Pro Analogue - Ensuring Accurate Measurements over Temperature

Thread Starter

dannybeckett

Joined Dec 9, 2009
185
Hi Guys,

This is quite an open ended question, and I'm being vague on purpose. Please no replies with "give more details of circuits" - there is no circuit at the moment.

I have a project on which involves measuring the impedance of a material using a sinusoidal signal at 18MHz. I plan to feed a stable excitation signal into the material (known f & V), and then determine Z by measuring the current flow. I need to achieve great stability and accuracy over a wide temperature range (-20 degrees to +120 degrees). All of the electronics in the circuit are going to be exposed to this range of temperature. I don't want the analogue measurement to be affected by resistors changing value etc as ambient temperatures change.

At the moment, current techniques involve profiling the measurement boards over temperature, then normalising the effect out via firmware look-up tables. What I'd like to achieve is a reading stable enough over temperature that this step is not required. And so, my question:

To any experienced analogue design engineers out there, what techniques do you employ to ensure accurate measurements over temperature? My current thinking is as follows:

> Achieve a rock-solid excitation frequency by using a low ppm xtal fed through a DDS
> Use something like a 2.5V voltage-reference IC that's stable over temperature to ensure a stable excitation voltage
> Currently unsure how to measure current with such a high excitation frequency, with stability over temperature. A very fast transimpedance amplifier with fast rectification and filtering?

Any hints/tips/help you guys can throw my way is greatly appreciated.

Dan
 
Last edited:

OBW0549

Joined Mar 2, 2015
3,566
Without having any idea how stable is "stable enough" (or whether your -20 to +120 degree temperature range is in Celsius or Fahrenheit), it's hard to determine whether you'll be able to eliminate profiling or not. Are you trying to keep readings within a few percent over the temperature range? Or a few parts per million? There's a big difference.

To any experienced analogue design engineers out there, what techniques do you employ to ensure accurate measurements over temperature? My current thinking is as follows:

> Achieve a rock-solid excitation frequency by using a low ppm xtal fed through a DDS
> Use something like a 2.5V voltage-reference IC that's stable over temperature to ensure a stable excitation voltage
> Currently unsure how to measure current with such a high excitation frequency, with stability over temperature. A very fast transimpedance amplifier with fast rectification and filtering?

Any hints/tips/help you guys can throw my way is greatly appreciated.
In addition to the above, especially using the most stable voltage reference you can find, you might consider building your circuit with low-tempco resistors. Most run-of-the-mill resistors are specified with a temperature coefficient of resistance of ±100 ppm/°C maximum, meaning that over a 0-100°C range their resistance will vary by as much as ±1%. Resistors with better stability are available, but expect to pay more.
 

Thread Starter

dannybeckett

Joined Dec 9, 2009
185
A bit of clarification, I'd like to achieve 1% error (2% if this is going to be too hard to achieve) of my full range over -20°C to +120°C. I've heard form other sources that using a ratiometric measurement may be more suited to this application, as opposed to relying on a stable voltage reference.
 

OBW0549

Joined Mar 2, 2015
3,566
A bit of clarification, I'd like to achieve 1% error (2% if this is going to be too hard to achieve) of my full range over -20°C to +120°C. I've heard form other sources that using a ratiometric measurement may be more suited to this application, as opposed to relying on a stable voltage reference.
Getting 1-2% stability over a 140°C temperature span is going to be quite a challenge, especially considering you're dealing with an 18 MHz RF signal. I've no experience in precision measurement at such high frequency, so I can't help you there.
 
Top