AD7689BCPZRL7
Product Specifications
AD7689BCPZRL7
Analog Devices — Analog & Mixed-Signal ICs
Product Overview
The AD7689BCPZRL7 from Analog Devices is a high-performance analog / mixed-signal IC providing precision signal acquisition, conversion, and conditioning. Analog Devices (ADI) is a world leader in high-performance analog, mixed-signal, and DSP ICs for instrumentation, healthcare, automotive, and communications.
Superb Automation partners with leading brands including Qorvo, TI, ADI, XILINX, NVIDIA, Intel, ST, MACOM, Renesas, and Coilcraft to provide complete BOM and supply-chain solutions.
Key Features
- High-resolution data conversion with excellent SNR and THD
- Wide dynamic range and high linearity across operating conditions
- Integrated voltage reference with low temperature drift
- Flexible digital interface — SPI, I2C, or parallel options
- Industrial/automotive temperature range with guaranteed specs
- Low power consumption with power-down modes
Key Specifications
| Resolution | bits |
| Channels | — |
| Sample Rate | kSPS |
| Interface | — |
| Supply Voltage | V |
| Operating Temp | °C |
| Package | — |
| INL | LSB |
| SNR | dB |
| THD | dB |
Applications
AD7689BCPZRL7: eight channels at 31 kilo samples each
The AD7689 is a 16 bit successive approximation converter with an eight channel multiplexer, a selectable internal reference, a temperature sensor, a channel sequencer and a one pole filter, all in a 4 by 4 millimetre package running from 2.3 V to 5.5 V. It converts at 250 kSPS and has no pipeline delay, which means the result of a conversion is available before the next one starts, and it draws 12.5 mW at 5 V while doing it. The typical job is a multichannel monitoring board where eight different sensors, supplies or thermistors have to be read in sequence and reported over an isolated serial link.
What the vendor page states
| Parameter | Value |
|---|---|
| Resolution | 16 bit, no missing codes |
| Channels | eight, single ended, differential with ground sense, or pseudobipolar |
| Throughput | 250 kSPS |
| Integral nonlinearity | 0.4 LSB typical, 1.5 LSB maximum |
| Dynamic range | 93.8 dB |
| Signal to noise and distortion | 92.5 dB at 20 kHz |
| Total harmonic distortion | minus 100 dB at 20 kHz |
| Reference | internal 2.5 V or 4.096 V selectable, external buffered up to 4.096 V, external up to the supply |
| Supply and logic | 2.3 V to 5.5 V analog, 1.8 V to 5.5 V logic interface |
| Power | 12.5 mW at 5 V and 250 kSPS, 3.5 mW at 2.5 V and 200 kSPS, 50 nA standby |
| Package and grade | 20 lead LFCSP 4 by 4 mm, minus 40 to plus 85 C, extended model to plus 125 C |
Throughput divided by eight
250 kSPS is an aggregate figure. Scanning all eight channels gives 31.25 kSPS each, and one conversion slot is 4 microseconds, so each channel gets 0.5 microseconds of the cycle. That is the number that decides whether the multiplexer can be used at full rate: the source driving each channel has to settle inside its share of the cycle, and the internal one pole filter helps but does not replace a buffer. The channel sequencer exists precisely so the host does not have to write a register between conversions.
Fifteen bits, not sixteen
A signal to noise and distortion figure of 92.5 dB is 15.07 effective bits by the usual arithmetic, and the 93.8 dB dynamic range is 15.29. The converter is 16 bit and monotonic, but the analog performance delivers about 15 bits, which is what an application should plan for. One count at a 4.096 V reference is 62.5 microvolts, and the 1.5 LSB maximum nonlinearity is 93.8 microvolts, or 22.9 ppm of full scale. Read the other way, a 2.5 V reference gives a 38.15 microvolt count and a finer resolution at the cost of 1.6 V of input range.
Energy per conversion instead of milliwatts
12.5 mW at 250 kSPS is 50 nanojoules per conversion, and 3.5 mW at 200 kSPS is 17.5 nanojoules, because the dissipation scales with throughput rather than sitting at a fixed value. A battery instrument sampling eight channels once a second spends almost nothing, and the 50 nA standby current at 5 V is 0.25 microwatts while it waits. The real power cost in that application is the reference and the buffer, not the converter.
Before the order goes out
The logic interface runs from 1.8 V and is separate from the analog supply, which is what allows the part to sit on a 5 V analog rail and talk to a 1.8 V processor without a level translator. The converter cross reference lists the pin compatible members of this family by channel count, the pin to pin cross reference covers the packages that share this footprint, and the analog and mixed signal category page holds the rest of the converter range.
AD7689BCPZRL7 on a 3.3 V rail: the reference choice sets the accuracy
The data sheet gives three reference paths for this part: the internal 2.5 V reference, the internal 4.096 V reference, and an external reference anywhere between 0.5 V and VDD. The supply floor moves with the choice. The internal 2.5 V setting needs VDD at 3.0 V or above, the internal 4.096 V setting needs 4.5 V or above, so a 3.3 V rail has one internal option and a 2.5 V rail has none of them. A board specified at 2.5 V runs from an external reference and inherits that part's initial tolerance, its drift and its turn-on behaviour, and a reference that is hard to find at short notice turns a converter order into a sourcing question, which is where our shortage recovery route usually starts.
