Help on identifying source of unwanted harmonics

Discussion in 'The Projects Forum' started by padhu84, Dec 15, 2009.

  1. padhu84

    Thread Starter New Member

    Dec 15, 2009
    21
    0
    Hi guys

    I have designed a small pcb for a data acquisition system involving NI-DAQ. This board is primarily housing a low noise pre-amplifier(Opamp AD8613 from Analog Devices) for the DAQ and its related power circuitry. I have used a dual-supply configuration for the opamp supplying +/-1.8V. The +1.8V comes from an LDO and the -1.8V comes from a Positive-to-negative converter(ICTL7660,TI-Switched Capacitor voltage converter). I have also added first order-RC filters to limit the response in the range of 100Hz to 10Khz.

    The main application of this board is to reduce the noise being coupled into the data-acquisition system. On analysis the frequency response at the output of the opamp, there are certain harmonics starting from f=3.3Khz. Subsequently, i can find them @ 2f, 3f, 4f etc. I am not sure about the source of these harmonics but they are definitely not desired.

    How to go about identifying the source of these harmonics and how to eliminate them? Currently i am looking at eliminating the switch-cap converter and instead directly feed from the battery. Any other suggestion? Please help me out.

    I have tried to squeeze in as much info as possible. Please let me know if any further details regarding the board is required
     
  2. padhu84

    Thread Starter New Member

    Dec 15, 2009
    21
    0
    The harmonics were observed when opamp input was grounded or fed with a signal from function generator.The opamp is acting as a non-inverting amplifier and the functionlity is all fine except for the harmonics at the specified frequencies.
     
  3. padhu84

    Thread Starter New Member

    Dec 15, 2009
    21
    0
  4. SgtWookie

    Expert

    Jul 17, 2007
    22,182
    1,728
    Your schematic is in .png format, which is good. However, it is far too small to be able to read it.

    Please make it at least 3x as large. You can attach it to a reply by using the "Go Advanced" button, and then "Manage Attachments"

    This is much better than linking to another site.

    Also, if you have the board layout, or a photo of your board, that would also be helpful.
     
  5. padhu84

    Thread Starter New Member

    Dec 15, 2009
    21
    0
    Hi there

    Sorry for uploading through a link earlier. Am new to the forum and i dint of this option. Thanks!!

    I have attached the schematics as u had mentioned. Please have a look.

    I would get you the layout as soon as possible. Just that i dont have the tool installed in my laptop and i need to open it in the univ lab system.
     
  6. SgtWookie

    Expert

    Jul 17, 2007
    22,182
    1,728
    OK, a few things right off the bat:
    1) You do not have 0.1uF bypass capacitors on the Vcc and Vee pins of the opamp.

    2) You've connected all of the inputs to the unused amps to ground via resistors. While one would think that should be OK, it could be causing problems, as each amplifier is operating open loop. So, for each unused amplifier:
    A) Connect the noninverting input directly to ground.
    B) Connect the output to the inverting input.

    These amplifiers are stable at unity gain. Connecting the output to the inverting input makes them unity gain voltage followers/buffers. Connecting the noninverting input to ground has the outputs following ground.

    3) Your low-pass filter has a small value resistor and a huge capacitor. Your response time will be terrible. You would be better off to increase the size of the resistor and decrease the size of the capacitor. You would be much better off to increase the number of orders in your filter; you would have far better response time and better filtering.

    As it is now, the 4700uF capacitor will have a LOT of parasitic inductance. This can cause an undesired LC resonance.
     
  7. padhu84

    Thread Starter New Member

    Dec 15, 2009
    21
    0
    Thanks for your response, SgtWookie.

    1. Missed it big time. Would rightaway connect the decaps.

    2. Actually, i thought about the unused inputs real hard but since no recommendations were given, i just thought i would be better off connecting it to GND through a pull-down resistor. My rationale was each opamp is going to act independent anyway though they are bundled in a single package. Didnt know much about the stability part. Anyway, would put all the unused opamps in voltage follower mode as you have suggested.

    3. The low pass filter capacitor is 4700pF and not 4700UF. Am looking at a cut-off of 10KHz so the corresponding R becomes 3.3K.
     
  8. russ_hensel

    Well-Known Member

    Jan 11, 2009
    818
    47
    Can you determine the switching speed of the supply? How does it relate to the noise? Temporarially replace it with a battery.
     
  9. SgtWookie

    Expert

    Jul 17, 2007
    22,182
    1,728
    Good. Ceramic would be OK, but metalized poly film would be better.

    That's true; they will all act independently. However, since there was no feedback loop from the output to the inverting input, they were all operating at maximum gain, which is 1000v/mV - when they have a 10k load. That means as little as 3.6uV p-p noise differential at the inputs would send the outputs slamming against the rails. Connecting the inverting input to the output gives 1v/1v gain.

    It really isn't obvious until it's explained.
    You really have to read the datasheet to find out if a particular opamp is stable at unity gain. Many are not.
    OK. The text on the schematic is really quite small, and my eyesight isn't what it used to be.

    Some other things that may help:
    Russ_Hensel brought up power supply noise. This is certainly a concern when using switch-type supplies; the 7660 is a switched-capacitor supply.

    It may help to add a fixed 1.8k Ohm load at the output of the 7660. This will only draw 1mA, but may prevent it from going into an intermittent cycle of charging, then waiting.

    Another item is to use a fixed 10 Ohm resistor between the +V/-V supplies and the Vcc/Vee pins of the opamp - the decoupling caps still need to be directly connected to the Vcc/Vee pins, though. The small resistance in conjunction with the bypass caps will act as a low-pass filter.
     
