EP2095501A1

EP2095501A1 - Transconductance amplifier with a filter for rejecting noise outside the useful band

Info

Publication number: EP2095501A1
Application number: EP08701490A
Authority: EP
Inventors: James Wei
Original assignee: Commissariat a lEnergie Atomique CEA
Current assignee: Commissariat a lEnergie Atomique et aux Energies Alternatives CEA
Priority date: 2007-01-16
Filing date: 2008-01-15
Publication date: 2009-09-02
Also published as: FR2911446A1; FR2911446B1; US8035422B2; WO2008090053A1; US20100176882A1

Abstract

The invention relates to a transconductance amplifier which supplies current variations (di=k.dv) upon receipt of voltage variations (dv). The amplifier includes a first MOS transistor (MN4), the drain of which supplies differential currents (I-di, I+di). The invention also includes an output stage having a second transistor (MP5) of the opposite type to the first. The source of the second transistor is connected to the drain of the first and the gate of the second transistor is biased at a constant voltage (Vref). In addition, the drain of said second transistor receives the current variations supplied by the first transistor that have to be applied to a sampling capacitance. The amplifier also includes a filter (FLT) with a frequency response centred on the central frequency (Fo) of the signals to be converted, having a very high impedance around said central frequency and a low impedance outside of the useful spectrum, whereby the filter is connected to the source of the second transistor (MP5) such as to divert current variations located in a frequency band that is outside the useful spectrum away from the second transistor.

Description

TRANSCONDUCTANCE AMPLIFIER WITH REJECTION FILTER

OF NOISE OFF USEFUL BAND

The invention relates to a transconductance amplifier, intended to provide current variations di when it receives voltage variations dv, and this with a desired conversion coefficient Gm called transconductance: Gm = di / dv The invention is applicable while particularly to perform certain types of sample-and-hold devices, more specifically those that operate by sampling a quantity of charges rather than a point value of voltage. Moreover, the invention applies not only to circuits intended to convert a simple voltage variation into a simple current variation, but also to differential circuits intended to convert a differential voltage variation into a differential current variation.

To set the context of this invention, it may be recalled that it is sometimes preferred to sample loads rather than voltages, to reduce the influence of clock noise (sometimes also called "jitter" or clock jitter) when we want to sample a high frequency signal under the control of a clock that defines the periodic sampling phases. By not integrating a voltage level into a sampling capacity but a current during a known sampling time, the influence of this clock noise is reduced. But then, since the input signal to be converted is generally in the form of a voltage (or more precisely of high-frequency voltage variations), a transconductance amplifier must be placed upstream of the one or more sampling capacitors. high quality that will very accurately convert voltage variations into current variations.

Samplers using a transconductance amplifier as an input stage are used, in particular, in high frequency telecommunications signal sampling applications for frequency transposition and then for analog-to-digital conversion of the signals. The frequency bands of telecommunications signals are very congested. A telecommunications channel uses a narrow band of frequencies and anything outside that band is annoying noise. So Generally speaking, the telecommunication signal transmitted on a given channel comprises a useful spectrum centered on a carrier frequency Fo. The carrier at frequency Fo is modulated in amplitude or in frequency and one of the purposes of the sampling is in particular to transpose the signal towards an intermediate frequency spectrum (centered on an intermediate frequency Fi lower than Fo) or even towards a spectrum in baseband (centered on a zero frequency).

However, it is known that the sampling, at a sampling frequency Fe, of a signal whose useful spectrum is centered on a frequency Fo produces on the one hand a useful spectrum in a transposed band centered on an intermediate frequency Fi = Fe - Fo, but also produces in this transposed band what are called folds of the noise spectrum; this folded noise comes not only from the noise present in the useful band centered on Fo, but also from the noise present in bands of the same width centered on Fi + Fe, or even on other bands still bound to the harmonics of Fe, such that a band centered on KFe + Fi or KFe-Fi, K being an integer. All these noises fold back to the new useful signal band centered on the intermediate frequency Fi, and the resulting noise is the sum of folded noises because the noises from the different bands are uncorrelated and remain uncorrelated after folding.

