Member #118891

Member Since: February 5, 2010

Country: United States

  • What could POSSIBLY go wrong?

  • A good cap for crystals that spec about 14pF loading capacitance.

  • The capacitance values shown in the schematic above imply a crystal with a load capacitance of about 16pF.

    A very common misconception is that the load capacitance spec somehow comes from the processor maker. The crystal is what drives this spec. Unfortunately the processor vendors tend to get beaten up on this point, with people asking them what the cap values should be..

    If the crystal specs 22pF load capacitance, then it wants to see the combination of the two caps in series, plus any parasitic capacitance, end up at 22pF. A good first shot would be ((Cload * 2) - 5) or ((22*2)-5) = 39pF. Try that value on your board, and verify oscillation frequency.

    Putting 22pF caps on a 22pF crystal is not correct design, and can leave you with an oscillator circuit that won't reliably start under some conditions, and it will be slightly off frequency.

  • The crystal data sheet tells all.

    For a crystal with 22pF load capacitance spec, you will want to use roughly 39pF caps. The crystal sees these caps as in series, so you need roughly twice the spec value, and then subtract about 5pF for stray capacitance, which will depend on layout.

    ((22*2)-5)= 39pF.

  • Why 22pF caps? With 20pF rated crystals, the cap should be closer to 39pF each, depending on the parasitic C on the particular board.

    22pF "works", for certain values of "work".

  • It works whether or not you're grounded.

  • Actually, I'm at this moment testing a few products here at EMITestlab.com We got another case of "Is that thing on" :) Dennis says "A few hundred bucks will give you a couple of scans from 30M to 1Ghz, that correlate to NIST".

    The misconception is that it's really hard to pass part 15 with common micro controllers and RF links. It can be, if your layout is bad, and you are using edges that are much harder than you need.

    BTW Ben, the fine for products that violate part 15 can be $10k per unit per day, plus recall and confiscation of non-compliant stock. CE is worse, and can include jail time.

    The "Ham Nube" you were talking about above, is an extra class licensee, as I am. A lot of RF and embedded systems guys are also hams. Without those pesky FCC/CE requirements, you probably wouldn't be able to use any of the radio spectrum in a constructive way.

  • Sorry Ben, you don't understand the problem, or the solution.

    Your Raspberry Pi is not an INTENTIONAL transmitter, but it's clocks and edge rates are high enough that its cables can be fairly effective antennas.

  • I agree with Andrew, KC6ETE here. I design low power radios and embedded systems professionally. I have seen in the small shops, an attitude of "that doesn't apply to me", or "This isn't a transmitter so I don't need to do the testing". :( If you're going to sell it, and it has a clock > 9kHz, and it's not a digital watch, then you'll at least need radiated emissions. If you intentionally transmit, then you need more testing.

    Some labs do charge a fortune for such testing, and when there's a problem you get a blank stare and a big bill. For those in the Boulder area, EMI Testlabs does inexpensive radiated and conducted tests, and when you hit a problem, Dennis can offer very good input on how to deal with it. He likes working with the little guys. His testing is done indoors inside a GTEM cell, so weather isn't a factor.

    Few designers in my experience, do enough up front with intelligent PCB layout and bypass design. They throw ferrites and shield cans and multilayer boards at the problem. It is actually pretty easy to do quiet designs on two layer boards in plastic enclosures, without spending any extra money on "fixes". I'm very happy to say that our most frequent comment at testing is "Are you sure that thing is on?"

  • The screwdriver is significantly changing the magnetic circuit, coupling more flux into the pickup.

No public wish lists :(