100 KHz Crystal Oscillator

[rcbuck]

I've been experimenting with a 100 KHz crystal oscillator with some success. I know these are really not crystals but are tuning fork oscillators at these frequencies.

I am using a Mouser part number 815-ABS25-100KT part. I can get the part to oscillate by adjusting the trim cap after powering the circuit up. I can bring it on frequency to exactly 100.000 KHz. However, if I power the circuit down and back up, the oscillator will not start without adjusting the trimmer until it starts. I can then adjust it back on frequency.

Circuit I am using is attached. Is there a better circuit that I can use?

[floobydust]

Some relevant threads:
https://www.eevblog.com/forum/projects/low-jitter-100khz-oscillator-for-ad/
https://www.eevblog.com/forum/projects/level-translate-the-output-of-a-tcxo-crystal-oscillator-module/

[Benta]

Increase R33 to 500k...1M.
The driving power for tuning fork crystals shouldn't be higher than around 1 uW. You're overdriving it.

[rcbuck]

floobydust, thanks for the links. Jitter is not important in my application. But the oscillator has to start reliably on power up. It looks like the 74LVC1GX04 that you referred to in the second link is probably the best approach.

Benta, I will try your idea and see if it works. However, according to several data sheets for different crystals, 47K would be the best value to be used for R33. That sets the high frequency cutoff at 125 kHz if C24 is set to 24 pF. Getting low frequency crystals to work is always a challenge. I have had problems with the 32.768 kHz ones in the past.

[Benta]

Main mistake: don't use Schmitt triggers for oscillators. Big no-no. Best solution is an unbuffered inverter:
74LVC1GU04 is ideal.
The output buffer you can leave as 74LVC1G14

[floobydust]

Just turning the trimmer cap can inject AC hum, enough for the oscillator to start (if the screw is not GND).

I suggest not using a Schmitt trigger inverter 74LVC1G14 for U6.
Consider NC7SU04 TinyLogic HS Unbuffered Inverter.  U7 is OK.

These circuits are difficult to troubleshoot as scope loading upsets things. I get them tuned and then they won't startup at low temperatures 

Make sure the load capacitances are optimal. Your math or model might be wrong. Try different values and see how it does.
These circuits are difficult to troubleshoot as scope loading upsets things. I get them tuned and then they won't startup at low temperatures. I usually hold the scope probe close with a large air gap to spy on things like startup time which helps me know if tuning or gain is improving or not.
Watchmakers use an inductive pickup coil to calibrate 32kHz oscillators. I've never gotten around to building one.

[schmitt trigger]

When I've experimented with tuning fork oscillators, my experience has been similar to what other posters have noted: ultra low drive is required for reliable startup and to prevent damage. I have also had success increasing the series resistor to 500K.

Additionally, you do require an unbuffered inverter. I've traditionally used one stage of a CD4049U as the oscillator itself, and the remaining stages as the buffer.

[Zero999]

Another good IC for crystals is the 74LVC1GX04.
https://assets.nexperia.com/documents/data-sheet/74LVC1GX04.pdf

[Benta]

Zero999, that's a nice one. I've added it to my list of "devices to keep in mind".

[rcbuck]

After studying the data sheet, I am going to use the 74LVC1GX04. That is probably the easiest and most reliable approach.

[floobydust]

ST app note AN5073 on SPDIF receivers, 2.2.1 the guy measures the open-loop gain of SN74LVC1GU04 at 36, and finds instability with SN74LVC1G04 with slow signals. Interesting to note.

[vk6zgo]

I've been experimenting with a 100 KHz crystal oscillator with some success. I know these are really not crystals but are tuning fork oscillators at these frequencies.

Then you "know" wrong -------100kHz crystals are as common as dirt!

I am using a Mouser part number 815-ABS25-100KT part. I can get the part to oscillate by adjusting the trim cap after powering the circuit up. I can bring it on frequency to exactly 100.000 KHz. However, if I power the circuit down and back up, the oscillator will not start without adjusting the trimmer until it starts. I can then adjust it back on frequency.

Circuit I am using is attached. Is there a better circuit that I can use?

Edit:-OK, I can see where you are coming from with the "tuning forks" terminology.
Yes, low frequency crystals are often cut into tuning fork form, but they are not tuning forks, & are several orders of magnitude more accurate & stable than real versions of those devices.
In any case, 100kHz crystals are not among this group, & are cut like other higher frequency crystals are.

As I pointed out, they are very common, as are stable oscillator circuits which use them---- Google for "100kHz calibrator" circuits.

[szoftveres]

I found so many people suggesting to use of unbuffered inverters, versus buffered (like the 74HC04).
I had similar experience; a low frequency pierce crystal oscillator I built ran just fine with 4069UB, but it didn't start at all with 74HC04.
Why is that? What makes the (presumably higher gain) buffered HCMOS ICs such a bad choice for building an oscillator? I would be curious about the explanation.

[edavid]

Edit:-OK, I can see where you are coming from with the "tuning forks" terminology.
Yes, low frequency crystals are often cut into tuning fork form, but they are not tuning forks, & are several orders of magnitude more accurate & stable than real versions of those devices.
In any case, 100kHz crystals are not among this group, & are cut like other higher frequency crystals are.

As I pointed out, they are very common, as are stable oscillator circuits which use them---- Google for "100kHz calibrator" circuits.

The device OP is using is definitely a tuning fork crystal.  It is much too small to be 100kHz AT cut.  (I have a few of those, and they are in oversize HC-6 cans about 2 inches long.  I don't think they have been a standard product for many years.)

[schmitt trigger]

I found so many people suggesting to use of unbuffered inverters, versus buffered (like the 74HC04).
I had similar experience; a low frequency pierce crystal oscillator I built ran just fine with 4069UB, but it didn't start at all with 74HC04.
Why is that? What makes the (presumably higher gain) buffered HCMOS ICs such a bad choice for building an oscillator? I would be curious about the explanation.

The way it was explained to me, is that unbuffered devices consist of a single NMOS/PMOS stage, which can be biased in "class A", necessary for the operation of this circuit.

I am sure other posters will add to this explanation.