Maker Pro
Maker Pro

Can I use one shared crystal?

S

SBS

Jan 1, 1970
0
Hi,
I have 2 IC (ATmega8 and an ST device) on the same
board where each them needs a crystal of 16MHz.

Can I use one parallel resonant shared crystal for both
devices?

What about the parallel load capacitance of the crystal?

Thanks and regards.
 
J

John Popelish

Jan 1, 1970
0
SBS said:
Do I really need to isolate the two oscillator circuits or can I
connect them as I have written above?

If I would have to use another logic chip, then I may use 2
crystal for an even better solution, don't I?

There are two problems to solve to use a single crystal to
drive two chips. The two inputs may not settle to the same
DC bias voltage. And the extra load of the second chip may
slightly affect the first. You solve the first problem by
coupling the output of the first oscillator to the input of
the second. You will have to find out by trial whether the
chip actually connected as an oscillator is loaded too much
by having its output coupled to the input of the second chip.

A problem free way to drive both chips from a single crystal
would be to build an independent oscillator with the crystal
and a cmos inverter, or use a canned oscillator. Use the
output of either of these to drive both clock inputs. If
the single inverter oscillator, I would capacitively couple
to the inputs to let each chip find its own bias point. If
the canned oscillator with a logic level output, direct
connection will work fine.
 
S

SBS

Jan 1, 1970
0
Donald ([email protected]) ha scritto:

::: Can I use one parallel resonant shared crystal for both
::: devices?

:: Yes, but you can not connect the two crystal pins together
:: on the two processor.

Then I can't connect both input (and so output) pins of the
devices together to one lead of the crystal?

:: Connect the crystal to one chip, and with a TinyLogic device
:: like NC7WZ17 from Fairchild or TI, connect the output of the
:: device with the crystal to the input of the other. ( leave the
:: output side open )
::
:: This will isolate the two oscillator circuits.

Do I really need to isolate the two oscillator circuits or can I
connect them as I have written above?

If I would have to use another logic chip, then I may use 2
crystal for an even better solution, don't I?

Thank you.
 
D

Donald

Jan 1, 1970
0
SBS said:
Hi,
I have 2 IC (ATmega8 and an ST device) on the same
board where each them needs a crystal of 16MHz.

Can I use one parallel resonant shared crystal for both
devices?

What about the parallel load capacitance of the crystal?

Thanks and regards.
Yes, but you can not connect the two crystal pins together on the two
processor.

Check the data sheets of each part. One pin will be an input, the other
an output.

Connect the crystal to one chip, and with a TinyLogic device like
NC7WZ17 from Fairchild or TI, connect the output of the device with the
crystal to the input of the other. ( leave the output side open )

This will isolate the two oscillator circuits.

good luck

donald
 
J

John Popelish

Jan 1, 1970
0
Anthony said:
Why can you connect the output of the canned oscillator directly to each
oscillator input without coupling caps?

Because the signal swing is large enough to saturate each
input, regardless of the exact zero input bias voltage of
either. The bias voltage is important only if the signal
swing is small and must be centered on that bias point to
achieve a full swing output.
 
A

Anthony Fremont

Jan 1, 1970
0
John said:
There are two problems to solve to use a single crystal to
drive two chips. The two inputs may not settle to the same
DC bias voltage. And the extra load of the second chip may
slightly affect the first. You solve the first problem by
coupling the output of the first oscillator to the input of
the second. You will have to find out by trial whether the
chip actually connected as an oscillator is loaded too much
by having its output coupled to the input of the second chip.

A problem free way to drive both chips from a single crystal
would be to build an independent oscillator with the crystal
and a cmos inverter, or use a canned oscillator. Use the
output of either of these to drive both clock inputs. If
the single inverter oscillator, I would capacitively couple
to the inputs to let each chip find its own bias point. If
the canned oscillator with a logic level output, direct
connection will work fine.

Why can you connect the output of the canned oscillator directly to each
oscillator input without coupling caps?
 
D

Donald

Jan 1, 1970
0
John said:
Because the signal swing is large enough to saturate each input,
regardless of the exact zero input bias voltage of either. The bias
voltage is important only if the signal swing is small and must be
centered on that bias point to achieve a full swing output.

What he ( John ) said.

If you don't like that answer given, please try it and let us know what
you've found.

donald
 
D

Donald

Jan 1, 1970
0
John said:
Because the signal swing is large enough to saturate each input,
regardless of the exact zero input bias voltage of either. The bias
voltage is important only if the signal swing is small and must be
centered on that bias point to achieve a full swing output.

