MCP604 Quad R-R opamp not behaving

[strobot]

Having a problem with this circuit

It uses an MCP604 opamp http://ww1.microchip.com/downloads/en/DeviceDoc/21314g.pdf

It is intended to be used with http://www.vishay.com/docs/29049/ntcle100.pdf 100K thermistors.

When all the thermistors are disconnected and Vcc = 3.28V

The outputs:
Pins 1, 7, 14 on the MCP604 = 2.57V  ... I don't understand this. 
Pin 8 = 3.27V Seems to be the only one that is normal.

The inputs:
Pins 3, 5, 10, 12 = 3.24V

The pulled-up thermistor connections all measure 3.27V.

With 4 10k 1% resistors connected to the inputs, the circuit is outputting:
0.987V, 0.983V, 0.987V, 0.984V (pins 1, 7 , 8, 14), as would be expected.

[SeanB]

What capacitors are you using for C2-C5? They could be leaky, or have surface contamination causing this.  Most likely, as you do have one good amp in the package that surface leakage is the cause of this, either across the capacitor or internal to it.

As well these are rail to rail output devices, but the inputs are 0V to ( in your case ) 2.08V range. The CM is Vs-1.2V, so that is probably causing it. The one going rail to rail is probably a better than guaranteed unit.

[oPossum]

From the spec sheet:

• Rail-to-Rail Output
• Input Range Includes Ground
Common Mode Input Range VCMR:  VSS – 0.3 to VDD – 1.2 V

So... You can't have inputs higher than Vdd -1.2V and expect normal operation.

[strobot]

Oh, man.  The common-mode input range is definitely the problem.  I wish I had read this!
Are you violating your op amp’s input common-mode range?
http://www.planetanalog.com/document.asp?doc_id=528175&site=planetanalog

I basically copied the thermistor circuit from this Microchip app-note http://ww1.microchip.com/downloads/en/AppNotes/00929a.pdf on page 15.
And didn't look deeper at the op-amps specs.  Why did they recommend an opamp with such a truncated common mode input range?

I'm going to try the LMV614 http://www.ti.com/lit/ds/symlink/lmv614.pdf
in place of the MCP604 since its available in the same SOIC-14 package.

EDIT: the capacitors C2, C3, C4, C5 are Murata GRM21BR71C105JA01L 0805 1uF 16V MLCCs
Interesting that you mention surface contamination.  I was having issue with flux residue affecting the resistance across the reference 23.2K 0.1% resistors and used an ultrasonic bath to try and get the boards as clean as possible.

[SeanB]

Surface contamination or cracked caps is indicated by the varying readsings between channels. If all the resistors are the same nominal value and the opamp is matched ( likely as they are on a common die) then only leakage and internal leakage in the cals will cause the 5mV variation you have. Short out the 100k resistor and try again, or replace the caps with film units to test.

[P_Doped]

"Why did they recommend an opamp with such a truncated common mode input range?"

I think that's a fair question to ask Microchip.  Their App Note shows a possible output of up to 5.0V with this configuration.  The supply voltage is only spec'ed to Operational Max of 6.0V with their specification table only holding up to 5.5V.  The maximum Vcm in the Electrical Specifications table is VDD-1.2V @25C, and I think that number will get worse with increasing temperature.

For this circuit, the highest output (and thus input) is needed for cold temperatures, where the common mode loss is not as bad.  With 6.0V supply, and possibly 5.5V, it might have actually worked.  But, for it to work, you really need a high supply.

I would ask Microchip what supply voltage is required for the opamp, since it's not specified in the App Note.  They probably will take the easy road out and say "maximum", but at least you'll make them think a bit.

[Baliszoft]

Try 6004 instead.

[strobot]

Surface contamination or cracked caps is indicated by the varying readsings between channels. If all the resistors are the same nominal value and the opamp is matched ( likely as they are on a common die) then only leakage and internal leakage in the cals will cause the 5mV variation you have. Short out the 100k resistor and try again, or replace the caps with film units to test.

Are you talking about the output voltages I reported with the 10k 1% resistors attached?  The output voltage variation is within the expected voltages given the 1% tolerancing of the resistors.

[SeanB]

Check all 4 using the same resistor, if you get the same reading on all 4 then there is no leakage. If it varies then you have leakage.