DE2720653C2 - - Google Patents

Description

The invention relates to a circuit arrangement to correct the voltage dependency of a semiconductor resistance according to the preamble of claim 1.

Circuit design is becoming more and more integrated Circuit components used. These semiconductor structures have changed in recent years as a result of multiple improvements in basic active and / or passive components further developed. In addition to the so-called active components, including transistors, Diodes etc. are understood to play the passive construction elements, especially resistors, play an essential role for many circuit types, especially for circuit applications related to analog circuits. To counter stands with the active components in integrated circuits to provide together can essentially two Techniques are used: one is the so-called compatible Hybrid technology in which the resistors in the form thin layers on the surface of an integrated half printed circuit boards are formed. The other possibility is the monolithic technology where the Resistors in the semiconductor substrate itself in the same How the active components are designed. in the the latter case, the resistors z. B. by a thermal diffusion process or by means of ion implant tion are produced.

Thin film resistors generally consist of one suitable metal, e.g. B. a nickel-iron-chromium compound or tantalum. These types of resistors offer a number  of merits, including a sheet resistance value in the range of 200 to 10,000 Ω / and a relative low temperature coefficient for resistance value (TCR) of the order of 3% between 0 ° and 75 ° C. The disadvantage, however, is that the thin layer often resisted additional procedural steps (Vacuum deposition, sputtering) require that mostly with those used to manufacture the semiconductor die not to combine self-used process steps are cash. This immediately increases the yield good circuits deteriorated. producer of therefore consider monolithic circuit concepts the thin film technology as relatively unattractive for Low cost mass production. The mono In contrast, lithic technology appears to be the most attractive most effective solution.

Diffused resistors are known per se. they generally consist of one simultaneously with the base or the emitter of a bipolar transistor Funding area. For example, a becomes conventional P-doped base region into a weakly N-doped Diffused epitaxial layer that spreads over a weak P-doped single crystal silicon substrate extends. The resistance value depends on a number of influencing variables from, e.g. B. the doping profile, the penetration depth of the Dopants, the length / width ratio of the diffuse dated zones etc. at the ends of the diffusion area then two ohmic contacts are formed, which are the off represent the connection of the resistor. The most positive Potential of the circuit is generally applied to the area the epitaxial layer, which is within the of the Isolation included area and the cons area includes. One is limited in such Resistance to the low sheet resistance (400 Ω /), the relatively large fluctuation range of the  Nominal values (± 20%) as well as regarding the high Temperature coefficient, due to which the resistance depending on the temperature can change significantly. On the other hand, speaks for these resistances her simple and well known Manufacturing route.

Resistors manufactured by means of implantation processes a particularly interesting alternative to diffused Resistances because they have significant advantages over them Offer. Boron implanted resistors in particular a large sheet resistance range (3 to 10,000 Ω / and above) and a low temperature sensitivity (Resistance change of 0.5% between 0 and 75 ° C). Furthermore, such resistors can be manufactured relatively precisely (± 2%). A more detailed description of this type of resistance can be found in an article by K. Rosendal with the title "Ion-Implanted Planar Resistors", in the magazine Radiation Effects, Volume 7, January 1971, pages 95 to 100.

A comparable representation of the three mentioned above Resistance types are provided by an article by H. H. Stellrecht entitled "Precision Ladder Networks Using Ion Implanted Resistors ", in the Wescon Technical Papers magazine, Volume 15, 1971, ref. 8080-28 / 2 and an article by Oosthoek, Den Boer and Hofker entitled "The Termal Properties of High Value Gallium and Boron Implanted Resistors in Silicon " Connection with the "European Conference on Ion Implantion" from September 7th to 9th, 1970.

However, little attention has so far been paid to the increasingly important problem of the voltage dependence of the resistors. The current-voltage characteristic of a resistor should normally be linear. However, this can actually only be achieved for implanted resistors with a very low sheet resistance in the order of at least 100 Ω / cf. in addition, for example, an article by John McDougall, inter alia, with the title "High Value Implanted Resistors For Microcircuits", published in the magazine Proceedings of the IEEE, Vol. 57, No. 9, September 1969, pages 1538 to 1542, and in particular Fig. 5 of this article. For implanted resistors with high resistance values which require high layer resistances (e.g. 20 KΩ /), the current-voltage characteristic becomes non-linear and drops, so that signal distortion of an input signal applied to it occurs. The reason for this is that the actual resistance value at a given voltage is higher than expected. In this case, the current-voltage characteristic corresponds almost entirely to that of a junction field-effect transistor (JFET). It should be emphasized that the value of the resistance depends primarily on the voltage V _epi applied to the transition between the resistance region and the epitaxial layer. To a certain extent, the value is still related to the voltage V ₁- V ₂ across the resistor. The above-mentioned distortion arises from the fact that the depletion layer on both sides of the PN junction expands with the voltage applied to the PN junction, the effective cross section of the resistance track being reduced and thus ultimately increasing its resistance value. This is a phenomenon known as pinch-off or pinch-off effect, particularly in the case of JFETs, but it is particularly critical for implanted resistors because they generally have higher specific resistance values than diffused resistors and consequently the resistance in thinner layers is achieved.

