PECL and Transmission line length

[bart]

Hello all, I am looking to transmit a 66MHz clock signal over a 50ft
length of 50ohm coax. I am considering using a Texas Instruments
TB5D1, which is capable of driving 50 ohm loads, but I am unsure as to
what the max switching frequency vs length of cable is for PECL. I
would appreciate any comments or suggestions as well. Thanks in
advance.

 

[Jim Thompson]

Hello all, I am looking to transmit a 66MHz clock signal over a 50ft
length of 50ohm coax. I am considering using a Texas Instruments
TB5D1, which is capable of driving 50 ohm loads, but I am unsure as to
what the max switching frequency vs length of cable is for PECL. I
would appreciate any comments or suggestions as well. Thanks in
advance.

Just use source termination.

...Jim Thompson

 

[Joerg]

Hello Jim,

Just use source termination.

That's one option but would lead to one reflection at the far end. More
EMI. What about termination at the end?

If there is a remote chance that the end of the coax might occasionally
be left dangling then source termination is the better option, of course.

Regards, Joerg

 

[John Larkin]

Hello all, I am looking to transmit a 66MHz clock signal over a 50ft
length of 50ohm coax. I am considering using a Texas Instruments
TB5D1, which is capable of driving 50 ohm loads, but I am unsure as to
what the max switching frequency vs length of cable is for PECL. I
would appreciate any comments or suggestions as well. Thanks in
advance.

The TB5D1 is a differential driver. Do you plan to use two pieces of
coax, or a twinax or something?

What receiver chip did you have in mind? Do you need to just send a
clock, or is there an associated data stream?

John

 

[[email protected]]

66 MHz isn't a particularly high frequency, so you probably don't have
to worry about any form of signal attenuation other than that due to
the resistance of the inner and outer conductors. Some 50R cable can be
quite resistive, so it might be worth checking out the specification of
the cable you have in mind.

50 feet of cable is roughly 75nsec of propagation delay, or six
wavelengths of your 66MHz clock, so it would be a very good idea to
terminate the cable at one end or the other. Terminating at both ends
does absorb any residual reflections twice as fast, but for a simple
66MHz clock this isn't really necessary - all you need to do is make
sure that any reflections aren't big enough to cancel out your clock
signal at the receiving end.

PECL-style outputs are emitter-followers, so they need pull-down
resistors to sink enough current to keep the output transistors turned
on and linear over the full signal voltage excursion, which is easier
to arrange with the terminating resistors at the receiving end.

Hope this helps.

 

[Fred Bloggs]

PECL-style outputs are emitter-followers, so they need pull-down
resistors to sink enough current to keep the output transistors turned
on and linear over the full signal voltage excursion, which is easier
to arrange with the terminating resistors at the receiving end.

PECL style GND and cable shield are -5.2V down -making that terminating
resistor quite a load.

 

[Joerg]

Hello Fred,

PECL style GND and cable shield are -5.2V down -making that terminating
resistor quite a load.

Probably has to be some kind of AC coupled deal. Else something might
rub through the outer jacket, shield touches chassis and... "Bzzzt".

Regards, Joerg

 

[[email protected]]

PECL is ECL run with the negative rail at ground and the positive rail
at +5V or +.3.3V (if you've got the right parts).

ECL terminating/pull-down resistors are normally returned to an
auxiliary rail set 2V lower than the the positive rail. If you don't
want to add a third power supply, you can put a second resistor in
series with the 50R terminating resistor to add enough DC resistance to
get the right current in the output transistor (around 5mA IIRR) and
by-pass it to ground with a capacitor to restore the terminating
impedance.

It is all spelled out in the Motorola ECL application notes.

 

[Fred Bloggs]

PECL is ECL run with the negative rail at ground and the positive rail
at +5V or +.3.3V (if you've got the right parts).

ECL terminating/pull-down resistors are normally returned to an
auxiliary rail set 2V lower than the the positive rail. If you don't
want to add a third power supply, you can put a second resistor in
series with the 50R terminating resistor to add enough DC resistance to
get the right current in the output transistor (around 5mA IIRR) and
by-pass it to ground with a capacitor to restore the terminating
impedance.

It is all spelled out in the Motorola ECL application notes.

It's also spelled out in the TI datasheet for this part, which , if
believed, indicates that a single ended drive from this part may not be
compatible with a DC coupled single ended PECL receiver, whereas a
differential output will be. Looks like it was designed with UTP in
mind, with the ultra-low skew "and all." As usual, the original
standalone question is pointless- what is the environment, what is the
receiver and coupling, CW or not etc...

 

[keith]

Hello Jim,


That's one option but would lead to one reflection at the far end. More
EMI. What about termination at the end?

You count on the reflection to make the signal whole. The termination is
at the source. It's a very effective low-power termination but is only a
good choice in point-to-point nets because the intermediate points on the
line see V/2. I don't believe EMI is any worse.

If there is a remote chance that the end of the coax might occasionally
be left dangling then source termination is the better option, of
course.

It's a better choice for most point-to-point nets.