LT1298 accuracy

john_s

As I'm getting closer to using Vref NOT equal to 5V I decided to start with this one...

Let's say we watch a 4-20mA current loop with a LT1298 12-bit A/D converter.
Suggested configuration includes 250 Ohm resistor that yields a 1 to 5V voltage drop across full current spec. Now, that allows for (4V of diff.voltage / (4095- ((1/5)*4095))) = 4/3276 = 1.22mV accuracy.

My question is:

Will I increase the accuracy (to 4V/4095=0.977mV) by first "expanding" 1-5V to 0-5V using a front-end OpAmp?

And the second one:

- How to apply a Vref and what is the optimal/suggested way of doing it with a LT1298?

- Will Vref = 4.095V be the optimal one as 1tick = 1mV accuracy with a Vin = 0 to 4.095V or there are some other preferred methods of dealing with voltages less than 5V span?

kwinn

John, if you are using a Vref of 4.095V I would use an op amp with a gain of 1, input the signal on the + input of the op amp, and offset the op amp by 1v to bring the signal to a 0-4.00V level. This way the 12 bit output of the ADC is directly proportional to the signal level in mV(ie 0000_0000_0000 = 0V, 1111_1010_0000 = 4.000V). Much easier to work with.

john_s

kwinn wrote: »

John, if you are using a Vref of 4.095V I would use an op amp with a gain of 1, input the signal on the + input of the op amp, and offset the op amp by 1v to bring the signal to a 0-4.00V level. This way the 12 bit output of the ADC is directly proportional to the signal level in mV(ie 0000_0000_0000 = 0V, 1111_1010_0000 = 4.000V). Much easier to work with.

In such a case - do you recommend to offset the opamp by exactly 1.000V ... or since its usually not rail-to-rail leave a bit of "slack" ?

And how much is that "slack" I wonder :-) in case I power opamp from 5V (Vcc)

p.s. A2D should be LTC1298 of course (not LT....)
btw. a very interesting observation about behaviour of LTC1298 with BS2 can be found here

http://forums.parallax.com/showthread.php?t=72993

p.s.2 When in doubt read the Data Sheet - which I did.... and found that I made a mistake in my original idea of adjusting Vref in LTC1298. You cannot do it as LTC1298 has both Vref and Vcc tied together. What I had in mind was related to its cousin LTC1286 that has a separate Vref pin The data sheet covers both in the same pdf and I was kind off tricked by NOT reading it...

I gues that covers my attempt to "adjust" Vref for LTC1298... or does it?


But the 1st question still remains valid.
What's the effect on accuracy when scaling Vin=1-5V to Vin=0-5V ?

kwinn

John, to answer question 2 first, the 1298 Vref is not adjustable. The power supply voltage Vcc is used as the reference and that also determines the span. At the minimum recommended 4.5V Vcc it would be 0 - 4.5V in 1.098mV steps.

If you use the 1298 at Vcc=5V and scale the signal to 0 - 5V the accuracy (actually the resolution) is 1mV per bit of the original 4V signal or 1.22mV per bit of the 5V signal.

I am assuming you are using a micro to control this ADC so I would suggest offsetting the signal so the ADC reads 40 - 50 when the input signal is 1V and amplify it (x1 for LTC1286 & 4.096V ref, x1.25 for LTC1298 @5V Vcc. This way the micro can tell if the input voltage goes below 1V (less than 40-50) and it can also subtract the offset from the reading.

john_s

wrote:

John, to answer question 2 first, the 1298 Vref is not adjustable. The power supply voltage Vcc is used as the reference and that also determines the span. At the minimum recommended 4.5V Vcc it would be 0 - 4.5V in 1.098mV steps.

Exactly - my mistake as I didn't read its data sheet for looooong time...

If you use the 1298 at Vcc=5V and scale the signal to 0 - 5V the accuracy (actually the resolution) is 1mV per bit of the original 4V signal or 1.22mV per bit of the 5V signal.

May I rather call it an increased 1mV accuracy (up from from 1.22mV) as the 12-bit resolution is what I cannot change indeed :-)

I am assuming you are using a micro to control this ADC so I would suggest offsetting the signal so the ADC reads 40 - 50 when the input signal is 1V and amplify it (x1 for LTC1286 & 4.096V ref, x1.25 for LTC1298 @5V Vcc. This way the micro can tell if the input voltage goes below 1V (less than 40-50) and it can also subtract the offset from the reading.

It's BS2 and after using LTC1298 in the same configuration for the last 12 years only now to discover what Tracy Allen explained in 2003 on all the errors it might have caused.

Well, better late than never :-)

john_s

kwinn wrote: »

John, if you are using a Vref of 4.095V I would use an op amp with a gain of 1, input the signal on the + input of the op amp, and offset the op amp by 1v to bring the signal to a 0-4.00V level. This way the 12 bit output of the ADC is directly proportional to the signal level in mV(ie 0000_0000_0000 = 0V, 1111_1010_0000 = 4.000V). Much easier to work with.

something like this?

kwinn

Yes, like that. The nice thing about having a micro between the output of the adc and the rest of the system is that you can replace the pots with precision resistors and let the micro calculate the corrections. All you need to do to calibrate it is input known voltages at or near the low and high ends of the range, and have the micro calculate an offset and multiplier to give the correct reading.

john_s

kwinn wrote: »

Yes, like that. The nice thing about having a micro between the output of the adc and the rest of the system is that you can replace the pots with precision resistors and let the micro calculate the corrections. All you need to do to calibrate it is input known voltages at or near the low and high ends of the range, and have the micro calculate an offset and multiplier to give the correct reading.

Eliminate pots - I like it a lot! I'll try to remember to implement something along that line in my next project.

Thanks for your help.