Time-Interleaved Analog-to-Digital Converter and Conversion Method Thereof
20220271765 · 2022-08-25
Inventors
Cpc classification
H03M1/123
ELECTRICITY
H04N25/75
ELECTRICITY
International classification
Abstract
Example embodiments relate to time-interleaved analog-to-digital converters and conversion methods thereof. One embodiment includes a slope analog-to-digital converter. The slope analog-to-digital converter includes a sample and hold stage configured to sample an analog input signal at a sampling frequency. The slope analog-to-digital converter also includes a comparator downstream to the sample and hold stage configured to compare the analog input signal to a slope signal. Further, the slope analog-to-digital converter includes a digital logic configured to receive a counter value corresponding to a voltage level of the slope signal and to sample the counter value based upon the comparison, thereby generating a digital representation of the analog input signal based upon the comparison. The slope signal is asynchronous to the sampling frequency.
Claims
1. A slope analog-to-digital converter comprising: a sample and hold stage configured to sample an analog input signal at a sampling frequency; a comparator downstream to the sample and hold stage configured to compare the analog input signal to a slope signal; and a digital logic configured to: receive a counter value corresponding to a voltage level of the slope signal; and sample the counter value based upon the comparison, thereby generating a digital representation of the analog input signal based upon the comparison, wherein the slope signal is asynchronous to the sampling frequency.
2. The slope analog-to-digital converter according to claim 1, wherein the slope signal is configured to be operable with a slope repetition period T.sub.slope, wherein the sample and hold stage is further configured to: sample the analog input signal over a sampling period T.sub.smp; and hold a voltage level of the sampled analog input signal over a hold period T.sub.hold, and wherein the hold period T.sub.hold is greater than the slope repetition period T.sub.slope.
3. The slope analog-to-digital converter according to claim 2, wherein the digital logic is further configured to sample the counter value based upon the comparison in a memory block.
4. The slope analog-to-digital converter according to claim 3, wherein the digital logic is further configured to sample the counter value in the memory block on a rising edge or on a falling edge of the comparator output, or wherein the digital logic is further configured to sample the counter value in the memory block exclusively during the hold period T.sub.hold, or wherein the digital logic is further configured to sample the counter value in the memory block at least once per hold period T.sub.hold.
5. The slope analog-to-digital converter according to claim 3, wherein the digital logic is further configured to add or subtract an offset value, preferably a fixed offset value, to the counter value.
6. The slope analog-to-digital converter according to claim 1, wherein the slope analog-to-digital converter is a single-slope analog-to-digital converter, a dual-slope analog-to-digital converter, or a multi-slope analog-to-digital converter.
7. A time-interleaved analog-to-digital converter comprising: an analog input signal; a slope generator configured to generate a slope signal having a slope repetition period T.sub.slope; a global counter configured to generate a counter value corresponding to a voltage level of the slope signal; and a plurality of slope analog-to-digital converters according to claim 1, wherein the plurality of slope analog-to-digital converters are configured to commonly receive the slope signal and the respective counter value, wherein the plurality of slope analog-to-digital converters are configured to receive the analog input signal successively in time, and wherein each of the plurality of slope analog-to-digital converters is configured to generate a digital representation of the analog input signal successively in time using the commonly received slope signal and the respective counter value.
8. The time-interleaved analog-to-digital converter according to claim 7, wherein the time-interleaved analog-to-digital converter further comprises a data aligner configured to combine the corresponding digital representation of the analog input signal from each of the plurality of slope analog-to-digital converters, thereby generating a digital representation of the analog input signal.
9. The time-interleaved analog-to-digital converter according to claim 7, wherein the time-interleaved analog-to-digital converter further comprises a clock generator configured to generate a plurality of clock signals, each having an identical clock period however different phases relative to one another.
10. The time-interleaved analog-to-digital converter according to claim 7, wherein each of the plurality of slope analog-to-digital converters corresponds to a conversion channel whereby the plurality of slope analog-to-digital converters are arranged in parallel to each other, thereby facilitating a plurality of conversion channels operable with a common slope signal.