Putting a filter at the input of the sample-and-hold device would eliminate some of the noise, but that would not eliminate the inherent noise of the input amplifier of the sampler. If we filter out, it is too late folded noises are already in the useful output bandwidth. In practice, it would be necessary to filter within the input amplifier of the sampler. Unfortunately, most input amplifier structures do not allow such internal filtering at the amplifier without degrading the performance of the amplifier.

As an example, the article by B. Nauta "A CMOS transconductance-C Filter Technique for Very High Frequencies" in IEEE Journal of Solid-State Circuits, vol 27 No. 2 pp 142-153 February 1992 describes a structure of transconductance amplifier which has good bandwidth characteristics but correlatively a loud noise and it does not easily make it possible to filter this noise. Another example is given in the article by M. Koyama et al., "A

2.5-V Active Low-Pass Filter Using All-n-p-n Gilbert CeIIs with a 1 -Vp-p

Linear Input Range "in IEEE Journal of SoNd State Circuits, 28, no.

12, pp. 1246-1253, December 1993. This is a two-stage amplifier that can support internal filtering but exhibits distortion.

In yet another example, T. Kwan and K. Martin., "An adaptive analog continuous-time CMOS biquadratic filter," in IEEE Journal of SoNd State Circuits, vol. 26, no. 6, pp 859-867, June 1991, the introduction of an internal filter would disturb the useful signals. It has been found that the structure of the output stage of a transconductance amplifier intended to be placed at the input of a sample-and-hold device can be modified to introduce an internal filtering which fulfills in the best manner possible a function of a transconductance amplifier. elimination of noise that is generated by folding during downstream sampling. The output stage is constituted according to the invention in folded cascode stage absorbing all the current produced by the transconductance amplifier and it is at this folded cascode stage that a filter is connected capable of deriving the current variations which do not occur. are not in the useful frequency band of the input signal to be sampled, without deriving the current variations that are in the useful band. The filter has a high impedance in the useful band centered around the frequency Fo of the carrier.

Accordingly, the invention proposes a transconductance amplifier comprising a first transistor of a first type providing current variations to be sampled having a useful frequency spectrum centered on a frequency Fo, characterized in that it comprises an output stage. having a second transistor of a type opposite the first, whose source is connected to the drain of the first, whose gate is biased at a constant potential, and whose drain receives the current variations which are provided by the first transistor and which must be applied to a sampling capacity, and in that it further comprises a frequency response filter centered on the central frequency Fo of the input signals of the amplifier, having a very high impedance around this frequency and a low impedance outside the useful spectrum, the filter being connected to the source of the second transistor so as to derive outside the second transistor currents variations that are in a frequency band outside the useful spectrum.

In practice, the current variations to be sampled are superimposed on a bias current equal to the sum of a first and a second constant current. The second transistor has its source connected to a first current source providing a current equal to the first constant current. Its drain is connected to a current output of the amplifier and to a second current source providing a current equal to the second constant current. The amplifier according to the invention is particularly intended to be placed at the input of a sample-and-hold circuit for the conversion into current variations of high-frequency voltage signals having a useful spectrum centered on a frequency Fo, before sampling of these current variations. The amplifier thus designed is simple or differential; if it is differential, it is constituted by two simple amplifiers and the filter is connected between the drains of the first respective transistors of the two simple amplifiers.

The filter can be a simple passive filter with inductance and capacitance. Thanks to the invention, a very high bandwidth and low noise sampler-blocker can be realized.

Note that the amplifier according to the invention can be realized with MOS transistors or with bipolar transistors, and therefore, throughout this description, it will be considered that the term "transistor" covers both the MOS transistors and the transistors. bipolar transistors and that the terms "source", "drain", and "gate", conventionally used for MOS transistors, must be interpreted as meaning "transmitter", "collector" and "base", respectively, if the transistors are bipolar. MOS transistors of the opposite type are then NMOS and PMOS transistors, opposite type bipolar transistors are NPN and PNP transistors respectively.

