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.

DAC37J84: Improving Noise Floor

Part Number: DAC37J84
Other Parts Discussed in Thread: LMK04828

Hi team,

Could you please suggest me how to improve the NF?

My customer has tried the things below, but it was not effective:

- Add decoupling cap on VCC

-  Stop inputting SYSREF after the start

- Decrease DACCLK input amplitude

DACCLK is from LMK04828 as PECL, so it is expected that phase noise should be as small as possible, but my customer considers to review the filter cut-off frequency.

Is there any possible way to get the NF better?



  • Hello Itoh-san,

    Could you please advise how the current NF is compared with the datasheet. Please advise the sampling rate and IF for good comparison. 


  • Hello Itoh-san,

    We are waiting for your reply. thanks.


  • Hello Mr.Kang

    I'm Sorry for lack of the information.

    I didn't measure NF directly.

    I measured  peak of signal which seted config47 as 0x0001 and config48 as 0x 0005 by useing Spectrum Analyzer .

    I changed Vcc decoupling cap ,dacclk phase noise,stop sysref then compared before changing peak of signal and after changing signal.

    But nothing changed.

    Sampling rate clk is 614.4MHz and IF is 138.3±5MHz.

    -Takeshi Miki 


  • Hi Takeshi-san,

    You are exercising the right steps to improve NF.

    I believe you are still outputting single CW tone when measuring the noise floor with the above configuration. It is important that the spectrum analyzer have correct noise floor enhancement setting to measure the noise floor properly. Many times, it is the spectrum analyzer that is limiting the noise performance, not the DAC. We have recently transitioned to measure noise floor via phase noise analyzer such as Keysight's E5052A/B or the R&S FSWP series

    Please advise how the noise floor performance compare to the datasheet. This is an important bench mark to start the debugging. It is possible that noise floor is already good enough when compared to our DAC specfication. 

    Please also review the following application note

    One quick way to improve the noise of the DAC is increase the sample rate while keeping the IF at the same rate. For every 2x increase in sample rate while keeping the IF at the same you can in theory achieve 6dB reduction in noise. However you are ultimately limited by the inherent DAC thermal noise limit. Basically there is a point where the 2x reduction in jitter is no longer helping because the thermal noise is the limit.

  • Hi Mr.Kang

    Sorry for late reply and thank you for explaining detail.
    Now we try to measure NSD as you say.

    We are useing MXA N9020A with preamplifer function and could see a small hill on  noise floor. its about -163dBm and BW looks 10MHz(IF= 138.8 ±5MHz) with average detect.

    But we cannnot  judge the result is right or not because first of all our test condition is right or not.

    so would you please show us  a test condition when you measured?

    (1) Did you use test circit like below fig?(our interface circuit doesnt use trans ,just use discrete passive component to be single 50Ω impedance)

    (2)what did you insert data  in RX*P/N?(we only use RX0P/N and RX1P/N pin but insert nodata) 

  • Hello,

    We cannot see your test circuit figure. The diagram is not displayed properly. We are using the standard EVM DAC output circuitry with the DAC output connected to standard 2:1 transformer to interface with the 50ohm load (of the test equipment).

    The standard test procedure for testing of the NSD is we inject 0dBFS tone (i.e. Figure 23 and Figure 24 of NSD vs. output frequency plots) at the output frequency. This data is transmitted over JESD204 link (over SRX pins)

    The Keysight MXA may not have as good of noise performance as the PXA. This is one thing you have to be careful. With also the pre-amp, you will have to notch out the output frequency tone with a notch filter to avoid pre-amp saturation. The notch filter will have a passband. If you can allocate the passband to be within 40MHz offset from the tone where there are no harmonics fold-back, then you can use this passband to measure the NSD.

    Change the center frequency of the MXA to the passband of the notch filter with the notch filter filtering out the fundamental tone output of the DAC. Turn on noise correction of the MXA and you should see the best performance of the DAC.

    Again, it is much easier if you can use the E5052 equipment to measure the phase noise directly at 40MHz offset. 


  • Hi Mr.Kang

    Thank you for your reply.

    I added  figure again.can you see figure and Is this right?

    We injected 0dBFS and as you say,signal is saturation .so we try to insert notch filter.

    We used PLL signal Analyzer and measured phase noise clock (generated from LMK04828) and it seems no problem.


  • Hi Takeshi-san,

    I can see the diagram and it is matching what we are using for the NSD measurement.

    Please use the PLL signal analyzer to measure the DAC output. You can derive NSD from there as well.