US 6,463,566 B1Grant

Dynamic double sampling charge integrator

Issue Date:2002-10-08

•9 Claims

•15 Drawing Sheets

Abstract

A charge processing circuit which integrates charge at an output node that is representative of an input charge provided at an input node. The circuit includes a precharge path coupled to the input node, the precharge path operable for setting the potential of the input node to a fixed precharge potential prior to introduction of input charge to the input node. A sensing path is coupled to the input and output nodes which is operable for returning the potential of the input node to the fixed precharge potential subsequent to introduction of input charge to the input node. A feedback element has an input coupled to the sensing path and the precharge path, the feedback element operable for setting said fixed precharge potential. In another embodiment there is provided a method of integrating charge at an output node of a charge processing circuit that is representative of input charge provided at an input node. The method includes replenishing charge at the input node; draining charge from the input node; stopping the draining of charge from the input node in response to a feedback element detecting when the input node reaches a fixed precharge potential; introducing an input charge on the input node from a source; draining charge from the input node onto the output node; and stopping the draining of charge from the input node in response to a feedback element detecting when the input node reaches the fixed precharge potential.

Metadata

Assignee

Massachusetts Institute of Technology

Inventor

Susanne A. Paul

Application Information

Application Number:US 09/497,985

Filing Date:2000-02-04

Priority Date:2000-02-04

Art Unit:7

Classifications

IPC:

G06F 1750

Field of Search:

327 91327 92327 93327 94327 95327 96327284327554323312348222348250348308

Patent Drawings (15 sheets)

Description

Background of the Invention

[0003] The invention relates to the field of charge integrating devices.

[0004] Charge generation is necessary whenever CCD signal processing circuitry must interface with voltage-domain signals. A variety of techniques are well known for generating charge packets from voltages. In all of these, an input diffusion is used to provide a source of electrons and the amount of charge that is collected from this source is controlled by either the source potential or the voltages on an initial set of register gates.

[0005] The surface channel fill-and-spill technique, shown diagrammatically in FIG. 1, is the most commonly used method of charge generation. FIG. 1A is a schematic diagram of a conventional fill-and-spill charge generator circuit 100, and FIGS. 1B-1E are associated potential diagrams for a second, third, fourth and first phase, respectively. In this approach, the input signal v_sis provided to gate G₂and a fixed reference level V_ris placed on G₁. During the fill phase in FIG. 1B, node f_gis pulsed low and charge is injected into the channel regions underneath G₁and G₂. During the spill phase in FIG. 1C, f_gis returned to a high potential so that excess charge from the output well spills back. The spill transition occurs rapidly at first, but then slows considerably as it nears completion. The long time constant associated with this spill transition limits the speed at which the fill-and-spill circuit can be operated. In addition to its limited speed, the fill-and-spill technique has the disadvantages that it provides poor linearity and does not perform accurate sampling.

[0006] Charge-domain signals can be used by non-CCD elements in two ways. The first method, referred to as destructive sensing, involves transferring the charge packet onto a receiving element. In this case, the charge packet is consumed by the operation. The second method, referred to as nondestructive sensing, involves creating a separate representation of the charge packet, without altering the original, so that multiple nondestructive operations can be performed on the same charge packet.

[0007] A schematic diagram of a conventional nondestructive floating gate amplifier circuit 200 is illustrated in FIG. 2A. FIGS. 2B-2E are associated potential diagrams for a first, second, third and fourth phase, respectively. The circuit includes a sensing gate G₄within a CCD channel. The sensing gate and the barrier that precedes it are not clocked. In preparation for receiving charge, node v_gis preset to the intermediate bias level V_hduring the time that the G₄channel is empty. The precharge is then turned off and v_gis left floating. During the next phase, a charge packet is transferred underneath G₄by lowering the voltages on G₁and G₂. As charge collects underneath G₄, it couples through the gate-to-channel capacitance and causes the potential on v_gto fall. The clocked v_gwaveform is then buffered to form the voltage-domain output v_b.

[0008] This approach to charge sensing has the disadvantage that nonlinearity is introduced by voltage dependence of its various capacitances. It also is not compatible with low voltage operation. because unclocked CCD gates, G₃and G₄, reside in the charge transfer path. Finally this circuit provides only a limited signal range before charge collects underneath G₃and is not sensed by the floating gate.

Summary of the Invention

[0009] Accordingly, the invention provides a charge processing circuit which integrates charge at an output node that is representative of an input charge provided at an input node. The circuit includes a precharge path coupled to the input node, the precharge path operable for setting the potential of the input node to a fixed precharge potential prior to introduction of input charge to the input node. A sensing path is coupled to the input and output nodes which is operable for returning the potential of the input node to the fixed precharge potential subsequent to introduction of input charge to the input node. A feedback element has an input coupled to the sensing path and the precharge path, the feedback element operable for setting said fixed precharge potential.

