Forum Discussion

Altera_Forum's avatar
Altera_Forum
Icon for Honored Contributor rankHonored Contributor
12 years ago

Using VHDL to generate 20Mhz sine wave and visualizing with a 10bit DAC

Hi all. I want to generate a 20MHz sine wave in VHDL but have no clue of how to do it. Also, with a device like EP3C10E144C8N, is it possible to achieve that frequency? Also, I intend to use 10-bit output , and use a 10-bit DAC to convert it. Would you please giving me some suggestions about the sine wave generator , and recommending any DIP (I cant use SMD) DAC device that could operate at that frequency? I used to use this CMOS DAC, http://www.analog.com/static/imported-files/data_sheets/ad7533.pdf, but it turned out that it is not fast enough for my case. Thanks for the help.

8 Replies

  • Altera_Forum's avatar
    Altera_Forum
    Icon for Honored Contributor rankHonored Contributor
    12 years ago

    The most simple way is to generate a 20 MHz square wave and extract the fundamental wave by a low-pass filter. Otherwise you need to generate a higher output data rate (e.g. 100 or 200 MS/s) to get a halfway continuous output signal. Data rates of several 100 MS/s aren't a problem for recent FPGA and DAC devices, but I fear you won't find a suitable DAC in DIP package. There are a few devices like ADV7120 (8 Bit, 80 MS/s). You may want to use a SOIC to DIP adapter board however.

  • Altera_Forum's avatar
    Altera_Forum
    Icon for Honored Contributor rankHonored Contributor
    12 years ago

    --- Quote Start ---

    The most simple way is to generate a 20 MHz square wave and extract the fundamental wave by a low-pass filter. Otherwise you need to generate a higher output data rate (e.g. 100 or 200 MS/s) to get a halfway continuous output signal. Data rates of several 100 MS/s aren't a problem for recent FPGA and DAC devices, but I fear you won't find a suitable DAC in DIP package. There are a few devices like ADV7120 (8 Bit, 80 MS/s). You may want to use a SOIC to DIP adapter board however.

    --- Quote End ---

    Wow, that sounds pretty neat and clever. If I am about to opt for quadrature modulation, is using square wave and low-pass filter still a valid option?
  • Altera_Forum's avatar
    Altera_Forum
    Icon for Honored Contributor rankHonored Contributor
    12 years ago

    --- Quote Start ---

    Wow, that sounds pretty neat and clever. If I am about to opt for quadrature modulation, is using square wave and low-pass filter still a valid option?

    --- Quote End ---

    QAM normally requires clean sine wave. A filtered approach will not produce a clean tone. A rectangular pulse has wide band that will not be filtered clean into a tone. You will need proper nco for that purpose.
  • Altera_Forum's avatar
    Altera_Forum
    Icon for Honored Contributor rankHonored Contributor
    12 years ago

    --- Quote Start ---

    QAM normally requires clean sine wave. A filtered approach will not produce a clean tone. A rectangular pulse has wide band that will not be filtered clean into a tone. You will need proper nco for that purpose.

    --- Quote End ---

    Oh, now I understand. How about Cordic algorithm?
  • Altera_Forum's avatar
    Altera_Forum
    Icon for Honored Contributor rankHonored Contributor
    12 years ago

    --- Quote Start ---

    Oh, now I understand. How about Cordic algorithm?

    --- Quote End ---

    First try simple solution. If your required tone is 20MHz and your sampling clock is multiple of 20 e.g. 100MHz then all you need is a cycle of sine data of 100/20 = 5 data points. This can be implemented as basic logic with a counter to read them sequentially.

    If your sampling clock divides with fractional result then you need a larger LUT (possibly altera NCO ip). people use cordic to avoid multipliers at the expense of more logic.
  • Altera_Forum's avatar
    Altera_Forum
    Icon for Honored Contributor rankHonored Contributor
    12 years ago

    --- Quote Start ---

    A rectangular pulse has wide band that will not be filtered clean into a tone.

    --- Quote End ---

    Both will need a filter, and with low sampling frequency of 100 MHz, the filter effort isn't so much lower than for a square wave.
  • Altera_Forum's avatar
    Altera_Forum
    Icon for Honored Contributor rankHonored Contributor
    12 years ago

    --- Quote Start ---

    Both will need a filter, and with low sampling frequency of 100 MHz, the filter effort isn't so much lower than for a square wave.

    --- Quote End ---

    I think you are talking about post DAC filter. For a proper NCO you only need the usual anti-image filter. For rectangular pulses you need a shaping filter for the tone's fundamental itself.

    However I am talking about QAM in digits. The normal procedure is to get a clean digital tone for modulation. The DAC output will then just need anti image filter.
  • Altera_Forum's avatar
    Altera_Forum
    Icon for Honored Contributor rankHonored Contributor
    12 years ago

    I just mentioned, that the filter effort won't be much different for 100 MHz sampling frequency. You can calculate yourself, you get 80 MHz first image frequency versus 60 MHz 3 harmonic, with similar energy. So the difference is probably one more filter order.