Crystal oscillator using 7404 is unstable

Question

I am currently building a small Z80 microprocessor system and require some kind of clock generation. It is not my first design: Earlier I used an Arduino pin that generated a slow clock or a 555 timer. However, earlier designs only included the Z80 and some RAM and ROM.

Now I want to expand upon that and include peripherals. That's why I chose a 3686.4 kHz Quarz, so that I can divide it later via a frequency divider to not only generate my processor clock but also my UART clock.

I did some research on the internet, read some books and schematics and found out that most use a pierce oscillator or a series resonant oscillator. For my design I want to use plain 74xx or 74LSxx chips, because I have them around and ready to use. Fortunately, most books for the Z80 and websites also built them using 74xx chips. So I tried to rebuilt them. My pierce design did not work at all, so I quickly changed my setup to a series resonant oscillator.

For my oscillator I use a single 7404 inverter that has a 100 nF decoupling capacitor attached between 5V and GND. Then I simply copy-pasted the circuit from "Build Your Own Z80 Computer" page 94. I added the capacitor after seeing it on many schematics and without it, it didn't start oscillating anyways. I attached my schematic below (it does not show the decoupling capacitor between 5V and GND)

Now when I apply some voltage to this circuit it indeed starts to oscillate. However, the frequency is not stable at all, seems to be 200 kHz below the stated one and the duty cycle is not at all 50%. You can see my logic analyzer output below.

Whereas the logical high state seems to be quite stable with one unstable cycle seeable at the right end of the graph, the off state varies quite a bit.

Frankly speaking, I do not know how to fix this as I have not much experience with oscillators. However, I have a feeling that the capacitor does have quite an influence. When I changed it from 100 pf to 22 pf the frequency was completely unstable. When I changed it to a higher value, I didn't see much difference at all. I also tried to exchange the 7404 with an 74LS04 as some schematics use it, but that didn't change anything.

If somebody can point out my mistake to me I would be very glad.

Additional Info: The data sheet of the crystal calls states CL=20pf and Rr < 120 Ohm.

Answer 1

Something you need to be careful with when using the Saleae device is that it takes samples, not continuous measurements. At 24MHz it is taking a sample every 41ns or so. The period for your 3.686MHz clock is 271ns. I'd expect to see some aliasing as the ratio between the two is 6.6:1.

An oscilloscope would be a better testing option.

Answer 2

This is just to teach you why your TTL oscillator fails in this Pierce series resonant mode, you have to see how TTL works.

Fixes

• change 100pF to 10nF to eliminate phase shift at 3.5MHz
• add 2.4k input to ground to pull output >> 2.0V towards 2.4V (opt.)

The impedance of 100pF is too high and = 1 k @ 3.5MHz so you get a bad 45 phase shift twice and the non-inverting loop is no longer in phase to satisfy the Barkhausen criteria to oscillate.. (Av>1 at 0 deg.)

This is the design of the TTL Inverter 7404 with 2 resistors added on input and coupling cap.

They changed the currents, resistor values and bias methods later with Schottky diodes but the current ratios were always 10:1 and the input threshold was/is always 2 diode drops or Vbe drops = 1.5V +/-0.1 depending on bias current and temp.

Parallel resonant XTAL OSC or XO's are best used with CMOS up to 20MHz and Series resonant transistor XO's above this f to avoid sensitivity to stray capacitance.

Answer 3

I looked up an old Z80 computer design from the 80ies (the mc CP/M-Computer). It also used an oscillator with 7404 gates:

The capacitor used there is 10nF. So I guess values in the pF range are much too small.

Answer 4

You have a series resonant oscillator, and the crystal is meant to be used with parallel resonant oscillator as it requires 20pF load capacitance. Either change to parallel resonant oscillator, or change the crystal to one meant for series resonance.

Answer 5

I have a 2 MHz crystal, and a 4 MHz, not your 3.odd MHz. With odd contacts on a breadboard, I first could not make your circuit work, but I have one that is almost like yours, just removing that capacitor altogether. And it works perfectly every time without that capacitor.

When I use you 100 pF I get a very dirty signal on the 2 MHz. If I increase the capacitance to 10 nF, as someone suggested it looks better. But then, the higher the capacitance there goes, at MHz frequencies the more it approximates just having no capacitance at all.

One commented under my earlier version of this answer said my circuit here doesn't work, but it does, I just had my analog scope settings wrong so I measured 4 times higher frequency. But that was just a newbie's operator error.

Frankly, I don't see the point of all this extra stuff if this here works much more reliable and simple. No capacitors to worry about frequency.

^{simulate this circuit – Schematic created using CircuitLab}

I don't know why everybody is showing the more complex and harder to get to work schematics. Possibly it has something to do with wearing out the crystal.