[0010] In accordance with another embodiment of the invention, there is provided a method of integrating charge at an output node of a charge processing circuit that is representative of input charge provided at an input node. The method includes replenishing charge at the input node; draining charge from the input node; stopping the draining of charge from the input node in response to a feedback element detecting when the input node reaches a fixed precharge potential; introducing an input charge on the input node from a source; draining charge from the input node onto the output node; and stopping the draining of charge from the input node in response to a feedback element detecting when the input node reaches the fixed precharge potential.

Brief Description of the Drawings

[0011] FIG. 1A is a schematic diagram of a conventional fill-and-spill charge generator circuit; and FIGS. 1B-1E are associated potential diagrams for a second, third, fourth and first phase, respectively;

[0012] FIG. 2A is a schematic diagram of a conventional nondestructive floating gate amplifier circuit; and FIGS. 2B-2E are associated potential diagrams for a first, second, third and fourth phase, respectively;

[0013] FIG. 3A is a schematic diagram of an exemplary embodiment of a DDS charge generator in accordance with the invention; and FIGS. 3B-3E are associated energy level diagrams for a fill phase, a spill phase, a collection phase and a sensing phase, respectively;

[0014] FIGS. 4A-4E are voltage waveforms for a sample-and-hold portion for the example of a linear input ramp for a DDS charge generator;

[0015] FIGS. 5A-5D are simplified linear models of a DDS charge generator in accordance with the invention, during each of the four phases: a fill phase, spill phase, collection phase and sensing phase, respectively;

[0016] FIG. 6A is a schematic diagram of an exemplary embodiment of a DDS wire transfer circuit; and FIGS. 6B-6E are associated energy level diagrams for a fill phase, spill phase, collection phase and sensing phase, respectively;

[0017] FIG. 7A is a schematic diagram of an exemplary embodiment of a DDS sensing circuit with charge output in accordance with the invention; and FIGS. 7B-7E are associated energy level diagrams for a fill phase, spill phase, collection phase and sensing phase, respectively;

[0018] FIG. 8A is a schematic diagram of an exemplary embodiment of a DDS charge sensing circuit with voltage output in accordance with the invention; and FIGS. 8B-8E are associated energy level diagrams for a fill phase, spill phase, collection phase and sensing phase, respectively; and

[0019] FIG. 9A is a schematic diagram of an exemplary embodiment of a DDS D/A and subtraction circuit in accordance with the invention; and FIGS. 9B-9E are associated energy level diagrams for a fill phase, spill phase, collection phase and sensing phase, respectively.

Detailed Description of the Invention

[0020] Speed and linearity are improved for a dynamic double-sampling (DDS) technique in accordance with the invention by turning off spill and sensing transitions before they enter subthreshold, and by sensing charge over a time-varying, rather than a static, cascode barrier. At the same time, the useful operating range is increased to nearly the full scale and the impact of spill and sensing clock jitter is reduced. Parasitic capacitor voltage dependence is decreased as well.

[0021] FIG. 3A is a schematic diagram of an exemplary embodiment of a DDS charge generator 300 in accordance with the invention. FIGS. 3B-3E are associated energy level diagrams for a fill phase, a spill phase, a collection phase and a sensing phase, respectively. The primary voltage-mode portion of this circuit is the sample-and-hold stage 302 formed from capacitors C₁and C₂and transistors M₃, M₄, M₂, and M₅. In addition, there is a feedback amplifier A₁(304) that computes the difference between f_gand the fixed reference voltage, V_r. A signal-dependent voltage source 306, which drives the bulk of M₅, is also included. Other circuit elements operate in the charge domain.

[0022] A series of precharge transistors connect capacitor C₂to the diffusion D₁and a series of sensing transistors connect it to an output well, G₅. Two gates, both labeled G₂₄, are shown on either side of an input diffusion D₂. These elements are shown as separate gates in the figure, but can also be implemented as a single gate. A secondary static sensing path 308, formed from G₈, G₉, and G₁₀, is located between the input diffusion and the output well. Its cascode gate G₈is tied to the fixed bias V_h′.

[0023] The following description of circuit operation references three figures. First, energy level diagrams for the charge-domain portion are provided in FIGS. 3B-3E. Second, voltage waveforms for the sample-and-hold portion are included in FIGS. 4A-4D for the example of a linear input ramp. Exemplary voltage waveforms for a DDS charge generator, signals v_i, v_m, and v_fare shown in FIGS. 4A-4C, and signals f_gand f_bfollow identical paths during phases 2 and 4 as shown in FIGS. 4D-4E.

[0024] FIGS. 5A-5D are simplified linear models of a DDS charge generator in accordance with the invention, during each of the four phases: a fill phase, spill phase, collection phase and sensing phase, respectively. Capacitor C_p1, included in this model, represents parasitic junction and routing capacitances on node v_f. Capacitor C_p2models capacitance of the cascode channel, underneath G₂₄and G₈, as well as parasitic junction and routing capacitances on f_g. Saturation and subthreshold modes of G₂₄are modeled by resistor R₂₄in series with a voltage source. The value of this voltage source is (A(V_r−f_g)−V_th), where A represents the amplifier gain. This simple model provides a qualitatively accurate description of circuit operation.

