Eliminating Power Line Noise in Sensitive Equipment: A Guide for Scientists

[0xDEADBEEF]

I am having noise of 120kHz(+-10kHz) and at least the first eight harmonics in my very sensitive equipment. It's not coming in from the input (it is independent from the input and I also suppress it with approximately 180dB) The frequency also seems to low for radio transmissions. So I suspect digital noise or the power line. If I find out it is the power line (230V 50Hz), what can I plug into clean it. Maybe something like a switched power supply with no transformation.

 

[mgb_phys]

I am having noise of 120kHz(+-10kHz) ... The frequency also seems to low for radio transmissions.

Possibly flyback transformer on a computer monitor - do you have any CRT in the building?

Try some ferrite cores on the power leads, 180Db is a lot to drop but you might be able to tune them for this narrow frequency band.
You could try running through a UPS - although the output waveform generator isn't necessarily all that clean - how smooth (sin wave wise) do you need the power?
If you do try a UPs you want one that is always running output from the battery, not a cheap pass-through model that just connects the input AC straight through.

 

[0xDEADBEEF]

Possibly flyback transformer on a computer monitor - do you have any CRT in the building?

We sure do.

Try some ferrite cores on the power leads,

OK, I'll try that. Can I put them around the cable, or do I need to put them around each wire?

180Db is a lot to drop but you might be able to tune them for this narrow frequency band.

I don't think I need that much, but this is what the low pass at the input is doing, so I am very sure that the input is not the source of the noise.

You could try running through a UPS - although the output waveform generator isn't necessarily all that clean - how smooth (sin wave wise) do you need the power?

Not very clean if the harmonics are not too high. I'd say anything below 1kHz will probably not make it through power supply stage of the equipment but who knows...

If you do try a UPs you want one that is always running output from the battery, not a cheap pass-through model that just connects the input AC straight through.

Any recommendations as to a brand or model? I guess the digital signal generation might produce digital noise if I am not careful.

 

[berkeman]

You have a lowpass filter with 180dB of attenuation, eh? Maybe on paper.

Your primary approach should be to figure out where the 120kHz noise is coming from. What is your overall setup? Can you run your instrument off of batteries temporarily to see if the noise goes away?

 

[0xDEADBEEF]

You have a lowpass filter with 180dB of attenuation, eh?

Or whatever a three pole factor ten cascade with 3dB@100Hz total will do... that entry point is blocked.

Maybe on paper.
Your primary approach should be to figure out where the 120kHz noise is coming from.

Well if the noise is on the power line, then I disagree. Everyone is saying "find the source", but in an institute I dislike that approach. We know in what range our signal is, I think we should treat the rest of the lab as evil and block anything that doesn't belong. What if I find out that the computer cluster one floor above is the source? How does it help? Man we found the tram 500m away as a source for one of our problems, shall we tell the city to stop using that route?
But I agree that it is useless to put on a aluminum hat and shield everything without identifying the entry point first i.e. checking the power line.

What is your overall setup?

A lock in with a lot of gain.

Can you run your instrument off of batteries temporarily to see if the noise goes away?

That might be a good idea. But since I don't have any 230V AC batteries on me currently, I'll have to check with the manufacturer how sensitive the lock in is to being opened and the voltages. On second thought: sometimes the noise is gone without a trace. Differential diagnosis is not straight forward. I think I'll check the power line first. If it carries my noise signal, that's all I need to know.

 

[mgb_phys]

You can try and remove the source. If you have a PC controlling the experiment try moving it further away or turning the monitor off, CRTs are a terrible source of 15-150Khz noise

You can try enclosing a small experiment in a cage. This sometimes takes a bit of trial and error a lot of tinfoil and some duct tape.
Also random changes to earthing can help, try earthing everything, isolating everything and earthing one item at a time. Noise reduction in physics experiments is not a science.