Too quick on the SEND button.

This is not a unknown problem.

I am sure many here have tried this before as well.

I tried it with 6502 processors years ago. The osc circuit would work
with each one seperatly but together, nothing, dead, natha.

I had an experienced friend look at my circuit, and he laughed his butt off.

He ask me if I have a clue what I was doing, I said "well, it should
work" and he laughed again.

So, the answer is "don't do it".

But please try it for your self.

donald
 
S

SBS

Jan 1, 1970
0
John Popelish ([email protected]) ha scritto:

:: There are two problems to solve to use a single crystal to
:: drive two chips. The two inputs may not settle to the same
:: DC bias voltage. And the extra load of the second chip may
:: slightly affect the first.


Exaustive explanation and good advices, thank you.
 
S

SBS

Jan 1, 1970
0
Donald ([email protected]) ha scritto:

:: I tried it with 6502 processors years ago. The osc circuit would work
:: with each one seperatly but together, nothing, dead, natha.
::
:: I had an experienced friend look at my circuit, and he laughed his butt
off.
::
:: He ask me if I have a clue what I was doing, I said "well, it should
:: work" and he laughed again.

:) Thank you.
 
C

colin

Jan 1, 1970
0
John Popelish said:
There are two problems to solve to use a single crystal to drive two
chips. The two inputs may not settle to the same DC bias voltage. And
the extra load of the second chip may slightly affect the first. You
solve the first problem by coupling the output of the first oscillator to
the input of the second. You will have to find out by trial whether the
chip actually connected as an oscillator is loaded too much by having its
output coupled to the input of the second chip.

A problem free way to drive both chips from a single crystal would be to
build an independent oscillator with the crystal and a cmos inverter, or
use a canned oscillator. Use the output of either of these to drive both
clock inputs. If the single inverter oscillator, I would capacitively
couple to the inputs to let each chip find its own bias point. If the
canned oscillator with a logic level output, direct connection will work
fine.

I dont see why one cant drive the other,
a usual arangement for one IC wich seem to have a cmos inverter between the
two pins
is to have ~ 20pf on both input and output pins
with the xtal between the two,
If the output of the IC doing the oscilating is conected to the other IC
input via a 100pf decoupling capacitor,
what can it do but load it by 10pf or so,
this can be compensated for by reducing the 20pf capacitor,
ofc the output of the 2nd IC is just left open.

you might find one ic or the other is better at driving the other.
and you might have to play around a bit with capacitances etc.
ofc you would need to make sure the track length to conect them is Very
short.
If its for even low volume production you would have to look into it a bit
deeper.

Colin =^.^=
 
S

SBS

Jan 1, 1970
0
colin ([email protected]) ha scritto:

:: I dont see why one cant drive the other,

Thank you Colin, can you draw a picture which explains
this situation, please?

I can't understand how to connect both devices to the single
crystal... :-(
 
N

Nico Coesel

Jan 1, 1970
0
SBS said:
colin ([email protected]) ha scritto:

:: I dont see why one cant drive the other,

Thank you Colin, can you draw a picture which explains
this situation, please?

I can't understand how to connect both devices to the single
crystal... :-(

As told before: A crystal oscillator has an input and an (inverting)
output. A crystal connected in be input and output will cause the
whole thing to oscillate at the crystal's resonant frequency.

If you want to use one crystal for 2 devices, you should place the
devices close together. Connect the crystal between the crystal
oscillator pins of device A, connect the crystal oscillator output
from device A to the oscillator input from device B with a small
capacitor (say 1nf).

The datasheets should tell you more about such a setup and which pins
are input and output.
 
J

John Popelish

Jan 1, 1970
0
Nico said:
Like this ??
Yes.

+------+
| | |
| | |
__|__ --- 0
\ / [ ] / \
\ / --- / \
0 | -----
| | |
| | |
+------+----()---+
1nf
CPU A CPU B

I don't think this will work.

I think the drive of CPU A output will be loaded too much to drive the
crystal and the input of CPU B.

Let me know how it works.

The OP should give it a try. Most modern CMOS chips have a very small
input capacitance.

The biggest risk may be other signals that capacitively
couple into the trace going to the second chip and
contaminate the wave being fed to the crystal. If this
trace is more than an inch long, I would probably shield it
with a pair of ground traces (grounded at the oscillator
end, not the receiver end).
 
