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.

AFE5818: HPF or DC compensation distortion?

Part Number: AFE5818

Hi,

We are experiencing problems with signal distortion when sampling "low frequency" signals in TGC mode with our own prototype hardware. We enable the LNA HPF (200 kHz cut-off) and PGA HPF. INP is ac-coupled with 15 nF and INM bypass is 15 nF. We don't use any active termination, but we do ac-couple the active termination pin with 1 uF so we can take advantage of the input voltage clamp function.

Injecting a 30 kHz CW test signal above a certain amplitude the sampled signal starts to run away in amplitude and distort (multiple harmonics appear). The signal measured at the AFE INP with oscilloscope is a clean sinusoid however, so there seems to be something going on inside the AFE. Disabling the LNA HPF and PGA HPF seems so "solve" the problem, but then we lose valuable ADC range due to the non-zero DC level.

It seems to us that most likely the HPF / DC-compensation circuits are causing problems. Increasing the bypass capacitor at INM from 15 nF to 1 uF didn't help by the way.

In more detail the AFE is initialized as in the datasheet power sequence including a hard reset, then the following registers are programmed:

0x000001 (soft reset)
0x010014 (must write)
0x032000 (14 bit lvds)
0x040011 (MSB first, 14 bit adc)
0x430000 (no clock skew)
0xc30018 (PGA HPF off, LPF 10 MHz)
0xc41000 (LNA HPF off)
0x150000 (HPF off),
0x020080 (sync pattern)

Here there is some LVDS timing/synchronization against the test pattern happening in our FPGA...

0x020000 (normal sampling)

0x000000 (ADD_OFFSET = 0)

0xc30008 (24 dB PGA, enable PGA HPF, 10 MHz LPF)
0xc4c200 (individual LNA gain, LNA HPF, 1.5 Vpp input clamp, no active term.)
0xc50200 (low noise and noise figure mode)
0xc60200 (DIS_CW_AMP=1, TGC mode)
0xc9aaaa (all LNA 12 dB gain)
0xcb0088 (EN_DIG_TGC=1, 200 kHz LNA HPF)
0xcd0000 (PGA_CLAMP_HALF=0, SUPRESS_HIGHER_HARMONICS=0, V2I_CLAMP=0, RED_LNA_HPF_3X=0)

For a few different input levels we measure clean sinusoids of ~30 kHz with an oscilliscope at the AFE inputs, but what we observe in sampled data is shown in the below plots (where I = AFE ch 16, Q = AFE ch 15 shown for part of the sequence sampled, while the FFT is done on the complete sequence).

Input ~10 mVpp:

image2019-3-15_13-30-32.png

Input ~35 mVpp (some harmonic components are starting to show more clearly):

image2019-3-15_13-30-48.png

Input ~100 mVpp (here there is a clear jump in signal level and harmonic content):

image2019-3-15_13-31-3.png

Is there something we are doing wrong in the register programming or are there some limitations in the HPF/DC-compensation circuits that we have failed to understand?

Thank you,

Christer

  • Hi Christer,

    How are you?
    Thanks for using AFE5818 device.
    Could you please let us know whether you are using AFE5818EVM and TSW1400EVM together?
    Also here are all the information including AFE5818EVM and Users Guide on the TI website.

    www.ti.com/.../afe5818evm
    www.ti.com/.../slou430a.pdf

    Then you can use the GUI and test and check if Ramp Test Pattern
    working on your AFE5818EVM setting.

    Thank you and best regards,
    Chen
  • Hi,

    We don't have the TSW1400EVM. We have produced our own hardware that includes an AFE5818. The test patterns look fine. The distortion problem is reproducible with signals in the ~30 kHz range, as per my description above, while e.g. a signal of ~500 kHz doesn't have this problem.

