Noisy ECG help

Discussion in 'The Projects Forum' started by EMFELAB, Mar 16, 2013.

  1. EMFELAB

    Thread Starter New Member

    Feb 11, 2013
    10
    0
    [​IMG]



    I have built this circuit on breadboard with the help of previous threads on this forum and example circuits in instrumentation amplifiers data sheets. Unfortunately I'm getting a very noisy signal (see the plot and spectrum). I tried mitigating this noise with a 60Hz notch filter thinking it was this type of noise the source of it but it resulted in no avail.

    The leads are shielded and twisted together so I think this is not the problem. So I wonder if there is anything I can do to increase the SNR??

    [​IMG]
    [​IMG]
     
    Last edited: Mar 16, 2013
  2. tindel

    Active Member

    Sep 16, 2012
    568
    193
    Do all of you op-amps have decoupling caps? They aren't shown in the schematic. Adding decoupling caps will help if you don't. I'd also try adding balancing resistors and caps on your non-inverting inputs to reduce offsets that you may be experiencing.

    What exactly are you trying to do with U2? Shouldn't the pin 5 on U1 just be tied to ground? It seems weird to me that you are controlling your ground with an op-amp as if your op-amp is going to give you a better ground node than ground. I'm a bit confused there.

    It looks like your output stage is right at 100Hz. So that's a good filter point. But then again - all of your noise seems to be below 100 Hz too.

    Breadboards are notorious for being bad in many ways - especially when you're measuring uV like you are. You could be getting ground impedance errors.

    What's your test setup like? Sometimes the problems are in the test setup, but at these low frequencies, I doubt it. However, if your scope is AC coupled, this very well could be the case - most scopes are AC coupled around 100Hz or so, so you'd be getting the full magnitude above 100Hz, but anything below that can be significantly reduced. I.E. Make sure you're setting your scope up to DC couple with an appropriate DC offset.
     
  3. KMoffett

    AAC Fanatic!

    Dec 19, 2007
    2,574
    230
    Shouldn't the input voltage sources be tied to the driven "right leg" ground and not to circuit common?

    Ken
     
  4. EMFELAB

    Thread Starter New Member

    Feb 11, 2013
    10
    0
    Thank you both for the replies. The following is a more complete schematic of the ECG front end.

    [​IMG]

    To answer some of your questions, U2 is supposed to be a high pass filter at 0.5Hz cutoff; the connection of the output of U2 to the reference of the INA was copied from example schematics of ECG, please see this link:

    http://forum.allaboutcircuits.com/showthread.php?t=19555

    As for decoupling capacitors , there are two 100nF caps at the negative and positive supply of the INA. Also, there are two 220pF at each input of the INA. Is there any other place where you think i should put more? Across both the 10 and -10V supply lines.?

    Regarding the test set up I'm using a DAQ from NI to sample the signals. Today I implemented a digital low pass filter (100Hz) and high pass (0.5Hz) but the noise does seem to persist at the low frequencies as you mentioned.

    [​IMG]
    [​IMG]

    Having tidied up some of the protruding wired in the breadboard I did notice some improvements int the signal, once I'm sure that the circuit is 100% correct I intend to
    solder it.

    The common move voltage should appear at the node CMV, this signal is them amplified and driven back to the body where it should set the CM of the input at an appropriate bias voltage for the INA.
     
  5. tindel

    Active Member

    Sep 16, 2012
    568
    193
    Ah - note that that high-pass filter is going to the negative voltage input of the INA118P, and the reference voltage is connected to ground. This may be the source of your problem.

    put a .1uF cap over all of your op-amp rails - not to ground, but across the rails.

    I guess my point here is that once you make a PCB - if you have a good ground plane, I expect some of your noise to go away. And when your measuring uV then the slightest amount of ground noise can cause problems, I believe that this is what U2 is trying to accomplish - filtering out dc ground noise.
     
  6. tindel

    Active Member

    Sep 16, 2012
    568
    193
    Any update? I thought this was a fun project you were working on... I'm curious if you got it working and what the problem was.
     
    EMFELAB likes this.
  7. GopherT

    AAC Fanatic!

    Nov 23, 2012
    6,043
    3,806
    You also have to make sure you have a good connection to skin. A good hand cream or, better (and almost the same as used in clinics) is KY Jelly. Just have a story ready for your spouse when they ask why the KY is on your workbench.
     
    EMFELAB likes this.
  8. KMoffett

    AAC Fanatic!

