Recommended Gain and Filter Type for EMG Data Acquisition: Sallen-Key or Butterworth?

Hello

I'm working on project called"design & implementation of EMG data acquisition",after I search on these systems i really confused about what the gain and the filter type that i have to use.

at the first stage i use instrumentation amplifier to amplify the signal, so how much the gain of this stage should be? and I use Band-Pass Filter in the second stage which is eliminating the unwanted frequency components, which type of filter is suitable to get a very clean EMG signal? Is sallen-key or butterworth good enough for that? and how much should the order of filter be?

the attached image for my design for now in Multisim, I will add proper ADC to it...

thanks.......

The signals you are measuring will be well known, look up the typical values, they vary somewhat depending on location etc, so that has to be taken into account.Some very general comments, for the best signal to noise, get as much gain as you can at the front end. If there is a possibility of saturating the front end, take the gain back a little and add variable gain in the next stage. If you aim for a signal of around a volt initially you should be fine.Again, the typical signal frequency components will be known, look them up. They will be band limited, so the main purpose of filtering will be to limit noise unless there are very specific reasons to limit the signal components.You have a misconception re filters, Butterworth is a class of filter, defined by its pole locations, Sallen Key filters are a second order active filter topology. You need to do some reading on the topic, there are a number of good tutorials on the subject on the web.Sallen Key can be any class of filter, Butterworth, Bessel, etc.Filter Topologies include Multiple feed back, Sallen Key etc.Sallen Key are easy to tune, but the interaction with the increasing output impedance of the op amp means their stop band performance at higher frequencies deteriorates unless it is taken into consideration.  There are a number of filter design programs around, TI Web Bench and  Schematica come to mind.For a wide band pass filter you normally have to cascade high pass and low pass filter sections.What you need to do is firstly to ascertain exactly what the amplitudes are you have to deal with, and how they are normally measured, and secondly, what the band width of interest is, armed with that information your design can proceed,cheers,Richard

Check the spelling of "theoretical" on your schematic. I agree with Richard's comments, and have some more to add.Your EMG signal is small, and the pickup leads make an open loop antenna, leaving you wide open to magnetic pickup of power-line, audio-frequency and radio-frequency (RF) interference, which overloads your instrumentation amplifier. This is a serious real-world problem.
Instrumentation designers place lots of gain ahead of filters because most active filters are noisy. Look up "noise gain" for why active filters are often noisy. RF designers minimize gain ahead of filters because RF designs deal with out-of-band signals much larger than the desired signal. RF designers also have access to passive filters which are quiet.
The filter in your schematic is pretty quiet (the noise source is the noise of the opamp multiplied by the noise gain of the circuit), which is good.
You could put a narrowband filter at the input, by replacing gain-setting resistor R1 with a series-tuned inductor-capacitor circuit. The resistance of the inductor serves as the gain-setting resistance. This filter is probably too narrowband for your application, judging by the filter in your schematic.

Modern 24-bit audio analog-to-digital converters (ADCs) have enormous dynamic range (and also built-in antialias filtering), so you could run your instrumentation amplifier at low gain and run the amplifier's output into an ADC and do the filtering digitally. I know this defeats the point of the exercise.
A follow-on exercise: If we want to build a true 32-bit 40-kHz-bandwidth audio ADC, with fullscale 32 bits above the thermal noise floor of a 50-ohm input, how big is the fullscale? The result shows why we won't see such monolithic parts anytime soon.

One of the principal reasons for using an instrumentation amplifier is that for a differential signal you get an high CMRR- in other words noise that appears at both inputs simultaneously is largely rejected. But I come from an industrial world where you have access to a differential signal. Look at the applications in the INA114 data sheet to see what I mean.
The EMG is in the medical sphere and if the human interface is anything like the EKG, itwill be decidedly single ended with the signal and common widely separated. I wonder in fact if the IA gives you any benefit other than  control of the gain with a single resistor. I found this abstract on your topic, and it looks like there are several problems in addition to the general ones detailed by Peter. I would suggest you purchase the article to give you greater clarity.
I would also question whether XFG1 is a fair representation of the EMG. As an outsider looking in,  I would think there should be a fairly high resistance between the COM and the ground  to account for skin resistance, to say nothing of the actual signal source which, based on the aforementioned abstract, is pretty complex.P.S. when you make the circuit in real life remember to add the supply decoupling capacitors!

There are some very good academic papers easily found with a quick search on the web which include  typical amplitude and spectral information. Also included is information on how the signals can be obtained and processed,cheers,Richard

I agree with Aubrey that you should put a resistor in series with XFG1, to simulate the series resistance of the EMG signal source. This resistance, multiplied by the instrumentation amplifier's bias current, could generate a significant DC offset.
An inductor at R1 is probably impractical at the low frequencies of interest here.Adding a capacitor in series with R1 is practical, and is useful if you do not need DC coupling. With the series capacitor, the DC gain is zero, so DC offsets (whether from the amplifier or from the EMG source) are not amplified.Even if the EMG is single-ended, a differential amplifier may be needed because the EMG ground is not the same as the analog-electronics ground. This is one component of the baseline noise discussed in the abstract Aubrey refers to above.The EMG ground can easily differ from the analog-electronics ground at frequencies too high for the instrumentation amplifier to differentially reject. Try placing a cellphone call with the phone near the EMG source or near the analog circuitry. Try placing the phone's charger near the EMG source or near the analog circuitry.
Richard, could you tell us the search terms you use to find the amplitude and spectral information?Thanks for giving us material to discuss.

"EMG signal processing" does it. Search using Google Scholar if you want the really technical stuff, there are a lot of papers and book extracts out there. Like any work detecting signals around the human body, it is common to use multiple electrodes, with a guard driven from the  output of the differential input stage. The "art of electronics 3rd edition" is worth looking at as well, follow through the info starting from section 5.15.2Our public library system gives us "free" access to thousands of professional journals throughout the world, so much so that it's sometimes overwhelming knowing where to start in setting up a search,
Cheers,Richard

Richard, thanks for the search term; I intend to look up some of the papers.If you build up EMG leads and instrumentation amplifier, I suggest you listen to the amplifier output through an audio amplifier and headphones or speaker. Turn the audio gain up until you hear something. Do you hear speech or music?If you have an audio spectrum analyzer, look at the instrumentation amplifier output for a spectral line at 19 kHz.
If you build up EMG leads, try connecting the leads directly to the vertical input of an oscilloscope with bandwidth of 150 MHz or more. What do you see?