ADS9120 16 位、2.5MSPS、15.5mW、SAR ADC，具有 multiSPI接口

7.3.1 Converter Module

As shown in Figure 36, the converter module samples the analog input signal (provided between the AINP and AINM pins), compares this signal with the reference voltage (provided between the pair of REFP and REFM pins), and generates an equivalent digital output code.

The converter module receives RST and CONVST inputs from the interface module and outputs the ADCST signal and the conversion result back to the interface module.

Figure 36. Converter Module

7.3.1.1 Sample-and-Hold Circuit

The device supports unipolar, fully-differential analog input signals. Figure 37 shows a small-signal equivalent circuit of the sample-and-hold circuit. Each sampling switch is represented by a resistance (R_s1 and R_s2, typically 30 Ω) in series with an ideal switch (sw₁ and sw₂). The sampling capacitors, C_s1 and C_s2, are typically 60 pF.

Figure 37. Input Sampling Stage Equivalent Circuit

During the acquisition process (in ACQ state), both positive and negative inputs are individually sampled on C_s1 and C_s2, respectively. During the conversion process (in CNV state), the device converts for the voltage difference between the two sampled values: V_AINP – V_AINM.

Each analog input pin has electrostatic discharge (ESD) protection diodes to REFP and GND. Keep the analog inputs within the specified range to avoid turning the diodes on.

Equation 1 and Equation 2 show the full-scale voltage range (FSR) and common-mode voltage range (V_CM) supported at the analog inputs for any external reference voltage (V_REF).

Equation 1.

Equation 2.

7.3.1.2 External Reference Source

The input range for the device is set by the external voltage applied at the two REFP pins. The REFM pins function as the reference ground and must be connected to each reference capacitor.

The device takes very little static current from the reference pins in the RST and ACQ states. During the conversion process (in CNV state), binary-weighted capacitors are switched onto the reference pins. The switching frequency is proportional to the conversion clock frequency, but the dynamic charge requirements are a function of the absolute values of the input voltage and the reference voltage. Reference capacitors decouple the dynamic reference loads and a low-impedance reference driver is required to keep the voltage regulated to within 1 LSB.

Most reference sources have very high broadband noise. The voltage reference source is recommended to be filtered with a 160-Hz filter before being connected to the reference driver, as shown in Figure 38. See the ADC Reference Driver section for the reference capacitor and driver selection. Also, the reference inputs are sensitive to board layout; thus, the layout guidelines described in the Layout section must be followed.

Figure 38. Reference Driver Schematic

7.3.1.4 ADC Transfer Function

The ADS9120 supports unipolar, fully-differential analog inputs. The device output is in twos compliment format. Figure 39 and Table 1 show the ideal transfer characteristics for the device.

The LSB for the ADC is given by Equation 3:

Equation 3.

where

Figure 39. Differential Transfer Characteristics

Table 1. Transfer Characteristics

DIFFERENTIAL ANALOG INPUT VOLTAGE (AINP – AINM)	OUTPUT CODE (Hex)
< –VREF	8000
–VREF + 1 LSB	8001
–1 LSB	FFFF
0	0000
1 LSB	0001
> VREF – 1 LSB	7FFF

7.3.2 Interface Module

The interface module facilitates the communication and data transfer between the device and the host controller. As shown in Figure 40, the module comprises of shift registers (both input data and output data), configuration registers, and a protocol unit.

Figure 40. Interface Module

The Pin Configuration and Functions section provides descriptions of the interface pins; the Data Transfer Frame section details the functions of shift registers, the SCLK counter, and the command processor; the Data Transfer Protocols section details supported protocols; and the Register Maps section explains the configuration registers and bit settings.

7.4.1 RST State

In the ADS9120, the RST pin is an asynchronous digital input. To enter RST state, the host controller must pull the RST pin low and keep it low for the t_{wl_RST} duration (as specified in the Timing Requirements: Asynchronous Reset, NAP, and PD table).

In RST state, all configuration registers (see the Register Maps section) are reset to the default values, the RVS pins remain low, and the SDO-x pins are tri-stated.

To exit RST state, the host controller must pull the RST pin high with CONVST and SCLK held low and CS held high, as shown in Figure 42. After a delay of t_{d_rst}, the device enters ACQ state and the RVS pin goes high.

Figure 42. Asynchronous Reset

7.4.2 ACQ State

In ACQ state, the device acquires the analog input signal. The device enters ACQ state on power-up, after any asynchronous reset, or after end of every conversion.

An RST falling edge takes the device from an ACQ state to a RST state. A CONVST rising edge takes the device from an ACQ state to a CNV state.

The device offers a low-power NAP mode to reduce power consumption in the ACQ state; see the NAP Mode section for more details on NAP mode.

7.4.3 CNV State

The device moves from ACQ state to CNV state on a rising edge of the CONVST pin. The conversion process uses an internal clock and the device ignores any further transitions on the CONVST signal until the ongoing conversion is complete (that is, during the time interval of t_conv).

At the end of conversion, the device enters ACQ state. The cycle time for the device is given by Equation 4:

Equation 4.

where

Table 2. Supported Commands

7.5.1 Data Transfer Frame

A data transfer frame between the device and the host controller is bounded between a CS falling edge and the subsequent CS rising edge. The host controller can initiate a data transfer frame (as shown in Figure 46) at any time irrespective of the status of the CONVST signal; however, the data read during such a data transfer frame is a function of relative timing between the CONVST and CS signals.

Figure 46. Data Transfer Frame

For this discussion, assume that the CONVST signal remains low.

