Maker Pro
Maker Pro

Time Domain Reflectometry

Hi!
I'd like design a pulse generator, for the fun of it, and because I need
to make some TDR work, together with my Tektronix TDS2024B oscilloscope.
I saw the circuit by Tomi Engdahl, which uses a 74AC14 (schmitt inverter),
but I am also tempted to use the Maxim DS1040Z-A15, which is a programmable
pulse generator (from 5ns to 15ns, through five 2.5ns steps).

My doubts are about driving the (50 ohm) cable:

what if I connect the chip directly to the cable? It is specced of being
capable of 50mA "short circuit current" which, if I am capable of doing
the basic math, means that the driver has an impedance of 100ohm (at 5V),
thus it is mismatched.

If I connect two of them in parallel I should get a good match for 50ohm,
right?

And what if I only use one driver? The signal will be smaller, or there
will be immediate reflections possibly destroying the IC?

Last but not least, some theory: what's the point in having ~picoseconds
rise time, when anyway you drive one hundred meters of 100pF/m cable with
a 50 ohm impedance in serie?
Before the impulse has traveled one meter, it has already been smoothed
out to ns-range rising/falling times.. and much worse after tens of meters.

Having a 1ns rise time is going to be worse than 10ps rise time just like
adding a couple of meters of cable, or am I mistaken? I don't really get
the point in having ps-scale impulses for such applications, even more
because of obvious price considerations.

About the 74AC14, Tomi's circuit uses 5 of them in parallel, with a 220
ohm resistor each, giving a "cumulative" output impedance of 50 ohm. But
what's the point in paralleling five buffers, to strenghten the current,
when you add resistors in serie at each output anyway?

And, if it makes sense, can I parallel even more cheap 74AC14's to improve
the pulse generator performance?

Tomi's circuit can be seen here:
http://www.epanorama.net/circuits/tdr.html

Thank you!
Andrea
 
R

Robert Baer

Jan 1, 1970
0
Hi!
I'd like design a pulse generator, for the fun of it, and because I need
to make some TDR work, together with my Tektronix TDS2024B oscilloscope.
I saw the circuit by Tomi Engdahl, which uses a 74AC14 (schmitt inverter),
but I am also tempted to use the Maxim DS1040Z-A15, which is a programmable
pulse generator (from 5ns to 15ns, through five 2.5ns steps).

My doubts are about driving the (50 ohm) cable:

what if I connect the chip directly to the cable? It is specced of being
capable of 50mA "short circuit current" which, if I am capable of doing
the basic math, means that the driver has an impedance of 100ohm (at 5V),
thus it is mismatched.

If I connect two of them in parallel I should get a good match for 50ohm,
right?

And what if I only use one driver? The signal will be smaller, or there
will be immediate reflections possibly destroying the IC?

Last but not least, some theory: what's the point in having ~picoseconds
rise time, when anyway you drive one hundred meters of 100pF/m cable with
a 50 ohm impedance in serie?
Before the impulse has traveled one meter, it has already been smoothed
out to ns-range rising/falling times.. and much worse after tens of meters.

Having a 1ns rise time is going to be worse than 10ps rise time just like
adding a couple of meters of cable, or am I mistaken? I don't really get
the point in having ps-scale impulses for such applications, even more
because of obvious price considerations.

About the 74AC14, Tomi's circuit uses 5 of them in parallel, with a 220
ohm resistor each, giving a "cumulative" output impedance of 50 ohm. But
what's the point in paralleling five buffers, to strenghten the current,
when you add resistors in serie at each output anyway?

And, if it makes sense, can I parallel even more cheap 74AC14's to improve
the pulse generator performance?

Tomi's circuit can be seen here:
http://www.epanorama.net/circuits/tdr.html

Thank you!
Andrea
1) Forget non-existant Maxim crap.
2) CMOS gaes can be paralleled to give whatever drive you want.
3) You can use a resistive "tee" for matching.
4) *NO* CMOS parts have picosecond rise, fall or delay times.
5) Try ECL type logic (offset the supplies if need).
6) Or try non-linear parts in the "tee" (eg: saturating L for the
input element).
7) Or try a snapoff diode.
8) Or an Esaki diode.
 
Hi!
I'd like design a pulse generator, for the fun of it, and because I need
to make some TDR work, together with my Tektronix TDS2024B oscilloscope.
I saw the circuit by Tomi Engdahl, which uses a 74AC14 (schmitt inverter),
but I am also tempted to use the Maxim DS1040Z-A15, which is a programmable
pulse generator (from 5ns to 15ns, through five 2.5ns steps).

