 13.
To eliminate the disadvantages mentioned of a
bandwidth limited to low frequencies and a limited dynamic range in the discharge of the steady current IIN, different variants to the circuit in figure 5 have been proposed, designed to extend the bandwidth of such amplifiers towards low frequencies without degrading performance in terms of noise.
One of these variants provides for a zero to be introduced, for example, to compensate for the frequency pole fp.
However, achieving constant gain in such amplifiers within the pass band is critically dependent, as a consequence, on meticulous calibration of the circuit components which introduce the zero.
Thus these components introduce a criticality such that to each operation to measure the current signal (iin) there must correspond a specific operation to calibrate said components.
Moreover, in many cases, there is a decrease in the gain of the integrator stage loop when the frequency increases, with a consequent reduction in the overall characteristics of the amplifier.
Another variant provides for example for the use of circuit configurations equipped with switches produced for example in MOS technology which allow discharge of the capacitance CI only when said switch
However, the operation of opening and/or closing the switch causes sudden variations in the gain of the loop of the circuit receiving feedback, thus altering the signal of the output terminal and introducing substantial disturbances in the output signal.
There is also the disadvantage, with the solution in figure 5, that the resistance Rf cannot be integrated into a single device together with the other devices constituting the measuring equipment because of its high value, and this in fact causes the measuring equipment to be less useful.
In view of the state of the technology described, the aim of the present invention is to produce an amplifier circuit capable of significantly extending the pass band towards low frequencies without in any way worsening the overall noise produced by the amplification stage or reducing its operating dynamic range.
According to the present invention, this purpose is achieved by means of an integrator circuit according to the characterizing portion of claim 1.
By means of the present invention it is possible to produce an integrator circuit equipped with an electrical network placed along the feedback loop
which makes it possible, for the same dynamic range value of the steady current IIN, to extend the available bandwidth by some orders of magnitude towards low frequencies or, for the same pass band, enables input steady currents IIN some orders of magnitude higher to be managed.
In fact it is possible to obtain a transimpedance amplifier circuit with a bandwidth which extends from tens of mHz to some MHz, keeping the noise introduced equal to that introduced by a conventional circuit, both having a resistance of equal value for feedback to the integrator stage.
Finally, since the gain of the amplifier is independent of the added network but depends solely on the relation between the two capacitances Cd and CI, the elements which compose the novel integrator circuit do not require particular precision characteristics, making the amplifier easy to produce despite their very high value and facilitating embodiment in an integrated device.
The characteristics and advantages of the present invention will become clear from the following detailed description of a practical form of embodiment of the invention, illustrated purely by way of non- limiting example in the appended drawings, in which:
- figure 1 shows a schematic block diagram of measuring equipment according to the known technology; - figure 2 shows an ideal form of embodiment of the schematic block diagram in figure 1; - figures 3A and 3B show graphs of the transfer functions of circuit elements of the measuring equipment in figure 2; - figure 4 shows an example of an input signal to the circuit in figure 2; - figure 5 shows a known embodiment of the schematic block diagram in figure 1; - figures 6A and 6B show graphs of the transfer functions of circuit elements of the measuring equipment in figure 5; - figure 7 shows a graph of the equivalent noise at the input of the measuring equipment in figure 5; - figure 8 shows a form of embodiment of the integrator circuit according to the present invention; - figure 9 shows a circuit diagram of a possible use of the integrator circuit in figure 8; - figure 10 shows a graph of the transfer function of the circuit diagram in figure 9; - figure 11 shows a graph of the equivalent noise at the input of the circuit diagram in figure 9; - figure 12 shows a first graph of the transfer
function of one of the circuit blocks illustrated in figure 8; - figure 13 shows a second graph of the transfer function of another of the circuit blocks illustrated in figure 8.
Referring now to figure 8, in which the elements already described are assigned the same numbers, the transducer 1 and the integrator stage 15 are shown.
The integrator stage 15 comprises, for example, the operational amplifier 7 which receives feedback by means of a first feedback line a and receives further feedback by means of a second feedback line P.
The first feedback line a comprises integration means 27 which in the case in question take the form of an integration capacitance CI, but could also be MOS transistors.
The second feedback line comprises an electrical network 16 in series with a network for outflow of the direct current y.
The outflow network y in the case in question in figure 8 takes the form of a resistance Rf with in parallel the spurious capacitor 9 Cp, but if it should be necessary to obtain the maximum possible dynamic range it is extremely advantageous to embody the outflow network y by means of active components, that
is by means of a MOSFET transistor, for example.
If it should be necessary to introduce the least noise it is extremely advantageous to embody the outflow network y by means of passive components, that is by means of resistances.
There is thus the advantage of being able to choose the configuration of the outflow network y to suit the specific requirements for which the equipment disclosed in the invention is intended.
It should however be noted that the resistance Rf may be non-precision and non-linear and therefore provision is also made for it to be embodied with MOSFET transistors operating below threshold where appropriate or it may be embodied using diodes.
