Multifunction DAQ

cancel
Showing results for 
Search instead for 
Did you mean: 

Getting incorrect voltage measurements overt resistor divider bridge


@Károly wrote:

I have a problem where I need to measure power on three separate lanes with different power supplies (12V, 5V and 3,3V), however, my PCIe DAQ can only measure voltages up to 10V. To get good measurements I need to know the exact voltage on all lanes as well as the voltage drop over some shunt resistors. As advised in: DAQ change input voltage range I have built up the following resistor divider bridges: bridge.png

I am measuring the voltage in single ended ground referenced mode and the voltage drops over the shunts in differential mode. Problem is that the 5V and 3,3V seem to provide sensible readings while the 12V is inaccurate. I am expecting voltage drops in the order of 0,01V and when I measure with finite samples I get a sinusoidal with a mean value of 0,1V so 10 times the expected. If I change to continuous mode it becomes even worse with a measured voltage drop of about 0,5V so 50 times the expected. If I change both ai5+ and ai5- to single ended ground referenced and subtract them in software I seem to get a reasonable value but I am unsure if it is actually correct since I don't know what causes the problem. Can anybody give me an insight?

 

 

 

 

 

 


 

 

 

 

 

As an electrical engineer I can say that your issue is two things. First is the current in the 10M voltage divider, you are expecting 12V/20Mohm = 600nA through your voltage divider to drive the input of the DAQ but you pick up noise or input offset current that is > 600nA in your circuit. Second is the input impedance on your DAC, it should be >> than 10M or the input offset current in your first gain stage will dominate the measurement you are trying to take with this particular circuit, but with these kinds of impedances you are more likely to pick up radio stations than your measurement if you are using test leads (antennas). 

 

The lesson here is: 

 

1) This is not the best circuit to do what you want, it works in SPICE but not in real life, check this out for some tips. 

 https://www.ti.com/lit/an/sboa347a/sboa347a.pdf?ts=1685961886417&ref_url=https%253A%252F%252Fwww.goo...

2) When you design a circuit, you need to think about the entire circuit, as in; what is going on inside the DAQ, if you attach it to your circuit, it will have some effect on it. 

 

Additionally, this is not a Labview question. There is one more thing you need to figure out to make this work, If the above info is not enough to get you going on that, I have an EE consulting business that can show you how to take these measurements : ) 

______________________________________________________________
Have a pleasant day and be sure to learn Python for success and prosperity.
0 Kudos
Message 11 of 28
(1,583 Views)

Thank you for the very detailed response,

 

Do you think if I use lower resistor values for the bridge (let's say 1M) it might work, or is it a totally lost cause and I must use an external amplifier?

The input impedance of the DAQ is specified as >1G so I assumed it would not be a problem as it is orders of magnitude bigger than that of my bridge.

Also thank you for the offer, however, this is a dissertation project that I was supposed to finish months ago to graduate from university but it is still not working, that's why I decided to consult these forums. As such I cannot afford to use a consulting service.

In your opinion then, is there any way to do what I want without adding many more external components?

0 Kudos
Message 12 of 28
(1,565 Views)

It seems like you were right. As others have suggested it might be worth moving the thread to the multifunctional daq category.

0 Kudos
Message 13 of 28
(1,563 Views)

I have move the thread.

 

 


@Károly wrote:

Do you think if I use lower resistor values for the bridge (let's say 1M) it might work, or is it a totally lost cause and I must use an external amplifier?

The input impedance of the DAQ is specified as >1G so I assumed it would not be a problem as it is orders of magnitude bigger than that of my bridge.

Also thank you for the offer, however, this is a dissertation project that I was supposed to finish months ago to graduate from university but it is still not working, that's why I decided to consult these forums. As such I cannot afford to use a consulting service.

In your opinion then, is there any way to do what I want without adding many more external components?


In the past, we used a total of 1MOhm divider and found that it did not work well for the DAQ.  We had to add voltage follower op-amp circuits for the divider readings.  Super simple circuit.

 

For the current shunt measurements, I found it better to use current shunt amplifiers that output a single ended value, such as this chip: TSC103IPT.


GCentral
There are only two ways to tell somebody thanks: Kudos and Marked Solutions
Unofficial Forum Rules and Guidelines
"Not that we are sufficient in ourselves to claim anything as coming from us, but our sufficiency is from God" - 2 Corinthians 3:5
Message 14 of 28
(1,552 Views)

Dear Henrik,

 

Thank you for the ideas and all the insight.

A few follow up questions that came to my mind. So if I understand it right you are suggesting two separate solutions. For the measurement of the voltage source keep the bridge just reduce the size of resistors to 1k (or 500k, which one would be better what do you think?).

