• State TI Thinks Resolved
  • Locked Locked
  • Replies 5 replies
  • Answers 3 answers
  • Subscribers 47 subscribers
  • Views 213 views
  • Users 0 members are here
Support feedback
Options
Options
Similar topics

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

DAC3171: DAC3171 14-bit D/A convertor (Creating AM modulation on the signal @-60 db relative to the carrier)

manu mathew1
Prodigy 20 points
Part Number: DAC3171
Other Parts Discussed in Thread: ADS4229, LP5910

Dear Sir,
We have an application where we are using the DAC3171 14-bit D/A convertor.
We have it working on our board, clocked with a 500 Mhz clock, steered by a Xilinx arxiv FPGA, rendering a 171 Mhz data signal at full scale.
Now, it is working correctly, with great performance. Only, we see AM modulation on the signal @-60 db relative to the carrier.
The AM modulation frequency is about 100 kHz, and this is the resonance frequency of the L-C circuits used for the power supply decoupling.
The DAC has 4 supplies {Vcc1V8_DAC_CLK, Vcc1V8_DAC, Vcc1V8_DAC_IOVDD, Vcc1V8_DAC_DIG} decoupled with a
Murata Ferrite bead BLM15EX471SN1D and a ceramic capacitor of 1 uF (+ an additional 1 nF). These circuits have a resonant
frequency around 100 Khz and we see the DAC amplitude modulate the output (-60 dB down) at 100 Khz.
The 3.3 Volt supply is direct from a regulator.
Note that we found this issue elsewhere on the board too. The 500 Mhz clock is produced by a 250 Mhz doubler, and likely has
some amplitude modulation, but no more than -50 dB.
Q.
- Are you sure this LC power supply decoupling is OK ?. We have a suspicion this problem is caused by the supply/DAC
combination.
- Would it be better to eliminate the ferrite beads and just use LDO regulators ?. (The board is full of LDO's anyway.)
- Are there some supplies more sensitive than others ?.
Q.
- Note: We have the ADS4229 on the same board, decoupled in the same way. Does it also have this sensitivity ?.
(I cannot measure).

  • 0
    Intellectual 420 points

    I am just another designer looking at the forum.

    The datasheet for the DAC3171, section 9, calls out three supply rails which are sensitive to noise.  The current requirements are in the single milliamp range for those rails.  You can afford significant filtering on those nodes, given the small currents.  Your power supply rail names don't correspond to the names on the data sheet, so let me reframe my comment:

    I would recommend a much larger ceramic capacitor after the ferrite bead to filter the low current inputs for CLKVDD18, VDDA18, and VDDA33.  I like to use BLM18RK102 ferrites and a 10uF cap to avoid the resonance for my low current feeds, typically supplying many 10s of mA.  If you model the BLM15EX471, you can find a proper capacitance value to eliminate the resonance.  A resistance in series with the ferrite bead may also be a simple way to help damp the resonance without notable voltage drop for 4mA class of consumption.

    If you have supply noise below the resonance of the LC filter from the ferrite inductance, it will come through without significant attenuation.   inexpensive LDOs like the LP5910 are good for getting rid of significant rail noise with a small footprint.  But if your problem is only with the resonance, a larger capacitor can be all you need to fully damp the resonance so it's no longer an issue.

  • 0
    TI__Mastermind 20130 points

    Hi Manu,

    One way to test out to see if the on board power supply is the issue, is to remove the LC filter on each supply domain and wire in a clean bench supply. If you can forward your schematics and possibly the layout we can look at the board for review.

    Another test to try is when the analog output frequency is moved to another location does the 100kHz spur follow the new frequency or stay put?

    Regards,

    Rob

  • 0 in reply to Rob Reeder
    Prodigy 20 points
    Hi Rob,
    Well, this is what I was afraid of.
    Can we have an updated datasheet? I mean:
    a. The oscillation cannot be avoided. [The Q for the ferrite is high @ low amplitude, low @ high amplitude.
         I deduce that from other circuits, that have uncontrolled oscillation in Spectre and controlled on the board.]
    b.Is it possible for TI to provides the data on the power supply rejection rate of
        their devices on this oscillation frequency.
      -> Means, there are 4 supplies, for each of them, the PSRR needs to be spec'ed around 100 Khz.
    Br
    Manu
  • 0 in reply to manu mathew1
    Intellectual 420 points

    The oscillation can most certainly be avoided.  Ferrite bead filters are LRC networks which can be underdamped (the oscillations you're seeing), overdamped (slower response to changes) or critically damped.  The ferrite bead has inductance and resistance at the lower frequencies which can work with the capacitance as-is or augmented with resistance from a fraction of an ohm to several ohms, depending on current requirements.  Your capacitance is much too low for the L and R in the ferrite bead.

    The data sheet, section 9, says which supplies are subject to noise.  The power supply rejection ratios for those rails over frequency are not included data, however.

    Simulate your power feed.  Eliminate the resonance.  You end up with solid rails.

    Sincerely,

    - John

  • 0 in reply to manu mathew1
    TI__Mastermind 20130 points

    Hi Manu,

    The most current datasheet is at the link below...

    Measuring the PSRR on our EVM won't give a good indication on what might be happening on your system board. Decoupling and PCB layout can all play a factor on the results. If you would like for us to look over your schematic and layout, please forward and we are happy to review.

    It would be best to follow John's advice below regarding the power supply to help eliminate the resonance.

    Regards,

    Rob