How do I use a 10Base T filter?

[Largo]

I should have a 10Mhz square wave generated by CMOS logic. My 50Mhz
scope shows overshoot, dampening or ringing on the peaks of the wave
which should be ~5 Volt p-p square. How can I minimize this
phenomenon? I looked at a 10Base T filter to "clean" this up. I use
RG-174 coax to port this signal to my application but not shure the
probe/scope or the coax is doing this. The application works ok
otherwise.

How do I get nice square wave?

Largo

 

[mc]

What kinds of circuitry do you have on both ends?

The scope *may* not be telling the truth. I presume you're using a 10:1
probe which has been adjusted on the scope's calibration signal...right?
Even so, a 50-MHz scope may not be an accurate judge of the squareness of a
10-MHz signal. Still, if it's mainly the scope's limitations, you'd expect
it to look like it's been rounded, not ringing.

 

[Tam/WB2TT]

I should have a 10Mhz square wave generated by CMOS logic. My 50Mhz
scope shows overshoot, dampening or ringing on the peaks of the wave
which should be ~5 Volt p-p square. How can I minimize this
phenomenon? I looked at a 10Base T filter to "clean" this up. I use
RG-174 coax to port this signal to my application but not shure the
probe/scope or the coax is doing this. The application works ok
otherwise.

How do I get nice square wave?

Largo

You don't mention what kind of CMOS logic. In general, it is incapable of
driving coax. The filter will do no good without a load resistor on the
output. To start out, remove the coax, and look at the CMOS chip with the
'scope. Make sure the probe is grounded somewhere within a couple of inches
of the chip.

Instead of the 10Base T, you might want to try a 33 Ohm series resistor with
something like 10 PF from the output side to ground. Depending on how much
drive your CMOS has, putting a 47 Ohm resistor in series with the coax input
might help if the coax is less than about a foot long.

Keep in mind, that in order to drive a 5V signal into 50 Ohm coax, and do it
right, your CMOS device would have to drive 5/50 or 0.1 Amp. You would also
need a 50 Ohm resistor across the output.

While you are at it, look at the VCC of the CMOS. If the chip does not have
a bypass capacitor on VCC, add a .1 - .33 before bothering to look.

Tam

 

[Largo]

You don't mention what kind of CMOS logic. In general, it is incapable of
driving coax. The filter will do no good without a load resistor on the
output. To start out, remove the coax, and look at the CMOS chip with the
'scope. Make sure the probe is grounded somewhere within a couple of inches
of the chip.

Instead of the 10Base T, you might want to try a 33 Ohm series resistor with
something like 10 PF from the output side to ground. Depending on how much
drive your CMOS has, putting a 47 Ohm resistor in series with the coax input
might help if the coax is less than about a foot long.

Keep in mind, that in order to drive a 5V signal into 50 Ohm coax, and do it
right, your CMOS device would have to drive 5/50 or 0.1 Amp. You would also
need a 50 Ohm resistor across the output.

While you are at it, look at the VCC of the CMOS. If the chip does not have
a bypass capacitor on VCC, add a .1 - .33 before bothering to look.

Tam

I will add, the 10Mhz comes off a pin (gate) of a 74AC86. It's
referenced to ground through 7 inches of RG-174 to an input pin of a
Xilinix XC9536 PLCC. The scope probe is Tek P6103B. The probe manual
describes high and low frequency calibration. I forgot if the Hi trim
tab was broken or if I didn't have the correct calibration generator
but remember reducing the amplitude of the effect by the low freq.
probe cal procedure.

Thank you for your input.
Largo

 

[Largo]

What kinds of circuitry do you have on both ends?

The scope *may* not be telling the truth. I presume you're using a 10:1
probe which has been adjusted on the scope's calibration signal...right?
Even so, a 50-MHz scope may not be an accurate judge of the squareness of a
10-MHz signal. Still, if it's mainly the scope's limitations, you'd expect
it to look like it's been rounded, not ringing.