[0025] Phase 1 is referred to as the fill phase. Signal v_mtracks the ramp on the analog input v_iand the intermediate node v_fis clamped to the bias V_c. When the sensing node f_gis pulled low, to V_p, the region underneath G₂₄is flooded with charge, and electrons consumed during the previous generation cycle are replenished. The charge packet underneath G₅, which was generated during the previous cycle, is transferred forward upon the transition of G₁and G₅to ground. Charge can not flow backward during this transfer because of the built-in threshold offset between barrier and storage gates.

[0026] Phase 2 is the spill phase. M₁is turned off and f_gis left floating. Initially, the voltage on f_gis much lower than that on V_rand the amplifier output f_bis saturated at its high level, V_h. After the potential on G₃is raised and the precharge path is enabled, electrons flow from f_gto the drain, V_d. Current flow is determined by the value of (f_b−f_g), the gate-to-source voltage of G₂₄. Initially, this voltage is large, G₂₄is in saturation, and the voltage on f_grises rapidly. However, as f_grises, currents are reduced and the process slows considerably. Once f_gapproaches V_r, the amplifier output falls and G₂₄transitions quickly through subthreshold and is shut off.

[0027] For simplicity in the analysis below, the amplifier is modeled with a linear transfer characteristic, and its saturation at high and low levels is ignored. The value of f_gas a function of time during the spill process is

\begin{matrix} f_{g} (t) = V_{p} ⅇ^{- (t - H) / τ_{2}} + (\frac{{AV}_{r} - v_{th}}{A + 1}) (1 - ⅇ^{- (t - H) / τ_{2}}) H \leq t \leq 2 H . & (1) \end{matrix}

[0028] A comparison between the time constant

\begin{matrix} τ_{2} = \frac{R_{24} (C_{p2} + C_{2})}{(A + 1)} & (2) \end{matrix}

[0029] for this operation with that in a circuit with fb held at a constant value shows that the time constant for the circuit is a factor of (A+1) times faster, and the incomplete portion of the transition in Eq. (1) is attenuated by a factor of e^A+1.

[0030] To a very good approximation, the precharge path is completely shut off for the remainder of the cycle withfg floating at a level of

\begin{matrix} f_{g} (2 H) = (\frac{{AV}_{r} - v_{th}}{A + 1}) . & (3) \end{matrix}

[0031] The final voltage on f_bis

\begin{matrix} f_{b} (2 H) = (\frac{A}{A + 1}) (V_{r} + v_{th}) . & (4) \end{matrix}

[0032] Nodes v_mand v_fare actively driven during this phase and are not impacted by the transition on f_g. At the end of phase 2 they have values

[0033] v_m(2H)=v_i(2H)=V_s (5)

[0034] and

[0035] v_f(2H)=V_c. (6)

[0036] Variable V_sis used to refer to the analog input v_iat the end of the sampling operation, at time 2H.

[0037] Phase 3 is referred to as the collection phase. G₅is raised in anticipation of receiving charge and the precharge path is closed by lowering G₃. Meanwhile v_m, which was previously tied to the analog input, is clamped to ground. The falling transition on v_m,

[0038] v_m(2H+Δ)−v_m(2H)=−V_s, (7)

[0039] is assumed to complete during a short time, Δ. It couples through C₁and C₂onto v_fand f_g. Node v_ffalls toward a final voltage of

[0040] v_f(2H+Δ)=V_c−V_sK₁, (8)

[0041] where the unitless constant K₁is defined as

\begin{matrix} K_{1} = \frac{C_{1}}{(C_{p1} + \frac{C_{2} C_{p2}}{C_{2} + C_{p2}} + C_{1})} . & (9) \end{matrix}

[0042] The voltage on f_galso falls from its precharged level to

\begin{matrix} f_{g} (2 H + Δ) = f_{g} (2 H) - V_{s} K_{1} \frac{C_{2}}{(C_{2} + C_{p2})} . & (10) \end{matrix}

[0043] The channel capacitance of G₂₄is included in C_p2because charge fills this region during both the fill and collection phases.