For a typical data transfer frame F:

1. The host controller pulls CS low to initiate a data transfer frame. On the CS falling edge:
   • RVS goes low, indicating the beginning of the data transfer frame.
   • The SCLK counter is reset to 0.
   • The device takes control of the data bus. As shown in Figure 46, the 20-bit contents of the output data word (see Figure 44) are loaded in to the 20-bit ODR (see Figure 40).
   • The 20-bit IDR (see Figure 40) is reset to 00000h, corresponding to a NOP command.
2. During the frame, the host controller provides clocks on the SCLK pin:
   • On each SCLK capture edge, the SCLK counter is incremented and the data bit received on the SDI pin is shifted in to the IDR.
   • On each launch edge of the output clock (SCLK in this case), ODR data are shifted out on the selected SDO-x pins.
   • The status of the RVS pin depends on the output protocol selection (see the Protocols for Reading From the Device section).
3. The host controller pulls CS high to end the data transfer frame. On the CS rising edge:
   • The SDO-x pins go to tri-state.
   • RVS goes high (after a delay of t_{d_RVS}).
   • As illustrated in Figure 46, the 20-bit contents of the IDR are transferred to the command processor (see Figure 40) for decoding and further action.

After pulling CS high, the host controller must monitor for a low-to-high transition on the RVS pin or wait for the t_{d_RVS} time (see the Timing Requirements: SPI-Compatible Serial Interface table) to elapse before initiating a new operation (data transfer or conversion). The delay, t_{d_RVS}, for any data transfer frame F varies based on the data transfer operation executed in the frame F.

At the end of the data transfer frame F:

• If the SCLK counter is < 20, it indicates that IDR has captured less than 20 bits from the SDI. In this case, the device treats the frame F as a short command frame. At the end of a short command frame frame, IDR is not updated and the device treats the frame as a no operation command.
• If the SCLK counter = 20, it indicates that the IDR has captured exactly 20 bits from SDI. In this case, the device treats the frame F as a optimal command frame. At the end of an optimal command frame, the command processor decodes the 20-bit contents of the IDR as a valid command word.
• If the SCLK counter > 20, it indicates that the IDR captured more than 20 bits from the SDI, and only the last 20 bits have been retained. In this case, the device treats the frame F as a long command frame. At the end of a long command frame, the command processor treats the 20-bit contents of the IDR as a valid command word. There is no restriction on the maximum number of clocks that can be provided within any data transfer frame F. However, as explained above, the last 20 bits shifted into the device prior to the CS rising edge must constitute the desired command.

In a short command frame, the write operation to the device is invalidated, however, the output data bits transferred during the frame are still valid output data. Therefore, the host controller can use such shorter data transfer frames to read only the required number of MSB bits from the 20-bit output data word. As shown in Figure 44, an optimal read frame for ADS9120 needs to read only the 16 MSB bits of the output data word. The length of an optimal read frame depends on the output protocol selection; refer to the Protocols for Reading From the Device section for more details.

7.5.2 Interleaving Conversion Cycles and Data Transfer Frames

The host controller can operate the ADS9120 at the desired throughput by interleaving the conversion cycles and the data transfer frames.

The cycle time of the device, t_cycle, is the time difference between two consecutive CONVST rising edges provided by the host controller. The response time of the device, t_resp, is the time difference between the host controller initiating a conversion C and the host controller receiving the complete result for conversion C.

Figure 47 shows three conversion cycles, C, C+1, and C+2. Conversion C is initiated by a CONVST rising edge at the t = 0 time and the conversion result becomes available for data transfer at the t_conv time. However, this result is loaded into the ODR only on the subsequent CS falling edge. This CS falling edge must be provided before the completion of the conversion C+1 (that is, before the t_cycle + t_conv time).

To achieve the rated performance specifications, the host controller must ensure that no digital signals toggle during the quiet acquisition time (t_{qt_acq}) and quiet aperture time (t_{d_cnvcap}), as shown in Figure 47. Any noise during t_{d_cnvcap} can negatively affect the result of the ongoing conversion whereas any noise during t_{qt_acq} can negatively affect the acquisition of the subsequent sample (and hence it's conversion result).

Figure 47. Data Transfer Zones

This architecture allows for two distinct time zones (zone1 and zone2) to transfer data for each conversion. Zone1 and zone2 for conversion C are defined in Table 3.

Table 3. Data Transfer Zones Timing

ZONE	STARTING TIME	ENDING TIME
Zone1 for conversion C
Zone2 for conversion C

The response time includes the conversion time and the data transfer time, and is thus a function of the data transfer zone selected.

Figure 48 and Figure 49 illustrate interleaving of three conversion cycles (C, C+1, and C+2) with three data transfer frames (F, F+1, and F+2) in zone1 and in zone2, respectively.

Figure 48. Zone1 Data Transfer

Figure 49. Zone2 Data Transfer

To achieve cycle time, t_cycle, the read time in zone1 is given by Equation 5:

Equation 5.

For an optimal read frame, Equation 5 results in an SCLK frequency given by Equation 6:

Equation 6.

Then, the zone1 data transfer achieves a response time defined by Equation 7:

Equation 7.

As an example, when operating the ADS9120 at the full throughput of 2.5 MSPS, the host controller can achieve a response time of 400 ns provided that the data transfer in zone1 is completed within 85 ns. However, to achieve this response time, the SCLK frequency must be greater than 188 MHz.

Note that the device does not support such high SCLK speeds.

Data transfer in zone2 can acheive lower SCLK speeds for the same cycle time. The read time in zone2 is given by Equation 8:

Equation 8.

For an optimal data transfer frame, Equation 8 results in an SCLK frequency given by Equation 9:

Equation 9.

Then, the zone2 data transfer achieves a response time defined by Equation 10:

Equation 10.