It's ringing. I have a P6103B 10X passive probe.
Thanks,
Largo

 

[Tam/WB2TT]

I will add, the 10Mhz comes off a pin (gate) of a 74AC86. It's
referenced to ground through 7 inches of RG-174 to an input pin of a
Xilinix XC9536 PLCC. The scope probe is Tek P6103B. The probe manual
describes high and low frequency calibration. I forgot if the Hi trim
tab was broken or if I didn't have the correct calibration generator
but remember reducing the amplitude of the effect by the low freq.
probe cal procedure.

Thank you for your input.
Largo

I have done that. Put something like a 27 to 47 Ohm resistor in series
between the output pin of the 74AC86 and the RG-174. Do not terminate the
output of the 174. You should get some slowing down of the rise and fall
times, but no ringing. Make sure the AC86 has a bypass cap on it; it will be
drawing heavy current during transitions.

Tam

 

[Largo]

I have done that. Put something like a 27 to 47 Ohm resistor in series
between the output pin of the 74AC86 and the RG-174. Do not terminate the
output of the 174. You should get some slowing down of the rise and fall
times, but no ringing. Make sure the AC86 has a bypass cap on it; it will be
drawing heavy current during transitions.

Tam

Hi Tam,
Is this a real or a measurement effect?

Largo

 

[Tam/WB2TT]

Hi Tam,
Is this a real or a measurement effect?

Largo

I think real, as based on measurements. I used both a 10X high impedance
probe, and a 20X passive probe feeding a 50 Ohm 300 MHz 'scope. The latter
probe only had about 0.8 PF of capacitive loading ( and 1K reistive
loading).

Another thing that sometimes works is to terminate output side of the coax
during transitions only. Load it with a 50 Ohm reistor in series with a 10 -
50 PF capacitor to ground. Obviously, you can't have a 50 Ohm DC load, or
you won't get 5V out of the circuit, plus, you will dissipate a lot of
power.

Tam

 

[Fred Bloggs]

I will add, the 10Mhz comes off a pin (gate) of a 74AC86. It's
referenced to ground through 7 inches of RG-174 to an input pin of a
Xilinix XC9536 PLCC. The scope probe is Tek P6103B. The probe
manual describes high and low frequency calibration. I forgot if the
Hi trim tab was broken or if I didn't have the correct calibration
generator but remember reducing the amplitude of the effect by the
low freq. probe cal procedure.

There will be no transmission line effects at your frequencies for a 7"
section of RG-179 and termination is not necessary. Winging a
C-distribution of 30p/ft gets you L=Cx50^2 or 75nh/ft for an equivalent
circuit of:
View in a fixed-width font such as
Courier.

 

[Fred Bloggs]

There will be no transmission line effects at your frequencies for a 7"
section of RG-179 ...

That should be RG-174, the calc is otherwise right.

 

[Largo]

I will add, the 10Mhz comes off a pin (gate) of a 74AC86. It's
referenced to ground through 7 inches of RG-174 to an input pin of a
Xilinix XC9536 PLCC. The scope probe is Tek P6103B. The probe
manual describes high and low frequency calibration. I forgot if the
Hi trim tab was broken or if I didn't have the correct calibration
generator but remember reducing the amplitude of the effect by the
low freq. probe cal procedure.

There will be no transmission line effects at your frequencies for a 7"
section of RG-179 and termination is not necessary. Winging a
C-distribution of 30p/ft gets you L=Cx50^2 or 75nh/ft for an equivalent
circuit of:
View in a fixed-width font such as
Courier.


.
.
.
. --[22n]--+--[22n]--
. |
. [18p]
. in-> | out->
. |
. ---------+---------
.

Using the fundamental result for a 2-pole critical damping R>2xsqrt(L/C)
applied to the two sections and assuming driver gate output of ~20 ohms
results in this:
View in a fixed-width font such as Courier.


.
. /-------------\
. |\ | | |\
. ----| >-[51]--|---------------|--[51]----| >---
. |/| | | |/|
. | \+-----------+/ |
. | | | |
. | | | |
. -----+---------' '--------------+---
.

You need to damp both sides of the line with ~50 ohms. Any ringing
observed is coming from a combination of probe effects and less than
ideal layout.

What would be the ideal layout?

