Ban ceramic resonators!

The title is meant to catch your eye, I'm not really suggesting that ;-)

I've made a post to the developers list on google groups but it appears to be stuck in moderation. If anyone could take a look at that I would appreciate it.

The post includes three suggestions for changes to future board versions. The one most likely to be of serious interest is removing the ceramic resonator on Uno/Mega boards (perhaps others too). By that I don't mean replacing it with a crystal, I mean getting rid of it entirely. I've done sample layouts for both Uno and Mega boards showing how this could be done. If you're interested either see my post on the developers list (if it ever gets moderated) or ask me more here.


Why? I've heard more complaints that the resonators aren't accurate enough, than suggestions that they be eliminated in favor of the internal oscillator (which in theory isn't quite accurate enough even for serial communications.) The layout need not change at all (although I guess you can get two extra IO pins if you eliminate the xtal.)

That's not my suggestion. Why not burn the CLKO fuse on the USB processor and use that 16MHz clock?

  • It is crystal controlled, thus solving the accuracy complaints (including from me).
  • It gets rid of two parts (the resonator and parallel resistor).

Seems like a win-win.

I have read comments that this was tried, but did not pass FCC limits. I have no knowledge of what was tried, but suspect that the clock was not routed as a top priority, but as an afterthought. With proper attention to trace length and AC grounding it should be possible to meet those limits. I have in fact run some E-M simulations indicating that my sample layout should work.

So, that's what I meant by saying a new layout is required. With the current positioning and rotation angles of the two processors I don't think this would be feasible.

I think this would be a real bonus on both sides -- better clock accuracy for users and smaller parts count where I suspect even pennies make a difference.

I don’t know if this would be worth the effort, since a crystal can be placed close to the MCU, is inexpensive, and it just works. I don’t think the USB-to-serial bridge should be a processor at all. It is an unnecessary complication, more expensive, and prone to accidental erasure and damage as many posts to the Installation & Troubleshooting forum have shown. An FTDI or similar chip as on the Redboard, Duemilanove, Metro 328, Nano, and many derivatives is just better. Or leave the USB-to-serial bridge off completely, as on the Pro Mini.
Krupski likes to replace the resonators with a crystal, and it is a very inexpensive hack. Discussed some here:

If even more inexpensive is good, those little cylindrical crystals are very cheap, and are also discussed in the same thread.

Well, it was just a suggestion.

Here are some observations:

  1. The ATmega16U2 is cheper than the FTDI FT232H -- $2.20 on Digikey for the ATmega16U2 in 5k quantities versus $2.60 for the FT232H in 3k quantity.

  2. FTDI may have shot themselves in the foot politically with their driver shenanigans trying to deal with counterfeit parts. I don't know if that's why Arduino made the switch, but it has left a bad taste in many mouths.

  3. I find that it is easier to burn the CLKO fuse on the USB processor and run a jumper wire. Yes, it will no longer meet FCC limits but that does not cause any trouble here around the house so I don't really care. But then, this is just my personal preference. Your mileage may vary.

  4. I am not privy to their cost models so I don't really know, but the volumes on this product are so large that a few cents may well be something they care about. From that viewpoint it may be attractive to get rid of a few more pennies in cost. Or not.

If they're not interested in this I totally understand...but just wanted to throw it out there to see.

Ah; that's what you had in mind. I never did understand why the two chips didn't share a clock from the beginning.

The ATmega16U2 is cheper than the FTDI FT232H -- $2.20 on Digikey

The FT232 is so 2008. You could substitute in a modern FT231x at about $1.50 (which doesn't need any crystal, either.) Or the Chinese CH340G seems to run about $0.50 (from "shady" sources...)

FTDI may have shot themselves in the foot politically

User-level users don't care. (and "Arduino" "may have shot themselves in the foot" with the whole .org vs .cc thing.)

the volumes on this product are so large that a few cents may well be something they care about.

I don't think that Arduino does much "cost reduction engineering." Perhaps they should.

Adafruit and Sparkfun both have cost/price-reduced Uno-like products (Metro and RedBoard, respectively) with significant re-design... Sparkfun in particular got rid of all the weird power-switching circuitry in favor of actually using the switch built into the power connector!

You can also notice that the Atmel ATmega328p Xplained Mini routes clock from the mEDBG chip to the 328p, instead of having a separate crystal/etc...

Thanks for the update on USB interface ICs. There's so much to keep up in this world with and I'm never up to date on everything.

Red board/Metro are nice too...and make a good point that Arduino could benefit from those cost engineering examples.