[0044] For optimum performance, the amplifier should remain saturated at its high level V_hduring the entire collection phase. To assure this condition, the input signal range is limited to

\begin{matrix} V_{s} > (\frac{v_{th}}{A + 1} - \frac{V_{r}}{A (A + 1)}) (\frac{C_{2} + C_{p2}}{C_{2} K_{1}}) . & (11) \end{matrix}

[0045] Meanwhile, the potential on G₉is high and the secondary sensing path through G₈and G₉is open. The voltage on f_grises toward the channel potential underneath G₈. Eventually G₈enters subthreshold and current gradually ceases. The rising transition on f_gduring this time is described by

\begin{matrix} f_{g} (t) = (f_{g} (2 H) - V_{s} K_{1} \frac{C_{2}}{(C_{2} + C_{p2})}) ⅇ^{- (t - 2 H) / τ_{3}} + (V_{h}^{'} - v_{th}^{'}) (1 - ⅇ^{- (t - 2 H) / τ_{3}}) 2 H \leq t \leq 3 H, & (12) \end{matrix}

[0046] where the time constant τ₃is

\begin{matrix} τ_{3} = R_{8} (C_{p2} + \frac{C_{2} (C_{1} + C_{p1})}{C_{1} + C_{2} + C_{p1}}) . & (13) \end{matrix}

[0047] The final phase, phase 4, is referred to as the sensing phase. The sensing path is enabled by raising the voltage on G₁. Immediately thereafter, any charge remaining on f_gfrom the collection phase is transferred to the output well. The resulting transition on f_gis similar to that during the spill phase. The amplifier output f_bis originally clamped at its high level, V_h. Current flow, determined by (f_b−f_g), slows as f_gapproaches V_r. Once fb begins to fall, G₂₄transitions rapidly through subthreshold and is shut off.

[0048] The value of f_gas a function of time is given by

\begin{matrix} f_{g} (t) = f_{g} (3 H) ⅇ^{- (t - 3 H) / τ_{4}} + (\frac{{AV}_{r} - v_{th}}{A + 1}) (1 - ⅇ^{- (t - 3 H) / τ_{4}}) 3 H \leq t \leq 4 H . & (14) \end{matrix}

[0049] Like the spill transition, the time constant, τ₄, for the sensing transition,

\begin{matrix} τ_{4} = \frac{R_{24}}{(A + 1)} (C_{p2} + \frac{C_{2} (C_{1} + C_{p1})}{C_{1} + C_{2} + C_{p1}}), & (15) \end{matrix}

[0050] is a factor of (A+1) times faster than that of a circuit where the f_bsignal is held at a constant bias.

[0051] Since v_fis floating during phase 4, it rises in response to the level

\begin{matrix} v_{f} (t) = v_{f} (3 H) + (f_{g} (t) - f_{g} (3 H)) \frac{C_{2}}{C_{1} + C_{p1} + C_{2}} 3 H \leq t \leq 4 H . & (16) \end{matrix}

[0052] In contrast to f_g, the value of v_fat the end of the sensing phase differs from its precharged value.

[0053] After the sensing path is closed, f_gremains floating for the remainder of the cycle at a level

\begin{matrix} f_{g} (4 H) = (\frac{{AV}_{r} - v_{th}}{A + 1}) . & (17) \end{matrix}

[0054] The feedback signal f_bremains at the voltage

\begin{matrix} f_{b} (4 H) = (\frac{A}{A + 1}) (V_{r} + v_{th}) . & (18) \end{matrix}

[0055] Current I_sis integrated in both the primary and secondary output wells over the sensing and collection phases. The sum of these two packets represents the final output,

\begin{matrix} Q_{o} = (f_{g} (4 H) - f_{g} (2 H + Δ)) (C_{p2} + \frac{(C_{1} + C_{p1}) C_{2}}{C_{1} + C_{2} + C_{p1}}) . & (19) \end{matrix}

[0056] Combining Eqs. 19, 14 and 10, and retaining only the most significant terms yields the result

\begin{matrix} Q_{o} = V_{s} C_{1} \frac{C_{2}}{C_{1} + C_{2} + C_{p1}} + ((\frac{{AV}_{r} - v_{th}}{A + 1}) (ⅇ^{- H / τ_{4}} - ⅇ^{- H / τ_{2}}) - V_{p} ⅇ^{- H / τ_{2}} + (V_{h}^{'} - v_{th}^{'}) ⅇ^{- H / τ_{4}}) (C_{p2} + \frac{(C_{1} + C_{p1}) C_{2}}{C_{1} + C_{2} + C_{p1}}) . & (20) \end{matrix}

[0057] When the assumptions H>>τ₂, H>>τ₄, and C_p1=0 apply, the result in Eq. 20 is simplified to

\begin{matrix} Q_{o} = V_{s} (\frac{C_{1} C_{2}}{C_{1} + C_{2}}) . & (21) \end{matrix}

[0058] This depends only on the sampled input signal and on C₁and C₂. Although v_th, A, and V_rdetermine the precharge and sensing values of this circuit in Eqs. 1 and 14, they do not impact the result in Eq. 21 because both precharge and sensing are performed with respect to these same values. Matching between different circuits does not depend critically on any of these parameters.

[0059] Spill and sensing transitions in a DDS circuit are, to a good approximation, performed to completion and the approximations H>>τ₂, H>>τ₄are nearly always valid. The values of f_gat the end of the spill and sensing phases are then equal, speed-dependent distortion can be eliminated, and parasitic capacitance, C_p2, has no impact on the output.

