1 chip 1KHz xtal osc?

[Mark Roberts]

I need a lowest possible parts count 1KHz square wave oscillator.

Must use crystal timing with accuracy as good as reasonably attainable
without use of oven..

To do this, I assume it is best to start with the highest frequency
crystal possible, as each divide down increases resolution.

Power requirements are open.

Can anyone please advise regarding the most suitable circuit design?

With appreciation,

Mark

 

[Tim Wescott]

I need a lowest possible parts count 1KHz square wave oscillator.

Must use crystal timing with accuracy as good as reasonably attainable
without use of oven..

To do this, I assume it is best to start with the highest frequency
crystal possible, as each divide down increases resolution.

Power requirements are open.

Can anyone please advise regarding the most suitable circuit design?

With appreciation,

Mark

I'd use a 74xx4040, I think it'll do 2^14 or 2^15, so use a 32768kHz, or
16384kHz depending.

Realize that "lowest parts count" and "as good as attainable" are
mutually exclusive -- a '4040 is _not_ a precision oscillator, so you'll
get some variations.

 

[David L. Jones]

I need a lowest possible parts count 1KHz square wave oscillator.

Must use crystal timing with accuracy as good as reasonably attainable
without use of oven..

To do this, I assume it is best to start with the highest frequency
crystal possible, as each divide down increases resolution.

Power requirements are open.

Can anyone please advise regarding the most suitable circuit design?

With appreciation,

Mark

The Epson SPG6851 is a true single chip solution with NO external
components required (it's a programmbale oscillator). BUT it is
offically discontinued according to Epson, so is not recommended for
new designs. But you can still buy it at Farnell (300-3097) and other
places.
Tempco ain't that great though.

There may be other programmable oscillators around, but I don't know
any off-hand that will do 1KHz.

Regards
Dave

 

[John Jardine]

I need a lowest possible parts count 1KHz square wave oscillator.

Must use crystal timing with accuracy as good as reasonably attainable
without use of oven..

To do this, I assume it is best to start with the highest frequency
crystal possible, as each divide down increases resolution.

Power requirements are open.

Can anyone please advise regarding the most suitable circuit design?

With appreciation,

Mark

A 74HC4060 (osc/divider) running a cheap 4.194304MHz Xtal using 2x22pf caps
and a 10Mohm resistor.
Take the 1kHz from pin 1

 

[John Fields]

A 74HC4060 (osc/divider) running a cheap 4.194304MHz Xtal using 2x22pf caps
and a 10Mohm resistor.
Take the 1kHz from pin 1

 

[John Fields]

I'd use a 74xx4040, I think it'll do 2^14 or 2^15, so use a 32768kHz, or
16384kHz depending.

Realize that "lowest parts count" and "as good as attainable" are
mutually exclusive -- a '4040 is _not_ a precision oscillator, so you'll
get some variations.

---
1. A 4040 is neither an oscillator nor a fractional divider

2. 2^14/1000 = 16.384, not something the 4040 can divide by

3. 2^15/1000 = 32.768, not something the 4040 can divide by

4. "As good as attainable with the lowest parts count" is achievable.

5. Wescott _Design_ Services???? Hmmm...

 

[Mike Harrison]

Not necesssarily, as long as the xtal is an integer multiple of the frequency.

I'd do it using a PIC12C508 or ATTiny11 - small, cheap, flexible. Parts are chip, xtal + 3 caps (one
decoupling).

This also has scope for software trimming, even temp compensation if you want to get really
adventurous!

 

[John Fields]

I need a lowest possible parts count 1KHz square wave oscillator.

Must use crystal timing with accuracy as good as reasonably attainable
without use of oven..

To do this, I assume it is best to start with the highest frequency
crystal possible, as each divide down increases resolution.

---
If your crystal is specified to perform to some level of accuracy
under some set of conditions, then regardless of the crystal frequency
or the number of divisions you do, the accuracy of the derived
frequency will be the same as that of the crystal, neglecting
artifacts introduced by the dividers.

 

[Tim Wescott]

A 4040 is only a divider; the 4040, 4020 and 4060 run together in my
mind -- I was thinking of the 4060 and should have mentioned that I
wasn't checking data sheets. Actually one of the 74xx4060's is
required; a 4060 is only good to 4MHz or so -- and no, I'm not checking
the data sheet for _this_ free advise either.

