Maker Pro
Maker Pro

Oscillator question

G

garyr

Jan 1, 1970
0
I needed a 24 KHz oscillator so I thought I would use the one shown in
Figure 5.30, p285 of A of E (schematic below, view with fixed font). My
first version, which used a SN74LVC2G04, would not oscillate. The second
version using 2 SN74HC1G04s did. My question is: why didn't the LVC version
oscillate? Both were made with SM parts on PC board.

+--/\/\--+-- -||--+
| 36k | 0.01 |
| \ |
| /1.88k |
| |\ \ |\ |
| | \ | | \ |
+ -| >0-+---| >0+-
| / | /
|/ |/
 
T

Tim Williams

Jan 1, 1970
0
You switched the RC.

Analogous to the frequency-domain statement made above, you need positive
feedback over a short period, and negative feedback over a long period
(DC).

As shown, you have two negative feedback loops. At best, you'll get
oscillation due to the propagation delay of the inverter (at a frequency
other than intended).

The canonical circuit is:

+--------+-----||----+
| | C |
\ \ |
/ Ri / R |
\ |\ \ |\ |
| | \ | | \ |
+-| >0--+----| >0--+---o out
| / | /
|/ U1 |/ U2

U1 has DC negative feedback through R, and positive AC feedback through U2
and C. Note Ri prevents overloading the input protection diodes on U1,
important for large C.

Other circuits are common, using different connections of resistors or
more inverters (often 3 total). None of these make any sense when two
works.

If you don't mind depending on a device parameter, you can also make one
with a single schmitt trigger inverter. The time constant corresponds to
the input hysteresis band, which will generally be smaller than the
approx. half supply worth in the above circuit.

Tim
 
R

Rich Grise

Jan 1, 1970
0
I needed a 24 KHz oscillator so I thought I would use the one shown in
Figure 5.30, p285 of A of E (schematic below, view with fixed font). My
first version, which used a SN74LVC2G04, would not oscillate. The second
version using 2 SN74HC1G04s did. My question is: why didn't the LVC version
oscillate? Both were made with SM parts on PC board.

+--/\/\--+-- -||--+
| 36k | 0.01 |
| \ |
| /1.88k |
| |\ \ |\ |
| | \ | | \ |
+ -| >0-+---| >0+-
| / | /
|/ |/



I dunno - last time I needed a clock, I made a pierce oscillator:
http://www.electronics-tutorials.ws/oscillator/crystal.html
(scroll down to "Pierce Crystal Oscillator.")

Cheers!
Rich
 
T

Tim Williams

Jan 1, 1970
0
Bill Bowden said:
Did you ever have problems with the input voltage to the left inverter
going above Vcc or below ground? Some circuits I see use a high value
resistor in series with the inverter input to limit discharge current
through the protection diodes. Maybe it depends on the size of the
capacitor? How much current can those protection diodes take?

A naked gate input would be best, since the protection diodes sink current
which messes up the time constant.

Analysis is simple: assume the switching threshold is in the middle, so
whichever state it's in, when the capacitor charges or discharges to +V/2
(which apparently occurs in t_1/2 = R*C*ln(2), assuming an initial voltage
of +V), it switches. That puts +V*3/2 on the input, so the cap is forced
down by +V/2 due to input diodes.

Input diodes typically handle <10mA, so this isn't a big deal for fairly
small caps (< 100pF?), but should have a resistor of say 1-10k for much
more. The RC time constant of the input pin now contributes to delay, so
you'll get a slightly lower frequency than expected (or maybe nothing if
you try running it near the gate's speed limit).

Tim
 
F

F. Bertolazzi

Jan 1, 1970
0
garyr:
+--/\/\--+-- -||--+
| 36k | 0.01 |
| \ |
| /1.88k |
| |\ \ |\ |
| | \ | | \ |
+ -| >0-+---| >0+-
| / | /
|/ |/

Warning: SPAM! ;-)

http://sourceforge.net/projects/fidocadj/

What about downloading a single Java program (no installation required, if
you don't like it, you simply delete it) that runs on any PC for free?
 
G

garyr

Jan 1, 1970
0
In what respect?

It makes the simple point that an odd number of inverters will always
oscillate, something that the OP might find useful.
 
T

Tim Williams

Jan 1, 1970
0
Bill Bowden said:
So we get an extra 2.5 volts on the input using a 5 volt supply when
the thing switches (7.5). Sounds about right. And if we use large
capacitors (>100pF), a resistor is recommended. I hadn't considered
the RC time delay of adding a resistor. Jim Thompson has an
interesting circuit using 3 inverters and a couple extra resistors to
avoid the diode problem. More parts, better performance.

http://analog-innovations.com/SED/CMOS-Osc-NoClip.pdf

Yes, so instead of the diodes messing up the time constant, or the current
limiting resistor adding propagation delay, you simply drive it with half
the amplitude. Works great. Caveat: Thevenin resistance has to be
considered, hence the big ratio of resistances. This is inefficient for
battery use.

I don't get the third inverter. You can get a fine source of negative
feedback from the first. The feedback is even stable if you use HCU
(necessary for quartz oscillators). Fanout or risetime is fine from HC,
which is buffered, but HCU could stand some sharpening.

I suppose if one were very serious about low power usage, you'd hack your
own inverter out of two discrete FETs, which *don't* have protection
diodes.

Tim
 
J

Jamie

Jan 1, 1970
0
ehsjr said:
Wrong. But, to be expected I suppose, when you rely solely on Wiki for
information.

Ed
Well, I know what a redneck school is, that has drivers ED!

On Monday, Wednesday and friday is drivers ED,
only because on Tuesday and Thursday, the car is needed for the
Sex education classes.

Have a good day..
 
Top