Post Go back to editing

ADA4945-1 THD behavior

Hello,

I used the ADA4945-1 as ADC front-end (LTC2380-24) with supply rails centered
to ADC 2.5V Vref ( Vcc= +7.5 / Vee = -2.5V for 10V total supply).

The FDA is configured as fully differential filter, with 2 cutoff frequencies.
2 small relays allow to select proper filter part. Because of the size of relays itself,
the FDA input trace are not  short as they would be.

So now i have two different questions about the ADA4945-1 :
---------------------------------------------------------------------------------
First, it is indicated in datasheet that exposed pad of the IC has to
be connected to negative rail. (but it not seem to be directly connected to it).
I have notice that too late and I connected it to ground plane instead.
So anyway, i don't have seen wrong behavior (current/voltage rails are ok).

Q1 ) Does it is a problem if exposed pad is connected to ground instead of VEE ?
(my Vee= -2.5V).

-----

I have made several THD measurements with the couple ADA4945-1 and LTC23800-24
and results are very promising.
Anyway i noticed that when i use "full power mode" (for lowest THD), the THD (H2) rise
much (~20-30dB more).
The FDA output doesn't seem to oscillate (i see nothing in 500MHz bandwidth), but if
i add a small amount (~500pF) of capacitor between FDA +/- inputs the THD remain ok.
I tried to reproduce the behavior in LTspice, but without success.
I presume that Spice model doesn't take THD behavior in account.

Q2) Do you have some suggestions about this ?
What can explain THD degradation ?

I hope you can help me to solve the issue.
Best regards.

Olivier

Parents
  • Hello,

    About my first question on the exposed pad, i have noticed after some reading,
    that both evaluation modules for the ADA4945-1 seem to have made same as me...
    The exposed pad is connected to ground, not to -Vee as writed on ADA4945-1 datasheet !
    So, must be connected to Vee or GND ?.....

    About my second question, i don't have explanation and i would really like to get some idea.
    If required, i can post schematics/layout.
    Please, i need help :-/

    Regards.

    Olivier

  • Hi Olivier,

    The noise performance will be best with the epad connected to the lowest potential (-Vs) so that’s where the epad should be connected.  On the first round of eval boards the epad was mistakenly connected to ground so when operated with dual supplies the noise will be slightly higher, on a single positive supply it won’t matter since ground is the lowest potential.  Later revisions of eval board will have the epad connected to Vs-.

    For your THD problem, the schematic and layout would help. 

    Regards,
    Goz

  • Hello,

    There is now near one month that is posted my message to ask some AD support about a critical issue i experienced.
    So, nobody here ?...

    Olivier

  • Hi Olivier,

    Apologies, can you show your distortion measurements with the full-power and low-power mode? What is your test set-up during measurement? It's really weird that at same set-up the THD performance worsens at full-power mode. How many units have you tested?

    Regards,
    Goz

  • Hello,

    So i explain all the issue in my previous message, and you ask for detailed measurements...
    I really think that it's not really what we can call efficient customer support.

    Anyway, i took some time these days to perform measurements in various conditions.

    I made measurements in both "low power mode" and "full power mode" of ADA4945-1,
    with normal schematic configuration (as my schematic/PCB sample previously posted),
    and then i repeated the measurements but adding a serial RC network between IN+/IN- of
    the ADA4945-1 (10 Ohms + 390pF).

    The signal source is an ultra-low THD 1kHz sine wave source with fully differential output.

    As you can show on FFT spectrum, specially in the "full power mode", the THD level is much
    worse than expected (datasheet specs) without the RC network.
    If the RC network is added, harmonics appear to decrease much.

    I suspect that my PCB routing add a small amount of capacitance between each IN+/- pins
    of the ADA4945-1 and then degrade it's THD behavior (bandwidth related ?).
    I have made some time domain measurements to check for instability of oscillation,
    but all signals are very clean with no trace of HF noise.
    I tried to reproduce also this behavior using provided AD SPICE model,
    but nothing is visible and no instability occur (THD behavior is not include in Spice model).

    So, my questions would be :
    - What AD think about my issue with the ADA4945-1 in this situation ?
    - What is the best way to compensate the capacitance at input pins ?
    - Can you confirm that ADA4945-1 is a voltage feedback type OPAMP ?


    I would be very grateful if AD help me efficiently to solve this.
    Best regards.

    Olivier


  • Hi Olivier,

    Please do this for me. Check for the response using Spectrum Analyzer without any input signal. This will help us check if the implementation, layout included is clean, no oscillations, etc. We suspect that you are having oscillations on your circuitry. By simulations, it won't be noticeable. However, it seems that the trace from the input pins going to RF and RG is around 4 cm or more. This is not good if we want to achieve low distortion. If we will able to verify that there are indeed oscillations, then you must redesign your PCB layout and minimize the mentioned trace length to remove oscillations.

    This behavior, one you have seen, are not repeatable on our eval boards.

    Regards,

    Jino

  • Hello,

    Thanks for the answer.
    You can show below the spectrum results with  no input signal, in both normal and low power
    mode of the ADA4945-1. There is no modification on schematics (no RC network).