The electrical network 16 has the function of introducing the attenuation"A" (or dually the gain "G") of the feedback line P.
This electrical network 16 may be implemented with an operational amplifier 17 receiving feedback by means of a feedback impedance 19.
In the specific form of embodiment in figure 8, the network 16 comprises a cascade of two operational amplifiers 17 and 18, each of which receives feedback by means of an impedance 19 and 20.
In particular, the impedance 19 is constituted by
a resistance 21 (Rzl) in series with a capacitor 22 (CAcl), as is the impedance 20, which is also constituted by a resistance 23 (RZ2) and by the capacitor 24 (CAC2) in series. For example, the values of the capacitors may be equal to 1 microFarad, while the values of the resistances may be equal to 2.7 MOhm.
It should be also noted that the resistances and capacitances in the circuit 16 may be non-precision and non-linear and therefore provision is also made for them to be embodied with MOSFET transistors operating in the below threshold ohmic zone.
The electrical network 16, in the specific form of embodiment, produces the attenuation"A" (or gain "G") as a function of the product of the gains of the operational amplifiers 17 and 18.
This attenuation"A"can be altered as desired to suit the specific requirements for which the circuit 14 disclosed by the invention is intended.
The interval within which"A"can vary is between 10-2 and 10-5.
The interval within which"G"can vary is between 1 and, ideally, infinity. Realistically"G"is between 1 and 107.
Moreover from the diagram in figure 8 it can be
seen that the inverting input terminal of the operational amplifier 17 is connected to the output terminal"out"of the operational unit 7 by means of a resistance 25 (Rp1), just as the output terminal of operational 17 is connected to the inverting terminal of the operational unit 18 by means of another resistance 26 (Rp2)- Between the output terminal of the amplifier 18 and the outflow network y is placed a filtering network 45 comprising a resistance 41 (Rh) with a capacitor 42 (Ch).
For example the value for the capacitor Ch is equal to 1 microfarad, while for the resistance Rh is equal to 4.7 kOhm.
This filtering network 45 is intended for filtering out at medium and high frequency the noise produced by the network 16 which, via the spurious capacitance Cp, might be injected as current into the node 7A.
The transfer function of the circuit 16 is illustrated in figure 12, in which the axis of abscissas gives the frequency expressed in Hz and the axis of ordinates gives the gain of the filter.
The pole 44 shown in figure 12 is that introduced by the resistance Rh and the capacitance Ch, while the
zeros 46 are those introduced by the feedback impedances 19 and 20.
Figure 13 shows the transfer function of the operational unit 17 only and of its feedback impedance 19, from which it can be seen that a zero is introduced at fz (due to the feedback impedance 19) and above all the attenuation"A".
It will also be noted that the frequency axis is cut with a gradient of-20dB, a value which may be increased to-40dB, for example, in accordance with the circuit of the electrical network 16.
The bandwidth"w"of this transfer function cannot be unlimited and therefore at the frequency f a pole or a zero is introduced.
For example, in accordance with the diagram of the electrical network 16, this # corresponds to the frequency of the pole introduced by the network 45.
The integrator circuit 14 just described in terms of its basic devices will now be illustrated in operation.
As previously described, the transducer 1 generates the current signal I composed by superimposing the DC signal IIN (10) and the variations iin (11)- The circuit 14 must be capable of detecting and
amplifying the fluctuation 11 in the current signal while at the same time preventing the direct current (or mean) component 10, IIN, of the current from leading to saturation of the amplifier 7.
Advantageously, the introduction of the network 16 is capable even at very low frequencies, by reason of the high attenuation achieved, of deactivating the path of the outflow network, that is it is capable of deactivating the resistance Rf in parallel with the spurious capacitance Cp, causing opening of the feedback line P.
In this way from the circuit node 7A, which operates as a virtual ground, the electrical network 16 is capable of drawing off a current equal to the mean value 10 (IIN) of the input current but lets through the AC component 11 (iin) in the first feedback line a, that is the electrical network 16 is capable of reading the output voltage V7B from the amplifier 7 by means of its components and processes it so as to generate a mean current equal to the mean value of the current IIN so as to cancel out entirely at the virtual ground point 7A the mean current generated by the transducer 1, in accordance with Kirchoff's laws.
Thus the electrical network 16 does not act on the AC component of the input current, that is the
signal 11, but processes the components at very low frequency from DC up to the frequency fmin, that is the signal 10.
Advantageously the electrical network 16 does not require calibration operations for a defined mean current generated by the transducer 1, but is capable of adapting dynamically to the value of the mean current generated by the transducer 1.
Thus the feedback line P of the operational unit 7 sees a high gain"G" (because of the presence of the cascaded operational units 17 and 18) equal to 107, for example, on the DC component IIN of the current signal, preventing said DC component 10 at low frequency from passing into the capacitor CI and assisting it to pass instead into the resistance Rf.