As an example, the host controller can operate the ADS9120 at the full throughput of 2.5 MSPS using zone2 data transfer with a 44 MHz SCLK (and a read time of 365 ns). However, zone2 data transfer results in a response time of nearly 800 ns.

There is no upper limit on t_read-Z1 and t_read-Z2, however, any increase in these read times will increase the response time and may increase the cycle time.

For a given cycle time, the zone1 data transfer clearly achieves faster response time but also requires a higher SCLK speed (as evident from Equation 5, Equation 6, and Equation 7), whereas the zone2 data transfer clearly requires a lower SCLK speed but supports slower response time (as evident from Equation 8, Equation 9, and Equation 10).

7.5.3 Data Transfer Protocols

The device features a multiSPI™ interface that allows the host controller to operate at slower SCLK speeds and still achieve the required cycle time with a faster response time. The multiSPI™ interface module offers two options to reduce the SCLK speed required for data transfer:

1. An option to increase the width of the output data bus
2. An option to enable double data rate (DDR) transfer

These two options can be combined to achieve further reduction in SCLK speed.

Figure 50 shows the delays between the host controller and the device in a typical serial communication.

Figure 50. Delays in Serial Communication

If t_{pcb_CK} and t_{pcb_SDO} are the delays introduced by the PCB traces for the serial clock and SDO signals, t_{d_CKDO} is the clock-to-data delay of the device, t_{d_ISO} is the propagation delay introduced by the digital isolator, and t_{su_h} is the set up time specification of the host controller, then the total delay in the path is given by Equation 11:

Equation 11.

In a standard SPI protocol, the host controller and the device launch and capture data bits on alternate SCLK edges. Therefore, the t_{d_total_serial} delay must be kept less than half of the SCLK duration. Equation 12 shows the fastest clock allowed by the SPI protocol.

Equation 12.

Larger values of the t_{d_total_serial} delay restrict the maximum SCLK speed for the SPI protocol, resulting in higher read and response times, and can increase cycle times. To remove this restriction on the SCLK speed, the multiSPI™ interface module supports an ADC-Clock-Master or a source-synchronous mode of operation.

As illustrated in Figure 51, in the ADC-Clock-Master or source-synchronous mode, the device provides a synchronous output clock (on the RVS pin) along with the output data (on the SDO-x pins).

For negligible values of t_{off_STRDO}, the total delay in the path for a source-synchronous data transfer, is given by Equation 13:

Equation 13.

As illustrated in Equation 11 and Equation 13, the ADC-Clock-Master or source-synchronous mode completely eliminates the affect of isolator delays (t_{d_ISO}) and the clock-to-data delays (t_{d_CKDO}), which are typically the largest contributors in the overall delay computation.

Figure 51. Delays in Source-Synchronous Communication

Furthermore, the actual values of t_{pcb_RVS} and t_{pcb_SDO} do not matter. In most cases, the t_{d_total_srcsync} delay can be kept at a minimum by routing the RVS and SDO lines together on the PCB. Therefore, the ADC-Clock-Master or source-synchronous mode allows the data transfer between the host controller and the device to operate at much higher SCLK speeds.

7.5.3.1 Protocols for Configuring the Device

As shown in Table 4, the host controller can use any of the four legacy, SPI-compatible protocols (SPI-00-S,
SPI-01-S, SPI-10-S, or SPI-11-S) to write data in to the device.

Table 4. SPI Protocols for Configuring the Device

PROTOCOL	SCLK POLARITY (At CS Falling Edge)	SCLK PHASE (Capture Edge)	SDI_CNTL	SDO_CNTL	#SCLK (Optimal Command Frame)	DIAGRAM
SPI-00-S	Low	Rising	00h	00h	20	Figure 52
SPI-01-S	Low	Falling	01h	00h	20	Figure 53
SPI-10-S	High	Falling	02h	00h	20	Figure 54
SPI-11-S	High	Rising	03h	00h	20	Figure 55

On power-up or after coming out of any asynchronous reset, the device supports the SPI-00-S protocol for data read and data write operations.

To select a different SPI-compatible protocol, program the SDI_MODE[1:0] bits in the SDI_CNTL register. This first write operation must adhere to the SPI-00-S protocol. Any subsequent data transfer frames must adhere to the newly selected protocol.

Figure 52 to Figure 55 detail the four protocols using an optimal command frame; see the Timing Requirements: SPI-Compatible Serial Interface section for associated timing parameters.

NOTE

As explained in the Data Transfer Frame section, a valid write operation to the device requires a minimum of 20 SCLKs to be provided within a data transfer frame.

Any data write operation to the device must continue to follow the SPI-compatible protocol selected in the SDI_CNTL register, irrespective of the protocol selected for the data read operation.

Figure 52. SPI-00-S Protocol, Optimal Command Frame

Figure 54. SPI-10-S Protocol, Optimal Command Frame

Figure 53. SPI-01-S Protocol, Optimal Command Frame

Figure 55. SPI-11-S Protocol, Optimal Command Frame

7.5.3.2 Protocols for Reading From the Device

The protocols for the data read operation can be broadly classified into three categories:

1. Legacy, SPI-compatible (SPI-xy-S) protocols,
2. SPI-compatible protocols with bus width options (SPI-xy-D and SPI-xy-Q), and
3. Source-synchronous (SRC) protocols

7.5.3.2.1 Legacy, SPI-Compatible (SYS-xy-S) Protocols

As shown in Table 5, the host controller can use any of the four legacy, SPI-compatible protocols (SPI-00-S, SPI-01-S, SPI-10-S, or SPI-11-S) to read data from the device.

