• State Resolved
  • Locked Locked
  • Replies 3 replies
  • Subscribers 79 subscribers
  • Views 1006 views
  • Users 0 members are here
Support feedback
Options
Options
Related

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.

MSP430F5XXX Overclock

Alessandro Fulignani
Intellectual 530 points

Other Parts Discussed in Thread: MSP430F5638

Hi,

In my application i often need speed. One month ago, i was a bit frustrated by my MSP430F5638 attached to a color LCD. So i tried to push up the CLK frequency..

(FLL ON+external XTAL 32768)

This family is specified at 20Mhz. My first attempt was at 25Mhz, and after a week of test with alla timers, SPI JTAG ports capture compare working, everything seems to works perfectly(No CRC erroers). Then i tried 30Mhz(!) And results were the same.

Well i thought i found a special CPU?  I did the same experiment on CPU taken from different lots...and the results were still with no errors. Now i have 50 boards running at 33Mhz by 4 days with all the peripherals engaged, and it is all ok. I also checked the ACLK with an oscillscope, just to see the jitter on the bus, and at 30Mhz it's almost perfect. Of course i will never sell my products, with the CPU outside its specification...

All this test(just for curiosity) have been made in a climate chamber in the range 0-70°

The MSP430F5xxx running at 30Mhz it's a very powerful machine !!

Does anybody did experiment on overclock?

does TI specifi this CPU at 20Mhz only for commercial purpose?

Alessandro Fulignani

  • 0
    Guru 227245 points

    Alessandro Fulignani said:
    does TI specifi this CPU at 20Mhz only for commercial purpose?

    Most 5x and 6x fmaily devices (except the 5438 non-A) are specified for up to 25MHz.

    It is possible that the USB controller memory interface, inlcuding the 2k internal, memory-mapped, dual-port ram, is only specified for 20MHz and instead of implementing a waitstate mechanism to the whole memory bus, the specs have been lowered generally. It wouldn't be nice if access to 2k of the 18k total ram cannot be accessed reliably. Did you test that?

    Depending on supply voltage, the digital circuitry (including the ACLK) usually runs faster if voltage is higher. There's some 'headroom' in all directions. However, not all circuitry can be assumed to respond wiht same performance. Exceeding the defined maximum speed, racing conditions may occur, retention times may drop, currents rise etc.

    Also, all the tests for overclocking are void for the next charge of CPUs, even if there was no new silicon revision. Just a different wafer may be enough.

  • 0 in reply to Jens-Michael Gross
    Intellectual 530 points

    Dear Jens-Michael Gross

    No in did not try the USB.  Right now i am just doing this test for curiosity. I did test on ram writing 0xaa and then 0x55, but only on 16K. I read and send data thru a crc verification, to a color display with a logic analyzer that keep trace on errors(0 after 5 days). The images shown on the display are sent from an external flash via SPI at 15Mbit!!! and alterned with 1 internal images that use 221K of 256. Every 100 images the hardware multiplier did a routine that use more or less all the possible combination. All the crc information, and a lot of other information are sent thru an RS485(RS232 MSP430). Of course every single bit that inform me about strange behavior has been set(SFRIE1). I think that in case of wrong bus timing, i should have seen some errors..anyway see this CPU running at 30Mhz it's impressive, expecially  in some application..
    I belive that with a 50-60Mhz version TI could send retired many competitor's CPU. And with a 32Bit MSP430, TI could fight also with cortex... 

    Thank you for your time Jens-Michael !,

  • 0 in reply to Alessandro Fulignani
    Guru 227245 points

    I just checked, and teh 55xx also has USb and is spoecified for 25MHz, so USB likely isn't the limiting factor. The 56xx has a DAC12 and an ADC12, but others (with 25MHz) have them too. ANother difference is the RTC_A vs. RTC_B, but again I don't see what could limit speed here.

    For the peripherals, I do not wonder that there is lots of headroom. Minimum supply voltage for 20 (25) MHz operation is 2.4V, and since CMOS operates faster at higher voltages, a supply of 3V or 3.6V should give additional 50% maximum speed. However, CPU, RAM and Flash are powered by the internal regulator (unaffected of VCC if above 2.4V) and it's surprising that they work so far above specs.

    Alessandro Fulignani said:
    I belive that with a 50-60Mhz version TI could send retired many competitor's CPU.

    Indeed, the core design has an impressive throughput. But the MSP was designed for ultra-low power, and with increasing speed, the power advantage over other other devices shrinks. Also, while the speed of ram and peripherals might be scalable, I fear that for the flash, waitstates would be necessary. (FRAM has the same problem, here 8MHz is the limit, mostly eating up the speed gain of >8MHz CPU clock)

    However, thanks for the interesting experiment. It was an appreciated distraction from the usual "I don't know what to do" posts :)

**Attention** This is a public forum