Noise measurement?

[ftransform]

So, I wanted to get a feel of the noise present on my bread boarded resistive sensor analog front end. I have available to me a rigol 1052e and a quality 6.5 digit multimeter.
The signal is about 0.5V, the rigol was claiming that the noise on this signal was 20mV. The Agilent meter in AC mode showed the noise to be <0.01 mV.

If this is just bread board noise then why is the agilent meter so much lower? Is this because the scope bandwidth is much higher?
The power supply for the sensor is a 0.5mA constant current source (bootstrap power supply made with 2 opamps and a 20hz filter) powered by Agilent power supplies. There is quite a bit of wires on the breadboard (the leads to the sensor are over a foot) but is it reasonable for the noise difference between the oscilloscope and the multimeter to be so great?

Also, does anyone know the vertical noise specifications for the rigol 1052e?

Measuring a 9V battery gives me a 10mV pp noise with the rigol when the ground clip is connected. If I use the little spring it drops to like 3-4 mV (hard to read).

[Simon]

The Rigol has a fair bit of noise of it's own, much of it picked up from the scopes PSU.

[AndyC_772]

The scope is measuring noise over a much wider bandwidth, so it will inherently read a much higher value. In any case, though, an oscilloscope isn't really a suitable instrument for accurately measuring the amplitude of a very low level signal - the noise of the scope's front end and the quantisation noise of the ADC will dominate.

[alm]

The DMM has a bandwidth of maybe 100 kHz, the scope 50 MHz. The bandwidth is about 500 times higher, so as first approximation (eg. white noise), you might expect the noise to be ~25 times higher (sqrt(500)). The exact figure of course depends on the power spectral density of the noise (it's unlikely to be flat up to 50 MHz). Noise is usually quoted for a certain bandwidth, eg. 1 MHz. If it's a temperature sensor, for example, you're unlikely to care about 10 MHz signals. You could add a low pass filter between scope and DUT, or even in the DUT itself.

The scope will also have a fair amount of inherent noise. To estimate it, you could short the probe at the tip (don't change the loop area between probe tip and ground lead) and measure the noise. The apparent noise (as shown on the scope) is the root of the sum of squares of the scope's noise and real DUT noise, both as RMS voltages. This assumes that these two noise sources are uncorrelated.

[Simon]

you can try on the rigol putting a steel shielding plate over the mainboard (taking care it still cools), i had some success with aluminium covered card.

[lewis]

Build a low noise, high gain noise preamplifier from a high performance op-amp to get accurate results over a specified bandwidth. This swamps the noise floor of the HP meter and Rigol giving you much better accuracy (assuming you need it).

Have a look at one of Jim Williams's app notes here: http://cds.linear.com/docs/Application%20Note/an124f.pdf

[alm]

Yes, a measurement preamplifier would be the next step, although in the case of DIY this moves the problem of building a low noise sensor amplifier to building a low noise measurement preamplifier .

A design for a measurement preamp, based on a Jim Williams design, is available here. I believe he also sells PCBs for the circuit.

[ftransform]

How about taking the rigol apart, putting it in a wooden box with a linear power supply?

Right now I'm debating weather to put a active filter between the differential amplifier and the ADC. Unfortunately just a capacitor of any substantial size just makes it ring.

I also have a tough time understanding some of the ADC"s characteristics. I assume the ADC will filter out a great deal of high frequency noise by itself right? What would this be under in the

3db bandwidth of my ADC is 600KHz, meaning that the agilent bench meters AC range measurement would be accurate it multiplied by 2? (the datasheet says the bandwidth of the Agilent meter measurement is 300KHz)

I might build that board it looks interesting..

[alm]

What's a wooden box supposed to accomplish?

[Simon]

the wooden box won't help, but a linear supply is not a bad idea, I'd try screening the main board first. (re calibrate the scope afterwards)

[ftransform]

the wooden box accomplishes a home brew feel. I should have clarified that the box was chosen for aesthetics.



I have found that a 300pF capacitor cleans up the signal nicely but does not cause out of control oscillation. I may try to include a 20hz LPF instead to see the difference.

[lewis]

So, I wanted to get a feel of the noise present on my bread boarded resistive sensor analog front end.