Table 5. SPI Protocols for Reading From the Device

PROTOCOL	SCLK POLARITY (At CS Falling Edge)	SCLK PHASE (Capture Edge)	MSB BIT LAUNCH EDGE	SDI_CNTL	SDO_CNTL	#SCLK (Optimal Read Frame)	DIAGRAM
SPI-00-S	Low	Rising	CS falling	00h	00h	16	Figure 56
SPI-01-S	Low	Falling	1st SCLK rising	01h	00h	16	Figure 57
SPI-10-S	High	Falling	CS falling	02h	00h	16	Figure 58
SPI-11-S	High	Rising	1st SCLK falling	03h	00h	16	Figure 59

On power-up or after coming out of any asynchronous reset, the device supports the SPI-00-S protocol for data read and data write operations. To select a different SPI-compatible protocol for both the data transfer operations:

1. Program the SDI_MODE[1:0] bits in the SDI_CNTL register. This first write operation must adhere to the SPI-00-S protocol. Any subsequent data transfer frames must adhere to the newly selected protocol.
2. Set the SDO_MODE[1:0] bits = 00b in the SDO_CNTL register.

When using any of the SPI-compatible protocols, the RVS output remains low throughout the data transfer frame; see the Timing Requirements: SPI-Compatible Serial Interface table for associated timing parameters.

NOTE

It is recommended to use any of the four SPI-compatible protocols to execute the RD_REG and WR_REG operations specified in Table 2.

Figure 56 to Figure 59 explain the details of the four protocols using an optimal command frame to read all 20 bits of the output data word. Table 5 shows the number of SCLK required in an optimal read frame for the different output protocol selections.

With SDO_CNTL[7:0] = 00h, if the host controller uses a long data transfer frame, the device exhibits daisy-chain operation (see the Multiple Devices: Daisy-Chain Topology section).

Figure 56. SPI-00-S Protocol, 20 SCLKs

Figure 58. SPI-10-S Protocol, 20 SCLKs

Figure 57. SPI-01-S Protocol, 20 SCLKs

Figure 59. SPI-11-S Protocol, 20 SCLKs

7.5.3.2.2 SPI-Compatible Protocols with Bus Width Options

The device provides an option to increase the SDO bus width from one bit (default, single SDO) to two bits (dual SDO) or to four bits (quad SDO) when operating with any of the four legacy, SPI-compatible protocols.

Set the SDO_WIDTH[1:0] bits in the SDO_CNTL register to select the SDO bus width.

In dual SDO mode (SDO_WIDTH[1:0] = 10b), two bits of data are launched on the two SDO pins (SDO-0 and SDO-1) on every SCLK launch edge.

In quad SDO mode (SDO_WIDTH[1:0] = 11b), four bits of data are launched on the four SDO pins (SDO-0, SDO-1, SDO-2, and SDO-3) on every SCLK launch edge.

The SCLK launch edge depends upon the SPI protocol selection (as shown in Table 6).

Table 6. SPI-Compatible Protocols with Bus Width Options

PROTOCOL	SCLK POLARITY (At CS Falling Edge)	SCLK PHASE (Capture Edge)	MSB BIT LAUNCH EDGE	SDI_CNTL	SDO_CNTL	#SCLK (Optimal Read Frame)	DIAGRAM
SPI-00-D	Low	Rising	CS falling	00h	08h	8	Figure 60
SPI-01-D	Low	Falling	First SCLK rising	01h	08h	8	Figure 61
SPI-10-D	High	Falling	CS falling	02h	08h	8	Figure 62
SPI-11-D	High	Rising	First SCLK falling	03h	08h	8	Figure 63
SPI-00-Q	Low	Rising	CS falling	00h	0Ch	4	Figure 64
SPI-01-Q	Low	Falling	First SCLK rising	01h	0Ch	4	Figure 65
SPI-10-Q	High	Falling	CS falling	02h	0Ch	4	Figure 66
SPI-11-Q	High	Rising	First SCLK falling	03h	0Ch	4	Figure 67

When using any of the SPI-compatible protocols, the RVS output remains low throughout the data transfer frame; see the Timing Requirements: SPI-Compatible Serial Interface table for associated timing parameters.

Figure 60 to Figure 67 illustrate how the wider data bus allows the host controller to read all 20 bits of the output data word using shorter data transfer frames. Table 6 shows the number of SCLK required in an optimal read frame for the different output protocol selections.

NOTE

With SDO_CNTL[7:0] ≠ 00h, a long data transfer frame does not result in daisy-chain operation. On SDO pin(s), the 20 bits of output data word are followed by 0's.

Figure 60. SPI-00-D Protocol

Figure 62. SPI-10-D Protocol

Figure 64. SPI-00-Q Protocol

Figure 66. SPI-10-Q Protocol

Figure 61. SPI-01-D Protocol

Figure 63. SPI-11-D Protocol

Figure 65. SPI-01-Q Protocol

Figure 67. SPI-11-Q Protocol

7.5.3.2.3 Source-Synchronous (SRC) Protocols

As described in the Data Transfer Protocols section, the multiSPI™ interface supports an ADC-Clock-Master or a source-synchronous mode of data transfer between the device and host controller. In this mode, the device provides an output clock that is synchronous with the output data. Furthermore, the host controller can also select the output clock source, data bus width, and data transfer rate.

7.5.3.2.3.1 Output Clock Source Options with SRC Protocols

In all SRC protocols, the RVS pin provides the output clock. The device allows this output clock to be synchronous to either the external clock provided on the SCLK pin or to the internal clock of the device. Furthermore, this internal clock can be divided by a factor of two or four to lower the data rates.

As shown in Figure 68, set the SSYNC_CLK_SEL[1:0] bits in the SDO_CNTL register to select the output clock source.

