Voltage measurements from op amps

Thread Starter

Metalfan1185

Joined Sep 12, 2008
170
I am currently working on a project of mine that has a section of it that pretty much is a precision DVM.

The goal is to use a Linear Technology LTC2440 24 bit ADC to take precise voltage readings (down to the uV is possible) and send them (via SPI) through a Microcontroller to an LCD. I am currently working to design the Analog front end stuff. I would like it to have a similar impedance to a Multimeter (10M Ohms). I accomplished this with the help of a user on a different forum that pointed me in the direction of a constant current voltage divider (9M, 900k, 90k, and 10k ohm resistors in series). I may change the ranges later but I'm trying to get something functional first.

I'm new to op amps so I was wondering if someone would like to help me with the signal path up to the ADC. (several ideas have came to me on this, the possibility of using opamps, Autorange vs Manual range, cascaded window comparators, ect.) I have a basic understanding of how op amps work but if this thing measures to 2 uV then its pretty sensitive stuff and I can already see its going to need additional measures against noise and stuff too.

If a standard set of ranges cant work, I would imagine using something like:

0-2.5V, (measuring sensors and stuff)
0-5V, (Mainly logic level monitoring)
0-12V, (DC motors and above logic devices)
0-15V, (Automotive Electrical measurements)
0-25V,
0-150V, Mains Level measurements
0-250V, Mains Level measurements
and 0-1000V Max range. May not be used much but just to have something for higher potentials.

I'm open to all input on this, I'm no master of electronics, just a long time hobbyist that may know a few things lol.
 

AnalogKid

Joined Aug 1, 2013
10,989
For any A/D converter of 10 bits or more, one of the major contributors to overall system inaccuracy is the A/D reference. For example, if you use an LM4040 with a 0.2% tolerance, and you don't have a 5 or 6 digit voltmeter to use to set a calibration adjustment that has a zero temperature coefficient, then your system is limited to a maximum equivalent accuracy of just 9 bits, no matter how many bits the A/D device actually is. For 2.5 V full scale, that is a basic precision of +/- 5 mV, not counting several other sources of error. It sucks. I know.

ak
 

Thread Starter

Metalfan1185

Joined Sep 12, 2008
170
Thanks for the reply guys.

As far as a schematic goes, it doesn't really exist. I mean i have all the other things all planned out on paper, the LCD interfaced with a microcontroller, the SPI interface and code for calibration adjustments and all that. Its just the front end that's missing. Everything up to the input of the ADC.


@AnalogKid I have given some thought to that as well. I would like this to be as precise as possible, and if not to an absolutely perfect reference point (I know these chips exist and can be in the $100's of dollars range...I will research some comparable ones) The project will have a graphing LCD that shows an oscilloscope-type output. This way even if one point measures say 2.56784 when a Professionally made meter at the same point measure's 2.56888 (probably a couple mV difference, but who knows?) Its more for monitoring a voltage over time to see changes in the voltages.

But also, I own a professional meter calibrator and have access to a 6 digit Fluke DMM that has been recently calibrated, so maybe its not so far-fetched an idea?

I could probably get something on paper for an idea, but its the approach I'm unsure of. I would say its safe to assume that for each element that is added (op amp, or any other semiconductor junction) there will be error, and its ideal to minimize that error wherever possible. You know?
 

crutschow

Joined Mar 14, 2008
34,285
Page 23 of the data sheet discusses buffering it the LTC2440 input.
(You should always thoroughly read the data sheet of any device you are using. It's generally the best source to show you how to properly use it.)

If you have a high accuracy voltmeter available for calibration purposes, then you can add some pots in the input attenuator for adjustment of the output readings versus input voltage.
 

Thread Starter

Metalfan1185

Joined Sep 12, 2008
170
Im unfamiliar with that buffering thing...I assume its like smoothing out the readings on analog pins? you read a few, average it, print average? Ill look into that as well.

