6

Related to: ESD protection of ADC input.

To protect an ADC input from out-of-spec voltages, including ESD, it is often recommended to use diodes connected to GND and Vcc. If I have a 1k series resistor on the ADC (ADS1158) input, is it OK to rely on the ADC's internal diodes, or is it worth adding the diodes as discrete components.

ADC input protection

Community
  • 1
Rocketmagnet
  • 26,933
  • 17
  • 92
  • 177

2 Answers2

6

The data sheet should cover this. Usually, if there are internal diodes specified they will be adequate for protection.

Note that diodes, internal or external, are intended to prevent IC destruction BUT not to prevent malfunction in normal operation if they conduct. Voltages should always be such during measurement that essentially no diode protection current flows. Keeping Vin inside the supply rails achieves this. Slightly outside MAY be OK, but with ADC especially, no guarantees.

Russell McMahon
  • 147,325
  • 18
  • 210
  • 386
  • Thanks Russell. The datasheet doesn't mention how much current the diodes are rated for. I just want to prevent damage to the device in case someone does something stupid. In normal operation the inputs won't go beyond the rails. – Rocketmagnet Sep 19 '12 at 11:22
  • In particular, the stray currents from the protection diodes can upset the analog parts of the circuit. In a multiple-input ADC like ADS1158, conducting the protection diodes in one channel can cause erroneous results in the others. – markrages Sep 19 '12 at 18:05
  • Why wouldn't external diodes prevent malfunction of the IC? Doesn't the extra-voltage go from the source to the supply rails? – m.Alin Sep 19 '12 at 22:50
  • @m.Alin - external diodes will usually prevent IC destruction. BUT they will not prevent malfunction in operation with any certainty. If Vin is outside supply rails functionality can be affected - ADC functions especially so. Problems occur because current that enters the substrate can reenter the circuitry at nodes that are usually not driven and most anything can happen - and sometimes does. A hidden node that is usually never driven may have an impedance of say 1 megohm to supply or ground, so that few microamps of substrate sourced current MAY turn on or off a phantom FET and wreak havoc. – Russell McMahon Sep 20 '12 at 03:10
5

Additional external diodes are an opportunity for extra noise and loading. So you might not want to put them in for that reason. While I agree with what Russell says NOT all ESD circuits necessarily work (even though they are represented by diodes) as indicated. Some of the best chip esd solutions run dv/dt clamps on the rails. If you slowly increase the rails you can easily over volt (EOS) gates etc. These clamps are great because they give less capacitive loading on pins, especially for sensitive inputs.

Of course the manf. won't necessarily tell you tell you what they are doing. You can test this by taking a part and looking at the I vs. V curves on a curve tracer to see if there actually are diodes in there.

placeholder
  • 29,982
  • 10
  • 63
  • 110
  • Very interesting. I will try this then. I'll slowly ramp up the voltage to higher than AVDD, and see if the pin clamps to AVDD. – Rocketmagnet Sep 20 '12 at 08:39
  • If you have a current limiter on the PS that will help or put a ammeter in line. Don't run it more than ~0.8 V hotter than the rail. OR you can do it unpowered with the Vdd tied to ground too. THat way you can reuse the part because you're not risking EOS (Electrical Overstress) – placeholder Sep 20 '12 at 16:45
  • But won't the 1k resistor be enough to limit the current? – Rocketmagnet Sep 20 '12 at 17:09
  • I mean use the current limiter to read the current. I'm thinking that if you lok at a diode curve you can limit the potential damage of EOS. But driving it hard with 1K and seeing clamping should work. – placeholder Sep 20 '12 at 17:28