The input of the PLCC is located on the primary circuit application
with 0.1 pin header for 10Mhz oscillator inputs. First prototype osc.
used a 3.3V Maxim econoscillator through 7 inches RG-174. Second, a
PLL with 74AC86 through 7 inch RGG-174. Although completely different
oscillators, I observe the same ringing effect. What's in common is
the coax and the scope probe.

I think my edge detector in the PLCC counted double off the rising
edge ringing in the econoscillator 10Mhz because it's three volt p-p.
The PLL 10Mhz is 5 v p-p.

Should I just use a "wire" between the primary appllication and
oscillator, omitting the coax completely?

Thank you,
Largo

 

[Fred Bloggs]

I will add, the 10Mhz comes off a pin (gate) of a 74AC86. It's
referenced to ground through 7 inches of RG-174 to an input pin of a
Xilinix XC9536 PLCC. The scope probe is Tek P6103B. The probe
manual describes high and low frequency calibration. I forgot if the
Hi trim tab was broken or if I didn't have the correct calibration
generator but remember reducing the amplitude of the effect by the
low freq. probe cal procedure.

There will be no transmission line effects at your frequencies for a 7"
section of RG-179 and termination is not necessary. Winging a
C-distribution of 30p/ft gets you L=Cx50^2 or 75nh/ft for an equivalent
circuit of:
View in a fixed-width font such as
Courier.


.
.
.
. --[22n]--+--[22n]--
. |
. [18p]
. in-> | out->
. |
. ---------+---------
.

Using the fundamental result for a 2-pole critical damping R>2xsqrt(L/C)
applied to the two sections and assuming driver gate output of ~20 ohms
results in this:
View in a fixed-width font such as Courier.


.
. /-------------\
. |\ | | |\
. ----| >-[51]--|---------------|--[51]----| >---
. |/| | | |/|
. | \+-----------+/ |
. | | | |
. | | | |
. -----+---------' '--------------+---
.

You need to damp both sides of the line with ~50 ohms. Any ringing
observed is coming from a combination of probe effects and less than
ideal layout.

What would be the ideal layout?

The input of the PLCC is located on the primary circuit application
with 0.1 pin header for 10Mhz oscillator inputs. First prototype osc.
used a 3.3V Maxim econoscillator through 7 inches RG-174. Second, a
PLL with 74AC86 through 7 inch RGG-174. Although completely different
oscillators, I observe the same ringing effect. What's in common is
the coax and the scope probe.

Where exactly are you placing the probe, its capacitance and GND lead
inductance can have a significant effect at these frequencies? Another
commonality is the PLCC input gate.

I think my edge detector in the PLCC counted double off the rising
edge ringing in the econoscillator 10Mhz because it's three volt p-p.
The PLL 10Mhz is 5 v p-p.

This could be anything like anomaly in the driver output under high
frequency loading.

Should I just use a "wire" between the primary appllication and
oscillator, omitting the coax completely?

No- the coax is much better than a wire because it is phase stable. The
SPICE sims show this is pretty well damped- going for a lower output
impedance at frequency:
View in a fixed-width font such as Courier.




. /-------------\
. |\ | | |\
. ----| >-[100]-|---------------|--[51]-+--| >---
. |/| | | | |/|
. | \+-----------+/ === |
. | | | 22p| |
. | | | | |
. -----+---------' '---------+----+---

To probe this circuit do this with a 1.2K R pushed into a coax cable:
View in a fixed-width font such as Courier.


..
..
.. /-------------\
.. | |
.. scope ----|---------------|---------.
.. Zin =50 | | |
.. \+-----------+/ |
.. | | |
.. scope | | |
.. gnd ------ | |
.. | |
.. | [1.2k]
.. /----------+--\ |
.. |\ | | | |\
.. ----| >-[100]-|---------------|--[51]-+--| >---
.. |/| | | | |/|
.. | \+-----------+/ === |
.. | | | 22p| |
.. | | | | |
.. -----+---------' '---------+----+---

 

[Largo]

I will add, the 10Mhz comes off a pin (gate) of a 74AC86. It's
referenced to ground through 7 inches of RG-174 to an input pin of a
Xilinix XC9536 PLCC. The scope probe is Tek P6103B. The probe
manual describes high and low frequency calibration. I forgot if the
Hi trim tab was broken or if I didn't have the correct calibration
generator but remember reducing the amplitude of the effect by the
low freq. probe cal procedure.


