moss () (12/22/90)
When designing a parallel mode oscillator like those used on most microcontrollers (80C31, 68HC11, etc.), it is necessary to specifiy a parallel crystal with a given load capacitance. I am trying to come up with a formula for determining the load capacitance. |\ | \ -----| >O------ | | / | | |/ | | | | | | | ----- | | | | | | | | +--| | | |--+ | | | | | | | | ------ | | | XTAL | | | | | C1 ------- ------- C2 ------- ------- | | | | | | ------- ------- ----- ----- --- --- For the circuit above, the load capacitance should be Cload = C1 + C2 + 2 * pin capacitance Now most design examples use C1 and C2 in the range of 20 to 33 pF, and pin capacitance is approx. 10 pF for most ICs, so Cload should be in the range of 60 to 100 pF. However, the standard Cload values for stock crystals (and the design examples I've seen) usually have Cload in the range of 20 to 40 pF. Has anyone explored this issue or seen a good explanantion of this discrepancy? I will be using crystals in large enough volumes that I can specify Cload to the manufacturer; however, I feel that I might be missing something in my analysis. Any help would be greatly appreciated. Barry Moss P.S. I already know that parallel and series resistors have the same internal composition, but that they resonate at slightly different frequencies when connnected series or parallel. -- ------------------------------------------------------- Barry Moss (604) 277-1511 Mobile Data International, 11411 Number Five Road Richmond, BC, Canada V7A 4Z3
mcovingt@athena.cs.uga.edu (Michael A. Covington) (12/22/90)
If the load capacitance is wrong, the frequency will be a *tiny* bit off, but the crystal will still oscillate (assuming the capacitance is not totally unreasonable). Actually, crystal load capacitance is something of a red herring because few people measure it very accurately -- not even the crystal makers. Are you building a 24.9999999-MHz 386? :) (I do mean a *tiny* bit off. Parts per million.)
stevem@specialix.co.uk (Steven Murray) (12/27/90)
Barry Moss writes, concerning load capaitors on crystal oscillators: >When designing a parallel mode oscillator like those used on most >microcontrollers (80C31, 68HC11, etc.), it is necessary to specifiy >a parallel crystal with a given load capacitance. I am trying to >come up with a formula for determining the load capacitance. Mr Moss - I can't really answer your question - I do not know why 22pf - 33pf capacitors are the ones that most often get used in this situation - but I do know that they are often the most suitable. A year or so ago I had to design a built-in modem for an 80C31 based product, and this required that we have the 80C31 oscillator be spot on in terms of frequency - the 80C31 oscillator is not really rated in terms of accuracy, and for best results with a switched capacitor modem, you want to be on frequency. (There were good economic reasons for using the 80C31's oscillator). Being a bit pragmatic I gave up on the books, got a load of different manufacturers 80C31's, crystals, a good frequency counter, some freeze spray and a heat gun, and lots of caps. In this particular situation 22pf on one side and 39pf on the other gave the best, stable, trimmed frequency. I had decided that I would be happy if I could consistently get 50ppm frequency accuracy in production, with maybe a few units out to 100ppm. I was getting 25ppm on the lab bench, so left it at that. Regards, Steven Murray -- Steven Murray uunet!slxsys!stevem stevem@specialix.co.uk I am speaking, but | If these are your opinions, then we are in agreement!! not for my employer.| Flames, spelling errors, complaints > /dev/null
johne@hp-vcd.HP.COM (John Eaton) (01/03/91)
<<<< < For the circuit above, the load capacitance should be < < Cload = C1 + C2 + 2 * pin capacitance < < Now most design examples use C1 and C2 in the range of 20 to < 33 pF, and pin capacitance is approx. 10 pF for most ICs, so < Cload should be in the range of 60 to 100 pF. However, the < standard Cload values for stock crystals (and the design < examples I've seen) usually have Cload in the range of 20 to < 40 pF. ---------- Cload is the capacitance as seen by the crystal. Cload = (C1+pinCap)*(C2+pinCap)/((C1+pinCap)+(C2+pinCap)) If you ignore the ground for a moment you will see that the two caps are in a series (not parallel) configuration as seen by the crystal. John Eaton !hpvcfs1!johne
verive@tellabs.com (Jeff Verive) (01/03/91)