  10. padhu84

    Thread Starter New Member

    Dec 15, 2009
    21
    0
    Thank you both for your suggestions.

    But I have ran into some problem, guys!! :(

    As suggested, i connected the decaps(0.1uF, ceramic) between V+ and GND && V- and GND. But on checking the circuit, the output is saturated at V-. This was not the case earlier. This is irrespective of the input applied to the Header P1. I shorted P1, connected an output of function geenrator, kept P1 open but still the output was saturated at V-. I removed the decaps but still there is no change. The output is saturated at negative supply.

    I didnt make any other change in the circuit. The gain of the opamp is set to approx 30( Feedback Pot set at 29K)

    Any ideas for troubleshooting? I am clueless y the behavior of the opamp changed :confused:
     
  11. padhu84

    Thread Starter New Member

    Dec 15, 2009
    21
    0
    Any idea guys??

    Has it got to do with the ceramic decaps which i added??Because the + to - voltage converter, TL7660, uses polarized cap at its output? Would using ceramic cap of 0.1uF in parallel to it(at the opamp power pin) cause any problem??
     
  12. SgtWookie

    Expert

    Jul 17, 2007
    22,182
    1,728
    Sorry I didn't see your response this morning.

    Check to see if you are still getting +V to the opamp.

    Did you connect the other three amps as I suggested?

    What are you measuring on their outputs?

    Did you try changing your inputs and outputs over over to another of the amplifiers in the IC? Don't forget, you have four amps in that one package.
     
  13. padhu84

    Thread Starter New Member

    Dec 15, 2009
    21
    0
    Hi SgtWookie

    Actually it was nothing but a soldering issue. The input was not coupling to the +ve input of the opamp due to improper soldering. But no idea how it happened since it was working properly earlier. Sorry for cribbing about it here linking it to the decaps:(.

    Anyway, my position now is.. as i have got the opamp running properly now, i will do 2 tests now

    (i) Isolate the + to - switching converter and instead directly feed the opamp from a DC RPS and see whether there is again any persistant harmonics mentioned in the earlier frequencies.

    (ii) Connect the filtering decaps back again and hope that it works with the + to - converter back on. I hope ceramic decaps would do, wouldn't it? Any issues with it being non-polar?

    And yes, i have put the unused opamps in voltage follower mode with + input grounded. I would like to mention one thing here. I have slightly tweaked the connections from wat is shown in the schematic. Instead of directly connecting the LPF output to the header P2, i have connected the filter to the OPAMP B, with it being a buffer. I had done this earlier itself and is working fine as a voltage follower. The reason i did was actually i had a 4.7uF and 3.3 ohm as a LPF and the opamp A was huffing and puffing. Consequently, the -Ve voltage dropped to -1.0V from -1.8V while functioning and was getting loaded. Thats when i brought the buffer in and also changed the C and R to 4700pF and 3.3K, as shown in the schematics.

    Now, i would update you as soon as i test with the above two cases, SgtWookie. Thanks for your support. Please keep encouraging me with more suggestions of yours:cool: !!
     
  14. SgtWookie

    Expert

    Jul 17, 2007
    22,182
    1,728
    Don't forget that you have 4 opamps total available.

    You will find that if you increase that 3.3k resistor to 10k, and increase your cap size, the loading issue will go away. It will slow the response time, but your opamp won't have to work so hard.

    If you want to keep the response time down, add more stages to the filter; ie: make it a higher order filter.
     
  15. SgtWookie

    Expert

    Jul 17, 2007
    22,182
    1,728
    What kind of settling time did you need for the output filter?

    The 3.3k resistor and 4.7nF (4700pF) cap don't do much filtering until the frequency gets pretty high.
     
  16. padhu84

    Thread Starter New Member

    Dec 15, 2009
    21
    0
    Its not a time-critical application. Rather more critical on noise performance. I wouldn't mind if the settling time is sort of big but i need to get rid of as much noise as possible. To be frank, i was not looking into the settling time until now.

    I did check on the frequency response for this and turns out to be pretty ok. The HPF is bang on with -3dB at 100hZ(as desired) but the LPF has -3db @ 8.5KHz(desired at 10KHz). Thinking adding another order or two would improve it closer to the desired value.
     
  17. padhu84

    Thread Starter New Member

    Dec 15, 2009
    21
    0
    Good news, guys!!I used a DC-RPS for negative supply and found tat the harmonics had vanished. So indeed the switch-cap converter seems to be the culprit.

    Now, that i have used it in this proto anyway, i would like to try rectifying it for this board by adding decaps. Since the harmonics are at pretty low frequencies(3k n multiples), would it make sense to add bypass capacitors of larger values than the usual 0.1uF?? say 4.7uF or even larger??

    Or any other suggestion. I would like to see the circuit working with this + to - converter chip albeit without those unwanted harmonics. I will change the circuit in my final design though.
     
  18. SgtWookie

    Expert

    Jul 17, 2007
    22,182
    1,728
    Try adding a LP filter between the output of the -v supply and the Vee of the opamp.

    Even a small amount of resistance will block a great deal of noise; if you have caps on both sides of the R.
     
  19. padhu84

    Thread Starter New Member

    Dec 15, 2009
    21
    0
    I am not sure how i will add it. oh wait. i had cut the trace to isolate TL7660. would just be fine if i somehow mount a small 3.3E inbetween. i guess so..

    And wat abt the cap value??
     
  20. SgtWookie

    Expert

    Jul 17, 2007
    22,182
    1,728
    Why don't you try a 10 Ohm resistor between the supply and the Vee pin, and use some decent-sized caps on either side?

    If the single 10 Ohm resistor isn't enough, add another in series with another cap to ground.
     
Loading...