Multifunction DAQ

cancel
Showing results for 
Search instead for 
Did you mean: 

Measuring Noise of 24 volt power supply

Solved!
Go to solution

I am trying to use a NI USB-6002 DAQ to measure the noise produced by a number of different 24 volt power supplies. The NI USB-6002 DAQ is only able to measure about +10 to -10 volts so I am thinking I need to use a voltage divider. I setup a voltage divider using four 5.1 MΩ resistors but I was getting a LOT of noise.

 

I simplified the circuit to use a 9 volt battery and only one resistor (as shown in the image below) and I was still getting about 550 mv difference between the maximum and minimum voltages over a 1 second period (i.e. noise). I am sampling 1000 times per second and have tried using RSE (single ended) and Differential modes.

 

 

image.png

 

If the resistor is removed, there is only 1 or 2 mv of noise.

 

We noted from the DAQ's datasheet that the Input impedance is >1 GΩ but I don't really know what that means.

 

Can you please let me know what I am missing? I don't think I should be getting practically any noise from a battery and a resistor. My end goal is to measure the noise of a number of 24 volt power supplies.

 

0 Kudos
Message 1 of 10
(4,898 Views)

Usually you choose AC input to measure noise.

Try something like this:

PSnoise.png

Diodes are 1n4148 ..  and limit the input noise  to ~ +-2V 

R2 limits inrush current

R1 Bias resistor (pull DC to Zero)

C cuts off the DC bias, voltage rating according to voltages monitored...

 

It's a high pass filter, removing the DC supply voltage .. cut off frequency is your homework 😉

 

 

For serious source noise measurements you should shield everything and twist the power supply conneting cables .... cables are antennas... 

If you connect a 9V battery you should read about the same as with a short cut ... the noise of the DAQ 😄

 

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'


0 Kudos
Message 2 of 10
(4,870 Views)

Henrik,

Maybe I am doing it the wrong way, but I think I want to measure both the AC and random change in voltage over say 1 second in order to relate that directly to a signal to noise ratio without filtering the signal. Since I want to measure the signal to noise ratio at a 24 volt level and only have a DAQ that can measure up to 10 volts, I need to reduce the voltage without changing the signal to noise ratio or shorting out the power supply.

0 Kudos
Message 3 of 10
(4,859 Views)

I'm still having difficulties with this. For now I am just trying to measure the voltage of a 9 volt battery with a resistor in the circuit. My goal is to get this working with 2 mv of noise or less. Then I will move on to the power supplies.

 

Any ideas?

0 Kudos
Message 4 of 10
(4,803 Views)

I would try something like this:

input divider.png

R1 to R3 should be equal value in a range of 100k to 470k , this will give you a 3:1 divider ...

R4 provide a I_bias path

C1 is a trim able capacitor  to compensate the 6002 input capacity (I didn't found in the spec 😕 ) the value range is a guess...

you can try other ones .. you will see if you need more or less capacity while you test the trimming.

The procedure is like the one for a 10:1 scope probe: Apply a square wave (~1kHz) and trim for a nice square wave captured with the 6002.

 

V1 is your DUT 😉

Use differential voltage measurement, see manual

Your approach with a single high value resistor is prone to noise 😉

 

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'


0 Kudos
Message 5 of 10
(4,796 Views)

I am using 220k for R1, R2 and R3. R4 is 500k.

When I use the circuit you suggested, I get about ~240 mv noise at best with the capacitor removed. Increasing the capacitance increases the noise. However, if I remove the bias resistor and connect the DAQ grounds together, the noise drops to 50 mv without the capacitor. In this case, adding capacitance decrees the noise to 28 mv (I'm not sure exactly what capacitance value I am using but it is less than 100 pF). So I think I just need more capacitance.

 

I increased the sample rate to 12500 Hz from 1000 Hz and noticed that the noise I am dealing with is at 1000 Hz.

 

When I switch to a 24 volt power supply, will I need to change any of the resistor or capacitor values?

0 Kudos
Message 6 of 10
(4,777 Views)
Solution
Accepted by topic author JonathanMATech

240mV is a lot of noise....

How noisy is your lab?

Differential setup?

 

Some things to try:

differential setup and 50kSPS samplerate

make a shortcut as close as possible (20mm cable) between ai+ ai- and 220k to AGND.