D

Donald

Jan 1, 1970
0
Nico said:
As told before: A crystal oscillator has an input and an (inverting)
output. A crystal connected in be input and output will cause the
whole thing to oscillate at the crystal's resonant frequency.

If you want to use one crystal for 2 devices, you should place the
devices close together. Connect the crystal between the crystal
oscillator pins of device A, connect the crystal oscillator output
from device A to the oscillator input from device B with a small
capacitor (say 1nf).

The datasheets should tell you more about such a setup and which pins
are input and output.
Like this ??


+------+
| | |
| | |
__|__ --- 0
\ / [ ] / \
\ / --- / \
0 | -----
| | |
| | |
+------+----()---+
1nf
CPU A CPU B

I don't think this will work.

I think the drive of CPU A output will be loaded too much to drive the
crystal and the input of CPU B.

Let me know how it works.

donald
 
N

Nico Coesel

Jan 1, 1970
0
Donald said:
Nico said:
As told before: A crystal oscillator has an input and an (inverting)
output. A crystal connected in be input and output will cause the
whole thing to oscillate at the crystal's resonant frequency.

If you want to use one crystal for 2 devices, you should place the
devices close together. Connect the crystal between the crystal
oscillator pins of device A, connect the crystal oscillator output
from device A to the oscillator input from device B with a small
capacitor (say 1nf).

The datasheets should tell you more about such a setup and which pins
are input and output.
Like this ??
Yes.


+------+
| | |
| | |
__|__ --- 0
\ / [ ] / \
\ / --- / \
0 | -----
| | |
| | |
+------+----()---+
1nf
CPU A CPU B

I don't think this will work.

I think the drive of CPU A output will be loaded too much to drive the
crystal and the input of CPU B.

Let me know how it works.

The OP should give it a try. Most modern CMOS chips have a very small
input capacitance.
 
W

whit3rd

Jan 1, 1970
0
Donald ([email protected]) ha scritto:

::: Can I use one parallel resonant shared crystal for both
::: devices?

:: Yes, but you can not connect the two crystal pins together
:: on the two processor.

Then I can't connect both input (and so output) pins of the
devices together to one lead of the crystal?

:: Connect the crystal to one chip, and with a TinyLogic device
:: like NC7WZ17 from Fairchild or TI

I think there's still a miscommunication here; the chip can accept
an external oscillator, OR oscillate a connected crystal, and
what is recommended is that ONE CPU oscillates a crystal
while the OTHER uses an external oscillator.

So, one poster thinks both CPUs are in oscillate-a-crystal
connection (and wants to connect two wires on each of the
two), while the respondent only connects two wires to
the crystal-oscilator chip, and one wire to the uses-external-
oscillator
chip.

The TinyLogic device is just used as a buffer, so the (possibly
sensitive) crystal wiring isn't connected to an antenna of
interconnect
wire. Any available gate would serve.
 
S

SBS

Jan 1, 1970
0
whit3rd ([email protected]) ha scritto:

:: The TinyLogic device is just used as a buffer, so the
:: (possibly sensitive) crystal wiring isn't connected to
:: an antenna of interconnect wire. Any available gate
:: would serve.

BTW, I need to use a crystal with nominal frequency
of 16 MHz. So the best soultion should be to use a crystal
for each device, shouldn't it?
 
A

Anthony Fremont

Jan 1, 1970
0
SBS said:
whit3rd ([email protected]) ha scritto:


BTW, I need to use a crystal with nominal frequency
of 16 MHz. So the best soultion should be to use a crystal
for each device, shouldn't it?

Possibly the easiest and least expensive anyway. Unless you have some
special need for them to run _exactly_ at the same frequency and phase.
 
R

Rich Grise

Jan 1, 1970
0
Nico said:
Like this ??
Yes.
+------+
| | |
| | |
__|__ --- 0
\ / [ ] / \
\ / --- / \
0 | -----
| | |
| | |
+------+----()---+
1nf
CPU A CPU B

I don't think this will work.

I think the drive of CPU A output will be loaded too much to drive the
crystal and the input of CPU B.

Let me know how it works.

The OP should give it a try. Most modern CMOS chips have a very small
input capacitance.

The biggest risk may be other signals that capacitively
couple into the trace going to the second chip and
contaminate the wave being fed to the crystal. If this
trace is more than an inch long, I would probably shield it
with a pair of ground traces (grounded at the oscillator
end, not the receiver end).

My "little voice" is also telling me that I'd use much less than
1 nf for the cap - maybe more like 33 ~ 120 pf.

Thanks,
Rich
 
Top