Because of the holidays, TI E2E™ design support forum responses will be delayed from Dec. 25 through Jan. 2. Thank you for your patience.

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.

AWR2944: Complex Multiplication Capability in Pre-Processing Block

Part Number: AWR2944

Tool/software:

 The problem I'm having is that when i copy data to the multiplication RAMs(0x06068000), only the LSB-24 bit has value,the HSB-8 bit is always 0. I've tried different ways(EMDMA,memcpy, HWA_configRam), all results are the same(the HSB-8 bit is 0). In addition, direct modification of the data at this address(the HSB-8 bit) will not work.The result is normal?  But if only change the destination data address to window RAMs(0x06064000), the HSB-8 bit has value, and the value is same to the input data.

Related Settings:

a) the mode of CMULT is set 0110b;

b) the input data ranges from -10000 to 10000;
c) the type of input data is same to the ramtpye;
  • Hi Huasong,

    I will check and get back to you in a week.

    Regards,

    Samhitha

  • Hi Huasong,

    Your observation seems to be coming from an invalid programming case of the Vector multiplication RAM. Please find the accurate memory layout of the HWA's Vector multiplication RAM.

    • This particular RAM has a word length of 24bits. Basically, by design, the 32-bit write drops the 8 MSBs when it gets placed on the memory and then pad 0s the next time you try to read (even from the HWA). Therefore, the top 8-bits of each 32-bit word you see in the memory will not be applicable and therefore there is no read or write operation possible here. The maximum value you can use per sample would be 0x00FFFFFF in a given 32-bit word. 
    • Additionally, there is one more correction. The I and Q bit width is 24 and not 32 and 16 for modes 6 and 7 as shown in the diagram.
    • In this 24-bit space, only a 21-bit scalar coefficient can be programmed (21 LSBs) as that is how the HWA interprets. This is the same format used by the scalar multiplication block. (Make sure your number is of max width 21-bits or les)

    Regards,

    Kaushik