I thought of that method, i was thinking of doing just that with some 10 turn pots. (I think i have some 25 turn ones as well). Do you think it would drift from there much after i set it? Im trying to think if a plausible method for switching ranges efficiently while preserving accuracy. I was thinking a rotary switch at first, then maybe a window comparator setup with that 4066 switch....idk im still thinking out loud at this point. Thanks for the reply!
 

Papabravo

Joined Feb 24, 2006
21,159
The purpose of buffering, among other things, is to prevent the impedance of the source and the impedance of the A/D input from affecting the result. A buffer device has a high input impedance and does not load the source.
 

crutschow

Joined Mar 14, 2008
34,285
The A/D requires a low source impedance for accuracy but you want a 10MΩ input impedance, so an op amp buffer can do that. It has nothing to do with "smoothing" readings or other analog functions.

You could use 10-turn trimpots to calibrate each attenuation value of the input (each attenuator step will have it's own error that likely will need correction).
Once you do a calibration it should stay in reasonable calibration for a year or more, unless you have critical requirements.
 

Thread Starter

Metalfan1185

Joined Sep 12, 2008
170
AH I understand what you mean now. Thanks @Papabravo. @crutschow that is just the advice I was searching for thanks! What op amp would you recommend? I have some of the basics laying around to play with, LM308N, LM741, MC1458, I have a LM336 and a REF02C for references but I havent assessed their specs yet for this application but I would assume noise and linearity are big time important. My requirements aren't exactly "critical" but ya know, you build something you want it to be as nice as you can get it (reasonably priced) lol.

Thank you for the feedback so far this is excellent!

***Edit Just found some TL072 and equivalents hanging out here as well.

The other thing Im thinking of is the negative scale. Im thinking I could either bias the input of the opamp and have it swing one volt or two in either direction....Or come up with a means of just directing the signal to a seperate inverting op amp once the voltage falls below the -Ref (GND).
 
Last edited:

crutschow

Joined Mar 14, 2008
34,285
I would go with the device they recommended in the data sheet.
You need an op amp with a very low input bias current (which goes through the 10MΩ input attenuator) and low offset to achieve good accuracy.
Certainly a 741 is not even close to meeting those requirements. :rolleyes:
The TL072 has very low input bias current but it has an input offset of up to 10mV.

If I read the data sheet correctly (Page 8), [you did read the data sheet, right?] it states that the input can go between plus and minus 1/2 the reference voltage, so you shouldn't need any special circuitry to measure negative voltages.
 

Thread Starter

Metalfan1185

Joined Sep 12, 2008
170
@crutschow it does seem that the device in the datasheet is what I need but its kinda pricey (averaging about $20 with different shipping/price combinations). I checked digikey and ebay, but im not entirely sure of any other good sources for stuff like this. I have also decided to use the precision reference used in the LTC2440 datasheet If I can (its decently affordable and very precise). Could anyone recommend either a more affordable op-amp than the LTC2051 for this application, or a cheaper source?

That section also had me browsing in the direction of learning about chopper amplifiers that's a new topic for me. I love learning this stuff hahaha.
Thanks!

Once I decide on these parts I can order them and start building the analog front end of this thing!
 

ErnieM

Joined Apr 24, 2011
8,377
I would never use a pot in a measurement system that had a microprocessor in the signal chain. Pots add another source of uncertainty or drift.

Flash or EEPROM memory can store a constant to correct readings. Load it once in a calibration cycle, use it repeatedly in a reading cycle.
 

Thread Starter

Metalfan1185

Joined Sep 12, 2008
170
An LMC6062 might be suitable for what you're doing. Digi-Key has them for about $3 in single quantities.
Excellent! I will look into this option more a little later. Seems like a nice device for the cost. Thanks!

I would never use a pot in a measurement system that had a microprocessor in the signal chain. Pots add another source of uncertainty or drift.

