Wideband operation with AD9162

Hi,

My customer is considering to use AD9162 to generate a wideband waveform (Center frequency: 4GHz, iBW: 2GHz)

I have following questions for this case. Could you answer those questions?

  1. Could you double check AD9162 can generate 4GHz +/- 1GHz signal?
  2. Is there any measurement result with same signal condition?
  3. For best flatness, I thought Mix mode is recommended for this case, is it correct?
  4. I think Nyquist Image should appear at DC to 2GHz. In this case, what is maximum spurious level due to image?

Best Regards,

Taiki

Parents
  • +1
    •  Analog Employees 
    on May 29, 2020 10:31 PM

    Hi Taiki,

    Thanks for reaching out. Below are my answers to your questions:

    1. Yes, the AD9162 can generate a 4 GHz +/- 1 GHz signal. However, your system requirements (particularly the iBW spec) are demanding, so the performance of the AD9162 will not be nearly as strong as the datasheet results which are taken over much narrower iBWs. A complex filtering scheme may also be necessary in order to mitigate aliases and images. You should carefully read Pages 65-70 of the AD9162 datasheet so that you understand the many implications and caveats associated with filtering, interpolation, and NCO operation. For example, you may assume that you can set fDAC = 5 GHz and fNCO = 4 GHz, and then you can easily shift a 2 GHz BW signal from baseband to 4 GHz. However, fNCO must be less than fDAC/2 when the FIR85 filter is disabled. You can enable the FIR85 filter which extends your fNCO range up to fDAC, but then your signal bandwidth will be limited to fDAC/4 since L = 1. You may think you can utilize the bandwidth from -fDAC/4 to fDAC/4 to achieve the full 2 GHz, but, when L = 1, the DAC output is limited to real data, meaning that the negative frequencies will simply be a mirror image of your positive frequency data.
    2. While the datasheet does not contain any measurement sets with your desired signal conditions, I used our AD9162 evaluation platform to generate an example data set for you. While the multitone data set may seem a bit unusual, it still illustrates the AD9162’s ability to generate wideband signals. In this case, note that the bandwidth of the signal is technically 1.8 GHz due to the 90% BW filters used for interpolation, and some of the signal magnitude roll-off may be due to cable loss rather than DAC performance.
    3. Mix mode offers the best flatness from 3 GHz to 5 GHz, but 2xNRZ mode would also work decently well. Keep in mind that Mix mode is effectively amplifying the alias of your input signal near fDAC.
    4. The position of the Nyquist image will vary depending on a variety of factors, the most important of which is fDAC. If using fDAC = 5 GHz, then, yes, you will see an alias from DC to 2 GHz. However, this alias from DC to 2 GHz would be the least of your worries—the major challenge would be the alias from 5 GHz-7 GHz which is nearly impossible to filter. Increasing fDAC to 6 GHz will shift the position of the aliases, making the filtering of nearby aliases more feasible. The two images included in this post should give you a rough idea when it comes to SFDR. Measuring from the highest point of each cluster, the nearest alias is approximately -16 dB down from the carrier, and the nearest spur (fDAC in this case) is about -30 dB down.

    Input Signal: 21 Tones from -1 GHz to 1 GHz, Evenly Spaced by 100 MHz
    DAC Settings: fDAC = 6 GHz, L = 3, FIR85 Enabled, 90% BW Filters, 2xNRZ Mode, NCO Enabled @ 4 GHz

    Input Signal: 21 Tones from -1 GHz to 1 GHz, Evenly Spaced by 100 MHz
    DAC Settings: fDAC = 6 GHz, L = 3, FIR85 Enabled, 90% BW Filters, Mix Mode, NCO Enabled @ 2 GHz

    A better solution for your customer may be the AD9166 (www.analog.com/.../ad9166.html). The AD9166 is currently Pre-Release.

    Let me know if you have any additional questions! Your question is quite open-ended, so there is much to discuss.

    echaykov

Reply
  • +1
    •  Analog Employees 
    on May 29, 2020 10:31 PM

    Hi Taiki,

    Thanks for reaching out. Below are my answers to your questions:

    1. Yes, the AD9162 can generate a 4 GHz +/- 1 GHz signal. However, your system requirements (particularly the iBW spec) are demanding, so the performance of the AD9162 will not be nearly as strong as the datasheet results which are taken over much narrower iBWs. A complex filtering scheme may also be necessary in order to mitigate aliases and images. You should carefully read Pages 65-70 of the AD9162 datasheet so that you understand the many implications and caveats associated with filtering, interpolation, and NCO operation. For example, you may assume that you can set fDAC = 5 GHz and fNCO = 4 GHz, and then you can easily shift a 2 GHz BW signal from baseband to 4 GHz. However, fNCO must be less than fDAC/2 when the FIR85 filter is disabled. You can enable the FIR85 filter which extends your fNCO range up to fDAC, but then your signal bandwidth will be limited to fDAC/4 since L = 1. You may think you can utilize the bandwidth from -fDAC/4 to fDAC/4 to achieve the full 2 GHz, but, when L = 1, the DAC output is limited to real data, meaning that the negative frequencies will simply be a mirror image of your positive frequency data.
    2. While the datasheet does not contain any measurement sets with your desired signal conditions, I used our AD9162 evaluation platform to generate an example data set for you. While the multitone data set may seem a bit unusual, it still illustrates the AD9162’s ability to generate wideband signals. In this case, note that the bandwidth of the signal is technically 1.8 GHz due to the 90% BW filters used for interpolation, and some of the signal magnitude roll-off may be due to cable loss rather than DAC performance.
    3. Mix mode offers the best flatness from 3 GHz to 5 GHz, but 2xNRZ mode would also work decently well. Keep in mind that Mix mode is effectively amplifying the alias of your input signal near fDAC.
    4. The position of the Nyquist image will vary depending on a variety of factors, the most important of which is fDAC. If using fDAC = 5 GHz, then, yes, you will see an alias from DC to 2 GHz. However, this alias from DC to 2 GHz would be the least of your worries—the major challenge would be the alias from 5 GHz-7 GHz which is nearly impossible to filter. Increasing fDAC to 6 GHz will shift the position of the aliases, making the filtering of nearby aliases more feasible. The two images included in this post should give you a rough idea when it comes to SFDR. Measuring from the highest point of each cluster, the nearest alias is approximately -16 dB down from the carrier, and the nearest spur (fDAC in this case) is about -30 dB down.

    Input Signal: 21 Tones from -1 GHz to 1 GHz, Evenly Spaced by 100 MHz
    DAC Settings: fDAC = 6 GHz, L = 3, FIR85 Enabled, 90% BW Filters, 2xNRZ Mode, NCO Enabled @ 4 GHz

    Input Signal: 21 Tones from -1 GHz to 1 GHz, Evenly Spaced by 100 MHz
    DAC Settings: fDAC = 6 GHz, L = 3, FIR85 Enabled, 90% BW Filters, Mix Mode, NCO Enabled @ 2 GHz

    A better solution for your customer may be the AD9166 (www.analog.com/.../ad9166.html). The AD9166 is currently Pre-Release.

    Let me know if you have any additional questions! Your question is quite open-ended, so there is much to discuss.

    echaykov

Children
No Data