• State Resolved
  • Locked Locked
  • Replies 4 replies
  • Subscribers 46 subscribers
  • Views 343 views
  • Users 0 members are here
Support feedback
Options
Options
Related

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.

DAC3484 FIR Interpolation Details

James Reis
Prodigy 30 points
Other Parts Discussed in Thread: DAC3484

I am looking for documentation describing the math asociated with how the DAC3484 applies the FIR interpolation filters, The description on pp. 57-60 of the data sheet is pretty sparse. For example, in 4:1 Interpolation:

(a) are the input data samples first replicated to the output rate (or are they zero filled between samples) prior to applying the FIR filter(s);

(b) which FIR filter(s), FIR0--FIR3, (with the appropriate interpolation kernel listed in Table-7) are applied at each level of interpolaton;

(c) is more than one FIR applied in cascade, depending on the amount of interpolation requested;

(d) what is meant by "Interpolation Composite Response", e..g., Fig 69? 

Essentially, I am looking for a computational math model for the signal processing implemented in the DAC3484 up conversion process that I can embed into a signal processing simulation.

Thank you for what ever assistance you may be able to provide.

--jim

  • 0
    TI__Guru* 78714 points

    Hi James,

    Regarding the interpolation filters:

    A) The input data samples are first zero filled between samples and the low pass filtered

    B) 2x interpolation uses FIR0 only. 4x interpolation uses both FIR0 and FIR1. 8x interpolation uses FIR0 to FIR2. 16x interpolation uses FIR0 through FIR3. 

    C) yes, you are correct

    D) On the datasheet, each FIR filter response is shown. The composite response shows the total response for 2x, 4x, 8x, and 16x interpolation. (i.e. for 4x interpolation, FIR0 and FIR1 composite is shown). 

    On the DAC3484 datasheet, there are FIR filter coefficients for the half-band filter. You should be able to plug these filter coefficients into programs such as Matlab or Simulink to simulate the total response.

    -KH

  • 0 in reply to Kang Hsia
    Prodigy 30 points

    Thank you for your cogent reply. Your response has been very helpful. Following up, I have been trying replicate the Magnitude Response of the FIR filters as shown in the datasheet, p.58, by inserting a FIR kernel (Tab. 7, p.60)  into a 2048 array, zero filling the unspecified elements; followed by a 2048 FFT; and as a last step generating and plotting the MAG spectrum. When I do this, my results agree with the the data sheet for FIR1 and FIR3, but DO NOT agree for FIR 0 and FIR2. My simulation was done in Excel2010. The work flow and simulation results for FIR0 nad FIR1 are referenced below:

    7506.DAC3484 FIR Filter SIM.pdf

    These sinc interpolation filters are well behaved, but I would like to know how the graphs in Fig. 64-67 were computed in general, and for Fig.64 and 66 in particular.

    Thank you.

    --jim

  • 0 in reply to James Reis
    TI__Guru* 78714 points

    Hello Jim,

    We used Matlab to plot the datasheet response of FIG64-67. 

    For instance,

    >> fir2 = [29 0 -214 0 1209 2048 1209 0 -214 0 29];
    >> freqz(fir2,1);

    I have confirmed again that FIR0 and FIR2 response agrees with the datasheet. Could you double check your FIR coefficient to make sure they match the ones on Table 7?

    Matlab reference:http://www.mathworks.com/help/signal/ref/freqz.html

    -KH

  • 0 in reply to Kang Hsia
    Prodigy 30 points

    Thank you, Kang, for your speedy reply. I will check your Matlab reference and the coefficients used in my computation to see if I can explain the differences.

    --jim