Blocking Out 100Hz EMI in the Lab

  • Thread starter gareth
  • Start date
  • Tags
    Noise
In summary, Gareth's problem is that he needs to record a signal at the same time as he is recording noise from an ion source. Every time he wants to record an ion burst, he has to record a burst of noise as well. He has tried various solutions, but has not been successful. He is considering changing the recording frequency of the ion source, but is worried that this will also cause the noise to occur at the same time. He is also considering trying to block out the noise with tin foil and sheetes of metal, but
  • #1
gareth
189
0
Hi,

I am using a simple circuit in the lab which detects current flow from an ion source. These are transient signals recorded at a rate of 100Hz. The problem is, the ion source is switching at the same freqeuncy at very high voltages. Needless to say there is a fair bit of 100Hz EMI. So every time I want to record one of these ion bursts I have to record a burst of noise also.

Here is what I'v done;

The signal is carried through coaxial cables, the shielding being ground. I have insulated the outside of the wires with copper tape which has helped immensly, but as soon as I plug in the voltage supply to the circuit (this is needed for the experiment) I get a big wack of noise again at 100Hz.

So is there an easy way of blocking out this noise? I've being messing around with tin foil and sheetes of metal but have not had great success.

Any experience here?

Many thanks
Gareth
 
Engineering news on Phys.org
  • #2
try changing the recording frequency if possible
 
  • #3
i've considered this, but I can't get away from the fact that the noise occurs at the precise time that I need to record the signal. I can't delay either the occurence of the event or the recording of the event.

thanks for the ideas though

any other ways around this?
 
  • #4
Copy exactly the same detector circuit connected to the same power supply but not to the ion source - subtract the readings?

Can you connect disconnect the actual ion source while still firing the power supply and do alternate noise and noise+signal readings and do a double correlated sampling approach?

Finding the best earthing scheme, adding extra sheilding, extra Earth's or removing locla Earth's is a matter of trial and error - there is a very good book called "Noise reduction in electronic systems" by Ott. According to Amazon it's still available.
 
Last edited:
  • #5
I've considered the post processing of the signals, but there it important information occurring during the noise which I don't want to play around with. The noise also varies from signal to signal.

I think I have access to this book and I will check it out, but I fear I might have tried most of it already!

I'll let you know the results
Thanks
Gareth
 
  • #6
Can you post a diagram of the system, including the grounding and shielding scheme? Can you post some typical data so we can see what the signal and noise look like?
 
  • #7
here is a simple schematic of the cct,

the shielding is connected to ground throughout the cct.

The ion source just acts as a current source in an otherwise open cct, this is shown as 'probe' in the cct diagram.
 

Attachments

  • probe cct.pdf
    15.8 KB · Views: 179
  • #8
Unfortunately, I cannot post any data, but the noise consists of a burst of spikes going from around -2 to +2 millivolts with a rise and fall of roughly a couple of ns. It lasts a couple of hundred ns in total.
 
  • #9
gareth said:
Unfortunately, I cannot post any data, but the noise consists of a burst of spikes going from around -2 to +2 millivolts with a rise and fall of roughly a couple of ns. It lasts a couple of hundred ns in total.

What is the bandwidth of the signal that you want to separate from the noise? What is the source impedance of the ion current source detector? Why do you show a series 50 Ohm resistor in each piece of coax in your model?
 
  • #10
The signals of the ion source range from several microseconds to around 100ns in duration, the source impedance would be negligable.

I placed the 50Ohm resistor in there to model the impedance of the co-ax, but I'm open to correction there.
 
  • #11
okay...i can see your representation of both coax cables and your circuit in the middle...

what bugs me is that your cicuit looks all passive...is this correct?

however, i can see your dilmemna...your source only produces a signal on the power supply's pulse...
...is your ion source an ion chamber? if so then you would need active amplification to get anything decent out of it...

otherwise...disconnect the ion source and then with the circuit going thru a load resistor, examine the noise...
have you grounded the ion sources case to your cables? it may be picking up the power noise as well...assuming it hasnt got a good grounding scheme inside...
ensure your circuit box is also foil wrapped and grounded...

is the power source via external cables to the ion source? try ensuring that they too are covered and grounded...otherwise they act like aerials.
 
Last edited:
  • #12
yes the circuit is passive, usually there is enough current for a reasonable signal to be detected.

The cct box is a metal box and is grounded ( ia also covered this with Cu tape to be sure) the noise is greatly attenuated with all the Cu tape, but when I plug in the 25V DC power supply it seems to pick up most of the noise. I suppose I need to shield this but Cu tap is a bit impratical here, would kitchen foil work?

