AD9102 SRAM Writing data rate calculation

Hi Manu,

The Double SPI Write mode is available only for the AD9106 and not for the AD9102 (see AD9102 FAQ8). However, the time it takes to write the entire SRAM (t_WSRAM) for the other SPI modes (3-wire and 4-wire) can be derived from the same expression, which would be [2 + (2 * 4096)] * 8/f_SCLK (note the exclusion of 2 as a multiplying factor for f_SCLK).

Proof:

No. of command bits: 16
No. of data bits: 16
No. of SRAM registers: 4096

t_WSRAM = t_{COMMAND CYCLE} + (4096 * t_{DATA TRANSFER CYCLE})
              = (16 * 1/f_SCLK) + [4096 * (16 * 1/f_SCLK)]
              = [16 + (16 * 4096)]/f_SCLK
              = [2 + (2 * 4096)] * 8/f_SCLK

where: f_SCLK = SPI clock rate (up to 80 MHz based on Tables 3 and 4 of the datasheet)

The above formula is true when the auto-decrement feature is used (i.e., the CSB bit stays low beyond the first data word and the SPI port automatically decrements the register address, allowing multiple writes to contiguous addresses) beginning at the 0x6FFF address space.

The time it takes to read the entire SRAM (t_RSRAM), on the other hand, would be more than twice as long since there is no auto-decrement feature for this operation. This means that the command cycle has to be repeated for all SRAM registers. Assuming that the SRAM registers are accessed consecutively with the CS bit set back to high right after the data transfer for exactly one SCLK cycle before it goes low again for the next read command, this value can be estimated through the following simplified equation:

t_RSRAM = 4096 * (t_{COMMAND CYCLE} + t_{DATA TRANSFER CYCLE} + 1/f_SCLK)
             = 4096 * [(16 * 1/f_SCLK) + (16 * 1/f_SCLK) + 1/f_SCLK]
             = 4096 * 33/f_SCLK

Please note that the auto-decrement feature and how it is applicable only for the write operation in SRAM are contrary to what is specified in the datasheet. Rest assured that document revisions are currently being worked on to include these corrections.

Hope this helps.

Regards,
Zaeefa

Hi Manu,

The Double SPI Write mode is available only for the AD9106 and not for the AD9102 (see AD9102 FAQ8). However, the time it takes to write the entire SRAM (t_WSRAM) for the other SPI modes (3-wire and 4-wire) can be derived from the same expression, which would be [2 + (2 * 4096)] * 8/f_SCLK (note the exclusion of 2 as a multiplying factor for f_SCLK).

Proof:

No. of command bits: 16
No. of data bits: 16
No. of SRAM registers: 4096

t_WSRAM = t_{COMMAND CYCLE} + (4096 * t_{DATA TRANSFER CYCLE})
              = (16 * 1/f_SCLK) + [4096 * (16 * 1/f_SCLK)]
              = [16 + (16 * 4096)]/f_SCLK
              = [2 + (2 * 4096)] * 8/f_SCLK

where: f_SCLK = SPI clock rate (up to 80 MHz based on Tables 3 and 4 of the datasheet)

The above formula is true when the auto-decrement feature is used (i.e., the CSB bit stays low beyond the first data word and the SPI port automatically decrements the register address, allowing multiple writes to contiguous addresses) beginning at the 0x6FFF address space.

The time it takes to read the entire SRAM (t_RSRAM), on the other hand, would be more than twice as long since there is no auto-decrement feature for this operation. This means that the command cycle has to be repeated for all SRAM registers. Assuming that the SRAM registers are accessed consecutively with the CS bit set back to high right after the data transfer for exactly one SCLK cycle before it goes low again for the next read command, this value can be estimated through the following simplified equation:

t_RSRAM = 4096 * (t_{COMMAND CYCLE} + t_{DATA TRANSFER CYCLE} + 1/f_SCLK)
             = 4096 * [(16 * 1/f_SCLK) + (16 * 1/f_SCLK) + 1/f_SCLK]
             = 4096 * 33/f_SCLK

Please note that the auto-decrement feature and how it is applicable only for the write operation in SRAM are contrary to what is specified in the datasheet. Rest assured that document revisions are currently being worked on to include these corrections.

Hope this helps.

Regards,
Zaeefa