Lower frequency means bigger and more expensive passives. The high
frequency *should* work if done properly. I would like to understand what
I've done wrong. Your catch of the missing capacitor is a promising
candidate.
I haven't tried to analyze failed parts - it is a good idea and I'll do it.
I think that the voltage surge on the SW node is a low-likelihood failure
mode, as it would indicate a serious design flaw in the switcher. Noise on
the analog supply rails which might throw regulation out of stability or
failure to detect an over-current condition is a more plausible explanation
IMO.

Appreciate your feedback. You might have solved the issue. I'll spin
another prototype rev soon and we'll find out.

On Sun, Feb 1, 2015 at 4:55 AM, Paul McMahon <[email protected]>
wrote:

> Comments inline...
>
> On Sunday, February 1, 2015 at 1:15:12 AM UTC-5, Ytai wrote:
>>
>> Thanks for your help. Inline.
>>
>> <[email protected]> wrote:
>>
>>> Ytai,
>>> thanks for the link.  I'm glad to see this is still active.
>>> I'm surprised you've narrowed it down to Load transients, since you had
>>> previously narrowed it down to spikes on Vin, and suspected long
>>> (inductive) power cables.
>>>
>>
>> We've previously observed voltage surges exceeding (for a microsecond or
>> so) the maximum rated input voltage. Since there were no other deviations
>> from the specs, we have attributed the failures to those surges. Regardless
>> of whether or not this was *the* problem, it is definitely *a* problem,
>> thus we have revved the board to include a protection circuit for this
>> case. After having verified that the surged are gone, we continued our
>> tests to verify that the problem has gone and discovered that it hasn't.
>> Moreover, we were able to demonstrate that we are presumably able to get
>> the TPS to fail without exceeding its maximum ratings. We are trying to
>> gather more data as requested by the TI engineers to find the cause and the
>> cure.
>>
>>
>>> I have three suggestions, in case you have the ability to test different
>>> scenarios.
>>>
>>> a) this one is a long shot, but it's easy, so could be worth a try.  The
>>> recommended design has a 0.1uF cap on Vin.  I don't know how this could
>>> cause the failure, unless there were a weird resonance building up inside
>>> the chip, which could be mitigated by the high-freq cap.
>>>
>>
>> Good point. I can see why this might explain the failure. I'll try that.
>>
>>
>>>
>>> b) I've had numerous bad experiences with switching supplies running at
>>> high frequencies. They are very sensitive to layout and load transients.
>>> I'm curious if the failure would occur at the low frequency setting.  To
>>> test this, connect FSW to PG or Vout.  Pretty tough with the QFN on IOIO,
>>> but maybe you have another vehicle to test it on.
>>>
>>
>> I don't have a PCB where that's possible. I'm also not sure what I'll do
>> with this information if I know. I'm been pretty careful about layout as
>> I'm very well aware of how tricky it can be (learned the hard way...), but
>> obviously fast switching circuits have a lot of advantages (much smaller /
>> cheaper passives).
>>
>>
>
> Well, the 'what you would do', would be to choose the lower frequency
> setting on a future version of the board.  Too bad there's no easy way to
> test it.
>
>
>>
>>> c) Although it should not be required for this type of synchronous buck
>>> converter, The fact that the output drops might indicate the low-side
>>> output FET has blown closed.  I also seem to remember the measured
>>> resistance between SW and GND was very low on a failed device.  That being
>>> the case, it might be worth trying to protect that FET with a Zener, so
>>> that any voltage transients arising from abrupt changes in current through
>>> the inductor are clamped.
>>>
>>
>> If you're referring to the graph posted on the TI forum, to me it looks
>> like the output capacitor being drained by the load as opposed to a hard
>> pull-down by a blown low-side FET. Zeners are not usually effective for
>> such protection, but rather two Schottkys (GND to SW, SW to Vin), but since
>> the datasheet doesn't call for them, I'm assuming the body diodes of the
>> FET and/or the synchronous operation of the high-side/low-side FET doesn't
>> require them.
>>
>
> I should have said Schottkys, not a Zener.  Did TI ever do a failure
> analysis of a failed part for you?  We really need to know if it's the
> input side or the output side that suffered the electrical stress.  Did you
> ever measure the resistance between SW and GND on a failed part?
> As for the body diodes, they may have become the current path in the case
> where the low-side FET didn't turn on as quickly as it should.  The idea is
> that the external Schottky could give an alternate path for the current,
> just until that lower FET turns on.
>
>
>
> --
> You received this message because you are subscribed to the Google Groups
> "ioio-users" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to [email protected].
> To post to this group, send email to [email protected].
> Visit this group at http://groups.google.com/group/ioio-users.
> For more options, visit https://groups.google.com/d/optout.
>

-- 
You received this message because you are subscribed to the Google Groups 
"ioio-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To post to this group, send email to [email protected].
Visit this group at http://groups.google.com/group/ioio-users.
For more options, visit https://groups.google.com/d/optout.

Reply via email to