On 03/17/2015 08:38 AM, Nicolin Chen wrote:
On Tue, Mar 17, 2015 at 08:24:27AM +0800, Xuebing Wang wrote:
- As pll4_main_clk is global for all 3 SSIs, I think that we can *not* use
8k for ssi1 and 44.1k for ssi2, right? Because 8k and 44.1k requires different pll4 clock, right?
True.
-- However, as 48k, 32k, 16k, 8k can use same pll4, thus different SSIcan use different sampling rate in this subset (48k, 32k, 16k, 8k). -- Same principle applies for 44.1k, 22.050, 11.025 subset of sampling rates.
Correct. You need to choose 48K-group or 44.1K-group if only using PLL4.
If you are so obsessed with full sample rate support in hardware level, you might need to do a trade-off like sacrificing video function -- you would then be allowed to use PLL5 to forge a rate for another rate group.
I forgot if PLL3, another possible parent clock of SSI, is suitable for one of the groups. You may also take a look at it.
Nicolin,
Thanks a lot for your help.
I am using an unused PLL3_PFD (454.7M) for SSI3, and it is working ok.
1) 454.7M is actually 454.736842. For 48k sampling rate, I am using (clk_pred = 1, clk_podf = 37) to generate ssi baudclk. -- 454.736842 / 1 / 37 = 12.290185 MHz
2) If I derive from PLL4, then ssi baudclk = 12.288000, which correspond to iMX6SL reference manual "Table 48-7. SSI Bit Clock and Frame Rate as a Function of PSR, PM, and DIV2". -- This is very accurate down to 1Hz
3) There is some deviation of baudclk (12.290185 vs 12.288000 MHz), which translates to the deviation of bit_clock of (3.072546 MHz vs 3.072000 MHz for 48k rate, 546Hz deviation).
Is this the most accurate clock we can get if we use PLL3_PFD? Thanks again.
Nicolin