In response to the question about crystal load capacitance - | ---- | crystal +----------| | | |----------+ | | ---- | | | | | |\ inverter | | | \ | +-----------| >()----------+----------[] output | | / | | |/ | ----- ----- ----- CL1 ----- CL2 | | | | +-------------+--------------+ | ----- --- - Given the above "typical" crystal oscillator, the values of CL1 and CL2 can be determined with a reasonable amount of precision. Before going into the procedure, let's lay down just a little ground work. In any oscillator, two criteria must be met (the Barkhausen Criteria) : 1) loop gain must be at least equal to 1.00; and 2) phase shift around the loop must be an integer multiple of 360 degrees. These criteria must both be met at the frequency of interest. Now, looking back at the circuit, we have an inverting amplifier and a feedback network. The feedback network is therefore a crystal in parallel with a series-capacitor pair (neglect the ground point for this analysis; it can be shown that there is a circulating current in the feedback network that does not "see" the ground.) Since our oscillator must have 360 degrees (or an integer multiple thereof), and the inverter provides 180 degrees, the feedback network must provide the other 180 degrees. For the feedback network to introduce a 180 degree phase shift, it must have the form of an L-C tank (given the above configuration), and we immediately see that the crystal takes the place of the "L" in the tank. With all that said, it is a fairly simple job to determine the values of the load capacitors CL1 and CL2. That is to say, the mathematics are quite straightforward, but there are practical problems with this circuit. Don't get me wrong, though; it's a good circuit - we just have to handle a few of "gotcha's". First of all, the inverter has a small value of capacitance associated with its input. For common logic families, this is typically in the range of 4-8 pF, so we must add this to CL1 (since it is actually in parallel with CL1). We will use a value of 6 pF for this purpose. Our output is used to drive some other circuitry, so we must also accommodate the capacitance associated with the inputs of the driven circuit. Usually we will choose to use another inverter to "buffer" the oscillator circuit; this way, the actual oscillator stage only has to drive a small, known load. The added capacitance is also about 6 pF, and we see that we must add it to CL2. There is also some stray capacitance associated with the wiring and/or printed circuit board traces, and this is usually only a few pF. Although this stray capacitance is distributed over the entire circuit, we usually treat it as if it were lumped into a 2-4 pF capacitance connected directly across the crystal. Finally, there is a capacitance associated with the output of the inverter, and this capacitance is a major source of frustration for the uninitiated. There are two "parts" to this capacitance; one is a physical capacitance due to the separation of conductors in the inverter's output stage and in the packaging, and the other is an effective capacitance which is due to the finite time required to propagate the signal through the inverter. The first capacitance is typically about 5 pF, and the latter, though highly frequency dependent, is generally taken to be in the range of 5-30 pF, with 15 pF a good compromise. These capacitances must be added to CL2. Now we can get down to determining the values of CL1 and CL2. One last caveat though - it has been empirically determined that excessively large or small values of capacitance can cause problems. I am not going to get into these problems here because they are too dependent on the technology (TTL, CMOS, NMOS, etc.), inverter configuration (compensated, de-compensated, Schmitt Trigger, etc.), and frequency. For most cases, however, we will be safe if we select values of CL1 and CL2 in the range of 10-40 pF. The most common scenario involves designing an oscillator given the crystal manufacturer's specified load capacitance, which we now know (from the discussion above) to be approximately given as 1 CL = 3 pF + ------------------------------------------------- 1 1 ------------ + --------------------------- CL1 + 6 pF CL2 + 6 pF + 5 pF + 15 pF We will commonly choose to make CL1 = CL2 (or CL1 a trimmer capacitor whose mid-range value is equal to CL2), so that given CL, it is fairly easy to calculate values for CL1 and CL2. For example, let's calculate CL1 and CL2 for a crystal whose load capacitance is specified as 18 pF : 1 18 = 3 + ---------------------- 1 1 ------- + -------- C + 6 C + 26 Solving this gives C = 17 pF. This is not a standard value, so 18 pF could be used for non-critical applications. If extreme accuracy is necessary, CL2 could be 15 pF and CL1 variable from about 10 to 50 pF, although actual values are much more likely to be determined after the basic circuit has been laid out on a printed circuit board (so that stray capacitances and other variables are made less variable.) In any event, the final values should match the calculated values within 10% to 20%. As is usually the case with high frequency analog circuitry, analysis is far more precise than is synthesis. I hope this has helped to de-mystify crystal oscillator operation. To be sure, this is a mere overview, but it addresses the major hurdles usually encountered in crystal oscillator design. One final note - the circuit and description above are for AT-cut crystals operating in parallel resonance and at their fundamental frequencies. While this may seem like a serious restriction on the utility of the above discussion, virtually all microprocessor crystals fall into this category. -- **************************************************************************** ** Jeff Verive | If they ever stop making those little candy flowers ** ** 259371048378 | for birthday cakes, I shall lose my will to live. ** ****************************************************************************