Measuring 50Ohm load terminator across frequency?

[John Larkin]

On Wed, 12 Nov 2003 22:46:31 -0500, Keith R. Williams
<krw@attglobal.net> wrote:

Hmm, a dual channel TDR? Is that anything like a VNA doing S12
in the time domain? ;-) Sorry!

Absolutely. It's called TDT, Time Domain Transmission measurement. If
you put a fast step into port1 of a 2-port box, and scope what comes
out of port2, that's the time domain equivalent to S21. The port-1
TDR, which you can do simultaneously, is (almost) just a Fourier
transform away from S11.

A Tek SD-24 (DC-20 GHz) sampling head does dual-channel TDR and is
cool for stuff like this; just turn on the TDR step of one channel and
use the other channel in pure sampling mode. If you attenuate the
step, you can use this to test the time-domain response of wideband RF
amplifiers and things like that.

John

 

[Ian Buckner]

"Paul Burridge" <pb@osiris1.notthisbit.co.uk> wrote in message
news:vhq4rvk5pqg6cc8v06n1a0h8hc79e2kcp7@4ax.com...

On Wed, 12 Nov 2003 08:41:31 -0500, Keith R. Williams
krw@attglobal.net> wrote:

In article <b0c4rvkrtocf94obmkn5f0jt37cld4a0ni@4ax.com>,
pb@osiris1.notthisbit.co.uk says...
On Tue, 11 Nov 2003 19:53:29 -0800, John Larkin
jjlarkin@highlandSNIPtechTHISnologyPLEASE.com> wrote:


[snip]
The 'stuff' is perfectly flat for a perfect 50 ohm termination,
spikes
up if the DUT is inductive, spikes down if capacitive, rings if
resonant. There is, in theory, enough info in the reflection to
fully
characterize the impedance.

Both the pulse gen and the scope need to have ~~ 1 ns response
to
resolve components close to 300 MHz. I have a Tek 11801 that has
about
20 ps TDR resolution, and it's great for characterizing the
impedance
of PCB traces.

I think there are some pretty good google-able TDR tutorials.

Interesting. So what does "TDR" stand for? time-domain something,
presumably?

Time Domain Reflectometer (or reflectometry). Think of a TDR as a
network analyzer that has had an inverse Fourier transform. ;-)
TDRs
are a rather useful piece of equipment.

Well I was going to buy a VNA at some point soon. Does that obviate
the need for a TDR or is it advisable to have a TDR in addition,
anyway? I mean, is there something the TDR can do that a VNA can't?

--

"I expect history will be kind to me, since I intend to write it."
-
Winston Churchill

All the VNA's I am familiar with have internal transforms available
either as an option or standard, so you can get the TDR as well.

Regards
Ian

 

[Paul Burridge]

On Wed, 12 Nov 2003 10:41:20 -0800, John Larkin
<jjlarkin@highSNIPlandTHIStechPLEASEnology.com> wrote:

If you plan to do RF (ie, sine waves, low amplitudes, Smith charts)
the VNA is much better.

Okay. Next question is: do you think I could handle one okay? I mean,
I don't know what half the knobs on my oscilloscopes do. Are they more
complicated than a 'scope?
--

"I expect history will be kind to me, since I intend to write it."
- Winston Churchill

 

[John Larkin]

On Thu, 13 Nov 2003 14:47:41 +0000, Paul Burridge
<pb@osiris1.notthisbit.co.uk> wrote:

On Wed, 12 Nov 2003 10:41:20 -0800, John Larkin
jjlarkin@highSNIPlandTHIStechPLEASEnology.com> wrote:

If you plan to do RF (ie, sine waves, low amplitudes, Smith charts)
the VNA is much better.

Okay. Next question is: do you think I could handle one okay? I mean,
I don't know what half the knobs on my oscilloscopes do. Are they more
complicated than a 'scope?

Get a copy of Chris Bowick's book, RF Circuit Design, and read, well,
all of it. It's an inexpensive and excellent paperback. Then read an
assortment of datasheets and appnotes on RF transistors and fets, like
the NEC/CEL parts, the ones with S-parameter tables and Smith charts.
If that all makes sense, and the math isn't too hairy for you, you
could probably handle a VNA; if not, no. Most ham radio guys who build
their own RF stuff do fine without a VNA.

John

 

[mike]

John Larkin wrote:

On Wed, 12 Nov 2003 22:46:31 -0500, Keith R. Williams
krw@attglobal.net> wrote:

Hmm, a dual channel TDR? Is that anything like a VNA doing S12
in the time domain? ;-) Sorry!


Absolutely. It's called TDT, Time Domain Transmission measurement. If
you put a fast step into port1 of a 2-port box, and scope what comes
out of port2, that's the time domain equivalent to S21. The port-1
TDR, which you can do simultaneously, is (almost) just a Fourier
transform away from S11.