Measures against such distortions can be found e.g. B. in the previously mentioned article by John McDougall. After that  the specific resistance value of the epitaxial layer Values greater than 100 Ω · cm increased. This can, however the electrical properties of neighboring active construction elements are adversely affected.

Another possibility could be the training of the depletion area within the resistance area limit and thus increase the nominal resistance value. This can be achieved by increasing the ion dosage, like that in the article by J. W. Hanson entitled "Ion Implanted N Type Resistors on High Resistivity Substrates ", in the journal J. Vac. Sci. Techn., Vol. 10, No. 6, November / December 1973, pages 944 to 947 and in particular is described in Fig. 6.

Another option is in the French patent font FR-PS 21 17 981 indicated. Although not a special hint given the voltage dependency of the resistors, afterwards a correction of the voltage dependence appears Implant neutral ions into the resistance area, preferably in the vicinity of the PN junction. The the main disadvantage associated with this is additional to see the required process step.

The invention has for its object with a circuit arrangement according to the preamble of claim 1 the voltage dependence of semiconductor resistors correct that an independent of the applied voltage Resistance value is achieved.

Find the features that are important for solving this task themselves in the claims. Summarized after  the invention a special setting of changing Potential difference between the resistance range and the surrounding epitaxial layer to the through such potential Changes either cause signal distortion if possible to keep small or by means of opposites of the same size To compensate for effects. To the epitaxial layer a potential that is determined in more detail below is created preferably the same changes as the resistance below whose voltage dependency is to be corrected. In other words, if the average value of the Potential difference between the resistance area and the Epitaxial layer constant over the length of the resistance remains, the resistance value remains constant, and the mentioned interference can be neglected.

The invention is based on execution examples explained in more detail with the aid of the drawings. It shows

1 shows a partial cross section through a typical, formed by ion implantation resistor, with its equivalent electrical circuit diagram.

Fig. 2 shows a first embodiment of the invention, in which the corrective measures are independent of the resistance to be corrected with respect to its voltage dependency;

Fig. 3A-D show further embodiments of the invention, in which the corrective action depending on the resistance to be corrected.

In Fig. 1 a conventional monolithically integrated resistor structure 10 is shown. On a single-crystalline semiconductor substrate, which is typically composed of a silicon body 11 of a first conductivity type, e.g. B. P ^- , there is a relatively thin (on the order of a few micrometers) epitaxial layer 12 which is relatively weakly doped (average resistivity 1 Ω · cm). An insulated region is formed in this epitaxial layer 12 of the conductivity type opposite to the substrate, the insulation being formed by the heavily doped insulation zones 13 which extend down to the substrate. By means of an ion implantation of a dopant of the first conductivity type, an elongated region 14 is then produced, which is to serve as a resistance region. This manufacturing step is generally carried out following the formation of the P-doped connection regions 15 , which were mostly provided simultaneously with the diffusion of the base regions of the bipolar transistors. The connection regions 15 heavily doped relative to the doping of the resistance region ensure that the resistance value defined by the resistance region 14 is determined only by the sheet resistance of this region. Simultaneously with the emitter regions of the bipolar transistors, a second heavily doped region 16 of the second conductivity type is generally provided as the connection region for the epitaxy layer. The two connections designated 17 and 18 are also formed for the resistance range. The second doping region 16 is also equipped with an ohmic contact 19 . The problems associated with the manufacture of implanted resistors are known per se, cf. in addition to the already mentioned prior art z. B. the US Patent 39 02 926. At the connections 17 and 18 of the resistor, the potentials V ₁ and V ₂ are applied. The connection 19 to the epitaxial layer can either remain open (generally undesirable) or be connected to a suitable voltage source while the most negative potential of the circuit is applied to the substrate in order to provide the PN junction 20 between the silicon body 11 and the Isolationsge 13 on one side and the epitaxial layer on the other side. This reverse voltage causes the electrical insulation of the semiconductor regions framed by the insulation zones. If necessary, the positive circuit potential can be applied to the terminal 19 , which ensures the electrical insulation of the resistance region from the epitaxial layer. This is to be used if several resistors are to be formed in the same epitaxial layer area. In any case, a suitable potential can be applied to the epitaxial layer via the connection 19 , depending on the circumstances.