Do you know the noise is coming in on the power line? I would have thought somebody building an expensive lock-in amp would do a decent job of power supply filtering.
A 230V AC battery can be cheaply obtained from a car battery and an inverter.
The inverter will be noisy but it won't be the same noise - if the original 120Khz goes away you at least know where the problem is. If not then it's probably being radiated in and picked up on a lead

 

[berkeman]

Well if the noise is on the power line, then I disagree. Everyone is saying "find the source", but in an institute I dislike that approach. We know in what range our signal is, I think we should treat the rest of the lab as evil and block anything that doesn't belong. What if I find out that the computer cluster one floor above is the source? How does it help? Man we found the tram 500m away as a source for one of our problems, shall we tell the city to stop using that route?

Yeah, I meant more like if it were a local source, as mgb is saying about nearby monitors, for example. I once had a problem that turned out to be the face of the LCD display on a battery powered oscilloscope. It never occurred to me that the LCD refresh signals would have such near-field capacitive noise associated with them. Took me a good long while to track the noise down to the oscilloscope itself!

You can check for local sources of noise by using a battery powered oscilloscope and walking around and holding the probe up next to other instruments (nothing on the probe, just floating) to see if you see the 120kHz characteristic noise. A portable spectrum analyzer is better, but not many people have those (we have two). Do you have access to any spectrum analyzers that include the 120kHz frequency range?

If the noise is coming in via the power line, there are certainly AC Mains filters available that you could put in-line with the power cord connection. They are commonly used where powerline communication signals are present, and interference is caused with other devices. I'll see if I can find some representative ones to post here...

That might be a good idea. But since I don't have any 230V AC batteries on me currently, I'll have to check with the manufacturer how sensitive the lock in is to being opened and the voltages. On second thought: sometimes the noise is gone without a trace. Differential diagnosis is not straight forward. I think I'll check the power line first. If it carries my noise signal, that's all I need to know.

Yeah, given that it's a lock-in and not some simpler device, battery powering may not be an option. mgb's idea of using an inverter with a different known switching frequency has merit, however.

 

[berkeman]

A google search for powerline filter turns up lots of different styles and sizes:

http://www.google.com/search?sourceid=navclient&ie=UTF-8&rlz=1T4GGLL_enUS301US302&q=powerline+filter

.

 

[f95toli]

Well if the noise is on the power line, then I disagree. Everyone is saying "find the source", but in an institute I dislike that approach. We know in what range our signal is, I think we should treat the rest of the lab as evil and block anything that doesn't belong. What if I find out that the computer cluster one floor above is the source? How does it help? Man we found the tram 500m away as a source for one of our problems, shall we tell the city to stop using that route?[/QUOTE]

I know people who sometimes do noise-measurements at night for precisely this reason (i.e no trams or trains)...

Anyway, I am a bit surprised you are having problems with a 120kHz signal if you are using a lock-in.
I assume you are using a battery driven pre-amp before the lock-in; have you checked whether or not the noise is present at the output of that?

Removing 120kHz shouldn't be very difficult; even a standard UPS should take care of that (although I must admitt I've never had any success working with UPS; the only "gadgets" I put on the mains side of my measurement equipment are isolation transformers to prevent ground loops).

What kind of lock-in are you using? If you need really low noise you should avoid the digital ones; there are still some analogue models out there.

 

[0xDEADBEEF]

Anyway, I am a bit surprised you are having problems with a 120kHz signal if you are using a lock-in.

Well the problem is in the amplitude not the frequency, but more about that at the end of the post.

I assume you are using a battery driven pre-amp before the lock-in; have you checked whether or not the noise is present at the output of that?

I have a preamp, but I use a power supply for that one, its noise is quite strong and on another frequency but the low pass kills it. On the other hand it is an interesting thought to think of the preamp as a noise probe...

Removing 120kHz shouldn't be very difficult; even a standard UPS should take care of that (although I must admit I've never had any success working with UPS; the only "gadgets" I put on the mains side of my measurement equipment are isolation transformers to prevent ground loops).

These are the type of things I am thinking about. I know that the cheap UPS that is somewhere unused in the server room doesn't filter, and I guess I didn't use the right keywords for my search at our electronics supplier.

What kind of lock-in are you using? If you need really low noise you should avoid the digital ones; there are still some analogue models out there.

It is a digital one, and supposedly it will measure nanovolts. But being digital is a problem, because the AD likes to start clipping whenever there is a strong signal present on whatever frequency. Actually most of the problems have been taken care off, by a combination of amplification, differential signaling, paranoid shielding, and the low pass, so I can drive the lock in at full amplification. But now I see these periodic noise spikes in an fft of the input signal, and they bug me.