Secondly, for the current measurement, move the shunt resistors to GND and still use a differential measurement between gnd and top of the resistor. The reason I originally tried avoiding this is because I am trying to measure power in a computer. I have been told that the digital chips in computers can tolerate fluctuations on the voltage rail but expect a good stable connection to ground to work correctly. Do you think placing the shunts on low side would cause problems? (Also on PCIe it is physically not possible to move the shunts to ground) EDIT: currents from all the voltage rails flow back to the power supply over the same GND cable so it is not possible to do low side measuring as I could not determine what proportion of the current came from which rail

 

Finally, could you recommend me a suitable buffer amplifier and instrumentation amplifier? Do you think it is possible to build up such a circuit on a matrix board without much trouble? Please bear in my that I am a computer science student trying to finish a dissertation project and I do not have enormous budget or industry skills, so probably the simplest way to make it work is the best even if it slightly sacrifices accuracy.

 

Thank you,

Károly

0 Kudos
Message 15 of 28
(1,552 Views)

Dear crossrulz,

 

Thank you for the answer and the suggestions. I can see that the component you linked is surface mount. Unfortunately I only have equipment to solder through hole components on matrix boards. I don't really have much experience selecting components and I am under extreme time pressure. I don't really wanna risk building up something that is not gonna work. Is there any chance you could help me choose suitable amplifiers if I tell you a bit more about my requirements? What are the different factors I need to know for picking the right components?

 

Thank you,

Károly

0 Kudos
Message 16 of 28
(1,531 Views)

@Károly wrote:

..

The input impedance of the DAQ is specified as >1G so I assumed it would not be a problem as it is orders of magnitude bigger than that of my bridge.

 


 You are correct, 1G >> 10M so it should work, yet it does not work. Why is this? 

______________________________________________________________
Have a pleasant day and be sure to learn Python for success and prosperity.
0 Kudos
Message 17 of 28
(1,513 Views)

There are two different issues goong on here. If you would have a steady input to a DMM then the 10 MOhm << 1G would be all that matters. But you also have a multiplexing ampliifier and the amplifier input is always having a capacitance (and even some inductance). Before the amplifier input can measure the real value the external signal must (dis)charge that capacitor from the voltage level of the previous multiplexed input value. Even with only 10pF the time constant gets really significant at an input impedance of 10 MOhm!!! 

Rolf Kalbermatter
My Blog
Message 18 of 28
(1,486 Views)

@rolfk wrote:

There are two different issues goong on here. If you would have a steady input to a DMM then the 10 MOhm << 1G would be all that matters. But you also have a multiplexing ampliifier and the amplifier input is always having a capacitance (and even some inductance). Before the amplifier input can measure the real value the external signal must (dis)charge that capacitor from the voltage level of the previous multiplexed input value. Even with only 10pF the time constant gets really significant at an input impedance of 10 MOhm!!! 


Don't spoil all the fun ; ) 

______________________________________________________________
Have a pleasant day and be sure to learn Python for success and prosperity.
0 Kudos
Message 19 of 28
(1,470 Views)

measure power on high speed IT stuff is another beast 😉

just because the current is usually heavely full of spikes and need a high bandwidth (high sample rate)

 

if it is just for a final measurement to compare two systems (old one and your new one) I would try to look for equipment to rent for a week or ask around in other departments if they already have a solution.

 

if it's for over all power, can you measure before the power supply?

 

you will need buffers and you can't go with a low side  power measurement (common ground). your bread board only help if you have old DIP amplifiers or go with breakout boards since the nowadays good (and available) parts are SMD.

Also the shunt connection (4 wire!) is critical (there are application notes on that toppic) and short connection to the buffer. Depending on the current the shunt heats up (low temperature coefficient needed) , a low value shunt (and yours are low, I would go for 10m or more) help in that direction, but result in very tiny voltages. tiny voltages are prone to noise (in the amp, in your wires as antennas, in all your wires and  connections due to thermoelectric voltages, ....)

 

crossrulz pointed to the TSC103IPT.  if you get that thing already soldered on a breadboard ( ask STM, maybe they have demo boards )

 

What currents do you expect to measure ? 

How fast do you need to measure (bandwidth needed)?

(the curent spikes will be higher, but think in mean power)

 

Something maybe as a first try:

You can throw some math on it,  if you replace your 10M with 1k ( even after the shunt, the additional load is known 🙂 ), make your shunts 10 to 100 times bigger (4 wire!, 1m is academic 😉 ) .    Use shielded twisted pair to connect to your DAQ (CAT5 cable is nice, shield one sided to AIGND) , use only differential measurements .

one point AIGND to powersupply GND connection (try 10k resistor) .

Finally you can add some capacitors after the shunt (a 100nF, 100uF combo) . this will mean out the current peaks. but without buffers your samplerate shouldn't be high anyway.

 

Feel free to post pictures and shematics 😄

 

 

 

 

 

 

Greetings from Germany
Henrik

LV since v3.1

“ground” is a convenient fantasy

'˙˙˙˙uıɐƃɐ lɐıp puɐ °06 ǝuoɥd ɹnoʎ uɹnʇ ǝsɐǝld 'ʎɹɐuıƃɐɯı sı pǝlɐıp ǝʌɐɥ noʎ ɹǝqɯnu ǝɥʇ'


Message 20 of 28
(1,387 Views)