What sort of sensor is it, and what is your application? Out of curiosity...

[ftransform]

It is just a RTD stimulated by 0.5mA.
All of this is probably overkill but I think it is a learning experience.
With the 300pF capacitor on the output of a unity gain differential amplifier according to the rigol my noise is now 10mVPP.
Whats kind of noise is reasonable for a breadboard?

[robrenz]

A differential measurement will also give much better results. Page 101 of this excellent HP handbook gives details of how to.  I have a Tek 7603 scope that I use with either a 7A22 or  7A13 differential vertical amplifers. The 7603 is analog but I can use the signal out connection of the scope to feed my Rigol. this is a very inexpensive way to get a very high performance differential probe for your digital scope. Yes it is a big unit but where else can you get 10µV/div differential probe for a couple hundred bucks? The 7A22 has separate upper and lower tuneable bandwidth filters to set what bandwidth you want to see.

[ftransform]

This is interesting... If I use my rigol's low pass filter I get virtually no noise if I put the cutoff to 20khZ.
However if I put my low pass filter to 500KHz I get the full noise.

What is causing noise between 20 and 500KHz?

I don't have a military submarine radio transmitter in my house......
And I know its not the mains because the signal was CLEAN at 20khz...

but according to the rigols FFT the noise peaks are at 1, 4 and 7 KHz.

What the fuck,.

[alm]

You can use FFT (there's a PC program floating around on this forum that can achieve a much better resolution by using the full record length) to get an idea of the frequency spectrum. Since it's beyond the audio range, it sounds like a switching regulator, for example a switching power supply or inverter for an LCD backlight. Could come from overhead fluorescent lights with an electronic ballast picked up by the long wires, or from something connected to your DUT. Try turning off all other equipment in the room (including lights). Does it still occur with the scope input shorted?

[ftransform]

I can't really get a FFT reading from the inputs shorted and turning off the lights/monitor does not help.

[ftransform]

What should the frequency scale be set to for measuring a FFT of a DC voltage?
1 Second?

[alm]

There is apparently a signal of interest somewhere between 20 kHz and 500 kHz, so I would set it so several periods of 500 kHz fit on screen. It's a trade-off between low frequency and high frequency resolution. DC itself is not very interesting on an FFT.

[SeanB]

Your computer power supply, the CFL lamps in the house and every switch mode power supply in the neighbourhood will make the noise. All coupled nicely into an antenna made from the mains wiring. If anybody has PLT equipment these is that as the dominant noise source from basically 100kHz to 1GHz.

[ftransform]

Well, what can I do about this voltage?

I can't put adequate filtering behind my differential amplifier because anything over 330 pF causes it to ring. I tried putting a voltage follower behind it with a capacitor but the large value of capacitance required caused an unacceptably high offset voltage.

What options do I have to deal with this problem?

Thoughts I had are:
1) Use the voltage follower + large capacitor and try to characterize the offset and account for it with a micro controller
2) Could an optical isolator be placed after my differential amplifier? I am unfamilar with these devices, can they be used to transfer a signal accurate to micro volts over?
3) Put it on a PCB in a shielded box and hope for the best
4) Find a differential amplifier that can handle a large capacitive load?
5) Filter the input signal better before it reaches the difference amplifier?
              -Right now I just have a 1 uF tantalum + 100nf ceramic capacitor between the inputs of the differential amplifier.
6) Is it possible to modify a unity gain differential amplifier IC to act as a filter without using another opamp?
7)If I simulate a capacitor using a general impedance converter will it still cause a op-amp to ring?
8)what other filter topologies should I consider?

[Simon]

Screen the rigols main board that will help, the noise is made by anything switching and remember that the screen is low resolution so if it can't quite make its mind up you get at least the 0 line +/- 1 pixel of error, thats 3 pixels already that is near legitimate.

[ftransform]

this is making me sad the temp keeps jumping around by 1c under a glass bowl

fucking noise.

I will try a delta sigma ADC maybe...

[ftransform]

ITS IMMUNE TO MY CAPACITOR SPAM!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

So I guess I am asking, how do I create a low pass filter with a low capacitance OR how do I prevent my differential amplifier from ringing.

[Simon]

maybe you should post a schematic and photo of your lashup