    Dec 19, 2007
    2,574
    230
    OK...had to try it. 1 cm puddle of K-Y on cardboard. Measured resistance across the puddle with a DMM. 800KΩ to 1MΩ. Not what I would consider conductive. Medical electrodes use a salt gel with silver/silver chloride metal electrodes.

    It may feel like, and look like K-Y, but it's not.

    Twenty three years in medical electronics.


    Ken
     
    EMFELAB likes this.
  9. GopherT

    AAC Fanatic!

    Nov 23, 2012
    6,043
    3,806
    Interesting. 20 years ago we measured KY to be 1% saline when I was a teaching assistant for an undergrad lab. 1% is about the same as wet biological tissue. It worked fine back then for this purpose. I haven't had to use it (for that purpose) since.
     
  10. EMFELAB

    Thread Starter New Member

    Feb 11, 2013
    10
    0
    I'm really sorry for the late reply. Basically I have been reading quite a bit about power supply bypassing and decoupling. I put capacitors across the supply rails, and to both supply rails and ground. All of these are 0.1uF

    Actually that is a mislabeling in the schematic, and the negative supply is going to pin 4 of the INA118 not 5 as you suggested

    I came to the conclusion that the noise source might not only be low frequency after all, the test set up was sampling at 1KHz which dictated a 500Hz range for the FFT. Using an oscilloscope some of the interference noise that I detected was up to 10KHz. So even after adding the capacitors the output signal stays pretty much the buried under noise :(

    I thought that the INA would get rid of the noise if this was present as common mode at both of the inputs; high frequency noise may be coupling at other places through the circuit??? or there is a mismatch of common mode at the input which is causing the HF noise to persist.

    I'm not sure that converting the circuit to single supply would help the SNR but I want to do it anyway. I plan to replace the ground connection of U2, U3 and U7 to Vs/2 and all the negative supply connections to ground. I'm afraid that it wont be as easy as this though; because of the swing limitations of this Op-amps with single supply connections.

    Thank you so much for your feedback
     
  11. tindel

    Active Member

    Sep 16, 2012
    568
    193
    A couple other possibilities that I thought about:

    It is possible that the A/D converter on that NI DAQ sucks - A/D's over 8-bits only ever use 8 bits! Do you have access to a analogue scope? This will tell you if you're really getting the noise that you think you're seeing without aliasing like digital scopes tend to do. You mention using a scope - I'm guessing it was a digital scope - did you see the same low frequency noise? was the FFT similar?

    A spectrum analyzer would be much better than FFT to determine really what frequencies you're having problems with. Spectrum analyzers are expensive though.

    I'm going to go back to grounding too:

    Remember that your currents have to return from where they came, completing a loop. The more direct the path back, the less grounding issues you'll have and the less noise problems you'll have. Also, at audio frequencies it is extremely important to have a star grounding system so you don't have currents from other places giving you ground voltages other than 0V.

    I'm guessing you have this on a solderless breadboard. Those things are awful, awful, awful for noise! They will pick up radiated noise and they also exacerbate grounding problems by design. You are showing that the concept is sound - I'd lay it out on a card while paying special attention to grounding. I bet you come out with a good product.

    Remember, you're amplifying very small voltages - in the few hundred uV range - it's difficult to do.

    Speaking of those kind of voltages, it might help to have a op-amp with a lower input off-set voltage. The ina118 is actually pretty hi 100uV/G * 10G = 1mV! That's huge! Maybe make that first gain stage like 2 to reduce your input offset voltage to 200uV. Then increase your other gain stages.

    Well, that's my ideas and ramblings - hopefully some of it helps.
     
  12. justtrying

    Active Member

    Mar 9, 2011
    329
    350
    I worked on a project just like this last year. Having caps at each op-amp helps, but the noise issue persisted until the circuit was moved to a PCB - short leads, good connections, problem solved.
     
  13. live5

    New Member

    Dec 27, 2012
    3
    0
    From what I can see is that you used the schematic from the datasheet in INA118. Though I would suggest a simpler circuit like the right leg feedback from the AD620 circuit which seems more balanced than this circuit. I also would suggest trying a lowpass filter to take out the noise.
     
  14. paulktreg

    Distinguished Member

    Jun 2, 2008
    611
    120
    Perhaps enclosing the whole board in a sheilded metal box like all commercial units usually do.
     
  15. EMFELAB

    Thread Starter New Member

    Feb 11, 2013
    10
    0
    Taking all your suggestions into consideration, the SNR improved considerably. My last resort thought, is to move the circuit to PCB.

    Thanks for all your help
     
Loading...