Conversely, when operating at higher frequencies, that is where the signal of interest, namely the AC component 11 of the current signal I, is located, it is the case that the feedback line ß, and in particular the electrical network 16, no longer amplifies the output voltage V7B of the operational unit 7 because the feedback line is open but produces an attenuation"A", equal to 10-6<A<10-1, for example.
In this way, the"out"terminal of the
operational unit 18 is at a fixed potential.
There is therefore a situation in which the outflow network y is placed between two points at almost fixed potential, that is in this outflow network y it is no longer possible for the input signal current iin to flow. Therefore the input current iin flows into the capacitor CI.
In other words, by means of the electrical system 16 it is possible to increase the rms value of the feedback resistance Rf when the frequency increases, so as to bring capacitance CI into play from the lowest frequencies, so as to extend enormously the operating bandwidth of the integrator without increasing its noise or decreasing the dynamic range.
It should also be noted that by increasing the values of Rp and reducing Rz in the circuit 16 it is possible further to reduce the minimum frequency fmin analyzable by the circuit disclosed by the invention, though to maintain the stability of the circuit disclosed by the invention it is necessary that in the vicinity of the minimum analyzable frequency fmin the gain of this filter 16 should be almost constant.
Referring now to the circuit illustrated in figure 9, in which the same numbers are assigned to the elements already described, an illustration is
given of an example of a possible application of the integrator circuit in figure 8 which is the subject of the present invention.
As already stated previously, the aim in view is not usually integration of the signal generated by the transducer but amplification of said input signal and where appropriate conversion of it into a voltage signal.
Consequently, as can be seen from figure 9, in series with the integrator circuit 15 disclosed by the invention there is a differentiator stage 29.
It should be noted that between this integrator stage 15 disclosed by the invention and said differentiator unit 29 an (optional) gain stage 30 has been inserted, capable of amplifying the signal generated by the integrator circuit 15.
The amount of the gain introduced by this stage 30 is linked to the ratio resulting from the feedback resistance 33 (R2) and the input resistance 32 (Ri) of the operational unit 34, that is a gain equal to (1+R2/Rl) is obtained.
It will also be noted that at the output from the operational unit 38, there is a further gain stage (also optional) 31, capable of extending the bandwidth of the differentiator stage 29 so as to place a
voltage signal Vour'at the output.
The differentiator stage 29 consists of an operational amplifier 38 (where appropriate in cascade with the gain stage 31) receiving feedback by means of a resistance 39 (Rfd) and a capacitor 40 (Cfd) connected to each other in parallel and having a capacitor 41 (CD) at the inverting input terminal.
In this way a circuit is obtained consisting of an integrator 15 cascaded with a differentiator 29 capable of operating within the entire bandwidth of the integrator 15.
Given that the bandwidth "#" of the integrator 15 is very large, for example with the values cited previously it extends from mHz to MHz, a circuit with a very large bandwidth is produced.
In fact, referring now to the graph in figure 10, it can be seen that the circuit in figure 9 has a pass band "#" three orders of magnitude greater than what is obtainable under the same conditions with the configuration described in figure 5, this value corresponding to the attenuation of three orders of magnitude of the feedback network ß in figure 12.
In particular, with reference to figure 10, in which the broken line again shows the transfer function of the circuit in figure 5 while the
continuous line shows the transfer function of the circuit illustrated in figure 9, it is quite clear what an advantage is offered by the presence of the integrator 15 disclosed by the invention.
In particular it can be seen that the transfer function of the transimpedance amplifier in figure 9 provides for the bandwidth # to extend, in the particular form of embodiment, over a frequency interval of 0. 5 Hz to 2.9 MHz.
The lower limit of this interval (namely fmin = 0. 5 Hz) is therefore the frequency above which the AC component of the input signal I passes into the feedback line a, that is it is the frequency below which the DC component of the input signal flows into the feedback line ß.
Therefore the electrical network 16 placed along the feedback loop ß deactivates the resistive path (namely Rf) in parallel with the integration capacitance (namely Cf) allowing the pass band of the circuit in figure 9 to be extended significantly towards low frequencies.
By increasing the low-frequency attenuation of the electrical network 16, it is possible to reduce further the minimum frequency fmin analyzable by the circuit. In fact it is possible to reach frequencies
of the order of a few tens of mHz, for example fmin = 10 mHz.
However, to maintain the stability of the circuit in figure 9 it is necessary that in the vicinity of the minimum frequency analyzable the attenuation supplied by the network 16 should be almost constant.
Moreover, considering figure 11, in which the equivalent noise of the circuit in figure 9 is given, it is easy to deduce that this electrical circuit 16 has in no way worsened the overall noise produced by the stage or reduced its dynamic range of operation, and this graph can be superimposed perfectly on the one shown in figure 7.
Clearly a person skilled in the art, for the purpose of meeting incidental and specific requirements, will be able to make numerous changes and produce numerous variants to the configurations described above, all of these being also included within the scope of protection of the invention as defined by the following claims.
Next Patent: METHOD AND DEVICE FOR SUPPLYING AN ELECTRIC CONSUMER WITH DIRECT CURRENT