Measure (capture 300ms of data .rmsnoise? FFT?)

replace the 20mm wire with a 60cm wire,  as a loop and twisted,  in both cases measure

buildup the voltage divider (small area, shielded (I like to recycle tinplate) )

shortcut the divider with 20mm and 60cm .. measure

 

A word to the capacitor : Since you want to measure noise you need some bandwidth (about 20kHz) , the 6002 has some input capactance, so together with the R1,R2 (440k)  you build a RC lowpass filter. The C1 build a high frequency path to compensate that RC filter. Apply a square wave at the input and see how the corners of the square are influenced by the C1 value. .. Basic EE 101 😄

 However if you only need a samplerate of 1kHs-10kHz ...ignore C1 ...   for 1kHz samplerate you can even put a 470pF capacitor between  AI+ and AI-  (assuming 220k for R1,R2) 😄

 

And no, 220k for R1-3 will load your 24V power supply with less than 40µA ... it's fine.

 

 

 

 

 

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'


0 Kudos
Message 7 of 10
(4,758 Views)

I added the 470pF capacitor between AI+ and AI- which worked for me. However, the bias resistor between AI- and AI GND made things worse still.

 

Thank you for your patience with my limited electrical knowledge Henrik.

0 Kudos
Message 8 of 10
(4,717 Views)

Hello Jonathan,

sometimes i am confused about the specs in those datasheets.

GOhm impedance sounds so nice and so promising for any measurement application.

But this is not the whole truth. Still there is a input capacitance in parallel and/or input bias current.

 

So in practice a settling time applies, mentioned in the manual

USER GUIDE NI USB-6001/6002/6003 page 18:

 

Multichannel Scanning Considerations
The NI DAQ device can scan multiple channels at high rates and digitize the signals accurately.
However, you should consider several issues when designing your measurement system to
ensure the high accuracy of your measurements.
• Use Low Impedance Sources—To ensure fast settling times, your signal sources should
have an impedance of <1 kΩ. Large source impedances increase the settling time of the
NI DAQ device and decrease the accuracy at fast scanning rates. For more information
about decreasing the source impedance of an analog signal, go to ni.com/info and enter
the Info Code rdbbis.
• Use Short High-Quality Cabling—Using short high-quality cables can minimize several
effects that degrade accuracy including crosstalk, transmission line effects, and noise. The
capacitance of the cable also can increase the settling time.
• Avoid Scanning Faster Than Necessary—Design your system to scan at slower speeds
to give the NI DAQ device more time to settle to a more accurate level when switching
between channels.

 

 

I am also into a measurement tasks using NI-DAQ 6259. My first experiences lead to the conclusion, that measuring voltage above the +/-11Volt require voltage dividers using precision resistors 0.1% tolerance.

 

- At first i was using a relation 5:1 by using 102 kOhm to 25.5 kOhm to be able to measure volage up to +/-50 VDC.

this worked up to sample rates of 10 kHz, then the measured voltage got somehow inaccurate, up to 1 volt error occured.

this was caused due to to too high values for those resistors leading to bad settling times. i didnt know, that those analog input channels are multiplexed to just one A/D-converter. And each time the DAQ is sampling those input channels, the input amplifier of the card has to adjust to another input voltage. A low input impedance is required to sample accurately at high sample rates above 10KHz.

 

- Thats why i replaced my input voltage devider to 1020 Ohm to 255 Ohm resistors, to achieve the same 5:1 ratio. This solved the problem of inaccurate measurement values at high sampling rates but lead to hot resistors, bec they only have 0.25 Watts each.

 

- Finally i found a good compromise using four of those 1020 Ohm precision resistors in series to one 1020 Ohm resistor (4 + 1 = 5 x 0.1% tolerance, 0.25 Watts) , which gave me again a ratio of 5:1, while they are not overheating and still precise enough (still much better than 1% resistors. This works fine. 

 

This way you could measure your 24 V power supplies at high sample rates without damaging the DAQ-card.

But it must be mentioned, that the power supply has a little more load and a leakage current of 5 mA will flow thru the voltage divider resistors.

I dont know if this is a problem for you.

 

Message 9 of 10
(4,452 Views)

Stupid question, but what point is there in measuring just the open circuit noise from the PSU? Surely you'd want to measure the noise under load, preferably under different load conditions, e.g. 0%, 10 %, 90%, 100%, 110% of rated current?

 

Also, you are assuming that the noise is generated in the PSU; however a major figure-of-merit for PSUs is the ability to filter out line noise from the input. I would suggest performing some kind of monitoring of the input as well, but be careful of mains voltages. You will need input fuses, attenuators, isolation amplifiers & a Professional Engineer to check the wiring.

 

Another thought:If this is a wide range PSU, will you get different answers in different parts of the input voltage range?

 

I think you are a bit limited with the USB6002; I am not sure that 50 kHz will be enough to resolve some of the nasties that may be on the incoming mains (e.g. noise from inverter drives, DOL induction motor on/off switching), or from the switching of the regulator in the PSU itself.

 

0 Kudos
Message 10 of 10
(4,371 Views)