 

[f95toli]

You could always try using pre-amp with a built in band pass filter before the lock-in.
Even the ubiquitous Stanford Research SR560 might work (I assume you have one of those; I don't think I've ever been to measurement lab that didn't) and it can be run off external 12V batteries. I keep a couple of car batteries in my lab that I use to power the pre-amps (although I have access to some very good low-noise supplies so I rarely have to use them).

I've found that as long as I put a SR560 first in the chain and use filter in the right place (and avoid ground loops of course) I can usually even get away with using a DSP lock-in.
I mainly use Stanford Research SR830/840 (depending on frequency), sometimes a SR530 and a for low-noise measurements a Signal Recovery 5210 (which is a good instrument as long as you don't use GPIB; if you do there is all sorts of digital noise)+occasionally some older models. All of these work well as long as you understand their limitations.


I am sure you already know this; but you should nearly always use external pre-amps even if you are using instruments which is suppose to be "low noise". Even home built pre-amps (something made from a couple of op-amps and a few batteries) are usually better than connecting your sample direcly to a mains powered instrument (there are of course exception). Also, anything connected to mains should be treated as a source of noise (which they are; even something as simple as a benchtop multimeter will send OUT noise via its input) and always assume that digital instruments will send out high-frequency noise (e.g. a 10MHz clock) both on their in- and output.

Low-noise measurements is something you learn how to do gradually; to a large extent it is more of a a craft than a science and the best way to learn is to work with someone who is more experienced. You also need to learn how different instruments behave; most people who do this a lot(like me) have "favourite" instruments (like the abovementioned SR560; although I still prefer the Signal Recovery 5113 simply because it is the model I used when I was a PhD student); that they tend to use over and over again simply because they've learned how to get the best performance out of them (and whey they move to new place they buy the same type of instruments for the new experiment, which is why some models have been around for a very long time).

 

[0xDEADBEEF]

Great suggestions. This will keep my money burn rate up. Much of the know how in the faculty has been lost, and I have source from other places, this helps a lot.

 

[0xDEADBEEF]

Btw I have never seen car batteries in a lab. Does anyone else use them?

 

[mgb_phys]

Btw I have never seen car batteries in a lab. Does anyone else use them?

Yes - until somebody makes a 12V power supply with microVolt noise levels that will run an experiment for a week and costs $30.

A car battery and a laptop is also a great way of avoiding ground loops on measuring instruments.
They are also used in a lot of high energy labs because it's easier and safer than trying to get a line voltage at several KV above ground.

 

[berkeman]

Either a car battery, or a sealed lead acid gel cell:

http://www.batterystation.com/gelcell.htm

We use them as mgb points out, and we also use them for portable testing, especially RF device testing.

 

[mgb_phys]

Yes we normally use the 12V 6Ah backup batteries from alarm systems.

 

[berkeman]

Yes we normally use the 12V 6Ah backup batteries from alarm systems.

The alarm systems don't work very well anymore, but the experiments in the labs work gangbusters! J/K :rofl:

 

[0xDEADBEEF]

the lock in is a long wave transmitter, the power line was not the problem.

Yay listen to radio lock in everyone.

 

[f95toli]

The following text is taken from the manual to one of the resistance bridges I use for thermometry (an AVS-47B, excellent instrument).

When we developed the AVS-46, our older bridge model, we were puzzled by a non-expected amount of noise in the preamplifier. Sitting in front of an oscilloscope
for hours, we tried to discover, what we had done wrong. The noise did not look normal: its amplitude behaved somehow differently than what we had seen previously. Then we got an idea to try to listen the noise instead of looking it with the scope. We connected some old earphones to the signal - and to our surprise, it was a short-wave broadcast from London!

Btw, the manual in question is actually very good and contains a lot of information about low-noise techniques in general (grounding etc). It might be well worth a look even if you are not interested in mK-thermometry.
It can be downloaded from Picowatts homepage

http://www.picowatt.fi/index1.html

My experiments are slightly more modern than yours, I tend to pick up TV-broadcasts...