Figure 68. Output Clock Source options with SRC Protocols

7.5.3.2.3.2 Bus Width Options with SRC Protocols

The device provides an option to increase the SDO bus width from one bit (default, single SDO) to two bits (dual SDO) or to four bits (quad SDO) when operating with any of the SRC protocols. Set the SDO_WIDTH[1:0] bits in the SDO_CNTL register to select the SDO bus width.

In dual SDO mode (SDO_WIDTH[1:0] = 10b), two bits of data are launched on the two SDO pins (SDO-0 and SDO-1) on every SCLK rising edge.

In quad SDO mode (SDO_WIDTH[1:0] = 11b), four bits of data are launched on the four SDO pins (SDO-0, SDO-1, SDO-2, and SDO-3) on every SCLK rising edge.

7.5.3.2.3.3 Output Data Rate Options with SRC Protocols

The device provides an option to transfer the data to the host controller at single data rate (default, SDR) or at double data rate (DDR). Set the DATA_RATE bit in the SDO_CNTL register to select the data transfer rate.

In SDR mode (DATA_RATE = 0b), the RVS pin toggles from low to high and the output data bits are launched on the SDO pins on the output clock rising edge.

In DDR mode (DTA_RATE = 1b), the RVS pin toggles and the output data bits are launched on the SDO pins on every output clock edge, starting with the first rising edge.

The device supports all 24 combinations of output clock source, bus width, and output data rate, as shown in Table 7.

Table 7. SRC Protocol Combinations

PROTOCOL	OUTPUT CLOCK SOURCE	BUS WIDTH	OUTPUT DATA RATE	SDI_CNTL	SDO_CNTL	#OUTPUT CLOCK (Optimal Read Frame)	DIAGRAM
SRC-EXT-SS	SCLK	Single	SDR	00h, 01h, 02h, or 03h	03h	16	Figure 69
SRC-INT-SS	INTCLK	Single	SDR		43h	16	Figure 70
SRC-IB2-SS	INTCLK / 2	Single	SDR		83h	16
SRC-IB4-SS	INTCLK / 4	Single	SDR		C3h	16
SRC-EXT-DS	SCLK	Dual	SDR		0Bh	8	Figure 73
SRC-INT-DS	INTCLK	Dual	SDR		4Bh	8	Figure 74
SRC-IB2-DS	INTCLK / 2	Dual	SDR		8Bh	8
SRC-IB4-DS	INTCLK / 4	Dual	SDR		CBh	8
SRC-EXT-QS	SCLK	Quad	SDR		0Fh	4	Figure 77
SRC-INT-QS	INTCLK	Quad	SDR		4Fh	4	Figure 78
SRC-IB2-QS	INTCLK / 2	Quad	SDR		8Fh	4
SRC-IB4-QS	INTCLK / 4	Quad	SDR		CFh	4
SRC-EXT-SD	SCLK	Single	DDR		13h	8	Figure 71
SRC-INT-SD	INTCLK	Single	DDR		53h	8	Figure 72
SRC-IB2-SD	INTCLK / 2	Single	DDR		93h	8
SRC-IB4-SD	INTCLK / 4	Single	DDR		D3h	8
SRC-EXT-DD	SCLK	Dual	DDR		1Bh	4	Figure 75
SRC-INT-DD	INTCLK	Dual	DDR		5Bh	4	Figure 76
SRC-IB2-DD	INTCLK / 2	Dual	DDR		9Bh	4
SRC-IB4-DD	INTCLK / 4	Dual	DDR		DBh	4
SRC-EXT-QD	SCLK	Quad	DDR		1Fh	2	Figure 79
SRC-INT-QD	INTCLK	Quad	DDR		5Fh	2	Figure 80
SRC-IB2-QD	INTCLK / 2	Quad	DDR		9Fh	2
SRC-IB4-QD	INTCLK / 4	Quad	DDR		DFh	2

Figure 69 to Figure 80 show the details of varoous source synchronous protocols. Table 7 shows the number of output clocks required in an optimal read frame for the different output protocol selections.

Figure 69. SRC-EXT-SS: SRC, SCLK, Single SDO, SDR

Figure 71. SRC-EXT-SD: SRC, SCLK, Single SDO, DDR

Figure 73. SRC-EXT-DS: SRC, SCLK, Dual SDO, SDR

Figure 70. SRC-INT-SS: SRC, INTCLK, Single SDO, SDR

Figure 72. SRC-INT-SD: SRC, INTCLK, Single SDO, DDR

Figure 74. SRC-INT-DS: SRC, INTCLK, Dual SDO, SDR

Figure 75. SRC-EXT-DD: SRC, SCLK, Dual SDO, DDR

Figure 77. SRC-EXT-QS: SRC, SCLK, Quad SDO, SDR

Figure 79. SRC-EXT-QD: SRC, SCLK, Quad SDO, DDR

Figure 76. SRC-INT-DD: SRC, INTCLK, Dual SDO, DDR

Figure 78. SRC-INT-QS: SRC, INTCLK, Quad SDO, SDR

Figure 80. SRC-INT-QD: SRC, INTCLK, Quad SDO, DDR

7.5.4 Device Setup

The multiSPI™ interface and the device configuration registers offer multiple operation modes. This section describes how to select the hardware connection topology to meet different system requirements.

7.5.4.1 Single Device: All multiSPI™ Options

Figure 81 shows the connections between a host controller and a stand-alone device to exercise all options provided by the multiSPI™ interface.

Figure 81. multiSPI™ Interface, All Pins

7.5.4.2 Single Device: Minimum Pins for a Standard SPI Interface

Figure 82 shows the minimum-pin interface for applications using a standard SPI protocol.

