Finding the Right Filter for -140 dB 60 Hz Attenuation

[m718]

Does anyone know how to filter 60 hz down to -140 db
all the notch filters I tried are about -60 db maximum but after I amplify the signal to 100db
the 60 hz becomes a problem again. I need a filter for this it doesn't have to be one stage and also I need the 100hz, maximum attenuation on 100 can't be more than 6db.
help with this problem would be very appreciated.

 

[f95toli]

Trying to filter out 60Hz is usually the wrong approach; it is much better not to get the 60/50Hz in there in the first place.
The first step is always to check your grounding and make sure there are no loops; once that is done you should make sure you are running all sensitive electronics (in this case the pre-amps) on batteries. Then you need to make sure there is no inductive pickup meaning you are using the right cables and -where needed- magnetic screening (such mu-metal) to avoid inductive pick-up.
Once you've all this you might start thinking about filtering. Although I must say I've never successfully filtered out mains interference; the kind of filter you need (which must for obvious reasons be active) tend to introduce so much extra noise in the circuit (e.g. 1/f noise) that is is not worth it (and you also need batteries etc for the filter; it rarely worth the extra effort).

60 Hz will always be a problem (unless you are are doing your experiment in a shed in the middle of nowhere using only battery operated electronics; which is what some people working with low-noise electronics actually do) and the best you can hope for is to make it insignificant.

 

[m718]

Trying to filter out 60Hz is usually the wrong approach; it is much better not to get the 60/50Hz in there in the first place.
The first step is always to check your grounding and make sure there are no loops; once that is done you should make sure you are running all sensitive electronics (in this case the pre-amps) on batteries. Then you need to make sure there is no inductive pickup meaning you are using the right cables and -where needed- magnetic screening (such mu-metal) to avoid inductive pick-up.
Once you've all this you might start thinking about filtering. Although I must say I've never successfully filtered out mains interference; the kind of filter you need (which must for obvious reasons be active) tend to introduce so much extra noise in the circuit (e.g. 1/f noise) that is is not worth it (and you also need batteries etc for the filter; it rarely worth the extra effort).

60 Hz will always be a problem (unless you are are doing your experiment in a shed in the middle of nowhere using only battery operated electronics; which is what some people working with low-noise electronics actually do) and the best you can hope for is to make it insignificant.

I can't prevent the 60hz, the sensor of this circuit is an inductive pickup sensor. I am using 2 opamp notch filters are they very noisy? would it be a lot less noisy if it was just RC notch filter?
and do these filters add drift to the signal?

 

[f95toli]

You can't make good passive notch filters that operate at 60 Hz; just calculate the component values you would need and you will see why.
I have no idea how noisy your filter is; it depends on the design and the op-amps; if there is enough gain before the filter stage it is probably not an issue (although this also means that you are amplifying the 60 Hz).

Would it be possible to re-design you measurement somehow to simply make in less sensitive to 60 Hz? E.g. by implementing a scheme where a lock-in amplifier could be used?

Also, is there any reason you can't simply screen the sensor with a magnetic screen (using mu-metal)? Note that there is simply no way to make "open" low noise setup using an inductive sensor in a normal building; there is far too much interference not only from mains but also from all types of electrical appliances, passing cars (or -even worse- mopeds with 2-stroke engines),trams etc.

I wasn't kidding when I wrote about the "shed in a forest" above; quite a few universities/institutes have places like that and they are used for precisely these reasons (specifically for e.g. biomag applications where SQUIDs are used to scan samples that can't be easily screened).

 

[TurtleMeister]

Another approach is to cancel it out rather than filter it out. Use two pickup sensors, one for the desired signal and the other for the noise. Mix them together 180 degrees out of phase. Adjust the gain of the noise amp until the 60hz is minimized.

 

[m718]

Can you explain the canceling out how would I mix them 180 out of phase ?



And how should I use a lock in amplifier?

 

[f95toli]

Another version of Turtlemeister's suggestion would be to use a single sensor but with a first- or second-order gradiometric pick-up coil.
But that assumes that you can use a gradiometer in you experiment (i.e. that you are only instersted in fairly localized field-variations).