11. The time-interleaved analog-to-digital converter according to claim 10, wherein the plurality of conversion channels are arranged in an array or a two-dimensional array.
12. The time-interleaved analog-to-digital converter according to claim 11, wherein the time-interleaved analog-to-digital converter further comprises a hierarchical sampling scheme configured to implement time-interleaving in at least two hierarchical levels.
13. The time-interleaved analog-to-digital converter according to claim 12, wherein the hierarchical sampling scheme is further configured to implement time-interleaving in one of at least two hierarchical levels vertically over rows of the two-dimensional array and to implement time-interleaving in one of at least two hierarchical levels horizontally over columns of the two-dimensional array.
14. A method for time-interleaved analog-to-digital conversion comprising: providing an analog input signal; providing a plurality of slope analog-to-digital converters according to claim 1; providing the plurality of slope analog-to-digital converters with a common slope signal and a respective counter value; receiving the analog input signal successively in time by the plurality of slope analog-to-digital converters; and generating, by each of the plurality of slope analog-to-digital converters, a digital representation of the analog input signal in time succession using the common slope signal and the respective counter value.
15. The method according to claim 14, wherein the method further comprises combining the corresponding digital representation of the analog input signal from each of the plurality of slope analog-to-digital converters, thereby generating a digital representation of the analog input signal.
Description
BRIEF DESCRIPTION OF THE DRAWINGS
[0027] Example embodiments are now further explained with respect to the drawings by way of example only, and not for limitation.
[0028]
[0029]
[0030]
[0031]
[0032]
[0033]
[0034]
[0035]
DETAILED DESCRIPTION
[0036] Reference will now be made in detail to various embodiments, examples of which are illustrated in the accompanying drawings. However, the following embodiments may be variously modified and the range of the present invention is not limited by the following embodiments. Similar entities and reference numbers in different figures have been partially omitted.
[0037]
[0038] Generally, to achieve high-speed signal conversion in, especially in DSP-based wireline receivers, all ADCs are implemented as time-interleaved SAR-ADCs. Typically, the SAR-ADC architecture demonstrates higher energy efficiency, as well as possesses the capability to be implemented in an advanced digital-centric technology node. Since its basic elements are a capacitive DAC, a comparator, and control logic, none of which requires high-performance analog transistor properties, and hence scaling into advanced nodes is relatively risk-free.
[0039] However, as described above, a disadvantage of the use of SAR-ADCs is their limited speed. Typical designs in advanced nodes reach conversion speeds up to ˜1 GS/s, which includes the use of a number of parallel time-interleaved channels, where interleaving factors of e.g. 64X are common. Future wireline communication requirements may include even higher data rates, so higher ADC sampling speeds (>100 GS/s), and thus higher interleaving factors. Since the speed of a SAR-ADC does not scale further with technology nodes, this brings along some challenges.
[0040] Since every individual SAR-ADC contains a capacitive DAC, it may incorporate a certain area to implement it. Because of the limited resolution that may be incorporated, this size is not substantial for one ADC, e.g. a lane height of 10 or 20 μm, but when adding many of those in parallel, the total ADC size becomes substantial. Apart from the economic cost of this area, it becomes problematic to distribute the high-frequency input signal over such long distances to the individual lane ADCs. The same goes for the digital outputs, since they all have to be routed over long distances to a central combined reconstructed output for further processing.
[0041] Another problem is the reference voltage distribution. The capacitive DACs in each lane needs a rather clean reference voltage in order to provide a clean residue signal for the SAR-ADC operation. These DACs also present a dynamic load to the reference buffer, which should have a low enough output impedance to keep the reference voltage steady. Therefore, on one hand making a reference buffer with low enough output impedance, and on the other hand distributing that over long distance to the ADCs is not very convenient.