Figure 82. SPI Interface, Minimum Pins

The CS, SCLK, SDI, and SDO-0 pins constitute a standard SPI port of the host controller. The CONVST pin can be tied to CS, or can be controlled independently for additional timing flexibility. The RST pin can be tied to DVDD. The RVS pin can be monitored for timing benefits. The SDO-1, SDO-2, and SDO-3 pins have no external connections.

7.5.4.3 Multiple Devices: Daisy-Chain Topology

A typical connection diagram showing multiple devices in a daisy-chain topology is shown in Figure 83.

Figure 83. Daisy-Chain Connection Schematic

The CONVST, CS, and SCLK inputs of all devices are connected together and controlled by a single CONVST, CS, and SCLK pin of the host controller, respectively. The SDI input pin of the first device in the chain (device 1) is connected to the SDO pin of the host controller, the SDO-0 output pin of device 1 is connected to the SDI input pin of device 2, and so forth. The SDO-0 output pin of the last device in the chain (device N) is connected to the SDI pin of the host controller.

To operate multiple devices in a daisy-chain topology, the host controller must program the configuration registers in each device with identical values and must operate with any of the legacy, SPI-compatible protocols for data read and data write operations (SDO_CNT[7:0] = 00h). With these configurations settings, the 20-bit ODR and 20-bit IDR registers in each device collapse to form a single, 20-bit unified shift register (USR) per device, as shown in Figure 84.

Figure 84. Unified Shift Register

All devices in the daisy-chain topology sample their analog input signals on the CONVST rising edge. The data transfer frame starts with a CS falling edge. On each SCLK launch edge, every device in the chain shifts out the MSB of its USR on to its SDO-0 pin. On every SCLK capture edge, each device in the chain shifts in data received on its SDI pin as the LSB bit of its USR. Therefore, in a daisy-chain configuration, the host controller receives the data of device N, followed by the data of device N-1, and so forth (in MSB-first fashion). On the CS rising edge, each device decodes the contents in its USR and takes appropriate action.

A typical timing diagram for three devices connected in daisy-chain topology and using the SPI-00-S protocol is shown in Figure 85.

Figure 85. Three Devices in Daisy-Chain Mode Timing Diagram

Note that the overall throughput of the system is proportionally reduced with the number of devices connected in a daisy-chain topology.

WARNING

For N devices connected in daisy-chain topology, an optimal command frame must contain 20 × N SCLK capture edges. For a longer data transfer frame (number of SCLK in the frame > 20 x N), the host controller must appropriately align the configuration data for each device before bringing CS high. A shorter data transfer frame (number of SCLK in the frame < 20 x N) can result in an erroneous device configuration, and must be avoided.

7.5.4.4 Multiple Devices: Star Topology

A typical connection diagram showing multiple devices in the star topology is shown in Figure 86. The CONVST, SDI, and SCLK inputs of all devices are connected together and are controlled by a single CONVST, SDO, and SCLK pin of the host controller, respectively. Similarly, the SDO output pin of all devices are tied together and connected to the a single SDI input pin of the host controller. The CS input pin of each device is individually controlled by separate CS control lines from the host controller.

Figure 86. Star Topology Connection

The timing diagram for N devices connected in the star topology is shown in Figure 87. In order to avoid any conflict related to multiple devices driving the SDO line at the same time, ensure that the host controller pulls down the CS signal for only one device at any particular time.

Figure 87. Three Devices Connected in Star Connection Timing Diagram

7.6.1 Device Configuration and Register Maps

The device features four configuration registers, mapped as described in Table 8.

8.1.1 ADC Input Driver

The input driver circuit for a high-precision ADC mainly consists of two parts: a driving amplifier and a fly-wheel RC filter. The amplifier is used for signal conditioning of the input signal and its low output impedance provides a buffer between the signal source and the switched capacitor inputs of the ADC. The RC filter helps attenuate the sampling charge injection from the switched-capacitor input stage of the ADC and functions as an antialiasing filter to band-limit the wideband noise contributed by the front-end circuit. Careful design of the front-end circuit is critical to meet the linearity and noise performance of the ADS9120.

8.1.2 Input Amplifier Selection

Selection criteria for the input amplifiers is highly dependent on the input signal type as well as the performance goals of the data acquisition system. Some key amplifier specifications to consider when selecting an appropriate amplifier to drive the inputs of the ADC are:

• Small-signal bandwidth. Select the small-signal bandwidth of the input amplifiers to be as high as possible after meeting the power budget of the system. Higher bandwidth reduces the closed-loop output impedance of the amplifier, thus allowing the amplifier to more easily drive the low cutoff frequency RC filter (see the Antialiasing Filter section) at the inputs of the ADC. Higher bandwidth also minimizes the harmonic distortion at higher input frequencies. In order to maintain the overall stability of the input driver circuit, select the amplifier with Unity Gain Bandwidth (UGB) as described in Equation 14:
Equation 14.
• Noise. Noise contribution of the front-end amplifiers must be as low as possible to prevent any degradation in SNR performance of the system. Generally, to ensure that the noise performance of the data acquisition system is not limited by the front-end circuit, the total noise contribution from the front-end circuit must be kept below 20% of the input-referred noise of the ADC. Noise from the input driver circuit is band-limited by designing a low cutoff frequency RC filter, as explained in Equation 15.
Equation 15.
where
• V_{1 / f_AMP_PP} is the peak-to-peak flicker noise in µV,
• e_{n_RMS} is the amplifier broadband noise density in nV/√Hz,
• f_–3dB is the 3-dB bandwidth of the RC filter, and
• N_G is the noise gain of the front-end circuit that is equal to 1 in a buffer configuration.
• Distortion. Both the ADC and the input driver introduce distortion in a data acquisition block. To ensure that the distortion performance of the data acquisition system is not limited by the front-end circuit, the distortion of the input driver must be at least 10 dB lower than the distortion of the ADC, as shown in Equation 16.
Equation 16.
• Settling Time. For dc signals with fast transients that are common in a multiplexed application, the input signal must settle within an 16-bit accuracy at the device inputs during the acquisition time window. This condition is critical to maintain the overall linearity performance of the ADC. Typically, the amplifier data sheets specify the output settling performance only up to 0.1% to 0.001%, which may not be sufficient for the desired 16-bit accuracy. Therefore, always verify the settling behavior of the input driver by TINA™-SPICE simulations before selecting the amplifier.