As I said the source of the noise is a high voltage switching unit which is around the size of a big PC tower. I am considering shielding this, but what is the best way to go about it? Cu tape is a no go as it is quite large. Foil? But I read in that book that was suggested to me that the cables running out of the shielded source will carry the noise also in their ground lines, so the ground lines need to be attached to the shield.

Any ideas?

And thanks for the posts!
 
  • #13
Ok, I think I've traced the bulk of the noise to the 25V power supply. This is a fairly standard DC power supply which seems to be picking up a lot of noise. Any suggestions on how to remedy this?
 
  • #14
gareth said:
Ok, I think I've traced the bulk of the noise to the 25V power supply. This is a fairly standard DC power supply which seems to be picking up a lot of noise. Any suggestions on how to remedy this?

See if you can put a capacitor across the output of the power supply. Try high and really low values such as 100 uf to 100 pf. Threading an EMI core or two through the leads of the power supply might help too.

If that doesn't work see if you can replace the power supply just in case.
 
  • #15
gareth said:
The signals of the ion source range from several microseconds to around 100ns in duration, the source impedance would be negligable.

I placed the 50Ohm resistor in there to model the impedance of the co-ax, but I'm open to correction there.

It sounds like the noise is a bit higher frequency than your signals, so you can at least get a few dB of attenuation with a lowpass filter on the signal.

As for shielding, no, kitchen aluminum foil will not help much, because you cannot make reliable electrical contact between the pieces, especially not for sealing up all the seams. Even with Cu tape, you need to solder the seams shut to ensure good Faraday shielding. And how you ground the shield varies with the application, and can make a really big difference in shielding performance.

The coax model that you show really isn't correct. The 50 Ohm Zo of coax comes from its inductance and capacitance per unit length. There is a small series resistance from the resistance of the wire, and a small parallel leakage conductance from dielectric loss.

The best ways to get rid of the noise depends on the situation, and on the noise coupling mechanism. If it's electric field noise, then Faraday shielding is best. If it's magnetic field noise coupling, then magnetic shielding and orientation are the best approaches. If it's conducted noise coupling, then filtering of the cables as suggested with ferrites can help a lot.

Can you post a picture or two, showing the apparatus and setup?

In the end, the best approach could be to convert the ion measuring device to battery power, add a preamp and filter to the detector, and transfer the waveform to the measuring device via fiber optics. In high noise enviromnents, that approach sometimes has to be used.
 
  • #16
Switch mode psus have no business being anywhere near lab equipement, however well you filter/shield them. We had one guy who went to the lengths of rebuilding the PC with linear supplies - although he was insane!
There is a lot to be said for running preamps from batteries, a 12v car battery will run most equipement for a long time and you would have a hard job building a cleaner supply!
 
  • #17
waht said:
See if you can put a capacitor across the output of the power supply. Try high and really low values such as 100 uf to 100 pf. Threading an EMI core or two through the leads of the power supply might help too.

If that doesn't work see if you can replace the power supply just in case.

I tried this, the capacitor makes little or no difference, and I have changed the power suply but this didn't imrpove matters. The power cable leads are coax and the shield is grounded, there is also Cu tape around them.


"The coax model that you show really isn't correct. The 50 Ohm Zo of coax comes from its inductance and capacitance per unit length. There is a small series resistance from the resistance of the wire, and a small parallel leakage conductance from dielectric loss."

OK, so the correct thing to model would be either the capacitor and inductor together, or the 50 Ohm resistor, but not both?

The orientation definatley effects the noise signal, i.e when I change the position of the cables cct or PSU the noise signal changes. Does this mean it's magnetic noise coupling?

Yes a DC battery solution would be nice, but unfortunatley I have to vary the voltage from around 0-30V in small increments (0.25V).

Thanks
G
 
  • #18
gareth said:
"The coax model that you show really isn't correct. The 50 Ohm Zo of coax comes from its inductance and capacitance per unit length. There is a small series resistance from the resistance of the wire, and a small parallel leakage conductance from dielectric loss."

OK, so the correct thing to model would be either the capacitor and inductor together, or the 50 Ohm resistor, but not both?

Look at the RLC model partway down this page:

http://en.wikipedia.org/wiki/Transmission_line

You model your piece of coax as that kind of transmission line. If it's not very long, then just use the lumped values for R, L, and C (you can measure these with an impedance analyzer). If the coax is any significant fraction of a wavelength long at your signal frequencies, you will need to consider termination strategies. That is why your input signal may need to be buffered before being driven down the coax, depending on the output impedance of the sensor. I also did not understand when you said that this Zout was "negligible". Does that mean very low Zout? (In which case you can use a back-termination scheme) Or does that mean a very high Zout? (In which case you will need to use a preamp to drive the coax)

gareth said:
The orientation definatley effects the noise signal, i.e when I change the position of the cables cct or PSU the noise signal changes. Does this mean it's magnetic noise coupling?