I'm happy for you that you're supernaturally perfect. When will you be
ascending to heaven?

2. 2^14/1000 = 16.384, not something the 4040 can divide by

Very true, but if you had troubled to look at my units you'd see that
16384kHz = 2^14 * 1kHz, which does divide nicely to 1kHz if you have a
14-stage counter like the 74xx4060.

3. 2^15/1000 = 32.768, not something the 4040 can divide by

Nor the 74xx4060. "I think it'll do" would certainly send _me_
scrambling for a data sheet.

4. "As good as attainable with the lowest parts count" is achievable.

"As good as attainable with the lowest parts count" is a null statement.
Zero is the lowest parts count, and nothing is what you achieve with
it. And that's as good as you're going to get with that parts count,
but probably not what he really needs. So you need to dig a little deeper.

5. Wescott _Design_ Services???? Hmmm...

That's right. And for paying customers I double or triple check before
I release my designs.

 

[John Fields]

Not necesssarily, as long as the xtal is an integer multiple of the frequency.

 

[Robert Lacoste]

I need a lowest possible parts count 1KHz square wave oscillator.
Must use crystal timing with accuracy as good as reasonably attainable

Why not a PIC10F clocked with a 32KHz crystal and 10 lines of assembly code
? A sub-1$ solution...

 

[John Fields]

John Fields wrote: A 4040 is only a divider; the 4040, 4020 and 4060 run together in my
mind -- I was thinking of the 4060 and should have mentioned that I
wasn't checking data sheets. Actually one of the 74xx4060's is
required; a 4060 is only good to 4MHz or so -- and no, I'm not checking
the data sheet for _this_ free advise either.

I'm happy for you that you're supernaturally perfect. When will you be
ascending to heaven?

---
Neglecting, for the moment, to chide you for not being able to take
correction, I will reply that it's not so much "when" as it is "if",
since I consider presumptuousness to often result in nasty surprises.
---

Very true, but if you had troubled to look at my units you'd see that
16384kHz = 2^14 * 1kHz, which does divide nicely to 1kHz if you have a
14-stage counter like the 74xx4060.

---
But, 16384kHz is 16.384MHz, and since you've already stated that a
4060 is only good for "4MHz or so" (while mentioning nothing about
74XX stuff in your original post) it now sounds like you're trying to
wriggle off the hook...
---

"As good as attainable with the lowest parts count" is a null statement.
Zero is the lowest parts count, and nothing is what you achieve with
it. And that's as good as you're going to get with that parts count,
but probably not what he really needs. So you need to dig a little deeper.

---
I think a lawyer or a doctor making that statement about pro bono work
would, at the very least, be severely reprimanded for that attitude
and the shoddy work it leads to.

You're obviously neither, but consider yourself seriously reprimanded
for advocating _pro mal_ work.

 

[Tim Wescott]

---
Neglecting, for the moment, to chide you for not being able to take
correction, I will reply that it's not so much "when" as it is "if",
since I consider presumptuousness to often result in nasty surprises.
---




---
But, 16384kHz is 16.384MHz, and since you've already stated that a
4060 is only good for "4MHz or so" (while mentioning nothing about
74XX stuff in your original post) it now sounds like you're trying to
wriggle off the hook...

From my first reply: "I'd use a 74xx4040".Hardly. I see far more systems get to production needing to be
completely re engineered over such null statements than I do because
someone got a chip number wrong. The latter problem shows up _really
early_ in the design process (when someone can't find oscillator pins
and asks, for instance). The former problem can make it all the way out
to the field before the problem is known ("hey, this thing looses a
minute a week!" "you said minimum parts count" "the customer needs a
minute a month!" etc.). Changing a part number before it even hits a
schematic is _way_ easier than a recall, or spending the next 10 years
struggling with unhappy customers.

The reason is because such heavily qualified statements come from you
assuming that your assumptions are the same as mine. If your
assumptions really are the same as mine then all will be well. In all
probability the OP just wanted something reasonably close, and the
oscillator in a '4060 would be more than good enough -- but one should
always check one's assumptions (as well as data sheets ).

---
I think a lawyer or a doctor making that statement about pro bono work
would, at the very least, be severely reprimanded for that attitude
and the shoddy work it leads to.