I think you need to tell us exactly what it is you are trying to measure before we can give you any more advice.

 

[m718]

I'm trying to measure magnetic fields (gaussmeter) but small field variations can't be measured with the 60hz there

 

[f95toli]

Yes, I understood that you were trying to measure a magnetic field; but what kind of signal are you trying to pick up?

 

[m718]

Yes, I understood that you were trying to measure a magnetic field; but what kind of signal are you trying to pick up?

not any signal in particular but I tested my meter to see if I can pick up any magnetic field from my mp3 player for example any magnetic field that the screen might be emitting.

 

[TurtleMeister]

Can you explain the canceling out how would I mix them 180 out of phase ?



And how should I use a lock in amplifier?

Sorry, I don't have any circuit examples, but you may be able to find something by searching "adaptive noise cancellation".

 

[f95toli]

not any signal in particular but I tested my meter to see if I can pick up any magnetic field from my mp3 player for example any magnetic field that the screen might be emitting.

If you are only interested in doing near-field measurements (where you can put one end of your sensor close to the object you want to measure) a gradiometer should work quite well (gradiometric sensors are e.g. used for non-destructive testing so they can be made to be very sensitive even in a noisy environment).
That is, if you are actually able to modify the pick-up coil for your gaussmeter(?)

The other alternative is to simply connect your sensor to a spectrum analyzer. I am guessing you are doing this at home meaning you don't have access to a real analyzer; but even a sound card might work if you amplify the signal a bit; you won't get rid of the 60Hz but you will be able to see all other frequencies as well (up to about 20 kHz if you use a sound card).

 

[Pumblechook]

You could boost the wanted signals with a well screened pre-amp right at the sensors.

 

[m718]

If you are only interested in doing near-field measurements (where you can put one end of your sensor close to the object you want to measure) a gradiometer should work quite well (gradiometric sensors are e.g. used for non-destructive testing so they can be made to be very sensitive even in a noisy environment).
That is, if you are actually able to modify the pick-up coil for your gaussmeter(?)

The other alternative is to simply connect your sensor to a spectrum analyzer. I am guessing you are doing this at home meaning you don't have access to a real analyzer; but even a sound card might work if you amplify the signal a bit; you won't get rid of the 60Hz but you will be able to see all other frequencies as well (up to about 20 kHz if you use a sound card).

I was thinking of the spectrum analyzer too I can see the signal very good on it using my soundcard but I will be sending the signal to a microcontroller for processing and it needs to be real time. Is there any circuit that will work like this?

 

[rolerbe]

How about an instrument amp configuration? Signal into one, and same size/length lead into the other that is left unconnected (or perhaps connected to ground at the signal source)? That way, the 60 Hz is common mode to the instrument amp, (many of which have in excess of 120DB CMMR). An off base idea?

 

[0xDEADBEEF]

I don't see why he shouldn't do active filtering: http://www.national.com/ms/LB/LB-5.pdf

Mod Note: National Application Note may be found here: http://g3ynh.info/circuits/Bob_Batey/LB-5.pdf[/color]

The twin ``T'' network is one of the few RC filter networks
capable of providing an infinitely deep notch. By combining
the twin ``T'' with an LM102 voltage follower, the usual
drawbacks of the network are overcome. The Q is raised
from the usual 0.3 to something greater than 50.

 

[rolerbe]

Excellent tech note find, 0xD. I'm sure I will use that some time in the future. Thanks!

 

[m718]

the 60hz is actually starting from 40 then peaks at 60 and goes down to 80 on my soundcard oscilloscope I was wondering if getting rid of 60hz with a high Q notch filter will automatically remove those frequencies or would they need another filter?

 

[Topher925]

A notch filter basically is a "T" filter. I have had some success with them when used for AC signals but not with DC. Instrument amps with high CMRRs work well also but can be expensive. You could try putting two of your sensors together and cancel out the 60Hz with a CMR filter just like a thermocouple does.

But, like stated before, the best way to get rid of 60Hz is just to remove the source of it.