If it's strictly orientation that changes the pickup, then yes, suspect B-field coupling. If changing the orientation also changes the parasitic capacitive coupling of the noise (between cables, for example), then the noise coupling can still be E-field in nature. One way to tell the difference is to take medium power inductor in the 1mH range that is made as an open slug, and hold it with your oscilloscope probe and ground clip. If you can see noise picked up by the slug, and that noise changes amplitude as you rotate the slug around in the air (especially if you can null out the noise at some orientation), the B-field noise is present.
 
  • #19
your power supple must be one of two types...
type one being a switcher which is producing the higher frequency noise.
or type two being the old type transformer with bridge and transistor type.

waht's suggestion of sticking a cap across the power lead is good in that you smooth the ripples but if there is noise only on one line, you are going to get it on both...

berkemans suggestion is where i would have headed also...noise is a terrible thing and one way of defeating it after having done all the shielding reasonably possible is to go active and use a differential amplifier on the front end...the common mode should wipe the noise down a bit, allowing you to see more of your signal...

however, first port of call is to try berks idea of a passive rc filter to knock that noise...
you can either gauge your values to suit the noise and take it from the signal line to ground directly or gauge it to the signal frequencies and do a serial throughput on your signal line...
 
  • #20
berkeman said:
Look at the RLC model partway down this page:

http://en.wikipedia.org/wiki/Transmission_line

You model your piece of coax as that kind of transmission line. If it's not very long, then just use the lumped values for R, L, and C (you can measure these with an impedance analyzer). If the coax is any significant fraction of a wavelength long at your signal frequencies, you will need to consider termination strategies. That is why your input signal may need to be buffered before being driven down the coax, depending on the output impedance of the sensor. I also did not understand when you said that this Zout was "negligible". Does that mean very low Zout? (In which case you can use a back-termination scheme) Or does that mean a very high Zout? (In which case you will need to use a preamp to drive the coax)

I'm having trouble here; in this case R is the resistance per unit length, rigtht? So if the line is short this can be neglected. Ok, then where is the impedance, and why is the resistance on a regular RG58 Co-ax 50 Ohms?

I read another thread about this and it said that the ratio of C/L (per unit length) determines the impedance. So what is the correct model? I would have to assume that I was correct in what I said earlier, that I should just leave out the 50 Ohm resistor in this case.

As for the Zout, the probe is a piece of wire, nothing more, so I assumed it's impedance to an alternating signal would be very little, so in my setup the hole thing is terminated to a 50 Ohm resistor on the scope, to account for the 50 Ohm impedance transmission lines. So if the impedance of the probe was high let's say, I would have to terminate the scheme with an equal amount of impedance to prevent reflections. Does this sound reasonable?

Moving onto the noise; so could I use one of those choke type thingies you find on most computer cables as an inductor to determine the nature of the noise. On a more fundamental level, what makes the distinction between the E-Field and B-Field noise, surely all nosie signals are comprised of a fraction of both?

Overall, I don't want to go down the road of actively or passively altering the signal, as the fact remains that there is data in there of which little is known about i.e frequency, duration, amplitude) so I guess shielding and orientation are my only friends left!

As you can see I have a lot of questions and your patience is appreciated.
 
Last edited:
  • #21
gareth said:
I read another thread about this and it said that the ratio of C/L (per unit length) determines the impedance. So what is the correct model? I would have to assume that I was correct in what I said earlier, that I should just leave out the 50 Ohm resistor in this case.

The intrinsic impedance of coax is

[itex]Z_0=\sqrt\frac{L}{C} [/itex]

and for RG58 this value is 50 ohms. For a proper match your source impedance has to look like 50 ohms when connected to the coax. And the other end has to be terminated to 50 ohms (somehow). This will insure an efficient power transfer across your coax, and eliminate any reflection that can affect your signal.

As far as those choke cores go, they would suppress EMI in a stand alone wire, not in shielded wires like a coax.

what makes the distinction between the E-Field and B-Field noise, surely all nosie signals are comprised of a fraction of both?

Indeed. A slight transient the E-Field will induce a B-Field and vice versa.

When you connect the scope directly to the power supply can you still pick up noise?
and what is the amplitude /div set on the scope?
 
  • #22
yes, there is indeed noise when the scope probe is connected directly to the power supply.

I have attached some screenshots of the noise to give you a better idea, as you can see the noise directly from the PSU is quite large. Whereas the noise from the CCt alone is of a manageable level, as soon as the PSU is plugged in it picks it up a lot more.
 