In Fig. 1 (right) is the electrical replacement circuit diagram of such an implanted resistor 10 is provided. The reference symbols given there correspond to those from the cross-sectional representation.

The resistance value R of such a resistor 10 can be expressed by:

R = R ₀ (1 + λ V _eff ) (1)

Mean:

R ₀ the actual resistance value determined by the type and size of the semiconductor material,
g the coefficient of voltage dependence and
V _{eff is} a voltage _resulting from the empirical relationship

V _eff = V _epi - V _RP + k V _R (2)

can be derived.

In the last equation:

V _{epi is} the potential applied to the epitaxial layer,
V _{RP is} the most positive potential across the resistor (in this example V _RP = V ₁ is assumed),
V _{R is} the voltage drop across the resistor, ie V _R = V ₁- V ₂ ( V ₁ and V ₂ are essentially variable potentials), and
k is a coefficient that depends essentially on the sheet resistance of the resistance region; it has been found that the value of k varies between about 0.4 and 0.6. An approximate value of 0.5 can usually be used. V _eff and therefore the value R of the resistance are essentially dependent on the variable potentials V ₁ and V ₂.

If V _{eff is} constant, R is independent of the applied potentials V ₁ and V ₂. Such a result can be obtained by applying a potential to the epitaxial layer with the resistance to be corrected for its voltage dependency that satisfies the following relationship:

V _epi = 1/2 ( V ₁ + V ₂) + V ₀ (3)

V ₀ is the voltage required to block the PN junction 20 ' between the P resistance region and the N epitaxial layer and the junction 20 already mentioned. If V _{epi is replaced in} accordance with equation (3), and if equation (2) is used, V _RP = V ₁, and k = 0.5, the following results:

V _eff = 1/2 ( V ₁ + V ₂) + V ₀ - V ₁ + 0.5 ( V ₁ - V ₂)

and thus

V _eff = V ₀ (4)

This shows that with this particular value of V _epi, the voltage V _{eff becomes} independent of V _R , namely the voltage across the resistor. Finally, for the resistance value R :

R = R ₀ (1 + λ V ₀) = h R ₀ (5)

Here h means a constant that is independent of V ₁ and V ₂.

The voltage to be applied to the epitaxial layer 12 must follow that according to the relationship in Equation (3). I.e. the epitaxial layer must be connected to a voltage which changes in the same way as the average voltage across the resistor to be corrected. Since the potentials V ₁ and V ₂ are present at the resistor, two approaches are possible, depending on whether the correction of the voltage dependency is independent of the resistor or more specifically independently of the voltage across the resistor or not.

The first solution is shown in the first exemplary embodiment according to FIG. 2, where the circuit means 21 for correcting the voltage dependency of the resistor 10 are related to the latter. The circuit means 21 for the correction mentioned essentially comprise a resistor 22 and the bias voltage V ₀. In order to meet equation (3), the voltage drop across the resistor must be 22 V _R. The voltages V ₀ + V ₁ and V ₀ + V ₂, ie the potential difference V ₁- V ₂, is therefore applied to the opposing stand 22 . These voltages V ₁ and V ₂ can be provided either independently of the voltages to the resistance to be corrected, or can be derived therefrom by suitable electrical circuits. For the correction of the voltage dependency, it is ultimately only necessary that a DC voltage of corresponding polarity is applied to one end of the resistor 22 . The resistor 22 is in turn equipped with a tap, for. B. with a center tap if you assume that the value k is about 0.5. Finally, the resistor 22 does not have to be accommodated on the same semi-conductor plate, it can be arranged, for example and advantageously, as a resistor implanted at the same time as the resistor 10 in an area isolated therefrom.

The further solution is represented by the exemplary embodiments shown in FIGS. 3A, 3B and 3D.

In the arrangement of Fig. 3A the voltages across the resistor 10 via a pair of amplifiers 33, 34, which may consist of emitter follower stages in the simplest case, applied to the resistor 22 '. These amplifiers have a voltage gain of 1 and mean practically no additional current flow. The circuit device 21 ' for correcting the voltage dependence also contains a DC voltage source 35 for V ₀ which is connected between the Mittenan tap of the resistor 22' and the terminal 19 .

