Go Down

Topic: Microchip crystal errata? (Read 4254 times) previous topic - next topic


Jun 29, 2014, 02:14 pm Last Edit: Jun 29, 2014, 02:58 pm by Constantin Reason: 1
Hi guys,

Having built a couple of boards over the years, I came across a (to me strange) configuration from a respected vendor. Specifically, if you look up the schematic of the MCP3911 ADC Evaluation Board (ADM00398), it features standard 0.1uF decoupling capacitors on the crystal for the MCP3911.

In my limited experience, the capacitors used with crystals are usually in the picoFarad range.  The crystal (X1) specified in the BOM is a standard Abracon 10MHz unit with a 18pF load capacitance rating, suggesting capacitors more in the 22-30pF range, not orders of magnitude higher...

So do you think MCPs specification of decoupling caps  for use with a crystal like this a likely error?


They are wrong.  That is impossible.


Jun 30, 2014, 06:22 pm Last Edit: Jun 30, 2014, 06:26 pm by Constantin Reason: 1
Hi Luis,  and thanks for the reply.

I posted the issue over in the MCP forums, have yet to get a reply. However, in another post , a user complained that the MCP3911 evaluation board was not producing the data he was looking for. I wonder if the MCP3911 on it was set up to run on the crystal-based clock vs. the MCU-based clock - it's a simple pin header with a jumper... and if the caps are all wrong, who knows if the crystal could even start oscillating?

FWIW, other reference schematics published by MCP (like the one for the tamper resistant power meter) show the kinds of pF load capacitors one would expect for a crystal. But, as with many cost-conscious designs, the power meter eschews the use of a crystal for the MCP3911 itself, using an IO line from the PIC MCU as a clock ticker instead.  I suppose that also has the advantage of you being able to vary the clock speed and hence the performance / power consumption of the MCP3911.

Anyhow, for now I am chalking up the crystal load capacitor specification in ADR00398 as erroneous unless someone has a good idea what a 0.1uF capacitor has to do with loading a tiny clock crystal!  XD

Thanks again for the reply. Constantin


I do that sort of things in my schematics.  You're placing a bunch of caps, and they're all in the same package, so you duplicate an existing cap and plunk it down all over, hoping that you'll remember to change the values to the "real" values before you turn it on.
(Since I build my own boards, I can usually correct at the last minute if need be.)   It's a bad habit.  It wouldn't surprise me TOO much to have a board handed off to a manufacturer without the final corrections.  A GOOD manufacturer would notice and call your attention to the potential problem.  But lately, "low cost" is more important/common than "good."


I left a message on their forum, but the crickets are deafening. While searching through other forum postings re: the MCP3911, I literally found dozens of messages that were left unanswered for weeks on end. Seems like that part of the web site has not managed to attain any kind of critical mass to make the community useful.

Today, I submitted an official trouble ticket with the heading "documentation error" and "moderate" importance. Let's see how quickly the official channels respond, but I am not holding my breath.


I would say that either someone has already, or will shortly have given Andrei Hapenciuc - and Bogdan Popescu - the traditional "kick in the pants".

Obviously a crystal could never work at all with those values and to surmise a combined fault whereby no-one ever noticed because they simultaneously got the clock settings wrong; well we hope that is even more improbable!   :smiley-eek:


:D  It took me a moment to realize that you must have meant the designer and checker of the schematic.  Yeah, assuming they're still with the company, etc. there might be a terse e-mail along the lines of 'what were you thinking'? A mechanical engineer finding an error in a reference design? The shame!  :smiley-eek-blue:

However, I wonder how much the folk at Microchip actually care. The trouble ticket is still 'pending'. I presume there must be a mountain of unanswered support tickets with 'the world is ending' importance rating ahead of mine.  That does not bode well for folk who have real issues with MCP products since unanswered questions = trouble for commercial projects.

All that said, the MCP3911 seems to be an odd duck in the MCP lineup. The MDAT pins in revision B of the chip were acknowledged as useless, noting a future fix. However, the Revision C of the MCP3911 datasheet (see Appendix A) mentions no MDAT fix in the silicon. So it's not clear to me whether the chips have been fixed or not. I won't worry about it, seeing that I don't use the MDAT pins, but this sort of sloppiness perhaps explains why competitors like Analog Devices enjoy such high market shares.


Here is the official reply from Mihai over there:

"The 100nf on the XTAL are a known issue. The error is on the schematic and on the board. So the user needs to replace them manually with lower value caps (~18pF)."

He also advised me to switch to the more modern MCP3910, which is pin compatible (but which looks harder to program).  I can't say I am terribly impressed:
A known issue that is not documented in the manual or the errata pages.
No apparent incentive / desire to fix the issues.
Knowingly selling a defective product.
Expecting a customer to hand-modify a development board (!!!) that allegedly is ready to use.
Technical support that takes 8 days to get back to a customer.


Well, they are 'usually' good with free samples... Spreading useless technology... This was the reason I quit using Microchip for anything important...

--> WA7EMS <--
"The solution of every problem is another problem." -Johann Wolfgang von Goethe
I do answer technical questions PM'd to me with whatever is in my clipboard


On the flipside, at least they replied. The MCP forums remain silent and MCP would likely do well to cull the dead stuff or the tumbleweeds will a bit too obvious.

Anyhow, he pointed me to the MCP3910, which indeed appears to be an improved version of the MCP3911. The chip model numbering chronology is a bit odd but an extra 0.3 ENOB, better THD, etc. is not bad. I'm  just not a big fan of the whole two-wire/4-wire thing that MCP implemented with the MCP3910, but as long as I don't accidentally activate the two-wire mode, I should be OK.

Curiously, the Errata page for the MCP3911 does not mention the possibility of upgrading to the MCP3910 as an option to enjoy functional MDAT pins. Odd!

Go Up