How close together do a controller and crystal need to be?

[Erick Anderson]

I designed a board for the 6-digit All Spectrum controller, which uses the Dallas TCXO chip. That's what goes in the DIP-14 socket in the picture. Right now they're as close to each other as possible. I'm thinking about redesigning the board to be a bit shorter, and moving the socket into the empty space at the right of the board would help. This would make the clock signal trace much longer, but is that actually a problem?

[Bill van Dijk]

As long as there is not something very noisy on the other side of the board you’ll be just fine.

 

Bill

--
You received this message because you are subscribed to the Google Groups "neonixie-l" group.
To unsubscribe from this group and stop receiving emails from it, send an email to neonixie-l+...@googlegroups.com.
To view this discussion on the web, visit https://groups.google.com/d/msgid/neonixie-l/1bbc978e-4883-4b0e-920f-d5f3ff4c4c2cn%40googlegroups.com.

[gregebert]

I'm assuming you are routing the output signal of an oscillator, not the crystal signals themselves.

The rise- & fall-times of the clock signal will determine how long the trace can be without termination. Faster edge-rates, say in the 2-3nsec range, will limit your trace to around 1 inch.

Signals propagate around 150psec/inch, and if the rise/fall times are about 10x (or larger) longer than the flight-time, then reflections should not have sufficient amplitude to cause false clocking.

In the example above, 1 inch of trace has a round-trip flight-time of 300psec. If the rise and fall delays are 3nsec or larger, you can safely use 1 inch of trace without using termination networks or controlled-impedance traces.

SPICE simulations are very helpful when deciding how to design clock lines when you cant satisfy the above rule.

[chuckrr]

A handy way I use, to remember the approximate speed of light, which is also the approximate

speed at which an electrical signal travels in a wire is just to think of it

as 1 nanosecond per foot.   Approximately.
 

To view this discussion on the web, visit https://groups.google.com/d/msgid/neonixie-l/bbe8ae0a-bb3c-4f4a-98f6-9f56bc1e1805n%40googlegroups.com.

[Paul Andrews]

Hmm. My recollection from high school physics was that the speed of propagation along a wave guide was around 90% the speed of light - presumably limited by the dielectric. Signals along a plain old wire, on the other hand, were more like 1/3 the speed of light. Now I'm going to have to double-check that for the first time in over 40 years!

[Adrian Godwin]

https://en.wikipedia.org/wiki/Velocity_factor

To view this discussion on the web, visit https://groups.google.com/d/msgid/neonixie-l/480dcaf7-ce2b-45be-b8cc-bb1ee79bc5a2n%40googlegroups.com.

[chuckrr]

Well even thoughe electricity flowing through a wire is somewhat slower than the speed of light, exactly how much slower

is something I do not know about.   What I do know is that the approximation of it being "around" the speed of light has served me

very well over the years.   One time on an an installation job at a telephone switching office where all previoisly installed

processor complexes had been installed using 50 foot coax cables to get the 12.352 mHzmaster clock to them all, this install

order called for the use of a 100 foot cable.   The result was that the newly installed processor complex was unable to synch

up with the rest of the switch.   Nobody knew why.   I calculated that at the speed of light, the 12.352 mHz clock signal would be

late at the new processor by around half of a cycle.     We talked this over with network operations and they confirmed my calcualtions

and then instructed the crew to change out ALL the clock cables in the entire exchange to 100 footers to match the new one.

There was lost of slack to stow, but it all came up just fine after that.  The install crew wanted to know how I managed to

know what the problem was.  I said, "simple gentlemen, it's the speed of light and our new intsall was getting the message rather late"
 

To view this discussion on the web, visit https://groups.google.com/d/msgid/neonixie-l/CALiMYruP%2BG1MwK5ivp-_SXgFPXO6AfnFPRAXNYpc68uFohPHmg%40mail.gmail.com.

[Charles MacDonald]

On 2020-12-27 6:27 p.m., chuckrr wrote:
> Well even thoughe electricity flowing through a wire is somewhat slower
> than the speed of light, exactly how much slower
>

> is something I do not know about.  this install

>
> order called for the use of a 100 foot cable.   The result was that the
> newly installed processor complex was unable to synch
>
> up with the rest of the switch.   Nobody knew why.   I calculated that
> at the speed of light, the 12.352 mHz clock signal would be
>
> late at the new processor by around half of a cycle.

it is legendary that Rear Admiral Grace Hopper, (the Computer genius who
is credited with the early computer language COBAL) used to had out
little pieces of wire, cut to the length that would take one Nanosecond
to travel. "This Gentlemen is a Nanosecond".

she was of course attempting to get computer designers to make the
interconnects shorter, as that would allow faster computers. as well as
explaining lag in satellite communications.

https://en.wikipedia.org/wiki/Grace_Hopper#Anecdotes

--
Charles MacDonald VA3CPY Stittsville Ontario
cm...@zeusprune.ca Just Beyond the Fringe
No Microsoft Products were used in sending this e-mail.

[David Forbes]

Don't worry about it. 

This is a CMOS 5V signal, I presume. It only has one load, at the far end of the trace. You do not say what the frequency is, so I will guess that it's on the order of 5-20 MHz. 

The signal integrity is important, to prevent ringing that would double-trigger the load circuit. A non-terminated transmission line will have a reflection, whose size is a function of the sharpness of the rising or falling edge, and the length of the trace.

The output transition time of the oscillator (the rise and fall time) is likely greater than 5 nanoseconds. As long as the trace length is a lot shorter than the transition time, you don't have to worry about reflection from the end of the line causing ringing. 

FR4 has a propagation delay of 2ns/foot, based on its dielectric coefficient of ~4.5. So a trace up to one foot long will 'swallow' the edge rather than causing ringing. 

I apologize for the slightly hand-waving analysis, but this isn't a critical case, so a detailed analysis isn't needed.

I recently designed a clock driver circuit board for a 4 GHz spectrometer, if you're wondering what I do in my day job that makes me think about clock traces.

On Sat, Dec 26, 2020, 4:53 PM Erick Anderson <hel...@gmail.com> wrote:

I designed a board for the 6-digit All Spectrum controller, which uses the Dallas TCXO chip. That's what goes in the DIP-14 socket in the picture. Right now they're as close to each other as possible. I'm thinking about redesigning the board to be a bit shorter, and moving the socket into the empty space at the right of the board would help. This would make the clock signal trace much longer, but is that actually a problem?

--