Other features and advantages of the invention will appear on reading the detailed description which follows and which is given with reference to the appended drawings in which: FIG. 1 represents an example of a known transconductance amplifier from which the present invention can be applied;

FIG. 2 represents the transconductance amplifier according to the invention;

FIG. 3 represents the amplifier according to the invention for a non-differential mode;

FIG. 4 represents an application of the transconductance amplifier of FIG. 3 to a simplified diagram of a differential sample-and-hold device.

The transconductance amplifier of FIG. 1 is a differential amplifier having two inputs E, E 'between which an input differential voltage (in small high frequency signals) V + dv, V-dv, to be converted into current is applied. differential I-di, l + di on two differential outputs in current S and S '. V is a common mode voltage on the inputs, I is an identical bias current on both outputs.

The amplifier is symmetrical since it works here in differential and therefore has two identical halves. The first half comprises a series of first and second MOS transistors (MP1, MN2) of the opposite type connected by their drains; the gate of the first transistor MP1 is connected to the input E; the source is connected to a constant current source I _B i as well as to a resistor of value R and to the drain of a third MOS transistor MN3 of the same type as the second; the source of the second transistor MN2 and that of the third transistor MN3 are grounded (Vss); the gate of the transistor MN3 is connected to the drains of the first and second transistors; and the amplifier further comprises a fourth transistor MN4 whose function is to copy to the output S the current flowing through the transistor MN3; the sources of the second and fourth transistors, MN2 and MN4 are therefore connected, their grids too. Finally, the gate of the transistor MN2 is biased by a fixed voltage V _Bias so that the transistor acts as a current source and maintains a very fixed current value, of value I _B 2, in the transistors MP1 and MN2. The other half of the amplifier consists of an identical set of four transistors, designated by the same references assigned the sign 'prime', and connected in the same way between the input E 'and the output S'; the current sources are identical in the two amplifier halves, the resistors as well as the constitutions of transistors. The resistors of value R are connected together and constitute a single resistor of value 2R shared between the two halves of the amplifier; this resistance of value 2R in fact connects the drains of the third transistors MN3, MN3 'of the two sets.

This scheme works ideally in the following manner: the transistor MP1 traversed by a constant current is a voltage follower; the small dv variations on its grid are fully reflected on its source. It is the same with MP1 ', with an opposite variation (-dv). The resistance of value 2R sees a variation of voltage 2dv at its terminals. It is traversed by a current variation di = 2dv / 2R = dv / R which can not flow neither in the branch l _B i nor in the transistors MP1, MN2, MP1 ', MN2' whose currents are set at I _B2 - The current variation 2dV / R can only flow in the transistors MN3 (in one direction) and MN3 '(in the other direction). It circulates in practice in the transistor MN3 a current l-di and in the transistor MN3 'a current l + di, where the common bias current I is simply IBI -IB2- The currents l-di and l + di are copied by the MN4 transistors and

MN4 'to constitute the differential output currents l-di = l-dv / R in one direction on S, and l + di = l + dv / R in the other direction on S'. The amplifier therefore has a transconductance di / dv equal to 1 / R for small signals; this transconductance is well controlled and very linear. The simplifying assumption is that the current copying is done with a factor of 1. It will be understood that this factor could be different by choosing a ratio of geometries different from 1 between the transistors MN3 and MN4.

Fig. 2 shows the block diagram of the present invention. The core of the transconductance amplifier of FIG. 1, surrounded by a dashed line, is used to form the circuit according to the invention. It is designated AMPC in FIG. 2 (in dashed lines as well) and it has two differential inputs E and E 'in voltage and two differential outputs in current S and S' as explained previously.

The heart of the amplifier is followed on each output by an output stage called "folded cascode stage", whose function is to restore a di current from the current l + di and a current -di from the current l-di, to apply these currents in principle to a sampling capacity. The entire transconductance amplifier, the sampling capacitor (s), and the sampling switches with their control circuits (not shown in FIG. 2) will then constitute a complete sampling circuit.

