cancel
Showing results for 
Search instead for 
Did you mean: 

HSI jitter.

st18
Associate II
Posted on August 12, 2008 at 16:37

HSI jitter.

5 REPLIES 5
st18
Associate II
Posted on May 17, 2011 at 12:41

A colleague was recently bitten by high jitter on some ATtinys - a process change meant that while meeting the spec sheet, the new parts made his device fail in the field.

8MHz - 1%.

What does this look like?

Is this (over 1 second) 8.040MHz +-1KHz, or 8.040MHz +-40KHz?

And yes, I'm wondering if I can use USB with HSI, and no external crystal.

Though other applications - for example R/C servo driving (the above failure in field example, causing excessive current draw due to the jitter)

lanchon
Associate II
Posted on May 17, 2011 at 12:41

hi,

I don't understand these parts of your post:

>8MHz - 1%.

>What does this look like?

>Is this (over 1 second) 8.040MHz +-1KHz, or 8.040MHz +-40KHz?

>Though other applications - for example R/C servo driving (the above failure in field example, causing excessive current draw due to the jitter)

please clarify them.

no, you can't use USB without a crystal on any microcontroller, the spec is too tight. (I mean for USB full speed; for low speed the spec tolerances accommodate alternatives to crystals, but the STM32 doesn't support low speed.)

lanchon
Associate II
Posted on May 17, 2011 at 12:41

BTW, the HSI is 8MHz +/-3% over full temp and at 3.3V (no voltage range given!), or 8MHz +/-240KHz.

st18
Associate II
Posted on May 17, 2011 at 12:41

To clarify.

The oscillator frequency spec is stated as 1% (at 25C).

The oscillator can meet this spec if

A) the frequency varies randomly due to lots of jitter by +-.25%, on a cycle-by-cycle basis, effectively acting as a 'spread spectrum' clock. On top of this is added thermal and voltage drift.

Every second the frequency will be wobbling around randomly over a 40KHz range.

B) The frequency over 1s varies by 1KHz, and changes due to thermal and voltage effects.

In the case of B) as I understand it, it might be plausible to first calibrate the clock speed using the RTC or other means, then use the USB device - for few seconds. (I'd not assume trimming with USB in use would work without more info on the PLL)

This is not solely an issue with USB of course.

It makes the HSI useless for some tasks involving relative timing, when the drift alone isn't a problem.

obtronix
Associate II
Posted on May 17, 2011 at 12:41

Quote:

On 12-08-2008 at 14:03, Anonymous wrote:

To clarify.

The oscillator frequency spec is stated as 1% (at 25C).

The oscillator can meet this spec if

A) the frequency varies randomly due to lots of jitter by +-.25%, on a cycle-by-cycle basis, effectively acting as a 'spread spectrum' clock. On top of this is added thermal and voltage drift.

Every second the frequency will be wobbling around randomly over a 40KHz range.

B) The frequency over 1s varies by 1KHz, and changes due to thermal and voltage effects.

In the case of B) as I understand it, it might be plausible to first calibrate the clock speed using the RTC or other means, then use the USB device - for few seconds. (I'd not assume trimming with USB in use would work without more info on the PLL)

This is not solely an issue with USB of course.

It makes the HSI useless for some tasks involving relative timing, when the drift alone isn't a problem.

No one can answer your question with any confidence.

The only way you can design like this (outside of spec) is to buy a part from one process and test it to your more strict specs and if it meets your spec make a life time purchase from the same process, otherwise you will be in the same boat as your colleague.

For a given process, the clock performance is much more stable then the spec implies, but STM has to take into account the different processes they use in different fab plants (as well as future fabs they may purchase or farm out).