[0060] None of the bias voltages V_d, V_c, V_h, V_r, or V_pinfluence the result in Eq. 21. Bias V_pshould be low enough that charge can fill the region underneath G₂₄during the fill operation. Any level, such as ground, that is less than V_h−v_this acceptable. Bias V_dshould be high enough that it can remove electrons from f_gduring the spill operation. It is typically held at the most positive supply. The level of V_cmust be high enough that v_f(3H) remains greater than zero for all possible values of V_s. Similarly, the levels of V_hand V_rmust be high enough that f_g(3H) , given by Eq. 12, remains greater than zero at all times.

[0061] The DDS charge generator 300 of FIG. 3A includes two blocks: a sample-and-hold front end, and a dynamic double-sampling core. The purpose of the sample-and-hold block is to produce a displacement charge proportional to the analog input voltage, v_i. The purpose of the DDS core is to integrate this displacement charge in a CCD output well. The DDS technique can also be applied to a number of other circuit functions, in addition to charge generation. In this approach, the sample-and-hold block is removed and an alternative source of electrons is provided to the DDS input. Examples of such circuits, including wire transfer, charge sensing, charge subtraction, and D/A conversion are presented hereinafter.

[0062] Wire transfer circuits provide an additional degree of flexibility by allowing charge packets to be transferred between nonadjacent or distant wells, via a metal line. To incorporate dynamic double sampling into a wire transfer circuit, the output from a CCD register is connected to the input of a DDS core as shown in FIG. 6A. FIG. 6A is a schematic diagram of an exemplary embodiment of a DDS wire transfer circuit 600. FIGS. 6B-6E are associated energy level diagrams for a fill phase, spill phase, collection phase and sensing phase, respectively. A CCD register 602, formed by gates G₁₁, G₁₂, and G₁₃, serves as the source of electrons. A DDS core unit 604 reconstructs incoming charge packets by integrating their electrons in a receiving well.

[0063] The operation of the circuit 600 is similar to that of the DDS charge generator 300. It is illustrated using the energy level diagrams of FIGS. 6B-6E. During the fill phase, a new supply of charges is provided to the circuit when f_gis pulled low to V_p. Gate G₁₂is off at this time to prevent charge from flooding backward into the source register. During the spill phase, f_gis reset high to a level of

\begin{matrix} f_{g} (2 H) = V_{p} ⅇ^{- H / τ_{2}} + (\frac{{AV}_{r} - v_{th}}{A + 1}) (1 - ⅇ^{- H / τ_{2}}) . & (22) \end{matrix}

[0064] The time constant for this transition is

\begin{matrix} τ_{2} = \frac{R_{24} C_{p2}}{(A + 1)}, & (23) \end{matrix}

[0065] where C_p2represents parasitic junction and routing capacitances on f_gas well as channel capacitance of gates G₂₄, G₈, and G₁₃.

[0066] During the collection phase, a quantity of charge Q_sis transferred onto f_gby lowering the potential of G₁₁. The potential on f_gfalls in response, and charge flows through the secondary sensing path, into the secondary receiving well G₁₀. The voltage on f_gis restored to a high level of

\begin{matrix} f_{g} (3 H) = (f_{g} (2 H) + \frac{Q_{s}}{C_{p2}}) ⅇ^{- H / τ_{3}} + (V_{h}^{'} - v_{th}^{'}) (1 - ⅇ^{- H / τ_{3}}) . & (24) \end{matrix}

[0067] During the sensing phase, charge, any remaining on the wire transfer node, flows through G₂₄to the primary output well, G₅. The final value of f_gis

\begin{matrix} f_{g} (4 H) = f_{g} (3 H) ⅇ^{- H / τ_{4}} + (\frac{{AV}_{r} - v_{th}}{A + 1}) (1 - ⅇ^{- H / τ_{4}}) . & (25) \end{matrix}

[0068] The time constant for the sensing transition, τ₄, equals that for the spill transition, τ₂.

[0069] Combining Eqs. 22, 24 and 25, and eliminating higher-order exponential terms, yields the resultant charge packet

[0070] Q_o=Q_s−(V_h′−v_th′+V_p)e^−H/τC_p2. (26)

[0071] In most cases, the approximation H>>τ₄is valid for this circuit and the first term, which represents a constant offset, is negligible. The result is then identically equal to the input Q_s, regardless of any parasitic capacitances.

[0072] Nondestructive charge sensing is used when a charge packet must be used multiple times. It replicates the original packet in either charge or voltage form, but does not alter the original in the process. The dynamic double-sampling technique can be applied to nondestructive charge sensing. FIG. 7A is a schematic diagram of an exemplary embodiment of a DDS sensing circuit 700 with charge output in accordance with the invention. FIGS. 7B-7E are associated energy level diagrams for a fill phase, spill phase, collection phase and sensing phase, respectively.

[0073] The circuit 700 is configured by connecting floating gate G₁₃from within a CCD register 702, to the input of a DDS core unit 704. Input charge is provided by coupling signal electrons from the floating gate channel to the DDS sensing node. These carriers are integrated in a CCD well by the DDS core unit to reconstruct the original charge packet.

