High frequency noise at the output of a 4th order filter

On a board, we have several identical channels of an analog signal generation chain. A single channel consists of an AD5452, followed by ADA4807-4. We use all 4 opamps on the ADA4807 package to construct a 4th order filter at a cutoff between 150-200khz. Our DAC's are driven by an FPGA.

AD5452 is in bipolar mode. It's reference voltage is 1.5V. AD4807 is driven by 1.5V and -1.5V.

In order to test the board, we generate a sine-wave around 60khz frequency from all channels. Our SPI speed is 50MHz. For a period of a sine-wave, we generate 32 samples only.

When we test the board, we see that *some* of these channels have a peculiar noise on the output, especially towards the negative part of the sine wave. Attached, you can see the waveform shape. If we zoom in to the bottom peak of the sine-wave, the "noise" frequency is approximatly 50MHz

Schematics of a single channel is below:

Also, we look at the output of the 3rd opamp in the above schematic, (to the left of R4), and the signal appears to be very clean at that point.

Our supplies appear to be clean as well. We tried adding a small capacitor (~100pf) at the output of the last op-amp, but it made the noise worse.

Any ideas what might be going wrong here?

  • It looks like it is trying to break into oscillation.

    --  I would never use a high frequency quad for an active filter, too many passives around a small IC.

       Big layout/crosstalk problem.   You have the output and input in adjacent op amps.  Switch to two duals.

    --  The thermal noise of a 10k resistor is 12 nV/rt-Hz;  the ADA4807 is 3.1nV, so you are thowing away

      performance/SNR.

    --  Is this a four layer board minimum??

    --  Did you remove all metal on all layers below the input pins??

    --  Did you use the Filter Wizard on our website and changed the default "low power" to "low noise"??

         All of your resistor values are too big.

    --  Adding a cap to any op amp output reduces phase margin and increases instability.

    Harry

  • - We picked the quad because of board real estate problems. Is there any application note to explain the issues that you've mentioned in your reply?

    - re: 10k resistor:   We used the resistor values from a filter design tool. Not sure at the moment whether it was the AD Filter wizard.

    - Board is 8L.

    - We did not remove all metal layers below all the input pins.

    - When we use the AD filter wizard now, we get the following for ADA4807:

    ... only R1B appears to be significantly smaller. Do you think the above is worth a try?

    Thanks

  • HA,

      A quad is not four op amps in one package, it is four functions in one piece of silicon.  At high 

    frequencies, you have coupling with stray C to substrate, through the bond wires, between

    the traces on the pc board, etc.  If you have capacitance on the inverting input of an op amp,

    you form a pole that leds to instability.  See:

    Planet Analog - Articles - Op amps: to dual or not to dual? (Part 1 of 2) 

    Planet Analog - Articles - Op amps: to dual or not to dual? (Part 2 of 2) 

    http://www.analog.com/media/en/training-seminars/tutorials/MT-045.pdf 

    I emailed the filter wizard author for comments, but he is out this week.

    With respect to real estate problems, If you make a tiny board, but it doesn't work, then what?

    "we did not remove all the metal...." you will have to do a board spin before you go into production.

    You could also try a slower op amp:  ADA4891-4, AD8618, AD8648, or ADA4084-4.

    Harry

  • 0
    •  Analog Employees 
    on Aug 2, 2018 3:46 PM
    This question has been assumed as answered either offline via email or with a multi-part answer. This question has now been closed out. If you have an inquiry related to this topic please post a new question in the applicable product forum.

    Thank you,
    EZ Admin