Interfacing a differential Sensor to a Differential ADC

Go To Last Post
5 posts / 0 new
Author
Message
#1
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Hello All,

Being a hopeless amateur at analogue electronics, I hope someone can give me a few hints on a problem that I have :)

I have a differential analogue sensor with the outputs swinging between 0.5V to 4.5V centered at 2.5V. That is, (4.5V - 0.5V = +4.0V) at one extreme and (0.5V - 4.5 = -4.0V) at the other extreme. The sensors are to be fed into a sigma-delta ADC. Signals are low frequency: about 400Hz BW.

Unfortunately, the sensors require buffering: the output impedance is not well specified in the datasheet and the ADC likes a low-impedance source. I can put the ADC in single-ended mode, but it would be nice to keep using the differential input.

I can do a simple voltage follower op-amp on each of the differential inputs, but it feels like there must be a better solution for a simple task like this. Plus there's the issue of matching op-amps etc. I suppose I could use an instrument amp, but ...

Can anyone recommend a good solution and/or part? Surely there must be something out there for a simple buffering problem like this :)

-- Damien

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Maybe look for op amps with differential outputs. e.g. LMP8350 or LTC1992

- S

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Lm733?

You can try putting caps on the adc inputs to lower the impedance.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Damien,

How much simpler can you get than two op-amp voltage followers? A standard 8-pin (MSOP, SOIC, DIP) and no external components, clean & simple.

However, if you also want your circuit to function properly and accurately, things will get a bit more complicated!

Depending on the specific ADC you are using you may find that the ADC itself will provoke noise & other sporadic behaviors from the op-amps. Typically, you need to be careful about which op-amp you use to feed the ADC input(s) as these inputs often draw very fast current pulses as a consequence of their front-end sampling process. There are special op-amps made just for this purpose, but you can get pretty good performance from many general purpose op-amps by adding an RC circuit between the op-amp(s) and the ADC input(s).

The R is connected from the op-amp output to the C, and the RC junction is connected to the ADC input. The other terminal of the C is usually connected directly to the ADC's analog ground input. In some case the feedback to the negative input of the op-amp is made directly from the op-amp output, in others the feedback is taken from the RC junction. The latter is the better choice, but some op-amps will oscillate with this configuration.

The C needs to be a high-speed type, ceramic is best. The value of the C should be as large as possible, which turns out to be not very large at all because of the ability of the op-amp to drive it without bursting into oscillation or similar ill action. The R is typically less than 100 Ohms because it will drop voltage wiht each gulp of current made by the ADC. This circuit generally defies calaculation and end up being a seat-of-the-pants tradeoff. Start with 47 ohms and a 4700 Pf cap and "tune" the values to get the "best" performance. The main ill effect you are looking for is op-amp oscillation. if you see this with your O-scope, decrease C and/or increase R.

Also, it is a good idea to have several different op-amps on hand to try in the circuit. My experience is that some work remarkably better than others in this situation. Of course, whichever op-amps you are using must meet the nominal operational requirements such as CM input range and output swing. If you use a standard 8-pin dual op-amp configuration, you will have about 4500 candidate op-amps to choose from!

Which ADC are you using?

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Sorry for the late reply :(

I was thinking something like the TI ADS1256. The rationale is that the sensors we are looking at have a noise specification of 5uV/rtHz, which, if we cut it down to 50Hz bandwidth is 35uV RMS noise, or about 17 bits of ADC. This ADC is capable of delivering greater than 18 noise-free bits at a sample rate of 200Hz (and well beyond...). The ADC has (optional) on-chip buffers, but they are only rated to VCC-2.0V, which is rather useless for the signals I have. The sensors are ratiometric to supply voltage (+5V).

Since the sensors are rather expensive, we are willing to cop a relatively expensive front-end (within reason!).

I have started to look at specific ADC drivers, though I am not yet sure exactly what I should be looking for.

On a related note: if I'm looking at a design with < 50uV sensor noise, how best to choose a power supply? A 5V low-noise LDO like the LT1762 is in the 20uV noise range. Or am I better off using a very low-noise voltage reference (e.g. REF5025: 3uV noise) and generating a 5V rail from a precision low-noise op-amp?

(Note that bias is a smaller consideration: The sensor will have at least 1.5mV bias, probably much more, and I will need to explicitly estimate the bias in the algorithm that uses the sensor data and hence. However, bias stability is something I am indeed interested in).

What a different world precision analogue is!

-- Damien