The observation about the limited range implies low res ADC a real good one.
I missed that .

But do an end to end error analysis to see if you are getting the accuracy desired.
IA costs money, and there is a temptation to roll your own IA and that leads to
some serious inaccuracies due to Aol differences between OpAmps and R ratio
accuracy and CM issues. And Vref onchip errors.

Regards, Dana.

Thanks Dana, you are an active member here, nice to learn from you

 

Keithley and Keysight (the newest HP) are the places to go. I basically remembered that HP and Fluke of yesteryear had a lot of stand-alone data loggers. For DC measurements, I used a lot of Keithley stuff - mainly electrometers.

Exactly KeepItSimple, I'm looking in their online sites for the appropriate device.
Thanks

 

You might also want to look at what National Instruments has to offer. National Instruments LabView is the defacto standard for controlling instruments. It's a visual programming language and a very different.

Stuff using the VXI bus might be appropriate.

 

Measurement Computing has DAQ SW and HW.

https://www.mccdaq.com/data-acquisi...MIjbjTgJ654AIVj4jICh2YqgujEAAYASAAEgKDCfD_BwE


Regards, Dana.

 

I did not read every post carefully but the questions that pop up in my head is what the ultimate purpose is of the system, and what is the intended goal, ultimately. It is one thing to work on a task but totally different to achieve the ultimate goal, which may be much simpler with a different approach. That concept allowed me to design machines that delivered the needed results while being entirely different from what our customers thought that they needed.

 

Measurement Computing has DAQ SW and HW.

https://www.mccdaq.com/data-acquisi...MIjbjTgJ654AIVj4jICh2YqgujEAAYASAAEgKDCfD_BwE


Regards, Dana.

Thanks Dana

 

I would not use an Arduino for the reasons I mentioned. It's only a 10 bit A/D. Fot a commercial application I would look to DATAQ Instruments or Measurment Computing or any of the other commercial lines of data acquisition units available. Those and others offer good commercial application equipment and include adequate software packages to get started. That includes data logging software.

Ron

I'm uncertain why anybody thinks a 10-bit ADC is inadequate. You don't run a signal that low into and ADC. You run it into an opamp, raise it's level, and then run that into an ADC. Very, very easily doable.

 

I'm uncertain why anybody thinks a 10-bit ADC is inadequate. You don't run a signal that low into and ADC. You run it into an opamp, raise it's level, and then run that into an ADC. Very, very easily doable.

Well I can't speak for anyone else but looking at the original post:

Hello everyone,
I searched in the net about a voltage recorder but unfortunately I can not find any with the following performance:
-Record DC voltage of the order of a few millivolts with a sensitivity of 0.1 mv
-Sampling frequency: 200 hz (if it is possible an adjustable frequency that can reach 200 hz)
I want to measure DC Voltage at the terminals of a wire that leads a DC current of 100 mA, the following diagram illustrates the assembly :

Thank you very much for any help.

OK, I did mention we can amplify the low level signal. That would seem logical but in that process we will also amplify any signal noise so we can amplify and filter as best we can. Discounting accuracy and with a focus only on the 10 bit A/D conversion. The best case resolution with 1024 quantization levels and a 5 Volt reference is going to be 5 / 1024 = 4.8 mV. That assumes a perfect 5 Volt stable reference. The thread starter mentions "Record DC voltage of the order of a few millivolts with a sensitivity of 0.1 mv". So they want to accurately resolve 100 uV. I likely mentioned oversampling as I generally do but in my opinion, especially based on what the thread starter subsequently mentions, a 10 bit A/D Arduino A/D solution is not a good choice. Even with a well filtered gain of 1000 that 100 uV becomes 0.1 volt.

Again, speaking strictly as my thinking but I don't see a 10 bit A/D as the best choice if constraints of budget are not an issue and this will be ongoing testing.

Ron

 

Well I can't speak for anyone else but looking at the original post:


OK, I did mention we can amplify the low level signal. That would seem logical but in that process we will also amplify any signal noise so we can amplify and filter as best we can. Discounting accuracy and with a focus only on the 10 bit A/D conversion. The best case resolution with 1024 quantization levels and a 5 Volt reference is going to be 5 / 1024 = 4.8 mV. That assumes a perfect 5 Volt stable reference. The thread starter mentions "Record DC voltage of the order of a few millivolts with a sensitivity of 0.1 mv". So they want to accurately resolve 100 uV. I likely mentioned oversampling as I generally do but in my opinion, especially based on what the thread starter subsequently mentions, a 10 bit A/D Arduino A/D solution is not a good choice. Even with a well filtered gain of 1000 that 100 uV becomes 0.1 volt.

