Be careful with your terminology. It claims to be free of _phase_ distortion, not distortion-free overall.
Yes, but it'll be subject to manufacturing variances. The reason they make the gain from the bandpass to the subtracter variable is to tune these out. Do you want to have to twiddle with every circuit to get the null deep enough?
Then I sincerely hope that your "EKG type" signal will never be needed to keep someone alive.
Google "virtual ground".
I'd also suggest that you Google "twin-T notch filter", or perhaps "active twin-T filter". You'll still have manufacturing variances to contend with, but at least if all of your capacitor values and all of your resistor values match, your null won't go away.
--
Tim Wescott
Wescott Design Services
http://www.wescottdesign.com
Save yourself some grief and try the configuration...
GyratorFilterNotch.pdf
on the S.E.D/Schematics Page of my website. ...Jim Thompson
--
| James E.Thompson | mens |
| Analog Innovations | et |
| Analog/Mixed-Signal ASIC's and Discrete Systems | manus |
| San Tan Valley, AZ 85142 Skype: skypeanalog | |
| Voice:(480)460-2350 Fax: Available upon request | Brass Rat |
| E-mail Icon at http://www.analog-innovations.com | 1962 |
I love to cook with wine. Sometimes I even put it in the food.
AOE also offers the "bridged differentiator" notch filter in figure 5.23. See also
formatting link
You tune the notch frequency with a single potentiomenter. In practice a se cond one is a good idea to trim out capacitor tolerances.
All of this class of notch filters work by introducing equal and opposite p hase shifts in two parallel paths - hitting + and - minus 90 degrees at the notch frequency, which probably isn't something the OP can live with. The theoretical answer is to add an all-pass phase correction network which int roduces equal and opposite phase shifts after the 50HZ has been notched out , but I've no idea how you'd do it.
My guess is that what the OP actually needs is a digital solution, where th e ECG is heavily over-sampled, then a digital phase-locked loop is used to detects the 50Hz content in the sampled data, and subtract it out. If the 5
0Hz content is absolutely stable - in phase and amplitude - this would work perfectly.
In reality, the ECG is taken from a patient, who can't be relied on to lie still and act as stable antenna for 50Hz mains pick-up.
other than near the notch frequency. That means the signals not close to the notch frequency are really at the same phase, no? As long as the null is sharp, the region with other than close to zero relative phase shift with be very small.
If the phase shifts are still a problem, a second notch filter designed with + phase shift where the first has a - phase shift and vice versa should bring the final phase shift to zero across the range and deepen the null further.
The problem is over-driving the front end which prevents the low signals of interest from ever reaching the digital domain.
The problem with a typical notch filter is the phase shift it gives to the signals it passes. A band pass filter will do the same thing. But if you subtract the output of a band pass from the original input signal it will give you a notch since at the center of the band the amplitude is (or should be) equal to the amplitude of the input making the difference near zero.
You don't get significant phase shifts because anywhere not close to the notch the band pass signal (which was phase shifted) is very small and does not impact the phase of the output.
The problem is getting the two signals very equal in amplitude at the center of the band pass in order to get the deepest notch. If the band pass filter is very narrow (high Q factor) it will be tricky locating it to the frequency you want. All the components in your circuit have tolerances which make it hard to get great filtering with this circuit.
The concept is simple, but a lot of words. In essence you need to create a fake ground by biasing the various signal nodes to a voltage at the mid point of your single supply. You can do this passively with a pair of resistors or actively by buffering the resistors with an opamp.
a second one is a good idea to trim out capacitor tolerances.
te phase shifts in two parallel paths - hitting + and - minus 90 degrees at the notch frequency, which probably isn't something the OP can live with. The theoretical answer is to add an all-pass phase correction network which introduces equal and opposite phase shifts after the 50HZ has been notched out, but I've no idea how you'd do it.
e the ECG is heavily over-sampled, then a digital phase-locked loop is used to detects the 50Hz content in the sampled data, and subtract it out. If t he 50Hz content is absolutely stable - in phase and amplitude - this would work perfectly.