8.1.3 Antialiasing Filter

Converting analog-to-digital signals requires sampling an input signal at a constant rate. Any higher-frequency content in the input signal beyond half the sampling frequency is digitized and folded back into the low-frequency spectrum. This process is called aliasing. Therefore, an analog, antialiasing filter must be used to remove the harmonic content from the input signal before being sampled by the ADC. An antialiasing filter is designed as a low-pass, RC filter, where the 3-dB bandwidth is optimized based on specific application requirements. For dc signals with fast transients (including multiplexed input signals), a high-bandwidth filter is designed to allow accurately settling the signal at the inputs of the ADC during the small acquisition time window. For ac signals, keep the filter bandwidth low to band-limit the noise fed into the input of the ADC, thereby increasing the signal-to-noise ratio (SNR) of the system.

Besides filtering the noise from the front-end drive circuitry, the RC filter also helps attenuate the sampling charge injection from the switched-capacitor input stage of the ADC. A filter capacitor, C_FLT, is connected from each input pin of the ADC to the ground (as shown in Figure 92). This capacitor helps reduce the sampling charge injection and provides a charge bucket to quickly charge the internal sample-and-hold capacitors during the acquisition process. Generally, the value of this capacitor must be at least 15 times the specified value of the ADC sampling capacitance. For the ADS9120, the input sampling capacitance is equal to 60 pF, thus it is recommended to keep C_FLT greater than 900 pF. The capacitor must be a COG- or NPO-type because these capacitor types have a high-Q, low-temperature coefficient, and stable electrical characteristics under varying voltages, frequency, and time.

Figure 92. Antialiasing Filter Configuration

Note that driving capacitive loads can degrade the phase margin of the input amplifiers, thus making the amplifier marginally unstable. To avoid amplifier stability issues, series isolation resistors (R_FLT) are used at the output of the amplifiers. A higher value of R_FLT is helpful from the amplifier stability perspective, but adds distortion as a result of interactions with the nonlinear input impedance of the ADC. Distortion increases with source impedance, input signal frequency, and input signal amplitude. Therefore, the selection of R_FLT requires balancing the stability and distortion of the design. For the ADS9120, limiting the value of R_FLT to a maximum of 10-Ω is recommended in order to avoid any significant degradation in linearity performance. The tolerance of the selected resistors must be kept less than 1% to keep the inputs balanced.

The driver amplifier must be selected such that its closed-loop output impedance is at least 5X less than the R_FLT.

8.1.4 ADC Reference Driver

The external reference source to the ADS9120 must provide low-drift and very accurate voltage for the ADC reference input and support the dynamic charge requirements without affecting the noise and linearity performance of the device. The output broadband noise of most references can be in the order of a few hundred μV_RMS. Therefore, to prevent any degradation in the noise performance of the ADC, the output of the voltage reference must be appropriately filtered by using a low-pass filter with a cutoff frequency of a few hundred hertz.

After band-limiting the noise of the reference circuit, the next important step is to design a reference buffer that can drive the dynamic load posed by the reference input of the ADC. The reference buffer must regulate the voltage at the reference pin such that the value of V_REF stays within the 1-LSB error at the start of each conversion. This condition necessitates the use of a large capacitor, CBUF_FLT (see Figure 38), between each pair of REFP and REFM pins for regulating the voltage at the reference input of the ADC. The effective capacitance of any large capacitor reduces with the applied voltage based on the voltage rating and type. Using X7R-type capacitors is strongly recommended.

The amplifier selected as the reference driver must have an extremely low offset and temperature drift with a low output impedance to drive the capacitor at the ADC reference pins without any stability issues.

8.2.1 Design Requirements

Design an application circuit optimized for using the ADS9120 to achieve:

• > 95-dB SNR, < –118-dB THD
• ±0.5-LSB linearity and
• Maximum-specified throughput of 2.5 MSPS

8.2.2 Detailed Design Procedure

The application circuits are illustrated in Figure 93. For simplicity, power-supply decoupling capacitors are not shown in these circuit diagrams; see the Power-Supply Recommendations section for suggested guidelines.

The input signal is processed through the OPA625 (a high-bandwidth, low-distortion, high-precision amplifier in an inverting gain configuration) and a low-pass RC filter before being fed into the ADC. Generally, the distortion from the input driver must be at least 10 dB less than the ADC distortion. The distortion resulting from variation in the common-mode signal is eliminated by using the OPA625 in an inverting gain configuration. The low-power OPA625 as an input driver provides exceptional ac performance because of its extremely low-distortion and high-bandwidth specifications. To exercise the complete dynamic range of the ADS9120, the common-mode voltage at the ADS9120 inputs is established at a value of 2.25 V (4.5 V / 2) by using the noninverting pins of the OPA625 amplifiers.

In addition, the components of the antialiasing filter are such that the noise from the front-end circuit is kept low without adding distortion to the input signal.