My doubts are about driving the (50 ohm) cable:

what if I connect the chip directly to the cable? It is specced of being
capable of 50mA "short circuit current" which, if I am capable of doing
the basic math, means that the driver has an impedance of 100ohm (at 5V),
thus it is mismatched.

If I connect two of them in parallel I should get a good match for 50ohm,
right?

And what if I only use one driver? The signal will be smaller, or there
will be immediate reflections possibly destroying the IC?

Last but not least, some theory: what's the point in having ~picoseconds
rise time, when anyway you drive one hundred meters of 100pF/m cable with
a 50 ohm impedance in serie?
Before the impulse has traveled one meter, it has already been smoothed
out to ns-range rising/falling times.. and much worse after tens of meters.

Having a 1ns rise time is going to be worse than 10ps rise time just like
adding a couple of meters of cable, or am I mistaken? I don't really get
the point in having ps-scale impulses for such applications, even more
because of obvious price considerations.

About the 74AC14, Tomi's circuit uses 5 of them in parallel, with a 220
ohm resistor each, giving a "cumulative" output impedance of 50 ohm. But
what's the point in paralleling five buffers, to strenghten the current,
when you add resistors in serie at each output anyway?

If you are driving a 50R cable, it is a good idea to drive it from a
50R resistive impedance, so that any reflection back from the cable
is absorbed in the source termination, rather than reflected back into
the cable, to be reflected back again and again.

If you take the small signal output impedance of each 74AC14 to be
about 30R, adding 220R in series gives a total output resistance of
250R per gate. Five 250R resistive impedances in parallel is 50R.

If you tried to make do with fewer gates, the gates would run out of
current when the signal was close to Vcc or Vss, and their output
impedance would be higher when the signal was in these regions.
And, if it makes sense, can I parallel even more cheap 74AC14's to improve
the pulse generator performance?

When you've got enough gates in parallel, adding more won't make any
difference to the performance, though the individual devices will run
cooler.
 
J

John Larkin

Jan 1, 1970
0
Hi!
I'd like design a pulse generator, for the fun of it, and because I need
to make some TDR work, together with my Tektronix TDS2024B oscilloscope.
I saw the circuit by Tomi Engdahl, which uses a 74AC14 (schmitt inverter),
but I am also tempted to use the Maxim DS1040Z-A15, which is a programmable
pulse generator (from 5ns to 15ns, through five 2.5ns steps).

My doubts are about driving the (50 ohm) cable:

what if I connect the chip directly to the cable? It is specced of being
capable of 50mA "short circuit current" which, if I am capable of doing
the basic math, means that the driver has an impedance of 100ohm (at 5V),
thus it is mismatched.

If I connect two of them in parallel I should get a good match for 50ohm,
right?

And what if I only use one driver? The signal will be smaller, or there
will be immediate reflections possibly destroying the IC?

Last but not least, some theory: what's the point in having ~picoseconds
rise time, when anyway you drive one hundred meters of 100pF/m cable with
a 50 ohm impedance in serie?
Before the impulse has traveled one meter, it has already been smoothed
out to ns-range rising/falling times.. and much worse after tens of meters.


No. A fast edge will propagate, as a fast edge, for great distanxes
along a transmission line. The line has distributed inductance as well
as distributed capacitance. Only non-ideal losses (dielectric and
resistive losses) soften up a propagating edge. A 1 ns edge can be
propagated through 100 feet of good coax.

Tomi's circuit looks fine.

John
 
C

colin

Jan 1, 1970
0
Hi!
I'd like design a pulse generator, for the fun of it, and because I need
to make some TDR work, together with my Tektronix TDS2024B oscilloscope.
I saw the circuit by Tomi Engdahl, which uses a 74AC14 (schmitt inverter),
but I am also tempted to use the Maxim DS1040Z-A15, which is a
programmable
pulse generator (from 5ns to 15ns, through five 2.5ns steps).

My doubts are about driving the (50 ohm) cable:

what if I connect the chip directly to the cable? It is specced of being
capable of 50mA "short circuit current" which, if I am capable of doing
the basic math, means that the driver has an impedance of 100ohm (at 5V),
thus it is mismatched.

If I connect two of them in parallel I should get a good match for 50ohm,
right?

And what if I only use one driver? The signal will be smaller, or there
will be immediate reflections possibly destroying the IC?

Last but not least, some theory: what's the point in having ~picoseconds
rise time, when anyway you drive one hundred meters of 100pF/m cable with
a 50 ohm impedance in serie?
Before the impulse has traveled one meter, it has already been smoothed
out to ns-range rising/falling times.. and much worse after tens of
meters.

