oscillator startup time

I don't understand your 7 million cycles at the end of a microsecond question. Do you mean 7000GHz? Or do you mean 7MHz? ???

Anyway, quartz crystals have a very high Q, more than 10^4 or 10^5. Their startup time is on the order of Q cycles, or possibly more than 5ms for a 15MHz crystal. That doesn't fit with your 43us / 8us timing requirement.

--
Off hand, I\'d say no.  Why don\'t you use a cheap clock oscillator
which runs all the time and just reset the counter?

Or an oscillator that can be gated on and off (they are available with enable pins) although that is virtually guaranteed to generate runt pulses at some time.

Cheers

PeteS

Don't do that, keep it running always, use Logic gates to switch it on/off .. That way, you can precisely control it via count of pulses if you wish using a counters. This will give you a perfect timing on the edges. something like 645 counts for the 43 us and 120 ct's for the 8Us.

i think i have the math correct there. i could be off by 1 decimal point. as far as the rest of your statement, i think you should read that again.

--
"I\'m never wrong, once i thought i was, but was mistaken"
Real Programmers Do things like this.
http://webpages.charter.net/jamie_5

I could. I just wanted to save one chip. Reading what you and other people answered, I believe it's the only way.

Thanks!

I meant that, even if the oscillator is slow to start up, I could at least get 7M cycles for the built-in counter to count by the end of the first microsecond. This would have given a minimal error at start up for my project.

*But* as far as I see, from you and other people who answered my question, it is not possible. Startup time is much longer. I will use a separate oscillator with a gate as a switch.

Thanks!

Octavio

Look at the High speed low pin count AVR's or PIC's, with careful coding, you can get your self a Xbase clock with the scaling generated signals you need from a single chip with the Xstyle Of course, don' expect that uC to do any other processing because it will up set the timing of course unless you use something like an AVR that can be calculated easy to account for the CPU usage of the clock. In any case, many of those uC have programmable counters that can output directly to a pin with on/off duty cycle programming and clock them self's directly from the base osc.

--
"I\'m never wrong, once i thought i was, but was mistaken"
Real Programmers Do things like this.
http://webpages.charter.net/jamie_5

For best results (not to say avoiding metastability) make sure it switches the signal on and off synchronously.

Cheers

PeteS

The project is more or less like an ansynchronous receiver with a clock 16 times the input data rate. So the clock doens't need to be synchronous as the error will be only few percent or 1/16.

Regards! Octavio

That's the problem: 7M means 7,000,000 cycles, and a 15MHz signal has only 15 cycles in one microsecond. Or did you mean at least 7 cycles in the first us, which would be like an immediate 7MHz signal?

A crystal oscillator starts up when the amplifier noise at 15MHz is captured by the crystal and saved as resonate energy. This saved signal is amplified and then again presented to the crystal, so its signal builds up. But until the amplitude becomes high enough for logic circuits to properly respond there is no output. Also, when the logic output first begins, after the long no-output delay, it has poor edge risetimes and isn't suitable to use for flip flops, counters, or microprocessor circuitry. That's why microprocessors have a long forced-reset interval. Otherwise they could perform incorrect operations, or crash, from the defective clock signal.

Many engineers use a gated oscillator circuit, which can start instantly, and is trimmable to a reasonable frequency accuracy, at least over a short time frame, like 43us (only 645 cycles).

If you need we can point you to such a circuit, or present one as an ASCII drawing.

--- "Resonate energy"???

Not true. When the crystal is hit with noise which causes its physical dimensions to change, it replies by "ringing" at its resonant frequency and generating a voltage when it relaxes. If the amplifier to which it's connected is biased around noise and has enough gain, then that voltage will be enough to cause a transition in the amplifier's output which will, if it's timed correctly, hit the crystal again, squeezing it and, once that edge is gone, causing the crystal to generate a voltage which will cause the cycle to begin anew.

-- JF

You don't think of that as the resonant energy building up in the crystal? The crystal's increasing voltage appears across it's resonate capacitance, with E = 0.5 C V^2. On the other half cycle this energy is stored as current in the crystal's inductance, E = 0.5 L I^2. That's increasing resonate energy, driving the increasing voltage appearing across the crystal. We're speaking, of course, of the very well-known electrical resonate LCR models of the crystal's piezo-mechanical action.

It is possible to get one-cycle turn-on times, *but* you then must accept a square-wave oscillator (well, as square as the IC will allow) and, in this case an overdriven crystal. I suggest you use a comparitor, using one input as the oscillator) gate.

One "runt" pulse at the end of the gate time, at worst.

--
OK.

Or if the precision doesn't need to be perfect, use a RC oscillator. It will start in one cycle. Alternatively add a crystal to calibrate the RC oscillator

Regards

Klaus

I think they mean that the gate edge should be in sync with the clock so that you don't get fractional pulses, which could confuse the counter.

Good Luck! Rich