ad7793 weirdness

[[email protected]]

I'm prototyping an RTD based temperature sensor using the analog
devices AD7793 24-bit ADC. I seem to be seeing some unexpected
results when feeding a small fixed voltage (340 millivolts) and
varying the gain of the AD7793. Reference voltage is fixed at approx
2.26 volts.

If I set the ADC for unity gain, I see a reading of 2,531,837...
If I set the ADC for 2x gain, I see a reading of 5,071,765... Which
is reasonable...
Now if I set ADC for 4x gain, I get an unusual reading of
6,116,603... Not double the 2x reading that one would expect.
If I then set the ADC for 8x gain, I see a reading of 12,050,792...
Which is about double the 4x reading, but not 8x the unity reading
that one would expect.

I've tried doing internal fullscale and zero calibrations without any
real change.

I did find a line in the datasheet which worries me: "When the in-amp
is active (gain >= 4), the common-mode voltage (AIN(+) + AIN(-))/2
must be greater than or equal to 0.5 V." Surely that is a mistake.
That requirement would render the gain function nearly worthless since
small input values is when you are going to want to use the higher
gains.

Can anyone help me figure out what I'm overlooking?

thanks.

 

[Joel Koltner]

I did find a line in the datasheet which worries me: "When the in-amp
is active (gain >= 4), the common-mode voltage (AIN(+) + AIN(-))/2
must be greater than or equal to 0.5 V." Surely that is a mistake.
That requirement would render the gain function nearly worthless since
small input values is when you are going to want to use the higher
gains.

It might not be a mistake: It's quite common to have some small signal sitting
on top of "some volts" of common mode voltage... and indeed, you can arrange
this yourself by just adding some bias resistors and coupling capacitors if
you have an AC signal (probably not for most RTD usages?) or using op-amps if
not (albeit with op-amps the analysis of what it does to very tiny signals can
take some effort).

I'd ask ADI for certain...

 

[Spehro Pefhany]

I'm prototyping an RTD based temperature sensor using the analog
devices AD7793 24-bit ADC. I seem to be seeing some unexpected
results when feeding a small fixed voltage (340 millivolts) and
varying the gain of the AD7793. Reference voltage is fixed at approx
2.26 volts.

If I set the ADC for unity gain, I see a reading of 2,531,837...
If I set the ADC for 2x gain, I see a reading of 5,071,765... Which
is reasonable...
Now if I set ADC for 4x gain, I get an unusual reading of
6,116,603... Not double the 2x reading that one would expect.
If I then set the ADC for 8x gain, I see a reading of 12,050,792...
Which is about double the 4x reading, but not 8x the unity reading
that one would expect.

I've tried doing internal fullscale and zero calibrations without any
real change.

I did find a line in the datasheet which worries me: "When the in-amp
is active (gain >= 4), the common-mode voltage (AIN(+) + AIN(-))/2
must be greater than or equal to 0.5 V." Surely that is a mistake.
That requirement would render the gain function nearly worthless since
small input values is when you are going to want to use the higher
gains.

Can anyone help me figure out what I'm overlooking?

thanks.

"The AD7792/AD7793 can be operated in unbuffered mode only when the
gain equals 1 or 2. At higher gains, the buffer is automatically
enabled. The absolute input voltage range in buffered mode is
restricted to a range between GND + 100 mV and AVDD – 100 mV. When the
gain is set to 4 or higher, the in-amp is enabled. The absolute input
voltage range when the in-amp is active is restricted to a range
between GND + 300 mV and AVDD - 1.1 V. Take care in setting up the
common-mode voltage so that these limits are not exceeded..."



Best regards,
Spehro Pefhany

 

[Arlet Ottens]

I'm prototyping an RTD based temperature sensor using the analog
devices AD7793 24-bit ADC. I seem to be seeing some unexpected
results when feeding a small fixed voltage (340 millivolts) and
varying the gain of the AD7793. Reference voltage is fixed at approx
2.26 volts.