[0042] Instead of trying to speed up the individual channels, the approach presented here proposes to use many more, but significantly smaller ADCs, such that the total area will be smaller. The smallest ADC is a slope ADC, often used in massively column-parallel image sensors as shown in
[0043] The slope signal V.sub.slope is generated by a slope generator 101 where the respective digital counter values are generated via a counter and control logic 102. Each column circuit 103, 104 operates as a slope ADC and is fed with the respective slope signal V.sub.slope and the respective counter values. Each column circuit 103, 104 comprises a comparator 105 that compares an analog signal 107 with the slope signal V.sub.slope. The comparison operation is illustrated in
[0044] Therefore, many parallel input voltages can be sampled simultaneously, and then compared to a common slope signal. This architecture can be heavily parallelized for a number of inputs. Each individual column 103, 104 contains only a comparator 105 and a memory 106, and all analog complexity and accuracy is shifted to the common slope generator 101. It can be implemented with high performance, as its associated power and/or area may be shared with many ADCs.
[0045] However, the big difference between an image sensor system and a time-interleaved ADC is that now all input signals to the ADC lanes are not synchronous anymore, but are shifted in time.
[0046] In
[0047] The slope ADC 200 further comprises a comparator 220 downstream to the sample and hold stage 210 and further a digital logic 230 downstream to the comparator 220. The comparator 220 comprises a first input 221 operably coupled to the output of the sample and hold stage 210, especially to the series capacitor 212, and is configured to input the sampled analog signal. The comparator 220 further comprises a second input 222 configured to receive a slope signal or an analog slope voltage. It is particularly to be noted that the slope signal is asynchronous to the ADC sampling frequency, i.e. the slope frequency runs asynchronously to the ADC clock with T.sub.slope<T.sub.hold.
[0048] The comparator 220 continuously compares the sampled analog signal to the slope signal, especially to the specific voltage level of the slope signal at a comparison instance, especially during the hold or conversion period. Upon a successful comparison, i.e. the case when the voltage level of the slope signal exceeds the voltage level of the sampled analog signal, the comparator 220 toggles its output.
[0049] On the other hand, the digital logic 230 comprises a first input 231 configured to input a counter value corresponding to the voltage level of the slope signal, especially corresponding to the monotonously increasing or decreasing slope voltage. The digital logic 230 further comprises a second input 232 operably coupled to the output of the comparator 220 and is configured to input the comparator output.
[0050] Moreover, the digital logic 230 comprises a local memory 240 and is further configured to store digital bits onto the local memory 240. The local memory 240 may be integrated with the digital logic 230. As a result, upon a successful comparison, the digital logic 230 receives the counter value on the rising edge or on the falling edge of the comparator output, indicative of the digital representation of the sampled analog signal for the specific comparison instance. The digital representation or value is then stored locally on the local memory 240. The digital logic 230 further comprises an output node 233 operably coupled to the local memory 240, through which the digital representation or value of the sampled analog input signal can be extracted and routed from the local memory 240.
[0051] In
[0052] The time-interleaved ADC 300 further comprises a plurality of slope ADCs 200 of
[0053] Furthermore, the time-interleaved ADC 300 comprises an input node common to the input nodes 201 of each of the plurality of slope ADCs 200 for receiving an analog input signal V.sub.IN. The respective clock input node 202 of the respective sample and hold stages 210 of the slope ADCs 200 are operably coupled to one of the set of outputs 321 of a clock generator 350. In this regard, each sample and hold stage 210, driven by the clock frequency of the clock signal, samples the analog input signal V.sub.IN in time succession over a sampling period T.sub.smp to generate a sampled analog signal V.sub.i (i=1, 2, 3, . . . , N), where the voltage level is stored or held in the capacitor 212 over a hold period or conversion period T.sub.hold.
[0054] For each slope ADC 200, the first input 221 of the comparator 220 is operably coupled to the respective sample and hold stage 210, especially to the series capacitor 212, and is configured to input the sampled analog signal V.sub.1. The second input 222 of the comparator 220 is operably coupled to the bus line 312 of the slope generator 311, through which the slope signal V.sub.SLOPE is globally fed.