    As you can see, all the spectrum bandwidth is very clean and no evident spurious frequency can
    be identified despite a large FFT averaging.
    In both normal and low power mode the spectrum is quite identical, with a bit more wide-band noise in full power mode.

    Note that the ADC after the ADA4945-1 is sampling at 1.536 MHz, with 8x SinC averaging filter.
    There is almost no anti alias filter at input, so even high frequency oscillations (>96k) will appear on FFT spectrum by aliasing. My own investigation in time domain to detect such instability has not been successful.

    Anyway, without redesign PCB layout, what would be the right way to mitigate the ADA4945 IN+/IN-  sensitivity to input to ground capacitance ? (compensation).

    Best regards.

    Olivier





  • Hi Olivier and all,

    If I may throw in my 2-cents on this:

    1. Mode pin: Your schematic shows the mode pin (pin 5) with a switch to select VAA (V+) or ground (EP which you did mention you had connected to ground but were thinking about changing to V-). According to the ADA4945 datasheet, full power (Mode = Vs+) and low power (Mode = Vs-) uses Vs- (not ground). Just wanted to make sure that you are complying with the Mode pin voltage requirements?

    2. Relay Effect / Trace lengths and routing: If the layout is a concern (as Jino has indicated), can't you "rework" (cuts and jumps) your board and eliminate the relay (and its traces) to confirm that you can lower your THD components? This is what I'd be doing if I were you because only you have the hardware on-hand than can prove (or disprove) what configuration gives you the best THD. I'd minimize the critical traces to the bare minimum just to verify that they are the culprit first before attempting any layout change / update. The EP connection (ground vs. V-) is also another thing to investigate with board rework (although that might be a bit more difficult).

    3. Input Source Harmonics: How certain are you that your differential source is not the source of some of these harmonics? I understand that Full Power mode of ADA4945 is your primary concern, but it'd be good to make sure these harmonics are not there to begin with. Have you looked at the source FFT somehow to verify?

    Regards,

    Hooman

  • Hi Olivier,

    In addition to Hooman's suggestions, can you also try sampling at slower rate and see if the harmonics fade away.

    Regards,
    Goz

  • Hi Olivier,

    Some more comments.

    1. Feedback type: I forgot to answer your question: Yes, the ADA4945-1 is a voltage feedback amplifier.

    2. Mode Input: On the schematic it looks like you have the ability to connect DGND to either VEE or GND, and the ability to connect MODE to either VCC or GND.  I believe the threshold for Mode is 1.2V above DGND, The Mode pin should work properly on your schematic if you have DGND connected to GND.  However, if you have DGND connected to VEE then I’m pretty sure the part would always be in higher power mode because connecting MODE to VCC would put it 10V above DGND and connecting it to GND would still put it 2.5V above the DGND level, higher than the 1.2V threshold.  Hopefully this makes sense.

    3. Looking for instability: A colleague of mine recommended that you also monitor the device supply current (both VAA and VEE or +Vs and -Vs) when you switch between Full Power and Low Power mode. I don't know how easy is that to do for you on your board because it'd mean you'd have to power the ADA4945 separately! There is about 3mA supply / quiescent current delta between the two modes. Furthermore, if there is any instability with the Full Power mode in your setup, that could also reflect on your current reading.

    You mentioned you looked for instability with a high BW scope. Have you tried to use a spectrum analyzer, in the upper frequency region, as well since those are much more sensitive? If you are sniffing for oscillations, I'd make sure your analyzer probe does not load the node (FET probe?) and try with and without signal and with varying input signal amplitudes.

    4. Eliminating the Filter as a culprit: Have you taken any THD readings with all filter components removed to create a standard gain of 1V/V FDA? I've seen capacitors wreak havoc on distortion even though you are specifying high quality C0G caps on your schematic. I would recommend you try that as it is fairly easy to do (removing the caps and replacing R24 and R30 (549ohm) with shorts).

    Regards,

    Hooman

  • Forgot to mention: I'd also experiment with adding a differential load to the FDA output pins (say 1kohm or 500ohm if the device can handle it) and noting the effect. Sometimes loading the output shows a significant change in response if there is any instability. I'd use minimum trace / lead lengths directly on the output pins. Better yet, two 500ohm series resistors tied to each complimentary output with the mid-point tied to VOCM for a 1k diff load returned to equivalent ground.

    Regards,

    Hooman

  • Olivier,

    I just noticed your VOCM buffer (LT6202) output is driving two large caps (C46 and C48 10uF and 100nF each). I'd use a series resistor (e.g. 100ohm) to isolate the LT6202 output from these large cap loads which are often cause for instability. You could just tie VOCM to a lab DC source if you wished (with close by decoupling cap to make sure it's not an issue).

    Regards,

    Hooman

Reply
  • Olivier,

    I just noticed your VOCM buffer (LT6202) output is driving two large caps (C46 and C48 10uF and 100nF each). I'd use a series resistor (e.g. 100ohm) to isolate the LT6202 output from these large cap loads which are often cause for instability. You could just tie VOCM to a lab DC source if you wished (with close by decoupling cap to make sure it's not an issue).

    Regards,

    Hooman

Children
No Data