Attachments

  • noise_CCT_and_probe_only.jpg
    noise_CCT_and_probe_only.jpg
    34.5 KB · Views: 302
  • noise_PSU_direct.jpg
    noise_PSU_direct.jpg
    34.4 KB · Views: 351
  • noise_whole_setup_together.jpg
    noise_whole_setup_together.jpg
    35.3 KB · Views: 360
  • #23
If you are getting a lot of noise from your power lead you may need to use shielded power input filtering with pass thru caps (specialized capacitor) between the filtering stages (sometimes more than one). Each filter stage is contained in its own shielding box internal to your circuit enclosure.

You could also be experiencing ground loop problems where current is flowing in the shield braid of of your coax. Careful attention to the layout of the bits and pieces is critical.
 
  • #24
gareth said:
yes, there is indeed noise when the scope probe is connected directly to the power supply.

Are you using a differential probe? Be very carefull when looking for noise using an oscilloscope, unless you are using a differential or (better) battery powered scope your probe is creating an additional path to ground which tends to lead to ground loops etc.
Hence, what you are seeing on the scope might have nothing to do with what you are interested in.

(maybe you already know this, it is just a very common misstake so I figured I should point it out).

Also, chokes CAN help even if you are using a coaxial cable. The idea of a core is to achieve "current equalization" in a conductor-pair circuits, essentially to "force" in the in- and out currents to be the same (as opposed to having the return current taking another path).
Anyway, the point is that can definately help to thread a coax cable (or any other pair) a few times through e.g. an ultraperm core. I use this method quite frequently, especially when I am using switched mode PSUs, and it usually reduces the noise quite significantly.

Also, from a practical point of view it is almost hopeless to filter out anything below a couple of hundred Hz. If you are having problems at those frequencies you really need to start thinking about proper grounding, using battery driven equipment (which is what I do whenever I do low noise measurements, I have a few car batteries in the lab) or simply use another measurement scheme.
 
Last edited:
  • #25
f95toli said:
Are you using a differential probe? Be very carefull when looking for noise using an oscilloscope, unless you are using a differential or (better) battery powered scope your probe is creating an additional path to ground which tends to lead to ground loops etc.
Hence, what you are seeing on the scope might have nothing to do with what you are interested in.

(maybe you already know this, it is just a very common misstake so I figured I should point it out).

No, I was not aware of this, but the connection was made with a co-ax cable, the outer shield being ground and the core being +ive. Regardless, when I hook the PSU up to the setup It pulls in a whole load of noise.

So filtering is out; I have signals at the same frequency

So is amplification I can't distort the true signal

I think I've done all the shielding I can do, unless I come across a large 10mm thick box of copper which I can put the noise source into, anyone have one?

Seems like I'm running out of options
 
  • #26
Those voltages are definitely ringing...your supply needs some kind of damping...
damping is usually provided via a snubber rcl network...
however...any chance of changing the power supply cable to the ion box?
see if that makes a difference...

ringing is usually caused by stray inductance or capacitance...
 
  • #27
Yes, it does look like ringing, but all instruments in the lab pick this up this noise (mind you they are all equipped with co-ax).

I don't think the cable will make a huge difference, it does not pick up any more noise when plugged into the setup at one end, but when the contacts touch the PSU the noise is much more prevalent.
 
  • #28
Then this noise must be on the powerline itself from some other kit...
how about using a filtered extension lead...that usually reduces line noise somewhat...
 
  • #29
gareth said:
Yes a DC battery solution would be nice, but unfortunatley I have to vary the voltage from around 0-30V in small increments (0.25V).

get a 36 volt battery and charger and adjust the voltage with a potentiometer driving a power transistor
 
  • #30
Is it possible to buy a DC PSU that is powered by battery? I've done a few searches on the internet but can't seem to find any. Anyone know-of/used such a device? Bearing in mind that it needs to have a varialble voltage of around 0-30V, the current drawn is fairly minimal so that's not a critical parameter.

Thanks
 
  • #31
Do you need to control it via a computer?
if not, 3x12V lead-acid batteries and one potentiometer would work.
Although personally I would use a single operational amplifier feed via a voltage reference instead, that way you can be sure that the voltage does not vary over time as the batteries are discharged.
Both circuits are very easy to build (the first one is just a single component+leads).
 

Similar threads

  • Electrical Engineering
Replies
3
Views
759
Replies
9
Views
4K
  • Electrical Engineering
Replies
12
Views
1K
Replies
13
Views
2K
Replies
31
Views
2K
  • Electrical Engineering
Replies
21
Views
11K
Replies
3
Views
1K
  • Electrical Engineering
Replies
1
Views
9K
Replies
11
Views
7K
  • Introductory Physics Homework Help
Replies
2
Views
5K
Back
Top