You're obviously neither, but consider yourself seriously reprimanded
for advocating _pro mal_ work.

And yes, I _am_ kicking myself for saying '4040 and not '4060,
particularly since I couldn't remember if it was a 14 or 15-stage and
that should have made me look (or at least warn the OP that he was
getting an off-the-cuff answer).

If I were to make the error of trashing someone for a fairly minor
mistake I'd be kicking myself much worse, of course.

 

[Max Hauser]

"Mark Roberts" in news:[email protected]...

I need a lowest possible parts count 1KHz square wave oscillator.

FYI, Linear Technology Corp. has a whole family of recent, handy precision
oscillator products with near-crystal accuracy. They can replace crystals
(AND their oscillators) in some spplications. Tiny packages, squarewave
ouptuts, freq is set by a combination of one external resistor and an
internal freq divider programmed by voltages you put on pins.

The LTC6900 covers 1kHz-20MHz and its data sheet is at

http://www.linear-tech.com/prod/datasheet.html?datasheet=857

(There are other products in the family. The original was LTC1799.)

 

[Reg Edwards]

Reliability is Quality versus Time.

ALL components MUST serve a recognisable purpose.

Any redundancy MUST be deliberately introduced and NOT have found its way
into the circuit by chance.

If a designer is unable to describe the purpose of each and every component
then he himself should be made redundant.

 

[Jim Thompson]

"Mark Roberts" in
FYI, Linear Technology Corp. has a whole family of recent, handy precision
oscillator products with near-crystal accuracy. They can replace crystals
(AND their oscillators) in some spplications. Tiny packages, squarewave
ouptuts, freq is set by a combination of one external resistor and an
internal freq divider programmed by voltages you put on pins.

The LTC6900 covers 1kHz-20MHz and its data sheet is at

http://www.linear-tech.com/prod/datasheet.html?datasheet=857

(There are other products in the family. The original was LTC1799.)

Amusing data sheet with "typos" induced by the conversion to PDF not
maintaining characters, e.g. "W" instead of capital omega.

To my mind +/- 1.5% is not very accurate.

...Jim Thompson

 

[Spehro Pefhany]

Why not a PIC10F clocked with a 32KHz crystal and 10 lines of assembly code
? A sub-1$ solution...

The PIC10F only has a crummy (by crystal standards) 1-2%-ish internal
RC oscillator for the clock. Plus, it's not available yet. And an
external BOR circuit is required in many applications. Other than
those details, it might be ideal.

Best regards,
Spehro Pefhany

 

[Tim Wescott]

Amusing data sheet with "typos" induced by the conversion to PDF not
maintaining characters, e.g. "W" instead of capital omega.

To my mind +/- 1.5% is not very accurate.

...Jim Thompson

Putting my marketing hat on for the moment (and not reading data sheets
_again_):

100ppm should be quite achievable with a crystal oscillator. So if you
have a 10MHz oscillator that's a 1kHz error, right?

Now if you have a 1kHz oscillator with 1.5% accuracy, that's only 15Hz,
right?

And 15Hz is _much_ smaller than 1kHz, right?

So they are getting better than crystal accuracy!

 

[John Woodgate]

I read in sci.electronics.design that Reg Edwards

Reliability is Quality versus Time.

ALL components MUST serve a recognisable purpose.

Any redundancy MUST be deliberately introduced and NOT have found its way
into the circuit by chance.

If a designer is unable to describe the purpose of each and every component
then he himself should be made redundant.

What about components that should be there, but aren't?

 

[Allan Herriman]

Rational number multiples of the frequency work (and are trivial to
implement) as long as some jitter can be tolerated. The output jitter
only needs to be at most one period of the input clock, but sometimes
poor implementations generate more.

This means that basically *any* off the shelf crystal frequency can be
used.

I guess it's time to plug my free fractional-N divider generator
program at http://fractional-divider.tripod.com/
You'll need an FPGA or CPLD though.

E.g. NTSC 4x colourburst xtal (14.31818MHz) requires a division of
715909/50 and this can be done with 21 flip flops (the program used
23) inside a tiny FPGA.

Assuming that there is already an XTAL osc and FPGA on the board, the
additional parts count is zero

Regards,
Allan.