Very low frequency noise problem

Hello, I've designed a device which can measure very low current(in nA range with 1M Ohm resistor) using a 16 bit ADC. The result data has an annoying noise with very low frequency about 6-10 milli Hz. I absolutely have no idea where it comes from. When I use upper ranges that have lower current measurement resistors(for example 100 Ohm) this problem disappears completely.
Can someone give me some suggestion how should I find source of this low frequency noise.
Thanks alot,
Regards.

i dont full understand what you mean with "upper ranges that have lower current measurement resistors" .. you mean higher frequencies?

if you look at the complete noise profile of your device, you'll see that lower frequencies (before the corner frequency) are dominated by flicker noise, and the higher frequencies are dominated by thermal noise. the psd of the noise is proportional to the resistance too. to higher resistor means higher noise.

you might want to consider some method of removing the noise, maybe by autozeroing or chopping.

Surely these low frequency noise comes from two reason:
1: 1/f noise
2: envromental noise(temperature drift, vibration, etc)
you need a high pass filter to filter this low frequency noise

Thanks alot xzp and rush 3k, The 1M resistor is paralleled by about 20 relays which I think may have a stray capacitance, can stray cap of these relays and the current measurement resistor be source of this error. for eliminating this noise HPF seems applicable but there are occasions that I need to measure DC current, then I guess HPF won't be the solution. xzp I read some of your previous posting and I thaught you may have experiences with similar problem. Can I do anything else apart from HPF to lower this noise.
Thanks

I think if you want to measure DC current, A capacitor in the feedback will help much. you can easily acquire by integral of this(set the detection period long enough to avarage the low frequency noise). you can read Bob Pease's letter on national's website, it will help you much. surelly the 20 relays will introduce more noise because of the contaction of each relay will bring the contact noise(1/f behavior), furhtermore, you should keep in mind that thermoelectric effect(the joint of the relay, as an example, the Cu-Pb/sn joint give 1-3uV/degree, and our room temperature will fluctuate 1 degree in tens of minutes). at low frequency, clean, dry your device. keep your detection system in good shielding and ground.
Good luck

fala said:

Hello, I've designed a device which can measure very low current(in nA range with 1M Ohm resistor) using a 16 bit ADC. The result data has an annoying noise with very low frequency about 6-10 milli Hz. I absolutely have no idea where it comes from. When I use upper ranges that have lower current measurement resistors(for example 100 Ohm) this problem disappears completely.
Can someone give me some suggestion how should I find source of this low frequency noise.
Thanks alot,
Regards.

Click to expand...

Fala: I like to have a look at your circuit diagram if possible. If it is copyright, perhaps you can send it to my email? If not, just post it here. It makes solving a problem a lot easier then all the guessing about your circuit....

If your node impedance is 1M, you'd better be shielding those relays, and bootstrapping if possible. Also, thermal EMFs can cause a lot of problems -- make certain that in addition to providing a stable temperature, you prevent air currents from blowing across the circuit -- this will show up as 1/f^n where n is something bigger than 1. (can't remember the typical range)
As xzp mentions, make sure your circuit is clean & dry (bake in an oven!) to prevent noise associated with surface leakage paths.
What ADC are you using?

Hello, xzp: Bob Pease was wonderful, it helped alot thanks. lladnar23: I'm shortening relay contacts as suggested by bob pease, is that enough? about air flow I had no idea about it can you introduce areference for that please? I don't understand how baking can help, because obviously after the board comes out of oven it will absord moisture so what's the benefit? do you mean I should use silica gel?
thanks.

Fala,

So typically I will wash & bake a board in the lab for testing to ensure that I don't run into any complications due to leakage, but if the design is sensitive to moisture & contamination, you would have to take precautions to ensure the correct operation on a long term basis. In the short term (several hours to a day), FR4 doesn't absorb moisture fast enough that you would have to worry about re-baking the circuit very often.
For the long term, one method is to use different materials - Teflon (PTFE) for example, is sometimes used instead of FR4 for the PCB. Also, if only a few nodes are sensitive to leakage, sometimes those nodes will be connected to a teflon standoff to eliminate contact with the rest of the pcb. You can see examples of this in commercial designs such as picoammeters, sample & holds, etc...
Another technique is to use conformal coating to protect the circuit from contamination/moisture.

As far as the relays go, I guess it really depends on the particular construction. I've had problems in the past with capacitive coupling into high impedance nodes, but upon second consideration, at 1M, it may not be as much of a problem (I was working with nodes requiring a leakage of <100fA which is much higher than 1M) If you're using reed relays with short leads, then the stray capacitance is probably OK.

I'll try to dig up some references for you on all of this...

The humidity envoromental will also deteriote the detection performance at low frequency,mainly due to the cell effect and leackage.
Baking help to remove the chemical dust and the mosture on the device surface.
at that low frequency measurement, you need to set up a damping system, etc.
air flow mainly change the temperature and humidity, and the vibration.
Radiation noise will also need to be considered (equilibrium in temperature will also induce noise due to the equilibrium thermal exchange, we call this radiation noise, you can read some paper on pyroelectric device)
I think you should read the fundamental book before you do the experiment <Noise reduction technique in electronic system>, it will help much.
Of course, The handbook provide by keithley on low level measurement also help you.
But if you have time, yo ucan read <Low noise electronic system design>, it is a classic book on low noise.
But I think the most effective way for the DC measurement is by using Lock-in method. you can read paper on fluxgate which is used to measure DC magnetic field. magnetic field also related to the DC current by amper's law.
Post your circuitry and device if possible(hehe, I think i may be in millitary usage, cause I don't think commercial usage will not take that low frequency)
Good luck

As xzp mentioned, the Keithley low level measurements handbook is a good read, and is available from their website.
Additionally, there is good information to be found in app notes on voltage references. Jim Williams makes some good points in LTC AN86, "A Standards Lab Grade 20-Bit DAC with 0.1ppm/c Drift"
Also read LTC AN82, "Understanding & Applying Voltage References".

Typically I do low noise experiments in a grounded cookie tin to serve as a faraday cage. I power the circuit with batteries (inside the tin) to prevent coupling of noise through the power connection, and to prevent ground loops. I also tape a layer of foam to the outside of the cookie tin to prevent temperature changes from ambient air currents, and I put the tin on a padded surface to prevent vibration and shock from coupling into the system. I wouldn't worry too much about magnetic fields as long as you don't have a CRT or a big transformer running right next to your setup.

thanks xzp & lladnar23, xzp: I definitely need more study, I’m not being allowed to publish the schematics. I'm too junior to design for military but I like to read your comment . Though I hate war I am always fascinated by high tech military stuff. especially advanced radars but the device is for scientific use. lladnar23 what a good idea to use cookie tin as faraday cage! thanks.