How to measure non differential signal with differential ADC?

[Michael Noone]

Hi - I need to measure a signal that will range from about 0-5 V with a
very high accuracy ADC. I was planning on using the TI ADS1252U
(http://focus.ti.com/lit/ds/symlink/ads1252.pdf) - as it's inexpensive,
fast, and accurate. Only one major question though - how do I connect a non
differential signal to differential inputs? It seems to me that every
single ADC I've looked at has differential inputs. In the datasheet for the
ADS1252U they show how to connect a bi-polar signal, but unless I'm getting
my terminology confused I don't believe that that is what I want. So - what
should I do? Thanks so much for your help!

-Michael Noone

 

[Spehro Pefhany]

Hi - I need to measure a signal that will range from about 0-5 V with a
very high accuracy ADC. I was planning on using the TI ADS1252U
(http://focus.ti.com/lit/ds/symlink/ads1252.pdf) - as it's inexpensive,
fast, and accurate. Only one major question though - how do I connect a non
differential signal to differential inputs?

If the signal is referenced to ground and positive, one plan might be
to ground the -Vin pin on the ADS1252. The device will measure the
voltage difference between the two pins, so don't choose your ground
unwisely. The signals have to be within the common mode range of the
converter, and I've not checked what that is- sometimes it varies with
the mode of the converter (eg. buffered vs. not etc.)

 

[hrh1818]

Page 7 of the application note gives this hint. "With a reference
voltage at less than half of VDD, one input can be tied to the
reference voltage, and the other input can range from 0V to 2 · VREF.
"

 

[Rob Gaddi]

If the signal is referenced to ground and positive, one plan might be
to ground the -Vin pin on the ADS1252. The device will measure the
voltage difference between the two pins, so don't choose your ground
unwisely. The signals have to be within the common mode range of the
converter, and I've not checked what that is- sometimes it varies with
the mode of the converter (eg. buffered vs. not etc.)

This will, however, cost you a bit of resolution, as the ADC is able to
swing from an input of +4 to -4 differentially, and grounding the input,
while the simpler solution by far, limits you to +4 to 0. Alternatively
you can use a good clean voltage reference and either a fully
differential amplifier such as the OPA1632 or two traditional op-amps to
let the signal use the entire differential range. This will improve
linearity somewhat and resolution by one bit, but it'll also cost a
couple dollars.

 

[Spehro Pefhany]

This will, however, cost you a bit of resolution, as the ADC is able to
swing from an input of +4 to -4 differentially, and grounding the input,
while the simpler solution by far, limits you to +4 to 0.

The ADC input range is to +/-Vref, limited by Vin(+) and Vin(-) >
-300mV. If the Vref and Vdd are (say) 5.100V, then the range is -300mV
to 5.1V, which covers the OP's 0-5V requirement with some room to
spare.

The LSB of a 24-bit ADC seldom means very much (you need to average
4096 readings on this part to have a hope of getting the 24th bit of
effective resolution).

This part doesn't seem to be spec'd very well for DC accuracy, despite
the suggested applications which require it.

 

[Rich Grise]

This will, however, cost you a bit of resolution, as the ADC is able to
swing from an input of +4 to -4 differentially, and grounding the input,
while the simpler solution by far, limits you to +4 to 0. Alternatively
you can use a good clean voltage reference and either a fully
differential amplifier such as the OPA1632 or two traditional op-amps to
let the signal use the entire differential range. This will improve
linearity somewhat and resolution by one bit, but it'll also cost a
couple dollars.

It also gives a negative binary output for negative inputs. See page 14-15:

http://focus.ti.com/lit/ds/symlink/ads1252.pdf

Cheers!
Rich

 

[Michael Noone]

The ADC input range is to +/-Vref, limited by Vin(+) and Vin(-) >
-300mV. If the Vref and Vdd are (say) 5.100V, then the range is -300mV
to 5.1V, which covers the OP's 0-5V requirement with some room to
spare.

The LSB of a 24-bit ADC seldom means very much (you need to average
4096 readings on this part to have a hope of getting the 24th bit of
effective resolution).

This part doesn't seem to be spec'd very well for DC accuracy, despite
the suggested applications which require it.

Ah! I had been worried about losing resolution if I ran it at it's "full
speed" - but I had looked through the datasheet and couldn't find
anything relating effective resolution to frequency. Now - as soon as
you've mentioned it - I see the graph was right in front of me the
entire time! Doh...

Ideally I'd like to get 20 bits resolution at 20khz, so I'm starting to
think that I should use a different ADC. The other chip I've been eyeing
is the TI ADS1271: http://focus.ti.com/lit/ds/symlink/ads1271.pdf. I've
looked through it twice now - and I still can't find anything about
frequency and resolution. Best I've been able to find is that it has
three modes - high resolution, high frequency, and low power - but I
can't find any numbers for those. Oh, and if 20 bits at 20khz is not
possible - I'd much rather sacrifice resolution than speed. Does the
ADS1271 look like a good solution? I see they have a diagram showing a
OPA1632 being used to connect a single ended interface input.

What do you think?

 

[Michael Noone]

If the signal is referenced to ground and positive, one plan might be
to ground the -Vin pin on the ADS1252. The device will measure the
voltage difference between the two pins, so don't choose your ground
unwisely. The signals have to be within the common mode range of the
converter, and I've not checked what that is- sometimes it varies with
the mode of the converter (eg. buffered vs. not etc.)

It was my understanding that with differential inputs the two inputs
should be the same magnitude and opposite signs. Is this not correct? It
wouldn't cause any problems for the negative input to be grounded?

-Michael

 

[Michael Noone]

This will, however, cost you a bit of resolution, as the ADC is able to
swing from an input of +4 to -4 differentially, and grounding the input,
while the simpler solution by far, limits you to +4 to 0. Alternatively
you can use a good clean voltage reference and either a fully
differential amplifier such as the OPA1632 or two traditional op-amps to
let the signal use the entire differential range. This will improve
linearity somewhat and resolution by one bit, but it'll also cost a
couple dollars.

Cost is not a serious worry right now. I mean sure - it's always great
to be the hero and find a solution using really cheap components - but
this will be replacing a multiple thousand dollar system, so a couple
dollars won't really break the bank

-Michael

 

[Spehro Pefhany]

Ah! I had been worried about losing resolution if I ran it at it's "full
speed" - but I had looked through the datasheet and couldn't find
anything relating effective resolution to frequency. Now - as soon as
you've mentioned it - I see the graph was right in front of me the
entire time! Doh...

Ideally I'd like to get 20 bits resolution at 20khz, so I'm starting to
think that I should use a different ADC. The other chip I've been eyeing
is the TI ADS1271: http://focus.ti.com/lit/ds/symlink/ads1271.pdf. I've
looked through it twice now - and I still can't find anything about
frequency and resolution. Best I've been able to find is that it has
three modes - high resolution, high frequency, and low power - but I
can't find any numbers for those. Oh, and if 20 bits at 20khz is not
possible - I'd much rather sacrifice resolution than speed. Does the
ADS1271 look like a good solution? I see they have a diagram showing a
OPA1632 being used to connect a single ended interface input.

What do you think?

There is a lot of good application data in "Techncial Documents" here:
http://focus.ti.com/docs/prod/folders/print/ads1271.html




Best regards,
Spehro Pefhany

 

[Steve]

How are Linear Technology's No Latency Delta Sigma converters in this
regard ? 10 Hz rate is fine for me.

http://www.linear.com/pc/categoryProducts.do