The folded cascode output stage which is connected to the output S comprises a transistor MP5 (PMOS) identical to the transistor MP1 of the AMPC core and whose source is connected to the output S, therefore to the drain of the transistor MN4. The function of this transistor is to eliminate a part of the current component I present on the drain of MN4. It is recalled that the component I is equal to IBI-IB2- The cascode stage therefore comprises current sources SC1 and SC2 which respectively establish currents I _B i and I _B 2 that can be subtracted from current I. The transistor MP5 is mounted so as to derive the current variations di from the AMPC core to an output Se which will comprise only the current di and not the current I-di.

The two current sources SC1 and SC2 of the same values I _B i and I _B2 that the sources of the AMPC core are connected in the following way: the source of MP5 is connected to the output S and also to the current source SC1 (I _B - _I ) which takes its power from Vdd. The drain of MP5 is connected to the current source SC2 (I _B 2) which directs its current to the ground Vss. The gate of transistor MP5 is at a fixed bias potential Vref.

The output of the cascode stage is a node Se and represents the output of the transconductance amplifier completed by its output stage; in a sampler, the output Se is intended to be connected to a sampling capacity via a switch or a system of switches.

The mounting and the numerical and geometric parameters are strictly the same for the other half of the amplifier; the elements bear the same references with the sign "prime".

The output stage operates in the following manner: since the output S supplies a current l-di which is equal to I _BI -I _B2 -CU, and because of the presence of the current source SC1 supplying the _B- 1 , the current flowing in the transistor MP5 is necessarily equal to I _BI - (I _BI -I _B2 -CU) so Iβ2 + di. And, because of the presence of the source SC2 which supplies I _B2 , the final output current of the transconductance amplifier is an outgoing current of value + di on the output Se. In the same way, it is an outgoing current of value -di on the output Se '. Having thus constituted two current outputs of di and -di value on the outputs Se and Se ', it is placed between the nodes S and S', that is to say within the transconductance amplifier itself. even since it is upstream of the final outputs Se and Se ', a narrow band FLT filter intended to eliminate the noise upstream of the sampling. This filter has the following properties: its impedance variation versus frequency curve is centered on the central frequency Fo of the signals to be converted; it has a very high impedance around this central frequency and a low impedance outside the useful spectrum of the signals to be transmitted, that is to say a low impedance outside a relatively narrow band around the central frequency. Such a filter can simply be a fully passive plug circuit (an inductor in parallel with a capacitance, satisfying the condition 2πl_CFo ² = 1). More sophisticated filters may be provided if a more rectangular filtering template than a single LC filter is desired. The filter is connected between the outputs S and S ', thus between the drains of the transistors MN4 and MN4', so as to derive from the transistor MP5 the current variations that would be in a frequency band outside the useful spectrum. In other words, the filter FLT tends to bypass the drains of the transistors MN4 and MN4 'for the non-useful frequencies (and in particular the noise situated outside the useful spectrum), so that the currents di and -di, for useless frequencies, tend to go from the drain of transistor MN4 to the drain of transistor MN4 'rather than to outputs Se and Se'. On the contrary, for the useful frequencies around Fo, the filter behaves like an open circuit and does not prevent the current variations di from being directed towards the output Se or Se '.

Note that this diagram is completely transferable to a non-differential mounting in which the transconductance amplifier uses only half of the diagram of Figures 1 and 2. Such an assembly is shown in Figure 3. The AMPC heart of the transconductance amplifier comprises half of the scheme of Figure 1, the elements affected by the 'premium' sign being deleted. There is a single input E to receive a voltage V + dv to be converted into current and a single output S to provide a current I-di where I is equal to IBI -IB2- The output stage of the transconductance amplifier only includes the transistor MP5 and the sources SC1 and SC2. The final output is the output Se providing a small signal current di where di = dv / R.

In this case, the FLT filter is connected between the output S and the low potential of the power supply (ground Vss). The operation is the same as in differential: for the useful frequency spectrum centered on the central frequency Fo of the signals to be converted, the filter has a very strong impedance and plays no role. For frequencies outside the useful band, the filter tends to short-circuit current di to ground, so that it does not exit through output Se.

