I've got a 74HC86 (quad XOR) and I'm using 2 of the gates (with one input
tied high) to make inverter oscillators. One crystal at 14.85MHz in a
SMT package was relatively easy to start up and stabilize. The 25MHz
crystal (in a short can package) was very sensitive to Rs (series resistor
between the inverter output and one leg of the crystal) and low supply
voltage. To get it to start reliably I ended up with Rs=110 @ 5V. The
whole thing is drawing about 25mA.

The problem I have is that the 74HC86 and 25MHz crystal seem to warm
up quickly and cause the 25MHz to drop by about 200Hz. This oscillator
isn't intended to be on for long periods of time, so I'd like to avoid
that. I've isolated those two parts by a combination of heating and
cooling components to see which contribute the most (the temp swing is
too small for me to feel, I'm just warming with a finger and cooling
with a puff of compressed air).

So if I'm putting 15mA+ into that crystal, is that unusually high?
What would be a typical amount for such an oscillator? Is there
anything I can do about the effects of chip heating (I'm assuming
the gate capacitance is going up as it heats)?

--
Ben Jackson

http://www.ben.com/

On Fri, 30 Jun 2006 15:48:03 -0500, Ben Jackson wrote:

I've got a 74HC86 (quad XOR) and I'm using 2 of the gates (with one input
tied high) to make inverter oscillators. One crystal at 14.85MHz in a
SMT package was relatively easy to start up and stabilize. The 25MHz
crystal (in a short can package) was very sensitive to Rs (series resistor
between the inverter output and one leg of the crystal) and low supply
voltage. To get it to start reliably I ended up with Rs=110 @ 5V. The
whole thing is drawing about 25mA.

The problem I have is that the 74HC86 and 25MHz crystal seem to warm
up quickly and cause the 25MHz to drop by about 200Hz. This oscillator
isn't intended to be on for long periods of time, so I'd like to avoid
that. I've isolated those two parts by a combination of heating and
cooling components to see which contribute the most (the temp swing is
too small for me to feel, I'm just warming with a finger and cooling
with a puff of compressed air).

So if I'm putting 15mA+ into that crystal, is that unusually high?
What would be a typical amount for such an oscillator? Is there
anything I can do about the effects of chip heating (I'm assuming
the gate capacitance is going up as it heats)?

From a stability factor point of view those "inverter" based
oscillators are amoung the worst. I doubt it's all crystal heating
and I bet there are parameter shifts in the "chip" as it warms up
operating at high frequency like that. Also If you drive the crystal
hard like they do (15ma!!!) bad things happen. Those circuits are OK
for micros and other less critical apps.

I suggest a discrete Bipolar or FET where you can control the
operating point of the device better.

Allison

On 2006-06-30, wrote:
From a stability factor point of view those "inverter" based
oscillators are amoung the worst.

Good to know. I was trying to reduce part count -- the quad XOR provides
the two oscillators plus one mixer.

and I bet there are parameter shifts in the "chip" as it warms up

The other parameter shift I was missing was droop in the 5V rail as
the 78L05 warmed up (rapidly!) and noise injected by a MAX202 charge
pump also on that 5V rail. I thought that regulator was just for an
(unused) LCD header, but I was wrong.

I suggest a discrete Bipolar or FET where you can control the
operating point of the device better.

If I can't tame this VCXO I will try that. This must be how Manhattan
prototyping branches off into "Ugly"...

--
Ben Jackson

http://www.ben.com/

On Sat, 01 Jul 2006 02:23:19 -0500, Ben Jackson wrote:

On 2006-06-30, wrote:
From a stability factor point of view those "inverter" based
oscillators are amoung the worst.

Good to know. I was trying to reduce part count -- the quad XOR provides
the two oscillators plus one mixer.

and I bet there are parameter shifts in the "chip" as it warms up

The other parameter shift I was missing was droop in the 5V rail as
the 78L05 warmed up (rapidly!) and noise injected by a MAX202 charge
pump also on that 5V rail. I thought that regulator was just for an
(unused) LCD header, but I was wrong.

Those added elements don't help. However I've used flavors
(both TTL and CMOS) of the circuit and stability is only ok at best.
One use was a simple freq counter (commercial purchase) and I
could never get it to stay zeroed on WWV, it was always 100hz or
more off every time I'd check it. I tried everything and finally
pulled that out and added a simple transistor circuit. After than the
error was usually less than a few hz.

I suggest a discrete Bipolar or FET where you can control the
operating point of the device better.