Having a 1ns rise time is going to be worse than 10ps rise time just like
adding a couple of meters of cable, or am I mistaken? I don't really get
the point in having ps-scale impulses for such applications, even more
because of obvious price considerations.

About the 74AC14, Tomi's circuit uses 5 of them in parallel, with a 220
ohm resistor each, giving a "cumulative" output impedance of 50 ohm. But
what's the point in paralleling five buffers, to strenghten the current,
when you add resistors in serie at each output anyway?

And, if it makes sense, can I parallel even more cheap 74AC14's to improve
the pulse generator performance?

Tomi's circuit can be seen here:
http://www.epanorama.net/circuits/tdr.html

You could use an emitter folower with a 50ohm resistor down to ground,
a very short high pulse will drive the cable hard,
when the pulse returns the transistor will be off and so just see the 50r
resistor.

for such a pulse the cable is not like a capacitance, untill the reflection
comes back from the other end
it looks like a 50r resistor.

Colin =^.^=
 
M

Mike Monett

Jan 1, 1970
0
John Larkin said:
No. A fast edge will propagate, as a fast edge, for great distanxes
along a transmission line. The line has distributed inductance as well
as distributed capacitance. Only non-ideal losses (dielectric and
resistive losses) soften up a propagating edge. A 1 ns edge can be
propagated through 100 feet of good coax.

What is good coax? My experience is 1ns edges won't go very far in RG-58U.
I forget exactly, maybe 10 ft.

TDR also has round-trip, which doubles the distance.

Regards,

Mike Monett
 
A

Anthony Fremont

Jan 1, 1970
0
Mike said:
What is good coax? My experience is 1ns edges won't go very far in
RG-58U. I forget exactly, maybe 10 ft.

TDR also has round-trip, which doubles the distance.

I don't think many people would consider RG-58 good coax. 9913, RG-213 or
RG-8X are my favorites for 50 Ohm cable.
 
J

John Larkin

Jan 1, 1970
0
You could use an emitter folower with a 50ohm resistor down to ground,
a very short high pulse will drive the cable hard,
when the pulse returns the transistor will be off and so just see the 50r
resistor.

That will be an asymmetric non-50 ohm impedance, and proper TDR needs
a true 50 ohm source; most electrical TDR is done with a step, not a
pulse. But you don't need a lot of voltage - 0.25 volts into 50 ohms
is standard - so you can pad down a 5-volt swing and get a very good
50 ohms.

John
 
M

Mike Monett

Jan 1, 1970
0
John Larkin said:

No fair. Not many people get a chance to see that. Especially a 100 ft
run:)

So you agree the ordinary everyday plain run-of-the-mill standard flexible
coax that everyone and loves knows won't transfer a 1ns edge very far?

Regards,

Mike Monett
 
G

Gavin Melville

Jan 1, 1970
0
Hi,

I've made rather a lot of these generators -- both fast edges, and
pulses. Over long (1000 metres) cables I personally prefer pulses. I
have attached two small PDF files (am I meant to do that here ??)
showing rather old versions of the circuits. The one using the S
family TTL has worked for 20+ years, and I've used it to find faults
at 750 m down a 75 ohm buried video cable. You need to check a test
length -- the velocity factor is never quite right.

The 2n2222 based pulse generator can generate any pulse length you
like. The circuit originally had a avalanche transistor -- 2n2369
from memory, and as long as you don't try a SOT23 2n2222 the can type
ones work well. Layout is everything.

The circuit came from a Jim Williams app note

http://www.linear-tech.co.jp/pc/downloadDocument.do?navId=H0,C1,C1154,C1002,C1223,P1209,D4150

Page 21.

A later version is here.

http://www.edn.com/contents/images/323017f3.pdf


The coax sets the pulse length, and it can be quite long, or even very
short.

Some related links.

http://groups.google.com/group/sci....ad/e05944df49770f56/46daca998cfb77b3?lnk=raot

http://www.linear-tech.co.jp/pc/downloadDocument.do?navId=H0,C1,C1154,D4183

http://www.i9t.net/fast-pulse/fast-pulse.html

As some others have commented the attenuation can be fierce -- so 20
volt pulses are good -- that way you might get 10mV back.

Steps work just as well, and are always easier to see. Also easier to
get the impedance from.
 
J

John Larkin

Jan 1, 1970
0
No fair. Not many people get a chance to see that. Especially a 100 ft
run:)

So you agree the ordinary everyday plain run-of-the-mill standard flexible
coax that everyone and loves knows won't transfer a 1ns edge very far?

