AD9208: Harmonic of Decimated FS mixes with NCO

Hi Dwijen,

I did not get any attachment. FYI, the virtual eval setup link is below

Virtual Eval | Analog Devices 

As you can see, it should be pretty clean, unless there is any sub-harmonic in your clock signal. What is your clock source?

Thanks

Umesh

Hi Umesh,

Thanks a lot for the reply. Sorry I missed on to attach image.

Our clock source is Keysight MXG N5182B Sig Gen. The spectrum is clean when NCO is not used. However when NCO is used I see this mixing component. I have verified this spur by changing Fs freq and NCO freq. Spur freq follows relations: Spur Freq = N*Fout-NCO

I have inserted image in below. 

Are you based out of Bangalore Office? If yes then we can arrange call or something to go over this issue.

Hi Dwijen,

Can you please give me some more information on your test setup? You are using the MXG to generate the tone. Are you using any filter in between the MXG and the ADC's input? If so what is it?

Can you also send me a screenshot of the same signal with the DDCs bypassed (full BW mode)? The setup for that is shown below. You can also find the setup in the wiki user guide.

Thanks

Umesh

HI Umesh,

Thanks for reply.

Yes I am using MXG for

MXG1 -> Generate 800MHz ADC input tone

MXG2 -> Generate 3GHz ADC Fs

MXG3 -> Generate Ref Clk  (375MHz for Full BW mode)

There are no filters between MXG and ADC input.

The Full BW Mode spectrum is given below as  you requested.

I have also tried putting 5% BW sharp cutoff BPF on the signal input. Input Sig Freq and BPF center freq is 800MHz.

I still see the spur at 48MHz. The ACE setup is same as that shared before.

48MHz = 3*375MHz - 700MHz

The BPF response is similar to that shown below. The below response is at 1GHz.

Hi Dwijen,

What does the full BW plot look like with the filter in line?

Also, what does the DDC spectrum look like when you remove the input signal (turn RF Off)?

I am afraid I dont fully understand the 48MHz = 3*375MHz - 700MHz math. Are you implying that this 425MHz tone aliases to 50MHz? I am also curious as to how you came to the 3*375MHz conclusion.

If you check the FFT spectrum, the 48MHz is tagged as "DC Frequency" in ACE. I am not sure if this is correct. Would it be possible for you to

- take an FFT (with DDCs on) with the 5% BPF between the MXG and analog input

- export the raw data and send it to me

Thanks

Umesh

HI Umesh,

Thanks for your reply. Please see my reply in line. I don't see option to attach file here on EZone. Can you provide your mail ID so I can send you raw data captured as you requested.

[Umesh]What does the full BW plot look like with the filter in line?

[Dwijen] See the snapshot in below

Full BW Spectrum with Filter Inline

[Umesh]Also, what does the DDC spectrum look like when you remove the input signal (turn RF Off)?

[Dwijen]See spectrum in below. Observe that Spur is still present at 48MHz

DDC Spectrum with RF Turned Off

[Umesh] I am afraid I dont fully understand the 48MHz = 3*375MHz - 700MHz math

[Dwijen] Umesh, My apologies for the typo in the above math. See in below updated math

48MHz = 2*375MHz (Second Harmonic of the Fout) - 702MHz (LO)

Here Fout is Fs/8. Second harmonic of the Fout is 750MHz. This beats with NCO set at 702MHz

To prove this I have provided another spectrum in below

I have changed the Fs to 2.991GHz. Fout = 373.88 MHz (Fs/8)

2*373.88 = 747.76MHz, NCO is set to 702MHz

Spur is seen at 45.76MHz = 747.76MHz - 702MHz

Hi Dwijen,

Thanks for the explanation. We here do not see this spur.

I am concerned that you are seeing this spur even when RF is turned off. What you see at 48MHz/3GSPS and 45.75MHz/2.991GSPS is the DC spur that has been rotated, and filtered. I have seen that in some cases this could be a possibility. To prove this, what we could do is to get the DDC setting to decimate by 16. And try the same thing (DDC spectrum without the RF).

FYI, we set the AD9208 in our lab and didnt see the spur that high. See below the full bandwidth and DDC spectrums.

Full Bandwidth

DDC

From our spectrum I have to believe this spur at 48MHz is either a harmonic or the DC spur. The ACE program correctly identifies the H3 Image as falling on -100MHz at 65dBc. This power matches the H3 power in the FBW mode plot.

We will continue to do more testing. However, we are going to be shutdown for the week and wont be back until after the New Year's.

Thanks

Umesh

HI Umesh,

Thanks for reply, Your always prompt response is really amazing and I appreciate same. 

I didn't quite understand following statement.

"What you see at 48MHz/3GSPS and 45.75MHz/2.991GSPS is the DC spur that has been rotated, and filtered".

Would you be able to provide some explanation on this DC spur rotation mechanism?

Regards,

Dwijen

Hi Dwijen,

Thank you for your kind words. I appreciate it.

In some cases where the frequency plan and the decimation ratio is such that the bandwidths are pretty wide, the DC spur can rotate around the decimated data and show up in the spectrum. This rotation is due to the fact that the NCO shifts the frequency and then the decimation filter filters it. If the shifted DC spur falls out of the filter band, the spur appears attenuated. If the shifted DC is in the filter band, then you wont notice any attenuation. My suggestion is that the spur you see is the former (DC spur outside of filter band).

The reason I suspect this is because of the DDC spectrum screenshot you sent with no RF. The spectrum should be clean but it still seems to have a -90dB spur at 48MHz. This could either be coming in from the clock, or the analog input, due to the frequency planning. I doubt this is the clock or analog since you have the analog input RF turned off, and as it it, you are using high quality signal sources. I will set this up in our lab when we return from our holiday shutdown.

Here is some additional information on how our DDCs work

What’s Up With Digital Downconverters—Part 1 | Analog Devices 

What’s Up With Digital Downconverters Part 2 | Analog Devices 

Please let me know if you have any questions.

Thanks

Umesh

Hi Umesh,

Thanks a lot for the explanation.

I have observed that,

In the snap shot you sent from your lab on Dec 22 has this spur present. However the amplitude you observe is lower than what we are observing. I have reattached the snapshot in below.

The amplitude of the 48MHz spur in the below snapshot is -98dBFS. The amplitude we are observing is around -94dBFS.

I have also observed that this spur amplitude is changing almost dB by dB with clock amplitude.

I have set clock power to be 10dBm on the MXG sig gen.

What clock power was used for the below snapshot in your lab?

Also is there a spec on the clock phase noise for AD9208?

Regards,

Dwijen