    Sincerely,
    Christer

  • My posted images look broken to me. I'll try to repost them again here:

    Input ~10 mVpp:

    Input ~35 mVpp (some harmonic components are starting to show more clearly):

    Input ~100 mVpp (here there is a clear nonlinear jump in signal level and harmonic content):

  • Just to make it clear again. This is in TGC mode and I and Q refers to channel 16 and 15 on the AFE. These are IF signals from an external mixer (not the AFE CW mixer). The FFT spectrum is the complex FFT(I + j*Q).

    Thanks,
    Christer
  • And the distortions are only seen in sampled data, not in the IF signals at the inputs of the AFE.

    Is there are problem with our register programming or is there a problem with the AFE DC offset compensation circuit?
  • So I suggest for you to try and reproduce the problem with your TSW-setup please program (after LVDS setup and such):

    0xc30008
    0xc4c200
    0xc50200
    0xc60200
    0xc9aaaa
    0xcb0088
    0xcd0000

    and then inject a 30 kHz 200 mVpp signal. We were able to reproduce the problem when connecting our FPGA platform (*not TSW1400EVM*) to a 5818EVM board. The sampled signal distorts in the same way. Perhaps you can reproduce it with TSW1400EVM?

    Thanks,

    Christer

  • Hi Christer,
    Thank you for let us see those plots.
    Yes, I will look into your questions and settings using AFE5818EVM and TSW1400EVM here
    to see if we can get the same results.
    Just at first for quickly checking, you can set the register of AFE5818
    by setting RAMP Pattern mode and then when you can receive the output data
    from your FPGA side to check if the captured data look like a ramp without any missing code or not
    (just easy check to make sure if this is the captured data working or not).

    Thank you!

    Best regards,
    Chen
  • Hi,

    Here is the RAMP output:

    Let me know how you are progressing.

    Thanks,

    Christer

  • Thank Christer,

    Thank you for showing me know ramp pattern test output.
    Looks like the captured output data work perfectly!
    Thanks!
    Also looks like you are using 14bit mode
    so the output Ramp data are from 0 to 16383. right?

    Thank you!

    Best regards,
    Chen
  • Yes, we use 14-bit mode and get a 14-bit output range.

    Are you able to test with TSW1400? We are very interested to see if you get or don't get distortion in the given test case with both LNA HPF and PGA HPF enabled.

    Thanks,

    Christer

  • Hi Christer,

    How are you?
    We are concerning about your question settings of running very low frequency (such as 30kHz) input signal to AFE5818 device.
    Yes, we also can see the same behavior such as higher harmonics
    when the input signal running from higher frequency such as 1MHz dropped down to lower frequency such as 200kHz.
    If you are using AFE5818EVM or you set your own designed board
    with Cbypass=15nF at INMx input pin.
    (Notice: our current AFE5818EVM board is designed by using 15nF right now.)
    In this case, the input signal may not work well
    at the lower frequency.
    Please refer this information to the AFE5818 Data Sheet at page 35.
    Because usually for the Ultrasound application using AFE5818 device,
    it does not use very low frequency signal to the input pin.
    If you need to run such low frequency input signal,
    you may need to change the Cbypass to 1uF.

    Yes, I will double check with the system engineer here
    and let you know soon.

    Thank you
    and best regards,
    Chen
  • Hi,

    Changing the Cbypass from 15 nF to 1 uF or programming LNA_HPF_PROG or RED_LNA_HPF_3X differently doesn't affect at what *input level* the harmonic distortion appears in terms of input voltage (Vpp). Changing Cbypass or HPF programming only affects the sampled level by the ADC.

    The only way we have found to get rid of the distortion is to completely disable LNA and PGA HPF. If either or both HPF are enabled there are distortion problems. Disabling the HPF is not an option as the resulting DC offset level is too high, will vary between individual chips and temperature.

    The datasheet talks about setting low cut-off frequencies (< 2 kHz) so to us it seems unreasonable that AFE5818 cannot handle a 30 kHz input signal within the input voltage range properly. That's why we were hoping we have incorrectly programmed the circuit.