A Tek SD-24 (DC-20 GHz) sampling head does dual-channel TDR and is
cool for stuff like this; just turn on the TDR step of one channel and
use the other channel in pure sampling mode. If you attenuate the
step, you can use this to test the time-domain response of wideband RF
amplifiers and things like that.

John

How well does this time/frequency domain conversion stuff work in
practice. I think I understand the theory, but what about actual practice?

It was 20 years ago, but I once tried to do some simple frequency
measurements with a TDR. The transform was EXTREMELY sensitive to
artifacts in the TDR and digitizer resolution. I gave up cause there
didn't seem to be any hope of making it work reliably. The fact that I
was trying to do it on a VIC-20 didn't help ;-)
I also tried to tweek out connector aberrations with frequency sweep
data. That also was a miserable failure.
Since then, I've always taken data in the domain of interest.
mike

--
Bunch of stuff For Sale and Wanted at the link below.
laptops and parts Test Equipment
Honda CB-125S
Color LCD overhead projector
Tek 2465 $800, ham radio, 30pS pulser
3T2 sampling sweep
Tektronix Concept Books, spot welding head...
http://www.geocities.com/SiliconValley/Monitor/4710/

 

[John Larkin]

On Thu, 13 Nov 2003 13:38:27 -0800, mike <spamme0@juno.com> wrote:

John Larkin wrote:
On Wed, 12 Nov 2003 22:46:31 -0500, Keith R. Williams
krw@attglobal.net> wrote:

Hmm, a dual channel TDR? Is that anything like a VNA doing S12
in the time domain? ;-) Sorry!


Absolutely. It's called TDT, Time Domain Transmission measurement. If
you put a fast step into port1 of a 2-port box, and scope what comes
out of port2, that's the time domain equivalent to S21. The port-1
TDR, which you can do simultaneously, is (almost) just a Fourier
transform away from S11.

A Tek SD-24 (DC-20 GHz) sampling head does dual-channel TDR and is
cool for stuff like this; just turn on the TDR step of one channel and
use the other channel in pure sampling mode. If you attenuate the
step, you can use this to test the time-domain response of wideband RF
amplifiers and things like that.

John




How well does this time/frequency domain conversion stuff work in
practice. I think I understand the theory, but what about actual practice?

It was 20 years ago, but I once tried to do some simple frequency
measurements with a TDR. The transform was EXTREMELY sensitive to
artifacts in the TDR and digitizer resolution. I gave up cause there
didn't seem to be any hope of making it work reliably. The fact that I
was trying to do it on a VIC-20 didn't help ;-)
I also tried to tweek out connector aberrations with frequency sweep
data. That also was a miserable failure.
Since then, I've always taken data in the domain of interest.
mike

Mike,

My experience here is limited, as I work mostly in large-signal time
domain, don't have a VNA, and seldom use s-params. My impression from
appnotes and stuff, and some limited experience, is that the
cross-domain stuff is generally mediocre. Scopes are usually lously
spectrum analyzers. A VNA can do narrowband sweeps with very high
dynamic range, so beats the hell out of a sampling scope when it comes
to small-signal RF stuff. I don't know how well a VMA does simulated
TDR.

Does anybody out there have a VNA that can to the TDR transform? Maybe
we could set up a couple of simple, easy-to-reproduce situations and
post measured waveforms to compare. I can do the TDR stuff on my
11801. Sounds like fun.

John

 

[Paul Burridge]

On Thu, 13 Nov 2003 08:03:41 -0800, John Larkin
<jjlarkin@highlandSNIPtechTHISnologyPLEASE.com> wrote:

Get a copy of Chris Bowick's book, RF Circuit Design, and read, well,
all of it. It's an inexpensive and excellent paperback. Then read an
assortment of datasheets and appnotes on RF transistors and fets, like
the NEC/CEL parts, the ones with S-parameter tables and Smith charts.
If that all makes sense, and the math isn't too hairy for you, you
could probably handle a VNA; if not, no. Most ham radio guys who build
their own RF stuff do fine without a VNA.

Thanks, John. At the risk of repeating myself (again) I *do* have
Bowick's excellent book and it's becoming increasingly dog-eared. In
fact I'm about to get stuck into it even further over the next few
weeks. I get the gist of what you say. I think I'll be okay in the
end!
--

"I expect history will be kind to me, since I intend to write it."
- Winston Churchill

 

[Keith R. Williams]

In article <cn16rvcaubaa9bd2euofj3h4ot2qmv46tf@4ax.com>,
jjlarkin@highlandSNIPtechTHISnologyPLEASE.com says...

On Wed, 12 Nov 2003 22:46:31 -0500, Keith R. Williams
krw@attglobal.net> wrote:

Hmm, a dual channel TDR? Is that anything like a VNA doing S12
in the time domain? ;-) Sorry!

Absolutely. It's called TDT, Time Domain Transmission measurement. If
you put a fast step into port1 of a 2-port box, and scope what comes
out of port2, that's the time domain equivalent to S21. The port-1
TDR, which you can do simultaneously, is (almost) just a Fourier
transform away from S11.

