# AD623-Unable to change gain or reference voltage

I have an air flow sensor (FS1012-1100NG) with expected differential output voltages ranging from -1mV to 20mV and I need to 1) know the precise value (mV) and gain to correlate V with airflow 2) make sure that value has long-term stability (order of days) and 3) end up with a voltage that makes the best use of my 12bit ADC.

I'm using an AD623 with a single supply (3.3V) with the FS1012 connected per the datasheet (see image).

• pins 1&8 Gain Resistor (I've tried 100ohm-1Mohm but no change in output/gain)
• pins 2&3 TP2+ and TP1+ (I've checked these and there is a differential voltage of about 2-3mV at max airflow)
• pins 4&7 Ground and 3.3V
• pin 5 Ref (Tested 0-3V using voltage divider, no change in output/offset)
• pin 6 Output (to digital multi-meter with ref to ground)

What am I doing wrong? Is there a better differential amplifier I should be using? Thank you in advance!

• Any INA with programmable gain shud work. BUT Vin MUST be within Vcm range near V+/2 so you do this with Rdivider pullup network. Aug 22, 2020 at 0:29
• Use an R network pack for best CMRR noise rejection Aug 22, 2020 at 0:36
• @brhans Yes all grounds are common. Aug 22, 2020 at 1:25
• @tlfong01 The AD623 is the diff. amp. How would you recommend measuring 0-20mV? Aug 22, 2020 at 2:34

INA126 specs for common mode range is undefined at Vs+=3.3, Vs-=0 yet it says works to Vs=3V I can see the Vcm only shifts down with Vs+ due to high side current source voltage drop. So I take 3.3/5 x100% of the Fig 22 mean voltage of 1.6V. this equals 3.3/5*1.6= ≈1.0V rounded down.

• using 100000x more current than input bias current to minimize offset from input bias current
• we get a 4 resistor solution with no loss of differential input. from Fig 14 Ib≈14 nA so you can make 1V with >1.4mA roughly even 1mA with this low impedance is OK.

thus Rpu=2.3k, Rpd=1.0k I=3.3/ 3.3k=1mA , Vth= 1/3.3 x 3.3V=1.0V. Pullup (pu) to 3.3V, Down(pd) to 0 V pref using lowest error tolerance if using gain=1000 to minimize DC out offset.

Then use a lower Zcm bias resistors for each INA input between Zsource and ZIn(INA)

## my preference

Use a 1.0 V LDO with RF cap and RC bias to both inputs at 1.0V. this gives better differential noise rejection and common mode line noise rejection. Using RC=<50ms and R= 50k, C=1uF

Somehow you must to define your nominal input voltage and output if you want unipolar from 0V or bipolar centred around midscale output. TBD

simulate this circuit – Schematic created using CircuitLab

• Where are you seeing this on the data sheet? I have a limited power supply of 3.3V so Vcm cannot get 3.75V above the ref pin Aug 22, 2020 at 15:21
• Sorry let me correct it, I re-read spec Aug 22, 2020 at 15:34
• I really appreciate your help! I'm not quite sure I understand the significance of the 1mA current since the thermophiles in my flow sensor (FS1012) have a resistance of 200KOhm. I've changed the setup of my wires to make the ground for my sensor and ref pins the same at 1V (like I think you are indicating) to both up my common mode voltage and improve the swing range. Is this what you meant? Any chance you could include a diagram to clarify anything I missed. Aug 22, 2020 at 17:50
• Thanks for including the Pile specs of 200k Aug 22, 2020 at 20:47
• I've attempted your setup but the output doesn't change when I vary the gain or ref voltage even though the input measurably changes. I've added this additional figure to my OP to show maybe a thermopile is more of a voltage source and so the offset voltage should come before the sensor. When I do this the gain finally impacts the output but the range seems very limited to whatever ref voltage I choose. The range improves when Vref is the same as the CM voltage of the thermopile. Why is this? How can I improve the range to get as close to rail to rail as possible? Aug 22, 2020 at 21:48

After much trial and error, I've concluded that a minimum offset voltage for both the reference and thermopile "virtual grounds" (aka TP1- and TP2-) needs to be +0.8V-1.0V above ground. The thermopile needs to be considered a voltage source and so the second figure in my OP is the basis for moving forward. I don't understand why this is but it is likely related to the topology of the In-Amp since ref directly tied to the gain shown in the block diagram for INA126. I also suspect there is a "deadzone" in the common mode graphs that is disregarded by using the term Vd/2.

I can say with confidence INA126 is the way to go and without @TonyStewartSunnyskyguyEE75 I wouldn't have come to an answer. Thank you for your help!