Overclocking SD16_A in MSP430F47187

Maxim Shtyrov91730 said:

My colleague built a schematic (on MSP430F427), where he overclocked Fm up to 1.79MHz. So, the questions are: Is this mode permissible? And what problems may occur, when smb overclocks the SD16(_A)?

First answer: no. Second: you won't get reliable results.
The SD16 is a delta-sigma converter. It works by integrating the input signal, comparing it to th ereference, emitting a comparison result bit and if this bit is 1, subtracting the reference form the input on the next iontegration/comparison loop. THe output bits are sent into a digital filter for discrimination. If you clock the meachnism faster than specified, the integration/comparison/feedback comes out-of-sync (racing conditions in the analog parts) and you'll get unpredictable results. Also, it may overload the reference´8whcih is constructed for low power, not for ultra-high-speed performance, so there is no reserve for higher speeds)

How do you overclock Fm by hardware? Feeding 1.79MHz into TACLK? If so, and you want to use this because of timing reasons, you'll need to use the /2 clock divider setting. Else simply use the internal ~1MHz.

Delta-Sigma converters aren't made for high-speed conversions. They are made for low-cost high-precision conversions.

You can get fast, cheap and precise ADCs. However..

fast and cheap is not precise
fast and precise is not cheap
cheap and precise is not fast

The SD16 falls in the third category. Teh ADC10/ADC12 into the first. The second category is reserved for expensive external devices. As it also usually requires a relatively high power, whcih doesn't fit into the MSP family at all.

To the third question: the internal reference and the SD16 analog circuitry suppresses voltage changes on AVCC quite good. It is sufficient to decouple AVCC from the CPU-caused  load ripples on DVCC. A separate supply/regulator isn't necessary. A larger improvement in accuracy can be achieved by using an external reference (which is, on the other hand, an expensive and power-consuming task compared to the internal reference, and requires careful layout to not make things worse rather than better)

Exhaustive explanation, Jens. Thanks a lot.

Concerning how did he overclock SD16: XT1 - crystal resonator 3,579545MHz (f_r), f_MCLK = f_r, f_ACLK = fr/2, f_m = f_ACLK. 

My colleague got reliable results, but... it seems to me, that he was very lucky (no temperature stresses - room temp only, and good specimens of chips).

Thanks a lot ones more.

Maxim Shtyrov91730 said:

Concerning how did he overclock SD16: XT1 - crystal resonator 3,579545MHz (f_r), f_MCLK = f_r, f_ACLK = fr/2, f_m = f_ACLK. 

Okay, so it's rather software (the clock system settings) than plain hardware overclocking :)I was tinking of something else (using TACLK with an externally generated signal)

Maxim Shtyrov91730 said:

My colleague got reliable results, but... it seems to me, that he was very lucky

Well, being outside the specs with the frequency does not necessarily mean being outside the specs for the result. However, it is not guaranteed that the specified performance is provided then. And the more you go outside the specs, the higher the chance that it doesn't work.
For experiments, I don't see a problem. If it works, it works. For a production release, I would stay inside the specs.