There will be no transmission line effects at your frequencies for a 7"
section of RG-179 and termination is not necessary. Winging a
C-distribution of 30p/ft gets you L=Cx50^2 or 75nh/ft for an equivalent
circuit of:
View in a fixed-width font such as
Courier.


.
.
.
. --[22n]--+--[22n]--
. |
. [18p]
. in-> | out->
. |
. ---------+---------
.

Using the fundamental result for a 2-pole critical damping R>2xsqrt(L/C)
applied to the two sections and assuming driver gate output of ~20 ohms
results in this:
View in a fixed-width font such as Courier.


.
. /-------------\
. |\ | | |\
. ----| >-[51]--|---------------|--[51]----| >---
. |/| | | |/|
. | \+-----------+/ |
. | | | |
. | | | |
. -----+---------' '--------------+---
.

You need to damp both sides of the line with ~50 ohms. Any ringing
observed is coming from a combination of probe effects and less than
ideal layout.

What would be the ideal layout?

The input of the PLCC is located on the primary circuit application
with 0.1 pin header for 10Mhz oscillator inputs. First prototype osc.
used a 3.3V Maxim econoscillator through 7 inches RG-174. Second, a
PLL with 74AC86 through 7 inch RGG-174. Although completely different
oscillators, I observe the same ringing effect. What's in common is
the coax and the scope probe.

Where exactly are you placing the probe, its capacitance and GND lead
inductance can have a significant effect at these frequencies? Another
commonality is the PLCC input gate.

In both oscillator prototypes, the probe hooks to the end of the 7 in
of coax. Same ringing both cases. So it's measuring the effects of
the coax? It's not connected to the PLCC during measurement. The
calibrated 10Mhz seems to be doing it's job being quite stable over a
48 hour period.

This could be anything like anomaly in the driver output under high
frequency loading.


No- the coax is much better than a wire because it is phase stable. The
SPICE sims show this is pretty well damped- going for a lower output
impedance at frequency:
View in a fixed-width font such as Courier.

It damps out in 1/2 the period and on both high and low peaks. I
wondered if this was detrimental to my counting circuit.
Largo

. /-------------\
. |\ | | |\
. ----| >-[100]-|---------------|--[51]-+--| >---
. |/| | | | |/|
. | \+-----------+/ === |
. | | | 22p| |
. | | | | |
. -----+---------' '---------+----+---

To probe this circuit do this with a 1.2K R pushed into a coax cable:
View in a fixed-width font such as Courier.


.
.
. /-------------\
. | |
. scope ----|---------------|---------.
. Zin =50 | | |
. \+-----------+/ |
. | | |
. scope | | |
. gnd ------ | |
. | |
. | [1.2k]
. /----------+--\ |
. |\ | | | |\
. ----| >-[100]-|---------------|--[51]-+--| >---
. |/| | | | |/|
. | \+-----------+/ === |
. | | | 22p| |
. | | | | |
. -----+---------' '---------+----+---

 

[Fred Bloggs]

It damps out in 1/2 the period and on both high and low peaks. I
wondered if this was detrimental to my counting circuit.

If it was bad enough to cause double counting in one case then it is
detrimental.

Largo

. /-------------\
. |\ | | |\
. ----| >-[100]-|---------------|--[51]-+--| >---
. |/| | | | |/|
. | \+-----------+/ === |
. | | | 22p| |
. | | | | |
. -----+---------' '---------+----+---

To probe this circuit do this with a 1.2K R pushed into a coax cable:
View in a fixed-width font such as Courier.


.
.
. /-------------\
. | |
. scope ----|---------------|---------.
. Zin =50 | | |
. \+-----------+/ |
. | | |
. scope | | |
. gnd ------ | |
. | |
. | [1.2k]
. /----------+--\ |
. |\ | | | |\
. ----| >-[100]-|---------------|--[51]-+--| >---
. |/| | | | |/|
. | \+-----------+/ === |
. | | | 22p| |
. | | | | |
. -----+---------' '---------+----+---