Mainly because of the amplifiers 23 and 24 , however, some disadvantages are associated with this exemplary embodiment due to the greater complexity and switching speed reduction.

A simplified embodiment is shown in Fig. 3B. As before, the task here is to reduce the fluctuations in the potential difference between the resistance area and the epitaxial layer, but in contrast to the solutions shown in FIGS. 2 and 3A, a compromise is made here. As is apparent from Fig. 3B, the terminal 19 to the positive voltage of V ₁ and V ₂ is connected. In other words, the voltage V _epi = V ₁ instead of the voltage equation (3). With V _RP = V ₁ and k = 0.5, we get according to equation (2)

V _eff = 0.5 V _R (6)

The resistance R is now defined by the relationship:

R thus increases with the voltage across the resistor. In this case, the correction of the voltage dependency is not complete, but only improved. Such a solution is perfectly acceptable if the potential difference V _R across the resistor is low and the value ( V ₁ + V ₂) / 2 is large. This solution is particularly attractive due to its simplicity. It can advantageously be combined with the solution according to FIG. 3C described below.

In the exemplary embodiment according to FIG. 3C, the resistor 10 with the value R is divided into two parts 36 and 37 , which are electrically insulated from one another and whose partial values are, for example, identical. The resistor 36 is connected with its connection 19 ' to the voltage V ₁, while the resistor part 37 and its ohmic contact 19''is the voltage ( V ₁ + V ₂) / 2. The new value of resistor 10 is now calculated as follows. The following relationships can be established for the partial resistor 36 with the value R ′ :

and

In the same way applies to the partial resistor 37 with its value R '' :

such as

V _R = V ₁- V ₂ applies here. The resistance value R ultimately results in:

Comparing this value with that of Fig. 3B or accordingly equation (7), it can be seen that by dividing the resistance into two equal parts, the value of the coefficient of V _R can be halved. In general it can be said that when the resistance R is divided into n equal parts according to the relationship:

the coefficient for the voltage dependence only comes into effect with the factor 1 / n . It should be noted that this technique can be advantageously applied to the embodiments of FIGS. 2 and 3A.

The exemplary embodiments according to FIGS. 3B and 3C are of interest for circuit applications with unipolar signals, in which e.g. B. the potential V ₁ is always greater than V ₂.

Is V ₁ relative to V ₂ both positive and negative, ie in the case of bipolar signals, the potential difference V ₁- V ₂ = V _R must be kept very low, e.g. B. to a few hundred mV or less, so that the then biased in the direction of passage between the implanted resistance region and the epitaxial layer is not worth mentioning current.

An even more advantageous exemplary embodiment for the case of bipolar signals is shown in FIG. 3D. The resistor 10 is again designed with a center tap, the potential of which can be applied to the epitaxial layer via the connection 19 ''' . This enables the following advantages to be achieved for operation with bipolar signals. On the one hand, there is a symmetrical arrangement for the positive and negative voltage periods and, on the other hand, a complete correction of the influence of the voltage dependency is achieved, since again:

In other words, a perfect correction of the influence of the voltage dependency is achieved in this case because the distortion caused by the resistor 38 is compensated for by an equal but opposite compensating distortion caused by the resistor 39 .

Claims (4)

1. Circuit arrangement for correcting the voltage dependency of a semiconductor resistor ( 10 ), the epitaxial layer ( 12 ) having two outer connections ( 17, 18 ) as a doping region of the first power type within isolated areas of a third outer connection ( 19, 19 ''' ) is formed from the second line type and at its outer connections ( 17, 18 ) different potentials V ₁ or V ₂ are present,
characterized by circuit means ( 21, 21 ' ) for applying a further potential V _epi = 1/2 ( V ₁ + V ₂) + V ₀ to the epitaxial layer ( 12 ), V ₀ being a constant DC voltage. 2. Circuit arrangement according to claim 1, characterized in that the circuit means ( 21 ' ) consist of amplifier ( 33, 34 ), and at least one further resistor ( 22 ) with center tap. 3. Circuit arrangement according to claim 1, characterized in that the third outer connection ( 19 ''' ) of the epitaxial layer ( 12 ) with a center tap of the semiconductor resistor ( 10 ) is directly connected. 4. A circuit arrangement according to claim 1, characterized in that the semiconductor resistor ( 10 ) is divided into n equal partial resistances and for each partial resistor there was a circuit arrangement for correcting the voltage dependency.