    Please have your system engineer confirm if this is a non-documented hardware issue or how to program the IC correctly to solve the issue.

    Thank you,

    Christer

  • Hi,

    Can you share your measurement results and settings so we can confirm that we're seeing the same amount of distortion?

    From your system engineer we'd like an explanation why the datasheet says "the signal chain of the device can handle signal frequencies as low as 10 kHz and as high as 50 MHz" (page 33), because there is nothing in the documentation otherwise suggesting there is performance degradation at lower frequencies.

    We're of course still hoping that we've programmed the IC incorrectly and you can tell us how to fix it.

    Thanks,

    Christer

  • Hi Christer,

    Thanks for giving us more detail information and description for AFE5818 device.

    Yes, we will share our measurement with you.

    please let us have about 3 days to do so.

    I have informed this message to our system engineer.

    We will reply to you soon.

    Thank you and best regards,

    Chen

  • Hi Christer,

    How are you?
    Before we can give you more measurement results,
    here are some suggestions from our system engineer:
    1. After you changed INM input pin with 1uF capacitor (originally: 15nF),
    on data sheet page 131, Table 72,
    Must set PGA_HPF_DIS = 1
    (also see page 15)

    2. also see page 132, table 74
    to make sure LNA_HPF_DIS (0=enable HPF, 1=disable HPF)

    and page 142, table 89
    to set LNA_HPG_PROG

    and page 143, table 91
    to set RED_LNA_HPF_3X

    Please follow below for setting:
    on data sheet page 23, figure 13, (also see page 15)
    finally set up LNA_HPF_PROG and RED_LNA_HPF_3X to let HPF
    (to be -3db cutoff = 16kHz)

    3. so far we don't have harmonics test report at low freq signal
    for example 30kHz, therefore there is no information for that.


    Thank you very much!
    Have a nice day!

    Best regards,
    Chen


    Due to the signal balance, after you changed the INM input pin
    with 1uF
  • Hi,

    We have not found any useful combination of LNA HPF disable/enable and PGA enable/disable or LNA_HPF_PROG or RED_LNA_HPF_3X. There is always significant distortion unless both LNA and PGA HPF are disabled (this is not a practical option).

    We have spent too much time trying different combinations. Please show us measurements and settings that doesn't cause distortion, e.g:

    * LNA HPF enable
    * PGA HPF enable
    * Vin = >200 Vpp at 10~100 kHz

    Thanks,

    Christer

  • Hi Christer,

    Yes, we also did very similar tests here are the results.

    Test signal at Low freq=113kHz.

    1) We tested with LNA_HPF=Disable, and PGA_HPF=Disable with very low freq input signal.

    The measured output signal is good.


    2) We tested with LNA_HPF=Enable (LNA_HPF_PROG='10', RED)LNA_HPF_3X=0)

    and PGA_HPF=Disable.

    The measured output signal shows big harmonics (distortions) (as the customer mentioned).


    Therefore, for running signal at very low freq (10-100kHz),

    please set LNA_HPF_DIS=1 and PGA_HPF_DIS=1.

    Thank you!

    Best regards,

    Chen

  • Hi,

    Nowhere in the datasheet does it state that the high-pass filter functions must be disabled for the specification in the datasheet to be valid, e.g. to work properly with 30 or 113 kHz signals. 

    On the contrary, from the datasheet:

    • Page 1: "Applications - Sonar imaging equipment"
    • Page 33: "This broad analog frequency range enables the device to be used in both sonar and medical applications."
    • Page 78: "In addition, the signal chain of the device can handle signal frequencies as low as 10 kHz and as high as 50 MHz. As a result of the device functionality, the device can be used in various applications (such as in medical ultrasound imaging systems, sonar imaging equipment, radar, and other systems that require a very large dynamic range)"

    This is not a trivial matter of just disabling the HPF. That is simply not a viable option in a real product that requires wide bandwidth and high dynamic range. 