The reference driver circuit, illustrated in Figure 93, generates a voltage of 4.5 V_DC using a single 5-V supply. This circuit is suitable to drive the reference of the ADS9120 at higher sampling rates up to 2.5 MSPS. The reference voltage of 4.5 V in this design is generated by the high-precision, low-noise REF5045 circuit. The output broadband noise of the reference is heavily filtered by a low-pass filter with a 3-dB cutoff frequency of 160 Hz.

The reference buffer is designed with the OPA625 and OPA378 in a composite architecture to achieve superior dc and ac performance at a reduced power consumption, compared to using a single high-performance amplifier. The OPA625 is a high-bandwidth amplifier with a very low open-loop output impedance of 1 Ω up to a frequency of 1 MHz. The low open-loop output impedance makes the OPA625 a good choice for driving a high capacitive load to regulate the voltage at the reference input of the ADC. The relatively higher offset and drift specifications of the OPA625 are corrected by using a dc-correcting amplifier (the OPA378) inside the feedback loop. The composite scheme inherits the extremely low offset and temperature drift specifications of the OPA378.

8.2.3 Application Curves

fIN = 2 kHz, SNR = 95.5 dB, THD = –118 dB

Figure 94. FFT with a 2-kHz Input Signal

Typical INL of ±0.25 LSB

Figure 95. Typical INL

9.2.1 NAP Mode

In NAP mode, some of the internal blocks of the device power down to reduce power consumption in the ACQ state.

To enable NAP mode, set the NAP_EN bit in the PD_CNTL register. To exercise NAP mode, keep the CONVST pin high at the end of conversion process. The device then enters NAP mode at the end of conversion and continues in NAP mode until the CONVST pin is held high.

A CONVST falling edge brings the device out of NAP mode; however, the host controller can initiate a new conversion (CONVST rising edge) only after the t_{nap_wkup} time has elapsed.

Figure 98 shows a typical conversion cycle with NAP mode enabled (NAP_EN = 1b).

Figure 98. NAP Enabled Conversion Cycle

The cycle time is given by Equation 17.

Equation 17.

At lower throughputs, cycle time (t_cycle) increases but the conversion time (t_conv) remains constant, and therefore the device spends more time in NAP mode, thus giving power scaling with throughput as shown in Figure 99.

Figure 99. Power Scaling with Throughput with NAP Mode

9.2.2 PD Mode

The device also features a deep power-down mode (PD) to reduce the power consumption at very low throughput rates.

To enter PD mode:

1. Write 069h to address 011h to unlock the PD_CNTL register.
2. Set the PDWN bit in the PD_CNTL register. The device enters PD mode on the CS rising edge.

In PD mode, all analog blocks within the device are powered down. All register contents are retained and the interface remains active.

To exit PD mode:

1. Reset the PDWN bit in the PD_CNTL register.
2. The RVS pin goes high, indicating that the device has processed the command and has started coming out of PD mode. However, the host controller must wait for the t_PWRUP time to elapse before initiating a new conversion.

10.1.1 Signal Path

As illustrated in Figure 100, the analog input and reference signals are routed in opposite directions to the digital connections. This arrangement prevents noise generated by digital switching activity from coupling to sensitive analog signals.

10.1.2 Grounding and PCB Stack-Up

Low inductance grounding is critical for achieving optimum performance. Grounding inductance is kept below 1 nH with 15-mil grounding vias and a printed circuit board (PCB) layout design that has at least four layers. Place all critical components of the signal chain on the top layer with a solid analog ground from subsequent inner layers to minimize via length to ground.

Pins 11 and 15 of the ADS9120 can be easily grounded with very low inductance by placing at least four 8-mil grounding vias at the ADS9120 thermal pad. Afterwards, pins 11 and 15 can be connected directly to the grounded thermal path.

10.1.3 Decoupling of Power Supplies

Place the AVDD and DVDD supply decoupling capacitors within 20 mil from the supply pins and use a 15-mil via to ground from each capacitor. Avoid placing vias between any supply pin and its decoupling capacitor.

10.1.4 Reference Decoupling

Dynamic currents are also present at the REFP and REFM pins during the conversion phase and excellent decoupling is required to achieve optimum performance. Three 10-μF, X7R-grade, ceramic capacitors with 10-V rating are recommended, placed as illustrated in Figure 100. Select 0603- or 0805-size capacitors to keep ESL low. The REFM pin of each pair must be connected to the decoupling capacitor before a ground via.

10.1.5 Differential Input Decoupling

Dynamic currents are also present at the differential analog inputs of the ADS9120. C0G- or NPO-type capacitors are required to decouple these inputs because their capacitance stays almost constant over the full input voltage range. Lower quality capacitors (such as X5R and X7R) have large capacitance changes over the full input voltage range that can cause degradation in the performance of the ADS9120.

11.1.1 相关文档　

相关文档如下：

• 《ADS9120EVM-PDK 用户指南》（文献编号：SBAU262）
• 《可实现最大 SNR 和采样率的 18 位、2MSPS 隔离式数据采集参考设计》（文献编号：TIDUB85）
• 《电压基准对总谐波失真的影响》（文献编号：SLYY097)
• 《REF60xx 集成 ADC 驱动器缓冲器的高精度电压基准》（文献编号：SBOS708）
• 《OPAx625 高带宽、高精度、低 THD+N、16 位和 18 位 ADC 驱动器》（文献编号：SBOS688）
• 《THS4551 低噪声、150MHz、精密全差分放大器》（文献编号：SBOS778）
• 《REF50xx 低噪声、极低漂移、高精度电压基准》（文献编号：SBOS410）
• 《OPAx378 低噪声、900kHz、RRIO、精密运算放大器零漂移系列》（文献编号：SBOS417）