Once I'd thought about this some more, I realized I'd done this
stuff 25ish years ago. I had a Tek Digital Processing System (A
Tek 7704 with an A/D sampler in the middle, hooked to a DEC PDP-
11/34) and was looking for interesting things to do with it in my
"spare" time.

I "found (in my livingroom ;-)" this "HBO filter" and decided to
analyze it. I used 7S11 and 7S12 (IIRC) plugins in the DPO and a
HP pulse generator to excite the widget. I then took the input
and output waveforms, averaged the hell out of 'em and did an
FFT/convolution to find what was inside (simply a sharp notch
filter).

A Tek SD-24 (DC-20 GHz) sampling head does dual-channel TDR and is
cool for stuff like this; just turn on the TDR step of one channel and
use the other channel in pure sampling mode. If you attenuate the
step, you can use this to test the time-domain response of wideband RF
amplifiers and things like that.

I found that an "impulse" as narrow as I could get it (and as
small - averaged thousands of times) was better than the TDR head
on the 7S11. It's been a long time since I've had the "need" to
make an HBO filter though. ;-)

(Oh, are the statute of limitations up yet? Honestly, it was all
about intellectual curiosity! *I* didn't manufacture them!)


--
Keith

 

[John Larkin]

On Thu, 13 Nov 2003 23:48:54 -0500, Keith R. Williams
<krw@attglobal.net> wrote:

I found that an "impulse" as narrow as I could get it (and as
small - averaged thousands of times) was better than the TDR head
on the 7S11. It's been a long time since I've had the "need" to
make an HBO filter though. ;-)

TDR heads classicly output a step.

The frequency-domain transfer function of a box is the Fourier
transform of the unit impulse time-domain response. Since a step is
the integral of an impulse, the TDR waveform transforms into the
integral of the frequency domain response. You can fix that by
differentiating the step response before doing the FFT, or by doing
the appropriate thrashing on the spectrum afterwards; this involves
scaling the amplitudes proportional to frequency and flipping sines
and cosines. Or something like that.

John

 

[John Woodgate]

I read in sci.electronics.design that mike <spamme0@juno.com> wrote (in
<3FB3F9D3.9060509@juno.com&gt about 'Measuring 50Ohm load terminator
across frequency?', on Thu, 13 Nov 2003:

How well does this time/frequency domain conversion stuff work in
practice. I think I understand the theory, but what about actual
practice?

It was 20 years ago, but I once tried to do some simple frequency
measurements with a TDR. The transform was EXTREMELY sensitive to
artifacts in the TDR and digitizer resolution.

Nothing has really changed: the sensitivity is inherent in the math.

I gave up cause there
didn't seem to be any hope of making it work reliably. The fact that I
was trying to do it on a VIC-20 didn't help ;-)

With better processors you can use tricks to disguise the grosser
errors, but they cannot be eliminated.

I also tried to tweek out connector aberrations with frequency sweep
data. That also was a miserable failure.
Since then, I've always taken data in the domain of interest.

Extremely wise. There should be a Society for the Prevention of Abuse of
the Fourier Transform.
--
Regards, John Woodgate, OOO - Own Opinions Only. http://www.jmwa.demon.co.uk
Interested in professional sound reinforcement and distribution? Then go to
http://www.isce.org.uk
PLEASE do NOT copy news posts to me by E-MAIL!

 

[Keith R. Williams]

In article <quo8rv4hmjda1rogljr8lgl2mbgvvg6oqm@4ax.com>,
jjlarkin@highlandSNIPtechTHISnologyPLEASE.com says...

On Thu, 13 Nov 2003 23:48:54 -0500, Keith R. Williams
krw@attglobal.net> wrote:


I found that an "impulse" as narrow as I could get it (and as
small - averaged thousands of times) was better than the TDR head
on the 7S11. It's been a long time since I've had the "need" to
make an HBO filter though. ;-)

TDR heads classicly output a step.

Sorry, I meant that an impulse from a high speed pulse generator (at
very low amplitude) turned out to be more useful than the much faster
step from the TDR head. The inputs were the same (the 7S11? is both a
sampler and a TDR, depending on what head one had in the horizontal
part).

The frequency-domain transfer function of a box is the Fourier
transform of the unit impulse time-domain response. Since a step is
the integral of an impulse, the TDR waveform transforms into the
integral of the frequency domain response. You can fix that by
differentiating the step response before doing the FFT, or by doing
the appropriate thrashing on the spectrum afterwards; this involves
scaling the amplitudes proportional to frequency and flipping sines
and cosines. Or something like that.

I divided the output spectrum by the input to get "what's in the box".
The Tek SPS systems were specifically set up to do such operations on
waveforms. The probes and fixtures were "taken out" of the equation in
much the same fashion. The setup compared quite nicely to a VNA on the
other side of the lab. The VNA operation was much simpler and faster
though.

--
Keith