[0055] The comparator 220 compares the sampled analog signal V.sub.1 to the slope signal V.sub.SLOPE, especially to the specific voltage level of the slope signal V.sub.SLOPE at a comparison instance, especially during the hold period or conversion period of the respective sample and hold stages 210. Upon a successful comparison, i.e. the case when the voltage level of the slope signal V.sub.SLOPE crosses the voltage level of the sampled analog signal V.sub.1, the comparator 220 toggles its output C.sub.1.
[0056] Accordingly, for each slope ADC 200, the first input 231 of the digital logic 230 is operably coupled to the bus line 214 of the global counter 211, through which the digital counter value D.sub.CNT of the corresponding slope signal V.sub.SLOPE is globally fed. The second input 232 of the digital logic 230 is operably coupled to the output of the comparator 220 and is configured to input the comparator output C.sub.1.
[0057] Although it is not explicitly shown, it should be understood that the digital logic 230 includes the memory block 240 for storing digital bits. Upon a successful comparison, the digital logic 230 receives the digital counter value D.sub.CNT, e.g. on the rising edge or on the falling edge of C.sub.1 based on whether the slope signal monotonously increases or decreases, thereby generating the digital representation of the sampled analog signal V.sub.1 for the specific comparison instance. The digital representation or value D.sub.1 is then stored locally on the local memory 240, through which the digital representation or value D.sub.1 of the sampled analog signal V.sub.1 can be extracted and routed via the output node 233 of the digital logic 230.
[0058] The time-interleaved ADC 300 further comprises a data aligner 340 operably coupled to the respective digital logic 230 of the plurality of slope ADCs 200, especially to the output node 233 of the respective digital logic 230 of the plurality of slope ADCs 200. The data aligner 340 combines the corresponding digital representation D.sub.1 of the sampled analog signal V.sub.1 from the respective local memory 240 of the respective digital logic 230 of the plurality of slope ADCs 200. This results in the complete digital representation DOUT of the analog input signal V.sub.IN, where the digital representation Dom is outputted from the data aligner 240, e.g. at an output node 341.
[0059] The time-interleaved ADC 300 moreover comprises a clock generator 350 configured to generate a plurality of clock signals, each having an identical clock period, however different phases relative to one another. The clock generator 350 is operably coupled to the time-interleaved sampling arrangement 320 so as to feed the respective clock signals to the plurality of sample and hold stages 210. In some embodiments, the clock generator 350 firstly receives a global clock signal driven by a clock frequency F.sub.s. Alternatively, the clock generator 350 may generate the global clock signal internally.
[0060] In either case, the clock generator 350 factorizes the global clock signal by an interleaving factor limited by the number of the plurality of slope ADCs 200, shifts the plurality of clock signals by one clock period with respect to each other thereby generating the plurality of clock signals that are out of phase by at least one clock period. Hence, each of the slope ADCs 200 operates on a frequency lower than the global conversion rate F.sub.s, which is defined by the interleaving factor. Although not shown in
[0061] In some embodiments, the data aligner 340 is configured to align the corresponding digital representation of the sampled analog signal from each of the plurality of slope ADCs 200 with the rising edge of the respective clock signal of the respective sample and hold stages 210. Therefore, at the end of the conversion period, the sampled or stored values are re-sampled so that the respective output is synchronous to the ADC clock.
[0062] In some embodiments, each of the slope ADCs 200 corresponds to a conversion channel for the time-interleaved ADC 300, e.g. channels 1 to N as shown in
[0063] Therefore, each conversion channel commonly receives the analog input signal V.sub.IN, operates in a time-interleaved manner, samples the analog input signal V.sub.IN in succession of time with respect to each other, and therefore generates the respective sampled analog signals. Furthermore, each conversion channel commonly receives the slope signal V.sub.SLOPE, i.e. one single slope signal, and the corresponding digital counter value, and performs signal conversion in parallel to generate the digital representation Dour of the analog input signal V.sub.IN.
