> ----------
> From: Mauro, Chuck[SMTP:
KILLspamChuck.MauroKILLspam
KLA-TENCOR.COM]
> Reply To: pic microcontroller discussion list
> Sent: Tuesday, January 27, 1998 12:14 PM
> To:
RemoveMEPICLISTTakeThisOuT
MITVMA.MIT.EDU
> Subject: Re: Internal Oscillators
>
> Charles,
>
> Thanks for bringing this up Charles - I think it's a great topic.
>
> From my experience with working directly with Microchip, I'd say that
> there is an excellent chance they will do this - eventually (measured,
> unfortunately, in several years of time - but the success of the
> 12CXXX
> family already should have convinced them to offer such a version in
> all
> the PIC families).
>
> I designed a very high volume product around the 12C509 (millions of
> units/yr., although, it hasn't shipped - yet -), and did EXTENSIVE lab
> testing on the internal oscillator - it's a pretty rock-solid design.
> I've even met with Microchip directly, and worked with the
> manufacturing/test engineers to better characterize and fine tune our
> oscillator requirements in our particular application. (BTW, don't
> expect to ever be able to talk to these guys yourself unless you're an
> incredibly large volume customer - and then only maybe...) I'm not
> saying that they changed anything for us, but we had a much more
> guarantee-able design after they "ran the numbers" for us (read: truly
> characterized the oscillator through extensive wafer testing).
>
> I firmly support you in your appeal to Microchip to increase the
> number
> of devices that contain an internal RC oscillator. I think everyone
> benefits. But - and everyone else out there should be aware of some
> important issues when applying this particular feature of any PIC -
> ESPECIALLY when it is first introduced to us (and even more
> importantly
> when devices are made available in "Engineering" type samples):
>
> 1) Any new chip they apply the internal oscillator technology to
> needs
> to be FULLY CHARACTERIZED by Microchip before I'd EVER use it. They
> typically SIMULATE the design and release preliminary data sheets with
> projected device characteristics... Things (osc. characteristics)
> will
> likely change after they build a few wafers and test for process
> corners... For me, after they had performed serious wafer test, I
> found
> that the projected osc. characteristics became WORSE... But,
> fortunately in my case, I was able to find a way around it.
>
> 2) Any of your designs need to seriously take into consideration:
>
> a) Temp range,
> b) Typical Vdd operating point and range (especially supply
> regulation).
> c) +/- 3 sigma process variation (at least) of the chip.
>
> These parametrics will affect the oscillator frequency more than
> anything else...
>
> 3) If you need to guarantee better than a 2.5%-3% tolerance on Fosc
> over
> ALL operating conditions, you're better off using a resonator or xtal,
> because you're sure to see the corners of their process tolerances in
> any kind of volume...
>
>
> To more directly address your original points:
>
> 1) I saved about 25-30 cents [includes all costs] per widget by using
> the internal RC osc. This includes all direct, indirect and typical
> hidden costs (p.c. real estate needed, pick and place costs, component
> stocking costs, improved reliability, etc.). Keep in mind that my
> volume is (was) between 20-30 MM units/yr for a several year period.
> Your savings will likely be even higher (perhaps as much as $1?).
>
> 2) My design was far more elegant because of the internal osc.
> design...
>
> 3) And reliable...
>
> 4) I also needed all the I/O I could get in an 8 pin device. This is
> the only way to maximize I/O capabilities of the 12CXXX family...
>
> 5) I also believed it may have helped in reducing my total RFI
> emmision
> signature for various reasons that I can't go into here...
>
> So, from a very real world example (sorry - can't disclose what the
> widget is or does - due to one of those annoying proprietary/trade
> secret clauses in an employee exit agreement), I can assure you the
> internal RC osc. is a great way to go if you do not have strict
> real-time timing issues. I suspect that a vast majority of
> personal/hobby/non mission-critical PIC projects could probably get
> away
> with using this type of cost-improving oscillator.
>
> I just wish the folks at Microchip would consider using a better (more
> temperature and voltage stable) internal osc. technology. I know for
> a
> fact that if they wanted to, they could tighten the tolerances to
> under
> 1%, and not necessarily at higher chip cost (I've researched the
> various
> fundamental methods, and it's quite doable).
>
> SO: HOW ABOUT IT MICROCHIP???
>
>
> Chuck Mauro
>