Choosing a crystal oscillator

[Dan Moos]

I'm looking for a crystal oscillator for my first go at a 6502 project. I plan to do 1 mHz for now, just to get something going.

If my project is 5 volt CMOS, is a TTL oscillator can OK?

Specifically, this :
https://www.amazon.com/gp/aw/d/B00B886Y ... ref=plSrch

[GARTHWILSON]

I don't see anything on the page saying that it puts out CMOS levels. If its logic high only meets TTL levels, the W65C02 is not guaranteed to work with it unless you put a 74HCTxx gate in between. The "T" in "74HCT" or "74ACT" means its inputs can work with the lower TTL-high voltage level, and its own output-high level will be close to 5V instead of the 3V or so of TTL. TTL is not able to pull up very high.

You'll find 2MHz and probably 4MHz too to be just as easy as 1MHz though, assuming you use something like 70ns memory instead of something old and slow like 250ns. You could get a can oscillator in the neighborhood of 8MHz and divide the frequency by two with a 74ACT74. The circuit is approximately in the middle of the page at http://wilsonminesco.com/6502primer/ClkGen.html in the 6502 primer. The higher-frequency oscillator will probably be cheaper, and you'll solve the above problem, and get a faster computer out of it, all at once.

[BigDumbDinosaur]

Dan Moos wrote:

I'm looking for a crystal oscillator for my first go at a 6502 project. I plan to do 1 mHz for now, just to get something going.

If my project is 5 volt CMOS, is a TTL oscillator can OK?

Specifically, this :
https://www.amazon.com/gp/aw/d/B00B886Y ... ref=plSrch

That item appears to be Jameco part number 27861. As far as the output of that oscillator goes, Jameco's page for it does not indicate whether its output is TTL or CMOS. I'd be cautious, as the 65C02 and 65C816 are somewhat fussy about the characteristics of the Ø2 clock signal. As Garth suggested, you could run its output through a 74ACT gate (AND, OR, etc.) and give the signal a boost. A another solution would be to use a 2 MHz oscillator and run it through a flip-flip, which will both strengthen the signal and assure exact 50-50 symmetry.

Two other points: firstly, depending upon your location, it may be a bit cheaper to order directly from Jameco's website. Secondly, a half-can oscillator has the same functionality but occupies about one-half the space of the full-can part you are looking at, plugging into a DIP8 socket or one of those made-for-half-can-oscillators sockets. Here's a suitable part at Mouser Electronics, but in a half-can package, and with output that almost swings rail-to-rail, more than satisfying CMOS input requirements (VOH is 4.5 volts minimum). I use the same part, though at a higher frequency, in my POC units. The below pic shows the Ø2 clock oscillator in POC V2, mounted in a compatible socket. The SOIC14 chip (U3) to the right of it is a 74AC74 flip-flip, which is used to condition the oscillator's output. Hence Ø2 is one-half the frequency of the oscillator.

Half-Can Oscillator in POC V2

[Dan Moos]

Thanks guys. I just ordered 4 of the oscillators that BigDumbDinosaur suggested. I also ordered some 2 mHz parts also.

Follow up. At what clock speed do I need to start worrying about transmission line effects/chip heat/propagation delay issues/yada yada? What's the "safe zone" for a hack job on a breadboard?

[GARTHWILSON]

The rule violations that will bite you are not really about MHz but about rise and fall times. Those of 74HC are slow enough that you won't have any problem with a wire-wrapped board as long as you keep parts close together and don't get a foot-square sprawl. If you do go big, the non-clock lines will be much more forgiving. If you don't cascade a bunch of logic levels and their delays, you probably won't need anything faster than 74HC, except maybe for Φ2 clock inputs, particularly with WDC's processors which want a very fast, clean clock input. You will always have to watch the timing diagrams though. Don't think that they ever just "go away."

[BigDumbDinosaur]

Dan Moos wrote:

Thanks guys. I just ordered 4 of the oscillators that BigDumbDinosaur suggested. I also ordered some 2 mHz parts also.

Follow up. At what clock speed do I need to start worrying about transmission line effects/chip heat/propagation delay issues/yada yada? What's the "safe zone" for a hack job on a breadboard?

I've long said that 8 MHz is possible with 74HC logic on a reasonably well-designed wire-wrapped board. Higher frequencies are possible on a wire-wrapped unit with meticulous construction and faster logic. Transmission line effects are more likely to get you with a solderless breadboard than with wire-wrapped construction. Just for your reference, electrical impulses travel about six inches per nanosecond, which means, to repeat a quote often heard around here, you can get away with murder at 1 MHz.

Reactive effects are the enemy of computer circuits, due to the high switching speeds that are involved. The longer the circuit paths the greater the series inductance and parasitic capacitance, and the greater the difficulty in maintaining good timing and acceptable waveform quality. Solderless breadboards are notorious for reactive problems, especially inductance from long curving wires going all over the place. Getting such a design to run in a stable fashion at more than a few megahertz is a real challenge.

Turning to glue logic, the more granularity it has the more important propagation delay becomes in how well the system will perform. The 6502 family does stuff on both phases of the clock. Generally speaking, the address bus is set up by the MPU during Ø2 low and the data bus becomes active when Ø2 is high. In other words, the time available for a read or write to occur is more-or-less one half the clock cycle time, 50 nanoseconds at 10 MHz, for example.

A little thought will show that in order to guarantee that an addressed device will reliably respond to read/write requests from the MPU, the device's chip select and read-enable or write-enable inputs must be asserted in enough time for the device to capture the data bus (write) or drive the data bus (read). If too much time passes from when the MPU puts the device's address on the bus until the device is ready for access the system will be unstable or may crash. Prop time through the glue logic, as well as overall system design, makes or breaks the timing and hence makes or breaks a design.