[0064] In
[0065] Along the first section (a) of
[0066] T.sub.slope=T.sub.rise where the slope repetition period is limited by the external slope frequency F.sub.SLOPE, as mentioned above. Analogous to this slope signal V.sub.SLOPE, a digital counter value DINT is globally distributed to all the channels i,j,k. The value of this counter is a representation of the respective voltage level of the slope signal V.sub.SLOPE, which is defined from 0 to 15, i.e. 15 levels, in consecutive counts for a 4-bit ADC.
[0067] In general, for N number of interleaved channels, each interleaved channel operates on a frequency N times lower than the global conversion rate F.sub.s. The conversion time per channel thus equals N times F.sub.s, and is segmented in a sampling time or period T.sub.smp and a hold time or period T.sub.hold. The sampling time is defined as time to sample the input signal in an input capacitor, e.g. the series capacitor 212, and the hold time T.sub.hold is defined as the time during which the sampled input voltage is held on that capacitor 212, and is compared to the global slope signal by the comparator, e.g. the comparator 220, in the channel.
[0068] For the proper operation, it may be that the slope repetition period T.sub.slope is smaller than the hold time T.sub.hold. The slope frequency is thus higher than the channel conversion frequency, and hence the slope timing runs asynchronously from the ADC clock F.sub.s. Due to the generation of a single global slope signal for all the ADCs instead of a plurality of slope signal dedicated to each respective ADCs, the timing of all N ADC channels cannot be made such that they all perfectly align with a complete rising or falling slope of the global slope signal, which is a prerequisite of a classical column-parallel slope ADC. However, with the proposed asynchronous but faster global slope signal, each channel will come across all possible voltages of the rising or falling slope at least once during its conversion time, although not in one consecutive monotonous rising or falling slope, but in segmented pieces, e.g. in two segmented pieces. Along
[0069] The digital logic 230 in each ADC channel i,j,k samples the value of the global counter D.sub.CNT in the local memory 240 exclusively during the hold time T.sub.hold and on the rising edge of the comparator output C since the slope signal is repetitively increasing. This relates to the instance when the rising slope of the slope signal V.sub.SLOPE crosses the sampled analog input signal V. Due to T.sub.hold>T.sub.slope, it may be that this will occur at least once per conversion cycle. In some situations, especially if T.sub.hold is much larger than T.sub.slope, this can occur twice per conversion cycle, however it will always be the same counter value D that is sampled. At the end of the conversion cycle, the stored values are re-sampled such that the output is synchronous to the ADC clock F.sub.s.
[0070] In some embodiments, the digital counter value D.sub.CNT is represented by a Gray code, in order to avoid time skew problems between the different bits when sampling its value in the local memory 240 of a channel. In this regard, a Gray-to-binary converter is incorporated into the digital logic 230 of the channel to convert the output to standard binary values for further processing. In addition to this, the digital logic 230 may add or subtract a fixed offset code to the value of the global counter D.sub.CNT to compensate for mismatches, e.g. in threshold voltage and in delay, in the comparator 220.
[0071] Turning back to
[0072] The third section (c) of
[0073] The fourth section (d) of
[0074] The above-mentioned examples show the correct operation of the proposed time-interleaved ADC 300 with a slope frequency asynchronous and faster than the ADC interleaved clock. Although the above-mentioned examples are illustrated with respect to a particular time-interleaved ADC implementation having 4-bit single-slope ADCs for each conversion channels, it should be noted that the underlying technique is compatible for ADCs operating with higher slopes, e.g. dual-slope ADCs or multi-slope ADCs. Therefore, it may be possible to implement a time-interleaved ADC comprising a plurality of multi-slope ADCs, while maintaining adequate conversion accuracy with a high interleaving factor, especially when operating on a global slope signal with a slope frequency asynchronous and faster than the ADC interleaved clock.
