is it possible? a crystal oscillator from SN74LS04 TTL hex inverter

[ali6x944]

is it possible to make a 32.786khz crystal oscillator from SN74LS04 TTL hex inverter ?
please help!
thanks.

[PA0PBZ]

[Ian.M]

I'd be very cautious about drive levels.  U1A input with R2 feedback means the crystal sees a fairly low impedance and you may well exceed the permissible drive power for smaller package crystals

[chris_leyson]

I wouldn't advise using TTL to drive a 32kHz xtal. If it's a tuning fork type crystal the motional resistance could be up to 50k ohm, secondly the drive level would be way to high, it's usually specified as 1uW max. You might get away with unbuffered CMOS inverters but getting the drive level down to 1uW is difficult.

[PA0PBZ]

I agree that it would be better to use a cmos version but that was not the question, and maybe it's a full size xtal, who knows 

[ali6x944]

it is a tuning fork crystal i salvaged out of an old clock  and i tried to use a cmos nand  74hc00n in the same configuration and it didn't work will,
it was acting strange, it was picking am transmissions and i was able to hear a talking through a laud speaker, but it was not that clear thew.
and it was not oscillating except when i put my fingers on the pins of the ic.

[Zero999]

The 74HC00 should be ideal for a crystal oscillator.

What circuit did you use?

The circuit posted above is for the TTL IC. CMOS ICs are slightly different.

[ali6x944]

how they are different?

[ali6x944]

because i literately used the same exact circuit but i add not a 10nf cap but two 20pf caps in earthier sides of the crystal.

[ali6x944]

and i did not use a coupling cap between gates

[uncle_bob]

Hi

The standard part to use always used to be the "un-buffered" version of the CMOS gates. A single inverter with about a 1 M ohm between output and input generally did the trick. Something like 20 or 30 pf to ground on both input and output kept things running properly. You may have a hard time finding UB (un-buffered) parts these days. Part numbers like CD4007UB still show up as available. The nice thing about "old" CMOS (CD series) is that the high frequency gain is not there. That's what messes you up doing 32 KHz with the LS, HC or (worse yet) AC gates.

Depending on the characteristics of the crystal you have (it's loss) you may still need two inverters in series to get enough gain. If so, do something like the schematic at the top of this thread.

Bob

[ali6x944]

Thanks a lot

[alsetalokin4017]

Using MCP14069UBCP:

(note: Chip is being supplied with only about 3.5 V in;  more voltage produces a distorted waveform with the component values I've used here;  this was originally a 20MHz xtal oscillator but I just decided to try it with a watch crystal for this test and didn't change any of the other components.)


Also tested, with same result, using CD4069CN and CD4069UBE.  Does _not_ work with 7404 variants I tried: SN74LS04J or plain 7404.

[ali6x944]

That is graet if am using UB-CMOS devices but  i don't have any,
The only gates i have are from the low-power schottky ones and a HC nand, and a HC or gates
I will post all datasheets in the form soon

[uncle_bob]

Hi

The 4000 series CMOS will interface directly to your HCMOS parts. There is no problem driving one with the other. The "normal suspects" will sell you a DIP package part for next to nothing and get it to you in a day or two.

The other alternative is to use a transistor (or two) rather than a gate.

Bob

[ali6x944]

bjt or jfet transistor?

[ali6x944]

https://drive.google.com/file/d/0B5vW-k7HbsL4MEtnZWo1dG4zTnM/view?usp=sharing
first

[uncle_bob]

bjt or jfet transistor?

Hi

BJT / JFET / MOSFET .... any  or all will work if you have enough of them. The old standard designs either used multiple bipolars (to do the impedance conversion) or a JFET and a bipolar.

Keep in mind that you will have something odd coming out of the circuit (like a clipped sine) that will then need to be converted to a square wave. A darlington BJT running in a colpits configuration is not a bad place to start. Cap from base to final emitter, cap from emitter to ground. Crystal goes from base to ground. Should see a nice signal on the emitter once it's biased up.

Bob

[ali6x944]

But i was just wondering if i could use tri-state invrting buffers to achive the oscilation instade of an a regular inverter?

[alsetalokin4017]

I tried the one-transistor circuit attached below from Bowden's Hobby Circuits but could not get it to work, probably because I didn't have the right trimmer capacitor and used a fixed 30pf instead. It just wouldn't start oscillating for me. Tried several different watch crystals too, that did work in the CMOS version, but no luck. Tried 2n3904 and BC337-25, no luck.

[uncle_bob]

But i was just wondering if i could use tri-state invrting buffers to achive the oscilation instade of an a regular inverter?

Hi

To make an oscillator work, you need to satisfy both phase and gain criteria. In order to do that, things need to be set up right. There is no guarantee that they will only be set up right at one frequency. By far the most common error with 32 KHz oscillators is using a device that *also* oscillates at a much higher frequency. Normally this is through some portion of the stray capacitance.

Bob

[ali6x944]

how?what? could you simplify more?
please I'm a noob at these things 

[uncle_bob]

how?what? could you simplify more?
please I'm a noob at these things 

Hi

The simple / easy / quick / cheap way to do it is to grab a 4000 series un-buffered gate for < $1 and wire that up. There are only a
a small number of parts involved and the chance of success is very high.

Doing a 32 KHz oscillator with discrete parts is a trial and error process. You very much need to know what you are doing. There are a much larger number of parts involved and more of them are critical. That's why nobody does it that way anymore. You *can* do it. It has been done that way many times. It's not the easy way. The other (minor) issue is that it's only going to work with the right parts and they are going to cost more than the gate .....

Bob

[Zero999]

The standard part to use always used to be the "un-buffered" version of the CMOS gates. A single inverter with about a 1 M ohm between output and input generally did the trick. Something like 20 or 30 pf to ground on both input and output kept things running properly. You may have a hard time finding UB (un-buffered) parts these days. Part numbers like CD4007UB still show up as available. The nice thing about "old" CMOS (CD series) is that the high frequency gain is not there. That's what messes you up doing 32 KHz with the LS, HC or (worse yet) AC gates.

If using unbuffered parts is not an option, then how about adding an input resistor and feedback capacitor to reduce the cut-off frequency to something more sensible?

[uncle_bob]

If using unbuffered parts is not an option, then how about adding an input resistor and feedback capacitor to reduce the cut-off frequency to something more sensible?

Hi

The problem there is that the spurious oscillation is across the package its self on the high speed gate. The next issue is that the R and C have added a bit of a phase shift / load inside your loop. That's going to mess things up as well.

Bob