wow

thanks

From cPhone


On Aug 8, 2011, at 7:38 AM, "JMGross" <[email protected]> wrote:

> Hi Eric,
> 
> the 1.8V in the datasheet are worst-case value. It may well be thaat you can 
> significantly go above 8MHz with 1.8V. However, I recommend increasing the 
> capacitance on teh VCore pin. The main problem is that the 
> internal voltage regulator reaches its limit for the stable core voltage if 
> the input voltage is too low.
> The core voltage is 1.4V(1.44V in LPM) when PMMCOREV is 0 (the default on 
> 5438A). On this voltage, operation up to 8MHz is guaranteed. If you exceed 
> 8MHz, this may mean that the flash or CPU needs more 
> than 1.4V for proper operation, it may also mean that the voltage regulator 
> cannot provide 1.4V anymore due to the higher current comsumption. Or both. 
> But it can also be none of them. 400kHz are almost 5%. 
> It may or may not work, I really cannot say. It probably will work with most 
> MSPs but perhaps won't with some.
> 
> However, you're asking about DCO and FLL. And that's another thing to 
> consider.
> The DCO most likely cannot provide 8MiHz.
> In fact, the DCO can only provide 256 different frequencies in the range of 
> 70kHz and 135MHz.
> It is not likely that 8MiHz is one of them. (or 8Mhz).
> So the DCO uses modulation patterns to switch between two frequencies. A 
> higher one and a lower one, to give 30 more different 'average' frequencies 
> between the 256 'real' ones.
> But this also means that the DCO will e.g. switch between 8MHz and 9MHz to 
> give your requested 8MiHz. So the cpu is clocked with 9MHz for some clock 
> cycles.
> (to get the worst case upper frequency and maximum error, you have to do lots 
> of math on the DCO frequency table, checking min/max frequencies and DCO tap 
> ratio etc.)
> Some time ago I calculated that worst case you may only use 23.7MHz on an 
> MSP430F5438A, so you can be sure that the DCO will never go above 25MHz on 
> any individual MSP.
> 
> The FLL is quite good. It increases and decreases DCO tap and modulation 
> setting so it gets n DCO ticks per each reference tick. With my 5438s I 
> didn't have problems with 115200Bd connections using the FLL for 
> a 16MHz DCO clock. With the internal REFO as reference.
> Since the maximum clock for the 5438 is 18MHz, I'm on the safe side :)
> 
> However, there are some more things to keep in mind:
> 1) the FLL does not change the RSEL setting. So you must still select the 
> proper RSEL range.
> 2) the crystal must be up and running before you can activate the FLL, otr 
> the FLL will try to adjust the DCO to x*0Hz = 0Hz. you can use the internal 
> REFO instead, ot let the FLL start on REFO until the crystal is 
> running, and switch its reference then (to save some time).
> 3) you don't know when the frequency has settled. The DCOFFG bit is cleared 
> as soon as the DCOx bits have been adjusted to >0 and are still <31. If 
> DCOFFG can be cleared, this only means that the DCO is not 
> adjusted against its limits. It does NOT mean that it has reached the final 
> value.
> All you can do is checking the changes and noticing if the change changes 
> direction. So the FLL is no longer increasing MODx and DCOx but swings 
> between two settings instead.
> you can speedup the process by detecing a safe (too low for sure) but non-0 
> start value for DCOx. Based on experience. It shortens the adjustment time, 
> since the FLL only adjusts one step at a time, so it takes 
> worst case 1024 reference ticks to reach the final position (32 DCOx and 32 
> MODx settings).
> 
> The FLL will take care of any required recalibration due to temperature drift.
> However, you cannot rely on this if you enter an LPM that switches the FLL or 
> the DCO or the reference off. In case the FLL is switched off, it is NOT 
> switched on during ISR execution. But disabling reference or 
> DCO will cause the FLL (if sitll enabled) to try to adjust the DCO up or down 
> during LPM.
> So you should explicitely disable the FLL before entering any LPM except 
> LPM0, and only re-enable it if you remain in active mode for more than 2 or 3 
> reference ticks.
> 
> JMGross
> 
> 
> ----- Ursprüngliche Nachricht -----
> Von: Eric Decker
> An: JMGross
> Gesendet am: 07 Aug 2011 07:22:53
> Betreff: Questions about 5438 FLL and DCO
> 
> Hi Jens,
> 
> I have a few questions that I hope you can answer.
> 
> We are using a 5438a (msp430f5438a) ti cpu.   We are running it at 1.8V and
> the docs say the max system frequency (dcoclk) is 8MHz (8,000,000).
> 
> How risky is it to run the DCOCLK at 8,388,608 Hz (8MiHz, binary megahertz)
> while staying at Vcore of 1.8V?
> 
> 
> 
> Also how good is the FLL?
> 
> On previous processors, 1611 and 2618 we had to periodically mess with
> calibrating the DCO to ACLK (off 32768 Xtal) by tweaking various DCO
> parameters.  If using the FLL I assume that we don't have to do that
> anymore.   That it just works.   Is that a reasonable assumption?
> 
> 
> thanks,
> 
> eric
> 
> -- 
> Eric B. Decker
> Senior (over 50 :-) Researcher
> 
> 

------------------------------------------------------------------------------
BlackBerry&reg; DevCon Americas, Oct. 18-20, San Francisco, CA
The must-attend event for mobile developers. Connect with experts. 
Get tools for creating Super Apps. See the latest technologies.
Sessions, hands-on labs, demos & much more. Register early & save!
http://p.sf.net/sfu/rim-blackberry-1
_______________________________________________
Mspgcc-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/mspgcc-users

Reply via email to