Figure 4 shows a schematic diagram of use of the transconductance amplifier to realize a sampler. The diagram considered as an example is a non-differential application to simplify the representation, but it is obviously transferable to a differential application using for example (but not necessarily) two separate sampling capabilities. The voltage to be sampled, dv in small signals, applied to the input E, is converted into a current di on the output Se by the transconductance amplifier which is that of FIG. 3. This current di is integrated in a capacitance d EC sampling for the duration of a first half-wave CLK of a pulse of a sampling clock and stores a sample of charges in the capacitance CE. During the duration of the next half cycle NCLK, the output Se is disconnected from the capacitance C _E and the latter is connected to a sampled voltage output Vech that can be exploited, for example applied to an analog-digital converter. To increase the sampling rate, one can very well predict, but it is not obligatory, that another sampling takes place during the following half-wave NCLK, in a capacity CΕ identical to the capacity C _E. Thus, the current di to be sampled is applied alternately to the capacitance CE and the capacitance CΕ and while the voltage of one is read and exploited, the other receives the current to be sampled. The capacities are reset after each operation in the case simple sampling without decimation, that is to say without adding series of N samples in the same capacity. The reset switches are not shown so as not to weigh down the diagram.

The transconductance amplifier according to the invention can be used in a sample-and-hold device for analog-to-digital conversion, in singles or differentials, and in particular in a sample-and-hold device providing two samples in quadrature phase at each sampling period ( which implies doubling the number of sampling capacities). It can also be used in a finite impulse response sampled filter in which a weighted sum of a succession of N di current samples is performed and samples are provided at a reduced frequency or decimation frequency which is the sampling frequency divided by N. It will be noted that the transconductance amplifier principle according to the invention has been described with reference to an amplifier core which is that of FIG. 1, but that it is applicable to other circuit diagrams. amplifier core, for example a scheme in which the transistor MN2 would have its gate connected not to a fixed bias voltage Vbias but to the gate of the transistor MP1. The only condition to be respected is that the transconductance amplifier core provides a current I-di or l + di whose component I is known to be neutralized by the current sources SC1 and SC2 of the folded cascode transistor circuit. It is desirable then that the source SC2 be replaced by an NMOS transistor whose gate is brought to the same potential as the gate of MP5, this potential being such that the current in the transistor thus added is the same current IB2 as that which forms a component of I, that is to say the same current I _B 2 as that flowing in the transistor MN2 of the AMPC core.

In all of the above, the MOS transistors can be replaced by bipolar transistors as indicated.

Claims

A transconductance amplifier comprising a first transistor (MN4) of a first type providing the current variations to be sampled, the amplifier being characterized in that it comprises an output stage comprising a second transistor (MP5) of a opposite type to the first, whose source is connected to the drain of the first, whose gate is biased at a constant potential (Vref), and whose drain receives the current variations (l-di, l + di) which are supplied by the first transistor and which must be applied to a sampling capacity, and in that it further comprises a frequency response filter (FLT) centered on the central frequency Fo of the signals to be converted, having a very high impedance around of this central frequency and a low impedance outside the useful spectrum, the filter being connected to the source of the second transistor (MP5) so as to derive outside the second transistor the variations of which are in a frequency band outside the useful spectrum.

2. Amplifier according to claim 1, characterized in that the filter (FLT) is also connected to a supply potential (Vss).

3. Differential transconductance amplifier characterized in that it comprises two simple amplifiers according to claim 1 for applying current variations to two sampling capacitors, and in that the filter (FLT) is connected between the drains of the first respective transistors (MN4, MN4 ') of the two single amplifiers.

4. Transconductance amplifier according to one of claims 1 to 3, characterized in that the filter (FLT) is a passive filter with inductance and capacitance.

Transconductance amplifier according to one of Claims 1 to 4, characterized in that the drain of the second transistor

(MP5) constitutes an amplifier output (Se) intended to be connected to through a switch system with at least one sampling capability.