Accuracy of CD4060 R/C-Oscillator?

Hello,
I've designed a simple R/C-Oscillator with a MC14060 (ON-Semi) C-MOS Logic IC. It's easy to calculate withe the formula f=1/(2.3*R*C)
But I can't find information on the tolerances for this frequency in the datasheet. I know it will be quite a lot. I guess 30 to 40%? But to really know if I can handle this, I need to know what exactly the manufacturer guarantees.
Do other manufacturers (TI, National, Intersil) have more information?
Thanks in advance,
Bernd

 

It's basically about the R/C exponential decay hitting the C/MOS gate
thresholds, which are typically guaranteed to be somewhere in the
interval between one third and two thirds of the difference between
the supply rails.

Schottky gates have more tightly specified thresholds, and a
comparator can be set up to be much more accurate again (to the point
where noise on the supply rail can become the decisive source of
uncertainty).

It's easy enough to do a worst case design. ln 0.333 is 1.099, ln
0.666 is 0.4065 while ln 0.5 is 0.693.

1.099 + 0.4065 = 1.5064

2 x 0.693 = 1.39

which is a 10% range. Claim +/-5% and you'll probably be in the right
ball park. The distribution of frequencies won't be Guassian -
there'll be more fast oscillators than slow ones.

For extra credit, throw in the difference between the output impedance
of a CMOS gate connected to the positive rail - the P-channel output
transistor - and the output impedance of the same gate connected to
the negative rail (the N-channel output transistor).

 

Tolerances: Start with the Rs and Cs, and apply those to the circuit
equation.
The circuit itself can do good or bad: a phase retard circuit feeding
the equivalent of a comparator adds in the uncertainty or tolerance of
the comparison level.
This is for starters.

 

This is true as far as it goes. For a more detailed discussion see

http://www.fairchildsemi.com/an/AN/AN-118.pdf

which brings out one of the nastier features of the oscillator circuit
recommended in the MC14060 data-sheet - the gate input goes outside
the rail, and if you rely on the MC14060's catching diodes to clamp
the voltage to the rails, you are injecting current into the ic's
substrate, which can be a bad idea.

If I get around to it. I'll model the recommended MC14060 oscillator
in LTSpice and post the schematic.

 

Here are the LTSpice models with a 50% threshold - best case - and a
67% threshold - worst case.

Version 4
SHEET 1 880 680
WIRE -32 160 -192 160
WIRE 64 160 -32 160
WIRE 176 160 128 160
WIRE 272 160 176 160
WIRE 384 160 336 160
WIRE -192 208 -192 160
WIRE -32 208 -32 160
WIRE 176 208 176 160
WIRE 384 208 384 160
WIRE -32 320 -32 288
WIRE 176 320 176 288
WIRE 176 320 -32 320
WIRE 384 320 384 272
WIRE 384 320 176 320
WIRE -192 432 -192 272
FLAG -192 432 0
SYMBOL cap 368 208 R0
SYMATTR InstName C1
SYMATTR Value 10n
SYMBOL res 160 192 R0
SYMATTR InstName R1
SYMATTR Value 43k
SYMBOL res -48 192 R0
SYMATTR InstName R2
SYMATTR Value 430k
SYMBOL Digital\\inv 272 96 R0
WINDOW 3 52 30 Left 2
SYMATTR Value Ref=2.5
SYMATTR InstName A1
SYMATTR SpiceLine2 Vhigh=5 Trise=40n Tfall=40n Rout=1k
SYMBOL cap -208 208 R0
SYMATTR InstName C2
SYMATTR Value 10p
SYMBOL Digital\\inv 64 96 R0
WINDOW 3 52 30 Left 2
SYMATTR Value Ref=2.5
SYMATTR InstName A2
SYMATTR SpiceLine2 Vhigh=5 Trise=40n Tfall=40n Rout=1k
TEXT -184 96 Left 2 !.tran 0 20m 0 10n uic
TEXT -192 64 Left 2 !.ic V(n001)=0

This oscillates at 1050Hz

Version 4
SHEET 1 880 680
WIRE -32 160 -192 160
WIRE 64 160 -32 160
WIRE 176 160 128 160
WIRE 272 160 176 160
WIRE 384 160 336 160
WIRE -192 208 -192 160
WIRE -32 208 -32 160
WIRE 176 208 176 160
WIRE 384 208 384 160
WIRE -32 320 -32 288
WIRE 176 320 176 288
WIRE 176 320 -32 320
WIRE 384 320 384 272
WIRE 384 320 176 320
WIRE -192 432 -192 272
FLAG -192 432 0
SYMBOL cap 368 208 R0
SYMATTR InstName C1
SYMATTR Value 10n
SYMBOL res 160 192 R0
SYMATTR InstName R1
SYMATTR Value 43k
SYMBOL res -48 192 R0
SYMATTR InstName R2
SYMATTR Value 430k
SYMBOL Digital\\inv 272 96 R0
WINDOW 3 52 30 Left 2
SYMATTR InstName A1
SYMATTR SpiceLine2 Vhigh=5 Trise=40n Tfall=40n Rout=1k
SYMATTR Value Ref=3.33
SYMBOL cap -208 208 R0
SYMATTR InstName C2
SYMATTR Value 10p
SYMBOL Digital\\inv 64 96 R0
WINDOW 3 52 30 Left 2
SYMATTR InstName A2
SYMATTR SpiceLine2 Vhigh=5 Trise=40n Tfall=40n Rout=1k
SYMATTR Value Ref=3.33
TEXT -186 98 Left 2 !.tran 0 20m 0 10n uic
TEXT -192 64 Left 2 !.ic V(n001)=0

which oscillates at 997Hz.

Less difference than my first sketch of a worst case analysis
suggested.

 

Interesting - if implausible - story. As far as I know, charge
carriers injected into the substrate via the protection diodes don't
do anything different from charge carriers that get there from
anywhere else.

Charge carriers that get into the gate oxide are another - but here
totally irrelevant - story.

If the MC14060 got messed up by current going through the protection
diodes, we'd have heard about it by now.

<snip>

 

Gnaw, that's "hot carrier injection", which has something to do with
excessive voltage drop through a channel, with a nearby thin-oxide gate.
Nothing to do with substrate diodes. High energy electrons jump the
oxide, causing gate leakage. If the gate happens to be a floating hunk of
metal (as is used in floating-gate EPROM structures), the gate keeps a
permanent charge.

Tim

 

Regarding injected charge, be advised that each time, some gets
trapped and the the accumulated charge acts as an input voltage and
*permanently* latches the gate; input voltage then is "ignored".
The fix? Anneal the damage in an oven; perhaps 135C will do the job.

 

As opposed to what's shown in the data sheet, and gives the right
answer when modelled with LTSpice.

Really? The Motorola data sheet shows just one input with hysteresis -
"Out 2" on pin 9 - which take the clock output from the oscillator and
cleans it up before feeding it into the counting flip-flops. It's
thresholds don't come into the calculation of the oscillator
frequency.

As you'd expect in a part where the recommended oscillator circuit
drove one of the inputs outside the voltage rails.

It's been done and posted - Jim has got me killed-filed so he doesn't
realise that he's made an ass of himself again.
The LTSpice model gives pretty much the right frequency - a bit too
fast with an exactly 50% threshold, and very slightly too slow with a
67% threshold.

<snip>