0
\$\begingroup\$

I built the following circuit featuring the HI7190, 24-bit ADC, straight off the datasheet: HI7190 test circuit

Since there is no available negative supply I built one using ICL7660, as AN9601 recommends. For VRef, as a test, I used a TL431. The actual measured VRef is 2.495V.

  • If I short Vin pins, the ADC gives a reading of around -10 to +10, which is to be expected because of the low quality reference and circuit construction. (This is still an impressive 6uVpp noise floor).

  • If I put a slightly discharged AAA battery, about 1.4V, between the input pins, I get a reading of about 3 500 000.

  • If i turn the battery around, the reading is about -3 500 000.

In both cases, Vinhi and Vinlo are floating. The battery terminals are connected to Vinhi and Vinlo and nothing else.

The problem I have is that the ADC reading does not match the actual values between Vinhi and Vinlo. The ADC has a voltage reference of (nominal) 2.5V. The readings of 3,500,000 would mean: (3500000 * 2 * 2.5)/(2^24) = 1.0431V. This is far from the 1.4V that are at the inputs. It's about 1.345x lower.

I have noticed this error is constant across all the range. I tried several voltages at the input and they are all off by the same factor.

I have measured the reference voltage, supply voltages, ground. It's all OK. I have even replaced the IC and the problem persists. The ADC has calibration registers. They are a all "auto calibrated" at power up. I have also overriden them by overwriting them like the datasheet suggests, but nothing seems to change.

Is there something I'm missing?

\$\endgroup\$
  • \$\begingroup\$ Try tying one side of the battery to ground. The input instrumentation amplifier may not like floating inputs. \$\endgroup\$ – John Birckhead Jan 4 '18 at 13:24
  • \$\begingroup\$ @JohnBirckhead how about the other input? You mean reference Vinlo and bat- to GND, and bat+ to vinhi? \$\endgroup\$ – hjf Jan 4 '18 at 13:27
  • \$\begingroup\$ As @John says each input has to be within the range of +/-3.2V with +/-5V supplies (probably a bit less headroom on the negative side in your case because of the unregulated charge pump). BTW, the reference should not affect the noise floor noticeably with inputs shorted. \$\endgroup\$ – Spehro Pefhany Jan 4 '18 at 13:41
  • 1
    \$\begingroup\$ Just try tying one side of the battery to VCM. It should not matter which. (Battery terminals still connected to the inputs) \$\endgroup\$ – John Birckhead Jan 4 '18 at 13:43
  • 2
    \$\begingroup\$ You're missing a link to the datasheet, which I added for you. Unfortunately, it does not discuss the details of the analog inputs, other than to say that the maximum leakage current is 1 nA. However, it's doubtful that they can operate without any return path for input leakage/bias currents, so tie at least one of them to Vcm, possibly through a large-value resistor (1M - 10M). \$\endgroup\$ – Dave Tweed Jan 4 '18 at 14:58

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.