“Chips will continue to swallow board functions” 
“Size, complexity, development time, and cost requirements are making individual components an expensive luxury”
As chip manufacturers continue to get better and better with integrating functionality on System on a Chip (SOC) devices, engineers are saying things like those listed above. It is true that the industry of making Integrated Circuits (ICs) is becoming more and more advanced every day. Some of the smartest people are behind the blazing rate of technology growth in that area, and given the ever-increasing market for portable devices it is no wonder IC manufacturers are investing in that direction.
A friend recently pointed me in the direction of 'Analog Front End' (AFE) chips that have become almost an entire stand alone circuit capable of interfacing directly with the physical world. With my love of analog electronics and experience developing precision measurement equipment, I was excited to look closely at the new technology advances.
The holy grail for chip manufacturers is to allow an engineer to design in the AFE IC, adding only user inputs/outputs and a microcontroller/FPGA, to complete the hardware development. It's certainly a tantalizing thought given the risk, time, and expense involved when designing analog systems from discrete components or even with analog ICs like op amps. But are we there yet?
To answer how close they are, I've read the specs of the AFE chips and compared them with a couple complete products that can be bought today. The AFE is Analog Devices' AD 5520, a Per Pin Parametric Measurement Unit/Source Measure Unit. This is very similar to Keithley's entry-level 2401 SourceMeter Instrument and the module-based National Instruments PXIe-4140 Source Measure module.
The datasheets for all of these devices are pages long, so I'll boil them down to the specs which most users will find important. Please consider these figures to be approximate as some massaging was required to give comparable figures.
| Specification | AD5520 | KI 2401 | NI 4140 |
| General Capabilities | |||
| Max Voltage Output | +/- 11V | +/- 20V | +/- 10V |
| Max Current Output | +/- 4mA | +/- 1A | +/- 100mA |
| Max Power Output | 44mW | 20W | 1W |
| Voltage Source | |||
| Gain Error | 1%, max1 | 0.17%, max1 | +/- 0.125% max1 |
| Offset Error | +/- 1mV | +/- 0.6mV | +/- 5mV |
| Current Source | |||
| Gain Error | 1%, max1 | 0.29%, max1 | 0.2%, max1 |
| Offset Error | <not spec'd> | +/- 0.6nA | +/- 5nA |
| Voltage Measurement | |||
| Gain Error | +/- 0.03% | +/- 0.012% | +/- 0.1% |
| Gain Error Tempco | 2 ppm/°C | 18 ppm/°C | 5 ppm/°C |
| Input Offset Error | +/- 5mV | +/- 0.3mV | +/- 5mV |
| Input Offset Tempco | +/- 15uV/°C | +/- 45uV/°C | +/- 1uV/°C |
| CMRR (Noise) | 73dB | 100dB | <not spec'd> |
| Current Measurement | |||
| Gain Error | 0.10% | 0.03% | 0.10% |
| Gain Error Tempco | 30 ppm/°C | 40 ppm/°C | 20 ppm/°C |
| Input Bias Current | +/- 1nA | +/- 60nA | +/- 500nA |
| Input Bias Current Tempco | +/- 50 pA/°C | +/- 9nA/°C | +/- 1nA/°C |
| CMRR (Noise) | 95dB | 100dB | <not spec'd> |
| Retail Price (in single quantities) | $33.46 | $3,120.00 | $4,699.00 |
Footnotes:
1. 'Max' designates spec over 0-70°C temperature range, compared to the stndard 25*C fixed.
Clearly these three are totally different products far from being interchangeable, especially since the IC will require additional components adding errors and cost while the completed units are ready to go. And after reading through the AD5520 spec sheet, there is still a large amount of analog development that is required to properly use the chip. So no, analog design isn't dead. But it's impossible to ignore the incredible progress and integration on the chip level, and easy to imagine the future possibilities.
Has anyone used one of these AFE-type chips? Are they as easy and as accurate as the documentation suggests? Are there any unexpected surprises that are found in development? Let us know in the comments!