What the internal reference costs in code
The 25 C limits in the reference table are 2.490 V to 2.510 V for the 2.5 V setting and 4.086 V to 4.106 V for the 4.096 V setting. On a 2.5 V full scale one 16-bit LSB is 38.1 uV, so the plus or minus 10 mV tolerance is about 262 LSB of full-scale error before a single conversion is taken. On the 4.096 V setting the LSB is 62.5 uV and the same 10 mV is 160 LSB. Both are arithmetic on the quoted limits, and neither is removed by the converter's linearity: the internal reference is a starting point for a ratiometric measurement, not a calibrated absolute one. The REF pin takes 10 uF of decoupling and the REFIN pin 0.1 uF, both close to the package, because one reference serves every channel in the sequence.
SINAD follows the reference. The data sheet quotes 92.5 dB at 20 kHz with a 5 V reference, 91 dB with the internal 4.096 V option and 88.4 dB with the internal 2.5 V option. In effective bits, for the same 20 kHz input, those are about 15.1, 14.8 and 14.4, so the loudest option and the smallest-scale option sit seven tenths of a bit apart.
1.7 MHz of front end behind a 125 kHz Nyquist limit
At 250 kSPS the Nyquist limit is 125 kHz, while the specified small signal bandwidth is 1.7 MHz in full bandwidth mode and 0.425 MHz with the quarter bandwidth setting in the configuration register. Anything between 125 kHz and 1.7 MHz folds into the result, so the network in front of INx is part of the accuracy budget rather than an optional extra. Aperture delay of 2.5 ns is negligible against a 4 us conversion period, so sample timing is not where a multi channel scan loses coherence.
Channel to channel leakage against one LSB
Crosstalk from an adjacent channel driven at 100 kHz is minus 125 dB typical and minus 120 dB minimum. On a 5 V full scale that is 2.8 uV and 5.0 uV, against the 76.3 uV one 16-bit LSB occupies on the same range, so the multiplexer is not the limit in an eight channel scan. What does limit the scan is the reference and the input network, and neither is channel specific. Conversion results are available immediately after the conversion with no pipeline delay, which is what allows a loop to close around the reading instead of around a delayed copy of it.
Ordering this code means ordering the whole reference decision with it. The related order codes in the same family are listed on the analog and mixed signal page, and the converter cross reference shows which neighbouring parts share the package and the pinout.
AD7689BCPZRL7: eight channels in 32 microseconds, and the reference that has to hold between them
The converter is a single successive approximation core behind an eight channel multiplexer, so the scan time, the per channel rate and the reference settling are the same question asked three ways. A buyer who reads the 250 kSPS headline and plans an eight channel acquisition at that rate has planned for 31 kSPS per channel, and the rest of the block is about what that means for the interface, the input network and the reference.
Two hundred and fifty thousand samples per second is four microseconds, shared eight ways
At the full rate one conversion takes 4 microseconds, so a complete eight channel scan takes 32 microseconds and each channel is updated 31,250 times a second. That is the number that belongs in the sampling specification, not the converter rate. It also sets the multiplexer settling requirement: the input network ahead of each channel has to charge the sample capacitor to the required accuracy inside the acquisition window that remains after the conversion, and at 4 microseconds per channel the window is short enough that a 10 kohm source with a 1 nF filter is already too slow. A 1 kohm source and a 100 pF filter gives a 100 ns time constant and settles to 16 bits in about 1.2 microseconds, which fits.
The serial interface is not the bottleneck: sixteen bits at 50 MHz is 0.32 microseconds
Reading sixteen bits over a 50 MHz serial clock takes 0.32 microseconds, which is a twelfth of the conversion period, so the interface is never the reason a scan runs slow. What does cost time is the configuration word and the channel change, both of which happen inside the same period. In a daisy chain the read of one device overlaps the conversion of the next, so the scan of eight channels on one die and the scan of eight dies in a chain have different timing budgets even though each die converts in 4 microseconds. The aperture delay of 2.5 ns is 0.06 percent of the conversion period, so sample timing is not where a multi channel scan loses coherence between channels.
Anything between 125 kHz and 1.7 MHz folds into the result
At 250 kSPS the Nyquist limit is 125 kHz, while the specified small signal bandwidth is 1.7 MHz in full bandwidth mode, which is 13.6 times the Nyquist frequency. A 400 kHz interferer therefore appears at 100 kHz in the result and cannot be removed by any digital filter afterwards. The quarter bandwidth setting narrows the front end to 425 kHz and cuts the window by a factor of four, at the cost of settling time; the other answer is an analog filter ahead of the multiplexer, sized for the slowest channel in the scan rather than the fastest. This is the reason a part with a 1.7 MHz front end is not automatically better than a slower one in a scanning application.
| Channels in the scan | Scan time | Per channel rate | Where it fits |
|---|---|---|---|
| 1 | 4 microseconds | 250 kSPS | Single channel fast loop |
| 4 | 16 microseconds | 62.5 kSPS | Vibration or power monitoring |
| 8 | 32 microseconds | 31.25 kSPS | Full scan, still fast enough for audio band |
The family and its interface options are compared on ADC DAC cross reference, the front end that drives it is covered on amplifier cross reference, and the reference and passives around it are quoted through one stop BOM sourcing.
Request Quote — In Stock, 100% New & Original, Full Warranty