Not an insoluble problem. Getting ECG's in the operating theatre, next to s urgeons who insist on using diathermy knives to minimise bleeding - the pla stic surgeon who recently took a BCC out of my cheek used one, giving very perceptible barbeque odours - means that the ECG electrode rigs are availab le with five-pole passive filters which clean out pretty much all of the hi gh-frequency noise.
HP and Hoffman-La Roche both published detailed papers on the subject back in the 1970's. I once got a technical inquiry from a friend from undergradu ate days whose job as an anesthesiologist was being made difficult by diath ermy knives and was able to dig the references out of the EMI Library and s end him copies.
Your A/D converters are going to need head room, but ECG signals aren't fas t so 20-bit sigma delta converters work fine, and can leave you plenty of h eadroom.
At least with an active twin tee and proper feedback, it will be fairly sharp. Precision resistors are relatively cheap; 2% caps are not too bad,and they can be matched with a simple bridge. Net result might be something repeatable enough for government work (as they say).
My point is that you aren't going to do all the filtering in the digital. There is no reason to not add as much analog filtering ahead of the amps and ADC that you can get to work reasonably. Even 20 bit ADCs can only do so much.... uh, 20 bits worth to be exact. I think that works out to $2.50.
It is just a simple two-pole resonator notch with all the properties of a single resonator, including phase rotation near the resonance frequency.
The single opamp resonator is known to be notoriously sensitive to component tolerances. Adding an opamp or two will help the situation. For designs, see the nearest textbook on opamp circuits.
e a second one is a good idea to trim out capacitor tolerances.
site phase shifts in two parallel paths - hitting + and - minus 90 degrees at the notch frequency, which probably isn't something the OP can live with . The theoretical answer is to add an all-pass phase correction network whi ch introduces equal and opposite phase shifts after the 50HZ has been notch ed out, but I've no idea how you'd do it.
e
he
d
ere the ECG is heavily over-sampled, then a digital phase-locked loop is us ed to detects the 50Hz content in the sampled data, and subtract it out. If the 50Hz content is absolutely stable - in phase and amplitude - this woul d work perfectly.
ls
to surgeons who insist on using diathermy knives to minimise bleeding - the plastic surgeon who recently took a BCC out of my cheek used one, giving v ery perceptible barbeque odours - means that the ECG electrode rigs are ava ilable with five-pole passive filters which clean out pretty much all of th e high-frequency noise.
ack in the 1970's. I once got a technical inquiry from a friend from underg raduate days whose job as an anesthesiologist was being made difficult by d iathermy knives and was able to dig the references out of the EMI Library a nd send him copies.
fast so 20-bit sigma delta converters work fine, and can leave you plenty of headroom.
Obviously not. You've got to filter whatever goes into an A/D converter so that there isn't much noise above half the sampling rate, otherwise it will get aliased into the bandwidth you are looking at - this is sci.electronic s.design, not sci.electronics.basics.
The point here is that notch filtering normally introduces inconvenient pha se shifts. If you do some of your filtering in the digital domain, getting rid of a specific stable frequency can be a bit easier.
I'm not talking about antialiasing. I'm talking about filtering the large 50 Hz component. That will be reduced as much as feasible to preserve the dynamic range of the ADC.
"Inconvenient" phase shifts can be minimized and what isn't eliminated
*can* be compensated in the digital domain. If your signal is swamped by a large interference signal, it will be too small to have useful dynamic range and can not be recovered.
so that there isn't much noise above half the sampling rate, otherwise it will get aliased into the bandwidth you are looking at - this is sci.electr onics.design, not sci.electronics.basics.
But you've got to have it in most real applications, so your point was some what redundant.
But doing the filtering with a notch filter introduces inconvenient phase s hifts, and anything that doesn't is also going to filter out the ECG compon ents which aren't quick.
phase shifts. If you do some of your filtering in the digital domain, gett ing rid of a specific stable frequency can be a bit easier.
Obviously, but 20-bits is about 120dB of dynamic range. 2^20 is 1048576. The decimal log of that number is 6.0206, and that translates into 120 dB.
I think you misunderstand the second op amp. It is just a 1:1 buffer with both an inverting and non-inverting input. So it acts as a subtractor passing the input signal with the notch subtracted out.
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.