If I set the ADC for unity gain, I see a reading of 2,531,837...
If I set the ADC for 2x gain, I see a reading of 5,071,765... Which
is reasonable...
Now if I set ADC for 4x gain, I get an unusual reading of
6,116,603... Not double the 2x reading that one would expect.
If I then set the ADC for 8x gain, I see a reading of 12,050,792...
Which is about double the 4x reading, but not 8x the unity reading
that one would expect.

I've tried doing internal fullscale and zero calibrations without any
real change.

I did find a line in the datasheet which worries me: "When the in-amp
is active (gain >= 4), the common-mode voltage (AIN(+) + AIN(-))/2
must be greater than or equal to 0.5 V." Surely that is a mistake.
That requirement would render the gain function nearly worthless since
small input values is when you are going to want to use the higher
gains.

Can anyone help me figure out what I'm overlooking?

It means that for higher gain settings, you must bias the inputs such
that the average is above 0.5V. For instance, you could have 1.000V on
AIN- and 1.340V on AIN+.

For RTD measurement, check fig. 21 of the datasheet. The Rref value must
be high enough such that (Iout1 + Iout2) * Rref > 0.5 V

 

[Spehro Pefhany]

It might not be a mistake: It's quite common to have some small signal sitting
on top of "some volts" of common mode voltage... and indeed, you can arrange
this yourself by just adding some bias resistors and coupling capacitors if
you have an AC signal (probably not for most RTD usages?) or using op-amps if
not (albeit with op-amps the analysis of what it does to very tiny signals can
take some effort).

I'd ask ADI for certain...

Anything *bad* in a data sheet can generally be taken as true..

Since generally you're running RTDs with constant current sources (one
source for 2-wire or 4-wire and two for 3-wire), and since RTDs are
generally isolated, a simple resistor to circuit ground, preferably
with a bypass, should suffice.

Best regards,
Spehro Pefhany

 

[Arlet Ottens]

It means that for higher gain settings, you must bias the inputs such
that the average is above 0.5V. For instance, you could have 1.000V on
AIN- and 1.340V on AIN+.

For RTD measurement, check fig. 21 of the datasheet. The Rref value must
be high enough such that (Iout1 + Iout2) * Rref > 0.5 V

By the way, if you're not using this 3-wire RTD configuration, you may
find the built-in bias voltage generator useful. See bits 14/15 of the
configuration register.

 

[[email protected]]

It means that for higher gain settings, you must bias the inputs such
that the average is above 0.5V. For instance, you could have 1.000V on
AIN- and 1.340V on AIN+.

For RTD measurement, check fig. 21 of the datasheet. The Rref value must
be high enough such that (Iout1 + Iout2) * Rref > 0.5 V

Thanks for all the quick responses... I am using fig 21 for the RTD,
but during this experimentation/testing, I just drove the excitation
current thru the reference resistor, grounded AIN- and connected to AIN
+ to the small variable voltage source.

So if I follow fig 21, using a 4.99K reference resistor, 210 uA
excitation current, and 100 ohm RTD, then the higher gains should work
fine?

 

[Spehro Pefhany]

By the way, if you're not using this 3-wire RTD configuration, you may
find the built-in bias voltage generator useful. See bits 14/15 of the
configuration register.

I don't see how that would help, unless he has a completely isolated
current source, which seems unlikely.

Best regards,
Spehro Pefhany

 

[Arlet Ottens]

Thanks for all the quick responses... I am using fig 21 for the RTD,
but during this experimentation/testing, I just drove the excitation
current thru the reference resistor, grounded AIN- and connected to AIN
+ to the small variable voltage source.

So if I follow fig 21, using a 4.99K reference resistor, 210 uA
excitation current, and 100 ohm RTD, then the higher gains should work
fine?

Yes. I'm using the device in a very similar way, and it works fine.

 

[John Larkin]

I don't see how that would help, unless he has a completely isolated
current source, which seems unlikely.

Best regards,
Spehro Pefhany

For a 2-wire rtd, make a voltage divider from any old Vref, with a
reference resistor and the rtd. Digitize the voltage across both and
do the math. Tweaks can do 3-wire and keep the common-mode voltage up.

But with a 24-bit adc, you can turn off the buffer and let one end of
the string be ground.

John