[0074] The operation of circuit 700 occurs as follows. During the fill phase, f_gis pulled low to V_p. This serves two purposes. First, it provides a new supply of electrons to the DDS core unit 704. Second, it presets the floating gate low and forces electrons, underneath it from a previous cycle, to be transferred forward.

[0075] During the spill phase, f_gis preset high to a level of

\begin{matrix} f_{g} (2 H) = V_{p} ⅇ^{- H / τ_{2}} + (\frac{{AV}_{r} - v_{th}}{A + 1}) (1 - ⅇ^{- H / τ_{2}}) . & (27) \end{matrix}

[0076] The time constant for this transition is

\begin{matrix} τ_{2} = \frac{R_{24}}{(A + 1)} (C_{p2} + \frac{C_{gc} C_{cs}}{(C_{gc} + C_{cs})}), & (28) \end{matrix}

[0077] where C_p2represents parasitic junction and routing capacitances on f_gas well as channel capacitance of G₂₄and G₈. The second term in this expression corresponds to effective capacitance of the CCD well when it is empty. The high potential on the floating gate at this time leaves it well situated to receive charge during the following phase.

[0078] At the beginning of the collection phase, the floating gate well contains no charge. Its storage capacity, which is proportional to the difference between the channel potentials of G₁₃and G₁₂, is approximately equal to

[0079] |Q|_max=(V_r−v_th1−4+v_th2)(C_gc+C_cs). (29)

[0080] Variables v_th1and v_th2refer to threshold voltages of the first and second level gates and 4V represents the high level of the φ_b2clock. A quantity of charge Q_sis transferred underneath the floating gate by lowering the potential of G₁₁. The floating gate potential falls temporarily in response and its storage capacity is reduced to approximately

\begin{matrix} {| Q |}_{\max} = (V_{r} + Q_{s} (\frac{C_{gc}}{C_{gc} C_{p2} + C_{cs} (C_{gc} + C_{p2})}) (\frac{C_{gc}}{C_{gc} + C_{cs}}) - v_{th1} - 4 + v_{th2}) (C_{gc} + C_{cs}) . & (30) \end{matrix}

[0081] When Q_sis large, this reduction in storage capacity may cause some electrons to be temporarily stored underneath the G₁₂barrier, where they are no longer sensed by the floating gate. However, as charge flows through the secondary sensing path, the voltage on f_grises, storage capacity is restored, and electrons are once again confined to the floating gate. At the end of collection, the level on the floating gate is

\begin{matrix} f_{g} (3 H) = (f_{g} (2 H) + Q_{s} \frac{C_{gc}}{C_{gc} C_{p2} + C_{cs} (C_{gc} + C_{p2})}) ⅇ^{- H / τ_{3}} + (V_{h}^{'} - v_{th}^{'}) (1 - ⅇ^{- H / τ_{3}}) . & (31) \end{matrix}

[0082] During the sensing phase, any charge that remains on f_gis passed to the primary output well, and the potential on f_grises to its final value of

\begin{matrix} f_{g} (4 H) = f_{g} (3 H) ⅇ^{- H / τ_{4}} + (\frac{{AV}_{r} - v_{th}}{A + 1}) (1 - ⅇ^{- H / t_{4}}) . & (32) \end{matrix}

[0083] The time constant for this transition is

\begin{matrix} τ_{4} = \frac{R_{24}}{(A + 1)} (C_{p2} + (C_{gc} + C_{cs}) {FK}_{eff}), & (33) \end{matrix}

[0084] The factor FK_effis used to translate storage capacity into effective capacitance seen by the gate.

[0085] Combining Eqs. 27, 31 and 32, and eliminating higher-order exponential terms, yields the resultant charge packet

\begin{matrix} Q_{o} = Q_{s} \frac{C_{gc}}{C_{gc} + C_{cs}} - (V_{h}^{'} - v_{th}^{'} + V_{p}) ⅇ^{- H / τ_{4}} C_{p2} . & (34) \end{matrix}

[0086] In most cases, the approximation H>>τ₄is valid and this expression can be simplified to

\begin{matrix} Q_{o} = Q_{s} \frac{C_{gc}}{C_{gc} + C_{cs}} . & (35) \end{matrix}

[0087] The output packet depends on capacitances of the CCD well, but does not depend on parasitics on the floating gate node.

[0088] The nondestructive sensing circuit 700 of FIG. 7A produces an output charge that is a replica of its input packet. A modified version of this circuit, shown in FIG. 8A, can be used to produce an output voltage instead. FIG. 8A is a schematic diagram of an exemplary embodiment of a DDS charge sensing circuit 800 with voltage output in accordance with the invention. FIGS. 8B-8E are associated energy level diagrams for a fill phase, spill phase, collection phase and sensing phase, respectively. The circuit 800 is configured by connecting floating gate G₁₃from within a CCD register 802, to the input of a DDS core unit 804.