Flash or EEPROM memory can store a constant to correct readings. Load it once in a calibration cycle, use it repeatedly in a reading cycle.
I have seen fluke meters that have trim pots for calibration, and also others that have a com port to hook up to a PC for calibration (through software). By your statement do you mean to check actual resistor values with a precision DMM and enter those values into the code and have the chip do the math to calclulate the correct value? Or do you mean have the components inside the device to provide a reference point and automatically adjust itself to that point before measurements are taken?
 

crutschow

Joined Mar 14, 2008
34,285
.........................
I have seen fluke meters that have trim pots for calibration, and also others that have a com port to hook up to a PC for calibration (through software). By your statement do you mean to check actual resistor values with a precision DMM and enter those values into the code and have the chip do the math to calclulate the correct value? Or do you mean have the components inside the device to provide a reference point and automatically adjust itself to that point before measurements are taken?
Typically, for software calibration, you input 0V and a known precise voltage to the A/D, and store those two A/D readings in EEPROM.
This is performed for each of the attenuator ranges you have.
Then those values are used to correct the subsequent measurement readings for gain and offset using linear interpolation.
This corrects, to a first order, for most errors between the input and the A/D output, leaving only the linearity of the A/D as an error source.

But note that, for this scheme to work properly, the microprocessor has to know the attenuator range setting.
This could be done by reading from a separate set of contacts on the range switch.
 
Last edited:

Thread Starter

Metalfan1185

Joined Sep 12, 2008
170
Typically, for software calibration, you input 0V and a known precise voltage to the A/D, and store those two A/D readings in EEPROM.
This is performed for each of the attenuator ranges you have.
Then those values are used to correct the subsequent measurement readings for gain and offset using linear interpolation.
This corrects, to a first order, for most errors between the input and the A/D output, leaving only the linearity of the A/D as an error source.

But note that, for this scheme to work properly, the microprocessor has to know the attenuator range setting.
This could be done by reading from a separate set of contacts on the range switch.
That makes sense. I was considering options for the range selection. Ideally I would like the range displayed on an LCD so the MCU will have to know which range it is on. As far as signal path goes I could make a hardware panel mount connector for each range, or Ideally have a switch that can switch between ranges as you said. I would need OV/UV protection in case its switched to a lower range while measuring a higher voltage than the ADC can take, but that should be there regardless. I was thinking of using something like a 74HC4066 switch chip to direct the input path to the ADC after the voltage divider. The datasheet says it works with 2V to 6V (no good) and has a on state resistance that seems to vary as the voltage changes (also no good), so that will affect linearity I'm sure. Maybe ill just see if I can add a 5PDT or 6PDT switch to the shopping list and use that for both the signal path and range ID to the chip. A good switch with clean contacts shouldn't have a very high resistance.

Thanks for the link to the linear interpolation page too, very helpful.
 

crutschow

Joined Mar 14, 2008
34,285
The resistance of the switch will have little effect if it is feeding a high impedance, like an op amp non-inverting input.
You could use an analog switch from the attenuator output as long as you protect each switch input from over voltage (such as with a series resistor and reverse biased diodes to the plus and minus supplies.
 

Thread Starter

Metalfan1185

Joined Sep 12, 2008
170
I apologize for the panned image, I don’t own a scanner so I had to take a picture of it and apparently my paper is a bit on the shiny side.

I believe this is pretty much what’s involved in the analog side so far although I know some elements we talked about are missing. That’s what I’m working through here. The diodes at the follower’s input are shottkey barrier diodes, I believe I found some with a really low Vf (about 200mV) from a different project but I can’t remember the part number. Is it ok to assume that if the current is low enough from the resistor divider tree then the diodes won’t be damaged by possible over/under voltages?

For the range selection I have another voltage divider that switches in different top side resistors, this I only need one analog pin on the MCU to read which range is selected.

The only calibration point on the diagram I drew here is the trimmer pot for the Reference voltage. We discussed adding possibly trim resistors in the divider as well or compensating for error in code. (I would think this could be done by measuring the ACTUAL resistor values with considerable accuracy and entering them into the code, have the code perform the calculations and then compare it to the voltage measured). Which method do you think would be more stable?
 

Attachments

Top