    Does this mean you are admitting that the AFE5818 is poorly or incorrectly designed and that you've failed in your verification? If this is true, you do understand that we will need to replace the AFE5818 at the cost of our customer? The datasheet is from 2015 and we have to drag this information out of you? This doesn't give us much confidence in choosing another ADC or other products from TI.

    Please advise.

    / Christer

  • Hi Christer,

    Yes, I will forward your suggestion to other engineers.

    I believe they will give you more idea to solve this problem.

    If you don't mind, could you please send me the email address?

    Thank you very much again!

    Best regards,

    Chen

  • I have made a Connect request.

    Thanks,
    Christer
  • Hi Christer,

    How are you?

    Here we have verified your request again

    including all the resistor settings and register settings as follows.

    And also you can see our measurement shown in the following photo.

    (All the correct settings are shown inside two red color boxes.)

    1) For very low input signal freq:

    The following settings must be required:

    make sure Connect INP with 0.1uF cap

    make sure Connect INM with 1uF cap

    Set LNA_HPF_DIS=0 (this can help reduce the offset difference from channel to channel)

    Set PGA_HPF_DIS=1 (this HPF is used for input high freq signal only)

    Here are two boxes shown:

    You can see if PGA_HPF_DIS=0, then the signal can generate higher harmonics

    so please turn off PGA_HPF to avoid this issue.

    After you disable PGA_HPF, you can see the output data working correctly.

    2) We used passive termination Rt=50 ohm (shown as the following plot).

    Please try it if you can.

    Hopefully this time you can see your measurement as good as we measured here.

    Thank you very much!

    Best regards,

    Chen

  • Hi,

    Thanks for the input.

    We have also concluded that the LNA HPF is bad but not quite as bad as the PGA HPF (which is really bad), but even the LNA HPF alone with PGA HPF disabled gives far from acceptable performance. There are of course two obvious ways to reduce the HPF distortion effect: turning off the HPF (LNA_HPF_DIS=1) or choosing an input signal frequency near or above the cut-off (100 kHz). That's what you just showed us, and by effectively bypassing most of the HPF of course the distortion is reduced.

    The expectation from any customer reading the datasheet will be that the HPF is functional and linear over the full LNA input range, because there is no information to suggest the contrary. The way things are developing we're losing our initial hope that it's a configuration error on our end. Now it looks like it's poor hardware design and even poorer documentation of those shortcomings.

    From the datasheet: "the signal chain of the device can handle signal frequencies as low as 10 kHz and as high as 50 MHz". So please show us a way to reach acceptable HPF distortion performace with a 10 kHz input signal.

    / Christer

  • Hi Christer,

    How are you?
    Our engineers has contacted with the application engineer
    who will reply to you very soon.

    Thank you very much!

    Best regards,
    Chen
  • Hi,

    We've managed to get further support through a TI FAE in our country, but I will leave a final comment for any other customer struggling with AFE5818.

    The concensus from us and TI (based on all measurements provided here and through direct contact) seems to be that the PGA HPF cannot handle any significant signal amplitude below its 80 kHz cut-off or it will go bananas (I say bananas because it's not a matter of simple "distortion", as the PGA HPF starts to generate a whole bunch of uncorrelated signal content). In conclusion this makes the PGA HPF rather useless and it must be disabled.

    However, I think the main reason we've struggled so long with AFE5818 is that based on the poor PGA HPF implementation we most likely mistook an external hardware fault on our main test board (perhaps a broken BGA solder joint) as another issue with the AFE5818 implementation.

    So, to use the AFE5818 in TGC mode we've concluded that one needs to:
    * Disable PGA HPF
    * Reduce the LNA HPF cut-off to a few kHz by: Cbypass = 1 uF, Cin = 100 nF
    * Implement an external HPF. We did this by a series capacitor isolated by an opamp stage (because increasing Cin directly resulted in unstable AFE5818 operation)

    Thanks,
    Christer