If I can't tame this VCXO I will try that. This must be how Manhattan
prototyping branches off into "Ugly"...

It do at that.

Allison

On 2006-07-01, wrote:
On Sat, 01 Jul 2006 02:23:19 -0500, Ben Jackson wrote:

I suggest a discrete Bipolar or FET where you can control the
operating point of the device better.

If I can't tame this VCXO I will try that. This must be how Manhattan
prototyping branches off into "Ugly"...

Well, this might be the best advice I ever got from Usenet. I fought
every parameter in the 'HC86 inverter oscillator and lost. The fact
that it works great at 14.85MHz and the elegance of using up the extra
gates lured me in. The whole thing is very voltage sensitive (around
1Hz/mV at the output of the gate, making it very sensitive to Rs as
well). This might be due to the propagation parameters of the gate
varying substantially with temperature and voltage.

I built a Colpitts oscillator with a plain old 2N3904 and it hasn't
moved by more than a few Hz in the hour it's been on.

Thanks!

--
Ben Jackson

http://www.ben.com/

Ben Jackson wrote:


Well, this might be the best advice I ever got from Usenet. I fought
every parameter in the 'HC86 inverter oscillator and lost. The fact
that it works great at 14.85MHz and the elegance of using up the extra
gates lured me in. The whole thing is very voltage sensitive (around
1Hz/mV at the output of the gate, making it very sensitive to Rs as
well). This might be due to the propagation parameters of the gate
varying substantially with temperature and voltage.

I built a Colpitts oscillator with a plain old 2N3904 and it hasn't
moved by more than a few Hz in the hour it's been on.

Thanks!

Ben Jackson

Hi Ben,

I have been following this interesting thread. Thanks for posting the data.

Could you do me a favor and post your 2N3904 schematic? I'd like to put it
in my SPICE Crystal Analysis program and see if it can verify the
performance you obtain. If you are interested, the paper is at

http://tinyurl.com/qpcoz

Regards,

Mike Monett

On 2006-07-04, Mike Monett wrote:
Ben Jackson wrote:

I built a Colpitts oscillator with a plain old 2N3904 and it hasn't
moved by more than a few Hz in the hour it's been on.

Could you do me a favor and post your 2N3904 schematic?

___ +5V
.--|33k|---+
| |
| |/
.-o---o----| 2N3904
| | |
| --- |
| --- .--o----
_-_100p| | |
|___| o---' .-.
- | | | 150
25Mhz | --- | |
| --- '-'
| 100p| |
o-----o------'
|
===
GND

The 33k/150 were chosen to get CMOS compatible output. The circuit will
oscillate with much higher values for each.

The two 100p might be slightly too high for the 18p crystal I'm using,
but I didn't tune those carefully because I want to pull the crystal
slightly, and in my actual circuit the grounded leg of the crystal is
actually in series with a variable capacitance to tune the VCXO.

(created by AACircuit v1.28.6 beta 04/19/05 www.tech-chat.de)

--
Ben Jackson

http://www.ben.com/

Ben Jackson wrote:

The 33k/150 were chosen to get CMOS compatible output. The circuit will
oscillate with much higher values for each.

The two 100p might be slightly too high for the 18p crystal I'm using,
but I didn't tune those carefully because I want to pull the crystal
slightly, and in my actual circuit the grounded leg of the crystal is
actually in series with a variable capacitance to tune the VCXO.

(created by AACircuit v1.28.6 beta 04/19/05 www.tech-chat.de)

Ben Jackson

Hi Ben, Thanks. What is the nominal setting of the cap in series with the
crystal? 18pf or so?

Do you have any info on the crystal, such as make and model?

Or any idea what the series resistance might be?

Would you like me to post the results to my web site?

Regards,

Mike Monett

Mike Monett wrote:

Would you like me to post the results to my web site?

To clarify - so you can see the waveforms?

Regards,

Mike Monett

On 2006-07-04, Mike Monett wrote:
Ben Jackson wrote:

The two 100p might be slightly too high for the 18p crystal I'm using,

Hi Ben, Thanks. What is the nominal setting of the cap in series with the
crystal? 18pf or so?

The "varicap" is actually a red LED, which G0UPL characterized as about
30p down to about 20p over 0..5V:

http://www.hanssummers.com/radio/varicap/index.htm

Do you have any info on the crystal, such as make and model?

It's a Digi-Key X231-ND, mfg p/n ECS-250-18-4-F. Digikey has the
datasheet. I think the ESR was 40 ohms.

Would you like me to post the results to my web site?

Sure.

--
Ben Jackson

http://www.ben.com/