[0089] In the configuration of circuit 800, the final well G₇that was otherwise used to store the result, is replaced by a capacitor, C₃. In this circuit, charge packets described by Eq. 35 are generated and stored in the wells of G₅and G₁₀. During phases 3 and 4, the output node v_ois precharged to V_c. During phase 1, the result is transferred onto v_oby lowering the potential of G₅and G₁₀. The potential on v_oafter all charge has transferred is given by

\begin{matrix} v_{o} = V_{c} + \frac{Q_{s}}{C_{3}} \frac{C_{gc}}{C_{gc} + C_{cs}} . & (36) \end{matrix}

[0090] One advantage of DDS charge sensing is that it has the same voltage requirements as a simple CCD register because its floating gates are clocked to the same levels as other CCD gates. Two other advantages of DDS charge sensing are that it provides greater storage capacity in its floating gate wells and permits a large output voltage swing. Signals are converted to voltages on the output v_o, rather than within the CCD register. Storage capacity is also improved because the floating gate is a virtual ground and output swing does not reduce storage capacity.

[0091] An operation that is commonly required in A/D conversion is subtraction of the output from a multiplying D/A from an incoming analog signal. DDS circuit techniques can be used to perform this function. In a DDS approach, an input quantity of positive charge, generated through capacitor elements, is combined on a wire with an incoming packet of negative charge, from a CCD register. The D/A full scale can be either a time-varying voltage or a constant reference voltage. Since diffusions are necessary on the wire, the operation is nondepleted and subject to thermal noise, coupling, and clock feedthrough. However, DDS subtraction has the advantage that its positive charge is formed through polysilicon capacitors. These capacitors provide good element-to-element matching and low voltage dependence, which is important if the D/A full-scale is a time varying signal. DDS subtraction also has the advantage that its incoming negative signal is not subject to nonlinearities due to capacitor voltage dependence.

[0092] A circuit for performing this function is shown in FIG. 9A. FIG. 9A is a schematic diagram of an exemplary embodiment of a DDS D/A and subtraction circuit 900 in accordance with the invention. FIGS. 9B-9E are associated energy level diagrams for a fill phase, spill phase, collection phase and sensing phase, respectively.

[0093] A CCD register 902, formed by gates G₁₁, G₁₂, and G₁₃, serves as the source of electrons. Its output is connected to the input of a DDS core unit 904. N-bit D/A conversion is performed by an array of (2^N−1) identically sized units 906, each one consisting of a clamp-and-sample circuit 908 in series with a capacitor C₁. The D/A elements provide positive charge to the input of the DDS core unit. The resulting combination of negative and positive charge is integrated in receiving well G₅. The DDS core of this circuit differs from that used for charge generation in that the secondary sensing path is eliminated.

[0094] The operation of this circuit is similar to that of the DDS wire transfer described earlier. During the fill phase, f_gis pulled low to V_pto provide a new supply of electrons to the circuit. Gate G₁₂is off at this time to prevent charge from flooding backward into the source register. Meanwhile the array of signals v_f(i), within the D/A, is forced low by the precharge through M₃.

[0095] The D/A precharge remains on during the spill phase. Node f_gis reset to a level of

\begin{matrix} f_{g} (2 H) = V_{p} ⅇ^{- H / τ_{2}} + (\frac{{AV}_{r} - v_{th}}{A + 1}) (1 - ⅇ^{- H / τ_{2}}) . & (37) \end{matrix}

[0096] The time constant for this transition is

\begin{matrix} τ_{2} = \frac{R_{24} ((2^{N} - 1) C_{1} + C_{p2})}{(A + 1)} . & (38) \end{matrix}

[0097] Capacitor C_p2represents parasitic junction and routing capacitances on f_gas well as channel capacitance of G₂₄and G₁₃.

[0098] During the collection phase, a quantity of charge Q_sis transferred onto f_gby lowering the potential of G₁₁. At the same time, the digital inputs w(i), which are in a thermometer code format, are used to selectively assert some of the signals v_f(i). When w(i) is high, a positive transition of V_cis introduced across C₁(i). When w(i) is low, the capacitor voltage remains unchanged. Electrons, from the source register, are cancelled by positive displacement charge, and f_gis restored to a higher voltage. The resulting potential on f_gfrom the combination of these competing effects is

\begin{matrix} f_{g} (3 H) = f_{g} (2 H) + \frac{Q_{s} + \sum_{i = 0}^{2^{N} - 1} w (i) C_{1} V_{c}}{((2^{N} - 1) C_{1} + C_{p2})} . & (39) \end{matrix}

[0099] During the sensing phase, the net charge on f_gflows through G₂₄to the output well, and the potential on f_grises to its final value of

\begin{matrix} f_{g} (4 H) = f_{g} (3 H) ⅇ^{- H / τ_{4}} + (\frac{{AV}_{r} - v_{th}}{A + 1}) (1 - ⅇ^{- H / τ_{4}}) . & (40) \end{matrix}

[0100] The time constant for the sensing transition, τ₄, equals that for the spill transition, τ₂.