Again, speaking strictly as my thinking but I don't see a 10 bit A/D as the best choice if constraints of budget are not an issue and this will be ongoing testing.

Ron

An arduino is not a good choice for a number of reasons, including noise and a less stable reference for the A/D conversion, and being essentially a toy with all of the problems associated with toys. Even with an amplifier, the basic system noise would be excessive. The other companies mentioned have been in the instrumentation business for quite a few years providing systems and software for customers who are very picky and demand equipment that is both stable and rugged. The ten-bits of the converter is not the problem, the rest of the concerns are the problem.
Just handling that hundred microvolt signal is a challenge, and doing it in a low-noise stable environment is not a simple thing, if accuracy is required.

 

Looking at the first post, it seems like the purpose of the measurement is to monitor the voltage drop of a connector under some condition. But it may also be intending to monitor the resistance of the crimped connection to the one part of the connector pair. Certainly that is an appropriate test to do, because at least one very big producer of a very large variety of crimp on connectors used to be very optimistic in their current ratings. We disqualified them based on total failures of the connection, not on a measured voltage drop.
The suitable instrumentation approach would be to get an amplifier designed for monitoring ammeter shunts, typically 100Millivolts full scale. OR use an amplifier designed for strain guages, which have an output that is very small. But I do wonder how rapid a failure is anticipated. Possibly the TS could share a bit of the application information with us?

 

An arduino is not a good choice for a number of reasons, including noise and a less stable reference for the A/D conversion, and being essentially a toy with all of the problems associated with toys. Even with an amplifier, the basic system noise would be excessive. The other companies mentioned have been in the instrumentation business for quite a few years providing systems and software for customers who are very picky and demand equipment that is both stable and rugged. The ten-bits of the converter is not the problem, the rest of the concerns are the problem.
Just handling that hundred microvolt signal is a challenge, and doing it in a low-noise stable environment is not a simple thing, if accuracy is required.

I think that is what I stressed through the thread.

Ron

 

Looking at the first post, it seems like the purpose of the measurement is to monitor the voltage drop of a connector under some condition. But it may also be intending to monitor the resistance of the crimped connection to the one part of the connector pair. Certainly that is an appropriate test to do, because at least one very big producer of a very large variety of crimp on connectors used to be very optimistic in their current ratings. We disqualified them based on total failures of the connection, not on a measured voltage drop.

Based on the original cartoon (drawing) that was sort of my take. One line of thinking was to just use a good quality DLRO (Digital Low Resistance Ohmmeter) with an interface. I have an old Agilent 34401 DMM which has a RS232 interface. My guess today is there are DLROs with a USB interface. Part of the everyday Agilent software allowed choosing a function and sample rate sending the output data to Excel. That or roll your own software and dump the data wherever you choose.

Problem was the original poster never got into detail as to exactly what they wanted to do leaving us to guess. If they wanted an automated test system they should have said so? Beats me?

Ron

 

Based on the original cartoon (drawing) that was sort of my take. One line of thinking was to just use a good quality DLRO (Digital Low Resistance Ohmmeter) with an interface. I have an old Agilent 34401 DMM which has a RS232 interface. My guess today is there are DLROs with a USB interface. Part of the everyday Agilent software allowed choosing a function and sample rate sending the output data to Excel. That or roll your own software and dump the data wherever you choose.

Problem was the original poster never got into detail as to exactly what they wanted to do leaving us to guess. If they wanted an automated test system they should have said so? Beats me?

Ron

Once again I need to remind almost everybody that my mind reading ability is VERY POOR, and so they need to explain things. I also have designed an ohm meter with full scale of about 2 ohms. It ran 10mA through the devices being checked, while being able to remain within specified accuracy with up to 100 ohms in any or all connection leads. AND it cost many hundreds of dollars less than the previous system that it replaced, in addition to being far more stable.