For various values of very far. People could make 200 GHz
oscilloscopes, but there's basically no way to get the signal to them.


The edge gets ugly. On, say, 100 feet of RG58, a steep part is still
transmitted, but its amplitude declines with distance and is followed
by drool.

_________
_____/
___/
__/ drool
_/
/
/
/ steep
/
/
-----------


I'll post some scope pics tomorrow maybe.

John
 
M

Mike Monett

Jan 1, 1970
0
John Larkin said:
The edge gets ugly. On, say, 100 feet of RG58, a steep part is still
transmitted, but its amplitude declines with distance and is followed
by drool.
_________
_____/
___/
__/ drool
_/
/
/
/ steep
/
/
-----------
I'll post some scope pics tomorrow maybe.

Thanks - that would be interesting.

I'm surprised it would go that far. Your RG-58 must be a lot better than
the cheap stuff everyone else gets.

Regards,

Mike Monett
 
M

Michael A. Terrell

Jan 1, 1970
0
John said:
Hardline.


16" wide brass waveguide.


--
Service to my country? Been there, Done that, and I've got my DD214 to
prove it.
Member of DAV #85.

Michael A. Terrell
Central Florida
 
P

Paul Hovnanian P.E.

Jan 1, 1970
0
John said:
Hardline.

But the OP might be stuck with a particular coax type for his
application. Remember, you TDR with the coax you've got, not the one you
want (or something to that effect).
 
P

Paul Hovnanian P.E.

Jan 1, 1970
0
Gavin said:
Hi,

I've made rather a lot of these generators -- both fast edges, and
pulses. Over long (1000 metres) cables I personally prefer pulses. I
have attached two small PDF files (am I meant to do that here ??)

For future reference: Your pdfs are not actually attached (which is not
allowed here). They are links to a web site, which is the preferred
method for posting graphics, binaries and such. Some of us who don't
have access to binary newsgroups can still see them.

If you must attach binaries, post to alt.binaries.schematics.electronic
and refer to the subject line here. We can all hop over there and try to
find them among all the jpegs of J.T.'s grandkids. ;-)
 
I

Ian

Jan 1, 1970
0
John Larkin said:
Hardline.

John
There's a 50 metre reel of Andrew's LDF5-50A in an adjacent cube,
that was used for TDR work.

Regards
Ian
 
J

Jim Yanik

Jan 1, 1970
0
No fair. Not many people get a chance to see that. Especially a 100 ft
run:)

So you agree the ordinary everyday plain run-of-the-mill standard
flexible coax that everyone and loves knows won't transfer a 1ns edge
very far?

Regards,

Mike Monett

Tektronix sold a lot of their 1503 long-range TDRs,measuring up to 50,000
ft of cable,IIRC. phone companies,cable companies.It used a 1/2 sine pulse.

IIRC,the TEK 1502 measured up to a couple of thousand feet,using a fast-
rise step pulse.
 
M

Mike Monett

Jan 1, 1970
0
Jim Yanik said:
Tektronix sold a lot of their 1503 long-range TDRs,measuring up to
50,000 ft of cable,IIRC. phone companies,cable companies.It used a
1/2 sine pulse.
IIRC,the TEK 1502 measured up to a couple of thousand feet,using a
fast- rise step pulse.

I used to own one. Can't remember if it was a 1502 or 1503. The
resolution was so poor on long cables that I gave it away to the
avionics company that maintained my Piper Malibu in San Jose.

The 1503A used a tunnel diode step generator with a risetime of
about 140ps. The B and C versions used a patented step generator
designed by a rather famous individual whose name escapes me now.

I tried to use his idea as a step generator in the Binary Sampler,
but ended up using a simple ecl logic signal from a 100EP16. A
simplified schematic omitting the 50 ohm output pad is shown here:

http://www3.sympatico.ca/add.automation/sampler/design.htm
Jim Yanik
jyanik
at
kua.net

Regards,

Mike Monett

Regards,

Mike Monett

Antiviral, Antibacterial Silver Solution:
http://members.spsdialup.com/[email protected]/index.htm
SPICE Analysis of Crystal Oscillators:
http://members.spsdialup.com/[email protected]/spice/xtal/clapp.htm
Noise-Rejecting Wideband Sampler:
http://www3.sympatico.ca/add.automation/sampler/intro.htm
 
R

Rich Grise

Jan 1, 1970
0
I don't think many people would consider RG-58 good coax.

Most Time Domain Reflectrometrists wouldn't consider 1 ns to be
an "edge". ;-)

Cheers!
Rich
 
Top