[0101] Combining Eqs. 37, 39 and 40, and eliminating higher-order exponential terms, yields the resultant charge packet

\begin{matrix} Q_{o} = (Q_{s} + \sum_{i = 0}^{2^{N} - 1} w (i) C_{1} V_{c}) (1 - ⅇ^{- H / τ_{4}}) - (\frac{{AV}_{r} - v_{th}}{A + 1} - V_{p}) ⅇ^{- H / τ_{4}} ((2^{N} - 1) C_{1} + C_{p2}) . & (41) \end{matrix}

[0102] The approximation H>>τ₄is nearly always valid for this circuit and the first term, which represents a constant offset, is negligible. The result is then identically equal to the difference between the input charge Q_sand the D/A displacement charge, regardless of parasitics on the wire transfer node, f_g.

[0103] A pair of DDS sensing circuits 700 can also be used to implement comparison. In this approach, two differential input packets are each replicated, using a DDS replicator circuit. The resulting packets are stored in CCD receiving wells and are translated to voltages at the input to a CMOS comparator during the following phase. This circuit can also incorporate differential amplification and common-mode rejection in the charge domain using the feedback techniques.

[0104] The resolution of this technique is greater than that for a configuration that performs charge-to-voltage translation on the floating gates. First, it supports a larger signal range so that the impact of comparator or amplifier input-offset voltages is reduced. Second, matching in the charge-to-voltage translation is improved because gains are determined by polysilicon capacitors and are only weakly dependent on CCD well capacitances.

[0105] Although the present invention has been shown and described with respect to several preferred embodiments thereof, various changes, omissions and additions to the form and detail thereof, may be made therein, without departing from the spirit and scope of the invention.

Claims

What is claimed is:

1. A charge processing circuit which integrates charge at an output node that is representative of an input charge provided at an input node, comprising:

a precharge path coupled to said input node, said precharge path operable for setting the potential of said input node to a fixed precharge potential prior to introduction of input charge to said input node;

a sensing path coupled to said input and output nodes which is operable for returning the potential of said input node to said fixed precharge potential subsequent to introduction of input charge to said input node; and

a feedback element having an input coupled to said input node and an output coupled to said sensing path and said precharge path, said feedback element operable for setting said fixed precharge potential.

2. The circuit of claim 1, wherein said output node comprises a CCD receiving well.

3. The circuit of claim 1, wherein said precharge path comprises a first gate coupled to said feedback element and a second gate coupled to a first clocked potential source, and wherein said sensing path comprises a third gate coupled to said feedback element and a fourth gate coupled to a second clocked potential source.

4. The circuit of claim 1 further comprising a source of said input charge, said source comprising a capacitor having an output coupled to said input node, and an input coupled to a circuitry that changes the potential across the capacitor.

5. The circuit of claim 1 further comprising a source of said input charge, said source comprising a CCD register that transfers charge onto said input node.

6. The circuit of claim 1 further comprising a first and second source of said input charge.

7. The circuit of claim 6, wherein said first source comprises a CCD register that transfers charge onto said input node.

8. The circuit of claim 7, wherein said second source comprises one or more capacitors having their outputs coupled to said input node and their inputs coupled to circuitry that changes the potential across said capacitors.

9. A method of integrating charge at an output node of a charge processing circuit that is representative of input charge provided at an input node, comprising:

replenishing charge at said input node;

draining charge from said input node;

stopping the draining of charge from said input node in response to a feedback element detecting when said input node reaches a fixed precharge potential;

introducing an input charge on said input node from a source;

draining charge from said input node onto said output node; and

stopping the draining of charge from said input node in response to a feedback element detecting when said input node reaches said fixed precharge potential.

Patent Citations (14)

Patent	Date	Inventor	Cited By
US3806772(A)	1974-04-01	Early	Applicant
US4209853(A)	1980-06-01	Hyatt	Examiner
US5247241(A)	1993-09-01	Ueda	Examiner
US5736757(A)	1998-04-01	Paul	Applicant
US5748235(A)	1998-05-01	Kondo et al.	Examiner
US5872469(A)	1999-02-01	Nestler	Examiner
US5969758(A)	1999-10-01	Sauer et al.	Examiner
US5982205(A)	1999-11-01	Vallancourt	Examiner
US6025935(A)	2000-02-01	Tseng	Examiner
US6118482(A)	2000-09-01	Clark et al.	Examiner
US6133954(A)	2000-10-01	Jie et al.	Examiner
US6147551(A)	2000-11-01	Hong	Examiner
US6320617(B1)	2001-11-01	Gee et al.	Examiner
US6366320(B1)	2002-04-01	Nair et al.	Examiner

Non-Patent Literature (1)

“A Distributed Floating-Gate Amplifier in Charge-Coupled Devices” by wen et al., ISSCC 75/ Wednesday, Feb. 12, 1975.Applicant