[0075] In
[0076] Because of the extremely small size of a slope ADC, the size of individual channels is also extremely small. Therefore, many of them can be interleaved in a manageable total area. In addition, because the speed of a slope ADC is much lower than the typically used SAR-ADCs, it may also incorporate a high interleaving factor to achieve sufficient aggregate conversion speed. Hence, a substantial amount of time-interleaved channels are implemented in a two-dimensional array as shown in
[0077] In particular, the time-interleaved ADC 500 comprises a hierarchical sampling scheme 520 that implements time-interleaving in at least two hierarchical levels. In this regard, the hierarchical sampling scheme 520 implements time-interleaving in one of at least two hierarchical levels vertically over rows of the two-dimensional array, i.e. a first interleaving rank, and to implement time-interleaving in one of at least two hierarchical levels horizontally over columns of the two-dimensional array, i.e. a second interleaving rank.
[0078] The first interleaving rank 520 is depicted vertically in the figure and splits the input signal V.sub.IN in N channels, each sampled by one of the N interleaved sampling clocks F.sub.Si, where i=1, 2, 3, . . . , N, at a frequency fs/N, spaced one sample Ts apart. The resulting signals V.sub.1, where i=1, 2, 3, . . . , N, are distributed horizontally to the second interleaving rank 211, shown in the block 510 of the example conversion channels, that is split into M channels, each sampled by one of the N×M interleaved sampling clocks F.sub.si,j, where i=1, 2, 3, . . . , N; j=1, 2, 3, . . . , M, at a frequency fs/(N×M). In this regard, the resulting signals V; are distributed directly through a distribution path 521 as well as optionally buffered through a buffered path 523.
[0079] The slope signal V.sub.SLOPE and its synchronous digital counter value D.sub.CNT are distributed to all channels of the N×M array through their respective routing paths 512 and 514. The functionality of each conversion channel 510 is the same as described along
[0080] In order to limit the amount of digital outputs routed horizontally towards the data aligner 340, a common bus 511 is implemented. In this regard, each of the outputs D.sub.i,j, where i=1, 2, 3, . . . , N; j=1, 2, 3, . . . , M, is sequentially placed on a data bus Di, controlled e.g. by its sampling clock to prevent data conflicts on the bus.
[0081] This two-dimensional arrangement is a convenient way to time-interleave a substantial number of N×M slope ADC channels in a limited area, while keeping sufficient routing of the input signal and clock limited, such that the associated parasitic will not grow out of bound, which would limit the aggregate conversion speed that can be obtained.
[0082] In
[0083] This may result in a smaller area per channel and effectively help keeping digital noise away from the analog sensitive blocks. Instead of a digital bus with the output words of every channel running from left to right, now each of the M comparator outputs C.sub.N,M are routed to the digital block 640. The slope signal V.sub.SLOPE is commonly distributed to the conversion channels 610 throughout the conversion matrix via a routing path 612. The digital counter value D.sub.CNT is now routed to the digital block 640 via a routing path 614, however commonly distributed to each respective digital logic 230 of the respective conversion channel 610.
[0084] The time-interleaved ADC 500 and the time-interleaved ADC 600 both depict the case where one single global slope is distributed to all ADC channels. In order to limit e.g. supply bounce caused by the increased area of the conversion matrix, which may have a signature with a frequency F.sub.SLOPE, it is also conceivable to use a limited number of similar but time-shifted slopes and their respective counter values, and assign each ADC channel to one of these. This may spread out the supply bounce more evenly in time, which will limit its impact. However, the functionality of the time-interleaved slope ADCs will not be affected by the implementation.
[0085] In
[0086] Embodiments herein can be implemented by hardware, software, or any combination thereof. Various embodiments may be implemented by one or more application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), processors, controllers, microcontrollers, microprocessors, or the like.
[0087] Although embodiments been illustrated and described with respect to one or more implementations, equivalent alterations and modifications will occur to others skilled in the art upon the reading and understanding of this specification and the annexed drawings. For example, instead of a single-slope ADC, a dual-slope ADC or a multi-slope ADC can be implemented at each conversion channel. Further, although single-ended implementations are shown for one or more embodiments, persons skilled in the art can understand modifications to differential implementations. In addition, while a particular feature may have been disclosed with respect to only one of several implementations, such feature may be combined with one or more other features of the other implementations as may be desired for any given or particular application.