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.

Noise coupled from switching regulator?

Other Parts Discussed in Thread: ADC10D1500, LP38513, LM25576, ADC10D1000

I am testing a board with an ADC10D1500 and I am getting low values of SINAD for some cases. 

I am using a 1.4GHz clock and the Demux-DES mode of the ADCs and I see distortion in the output data that may match with a frequency of ~300KHz. This frequency could be the clock in the switching regulator. I use a LMX25576 followed by a linear regulator (LP38513).

I followed the layout recommendations in ADC datasheet regarding ground a power planes.

Do you think that this clock in the switching regulator can be the root cause of the distorted signal and, hence, low SINAD? If so, any recommendation to erase this effect? Is it more likely to be coupled through the supply or through the ground planes? 

  • Hi Sergio,

    It is very likely that the 300KHz noise from switching regulator is coupling with your signal and making your SINAD look bad. You want to separate the power supply section from the analog input section of the ADC. Also make sure your analog input traces on board going to the ADC are as short as possible. For reference you can look at ADC10D1500 evm board . Here is the link for the evm. Schematic and layout files are under Design file section.
    www.ti.com/.../adc10d1500rb board&tisearch=Search-EN-Everything

    Also here is the link to app note which talks about general guidelines about layout and schematic for giga sample ADC.

    www.ti.com/.../litabsmultiplefilelist.tsp

    Regards,
    Neeraj Gill
    HSCC
  • Hi Neeraj,

    Thank you for your recomendations, we already followed that AN when designing the PCB and the regulators architecture choice is based on the EV. Unfortunatelly there is not much spare area in the PCB to separate more the switching regulator and the ADC. I am going to try replacing the choke between the switching regulator and the linear regulator (BLM18SG121TN1) as I have measured again the SINAD with an external voltage source driving the linear regulator and that noise seems to be removed (switching regulator was still working but its output was not used as input for the linar regulator). Do you have any recommendation to replace this component?

    I have another question about the SINAD. We use the DESIQ configuration, the same input signal is provided to the two channels (I and Q) of the ADC10D1500. As I told you before we followed the recomendations and the layout is symmetric from the input balum to the inputs of the ADC10D1500 (the architecture is based on another AN from TI "Driving GSPS ADCs"). If I generate a sinusiodal input of 100MHz and calculate the SINAD and ENOB for the 4 outputs separately (DI, DId, DQ, DQd) I get an ENOB over 8 bits for the 4 channels (in this case the equivalent sampling frequency would be fs/4 (750MHz), enough to convert an input of 100MHz). However, when I merge the four channels in an unique vector and perform the same calculations I lose around 1 bit in the ENOB. Do you think this could be due to different offset and/or gain between the four output channels? If so, is there a way to solve it via registers or calibration?


    Thanks in advance.

    Sergio

  • Hi Sergio,

    Sorry for the late reply, I had some laptop issues.
    If you look at the ADC10D1500 BOM you will see we have used BLM18SG260TN1D on that.

    Regarding your ENOB issue I would recommend looking at the FFT of all 4 channel combined and seeing what kind of spurs are affecting your performance. if you have offset mismatch you will see spurs at Fs/2 and Fs/4 and at DC. if you have gain mismatch and timing skew between the composite ADC you will see a spur at Fs/2-Fin. Adjust offset, gain accordingly.

    Please refer to following app note for detailed information.
    www.ti.com/.../slaa617.pdf


    Regards,
    Neeraj Gill
  • Thank you Neeraj. I have that AN, I am going to check it again.


    About the choke, there is a BLM18SG121TN1D choke between the LM25576 and the LP38513 and a  BLM18SG260TN1D at the ouput of the LP38513. As I have checked that providing  the LP38513 with a clean supply voltage solves the problem, I am focusing on "cleanning" the supply at that point.

    I opened a question in the non-isolated DC/DC forum () beacuse I am considering the possibilty to increase the L of the buck converter in order to reduce the ripple. Maybe this could help.

    Regards,

    Sergio

  • As a quick example. Here are the results for a 100MHz input signal sampled at 2.8GHz (demux and DES mode). Only one channel (I) and the four channels (I, Q, Id, Qd).

  • Hi Sergio

    The spurs that you are seeing on the EVM look larger than normal. Please make sure your ADC is calibrated after to warms up. Please also make sure the signal source you are feeding into the ADC is clean please use bandpass filter to clean the signal before it is fed into the ADC.

    I don’t have ADC10D1500 board with me so I am making these measurements on ADC10D1000 in DES mode.

    Here is the filtered signal but the ADC has not been calibrated after power up and ENOB is about 8.0 bits.


     

    Here is the unfiltered signal applied and you can see ENOB is 7.4. The ADC has been calibrated


     

    Here is the measurement done after part has been calibrated and input signal is filtered with a band-pass filter. As you can see I am getting the ENOB of 8.4.

     

    Regards,

    Neeraj Gill

     

  • Thank you Neeraj,


    I was not filtering the input signal. I use directly the output of an Agilent signal generator. I am going to repeat the measurements following this procedure:

    - Power-on calibration

    - Wait some minutes to allow the self-heating of the ADC

    - Select the operating mode

    - On-demand calibration without input signal

    - Apply the input signal (with filter if possible)

    - Perform measurements

  • Hi Neeraj,

    Can you tell me the reference of a recommended filter for these measurements? (the one you are using if it is possible).

    Regards,

    Sergio
  • Hi Sergio,

    Sorry for late reply, I was out on month long vacation.
    I was using Trilithic tunable bandpass(5MHz) filter for my measurements.

    Regards,
    Neeraj Gill
  • Thank you, Neeraj.


    PS: I hope you enjoyed your holidays.