On Mon, 02 Jan 2006 16:10:02 -0800, dansawyeror
wrote:

The coupler measures
about -14 dbm.

What does this mean? Are you trying to tell us that the power on the
coupler port is 14dB less than the through power? What has dBm got to
do with it?

You didn't report the power in the coupler port with a s/c and / or
o/c at the measurement plane. Did you perform this cal?

(The phase angle and reflected ware very touchy. It was almost impossible to
adjust by changing frequency. It was easier to 'adjust' it by sitting very still
and moving my arm.)

That is understandable. How much coax between the A probe and the
load, and the B probe and the load... how many degress does this total
electrical length change for a 1% change in frequency? Does that
explain some of the phase sensitivity?


The antenna is a copy from the ARRL handbook. It is a 4 inch segment, a 1 inch
long by 3/4 inch diameter 5 turn coil, and a 4 inch tip. It is mounted over a 2
foot square aluminum plate. This antenna should have an input impedance less
then 20 Ohms.

Don't you need to measure some "known" loads. Why not try a 50 ohm
load tee'd to a s/c stub (quarter wave at a known frequency) and see
if you get the predictable results at different frequencies around
resonance. Then try two 50 ohm loads in parallel with the stub. (339mm
of RG58C/U should have a Z of around 6000+j0 ohms at around 146MHz, at
half that frequency it should be 0.85+j50, etc...


How can it measure very close to 50 Ohms? Is there something wrong with this
analysis?

See if you can trust your measurements on known loads before wondering
why the unknown load isn't what you expect when using unknown
measurement technology... too many unknowns.

Owen
--

On Tue, 03 Jan 2006 06:42:53 GMT, Owen Duffy wrote:

(The phase angle and reflected ware very touchy. It was almost impossible to
adjust by changing frequency. It was easier to 'adjust' it by sitting very still
and moving my arm.)

...Does that
explain some of the phase sensitivity?

Hi Owen,

Being very touchy, especially to the specifics of sitting very still,
sounds like classic common mode problems.

73's
Richard Clark, KB7QHC

On Mon, 02 Jan 2006 23:06:43 -0800, Richard Clark
wrote:

On Tue, 03 Jan 2006 06:42:53 GMT, Owen Duffy wrote:

(The phase angle and reflected ware very touchy. It was almost impossible to
adjust by changing frequency. It was easier to 'adjust' it by sitting very still
and moving my arm.)

...Does that
explain some of the phase sensitivity?

Hi Owen,

Being very touchy, especially to the specifics of sitting very still,
sounds like classic common mode problems.

Yes it does Richard. I saw Roy's response regarding isolation of the
feedline, and it is a valid comment. My comment was towards the
reported frequency sensitivity... until the effect of the propagation
delay is removed from the results, the underlying impedance is
obsured.

Owen
--

On Mon, 02 Jan 2006 21:56:37 GMT, Owen Duffy wrote:

On Mon, 02 Jan 2006 13:34:54 -0800, dansawyeror
wrote:

Owen,

That idea seems to work. I set it up and then looked and the 'interference', the
change in phase based on changing the pad. Zero pad showed several degrees phase
shift from -40 dbm, 10 dbm showed small shift, and 20, 30, and 40 were all about
equal. I decided on 20 dbm as a practical base.

This doesn't make sense... are you using "dbm" to mean decibels of
attenuation, usually written "dB".

The units "dBm" are usually written to qualify a power level with
respect to one milliwatt.

The attenuator on your sig gen might be marked in dBm, but that
applies to the combination of the oscillator, possibly its level
meter, and the attenuator as a system.

Using the wrong terms for things is often a result of a concept gap!

A 20dB attenuator will reduce the effect of the reflected component to
about the same level as you would expect from a practical directional
coupler, more attenuation is better if you have the power from the sig
gen and the VVM probe chan can operate at the higher input level.

Owen

Let me pick a nit or two.

More attenuation is not necessarily better. In theory the improvement
in source match is two times the attenuation, so a 10 dB pad improves
the return loss to no less than 20 dB, even with a zero ohm source,
and with a decent source match of RL = 10 dB or so, is as good as you
need. (I know you know this already)

I say this because it's very likely that the return loss of the
attenuator isn't any better than 25-30 dB, regardless of its
attenuation. For example Narda makes a "precision" Type N attenuator:

http://www.nardamicrowave.com/east/P...dPrecision.pdf

Note the VSWR spec, 1.15 at low frequency. That's a 23 dB RL. So
although a 20 dB pad in theory provides a minimum 40 dB RL, the actual
RL can be as little as 23 dB.

Manufacturers have to work really hard and typically use a precision
connector like 3.5mm or 7mm to build a 40 dB RL termination although
Anritsu will sell you a 40 dB RL type N termination for -only- $650
USD.

Also, and this goes back a post or two, where you suggested that if a
pad is used between the generator and the input to the coupler, the
"A" probe (reference/incident) should be between the generator and the
pad.

This is contrary to what I tried to recommend earlier when I said:

"It would also be nice it you had a 6 - 10 dB pad between the
generator and the directional coupler (DC); located right at the DC.
You want the source match to be set right there and the A probe to
sample right there."

Let me offer this thought experiment:

If you had two directional couplers, such that one could be used to
sample the forward signal and the other the reflected, would you place
a pad between them to isolate the generator from the effects of the
load?

On Tue, 03 Jan 2006 08:25:48 -0700, Wes Stewart
wrote:

....
Owen

Let me pick a nit or two.

More attenuation is not necessarily better. In theory the improvement
in source match is two times the attenuation, so a 10 dB pad improves
the return loss to no less than 20 dB, even with a zero ohm source,
and with a decent source match of RL = 10 dB or so, is as good as you
need. (I know you know this already)

I say this because it's very likely that the return loss of the
attenuator isn't any better than 25-30 dB, regardless of its
attenuation. For example Narda makes a "precision" Type N attenuator:

http://www.nardamicrowave.com/east/P...dPrecision.pdf

Note the VSWR spec, 1.15 at low frequency. That's a 23 dB RL. So
although a 20 dB pad in theory provides a minimum 40 dB RL, the actual
RL can be as little as 23 dB.

Manufacturers have to work really hard and typically use a precision
connector like 3.5mm or 7mm to build a 40 dB RL termination although
Anritsu will sell you a 40 dB RL type N termination for -only- $650
USD.

Also, and this goes back a post or two, where you suggested that if a
pad is used between the generator and the input to the coupler, the
"A" probe (reference/incident) should be between the generator and the
pad.

This is contrary to what I tried to recommend earlier when I said:

"It would also be nice it you had a 6 - 10 dB pad between the
generator and the directional coupler (DC); located right at the DC.
You want the source match to be set right there and the A probe to
sample right there."

Let me offer this thought experiment:

If you had two directional couplers, such that one could be used to
sample the forward signal and the other the reflected, would you place
a pad between them to isolate the generator from the effects of the
load?

No, of course not... it just adds another source of error, and
increases the gap between the measurements being made on both probes.

But Dan does not have two directional couplers. To my mind, if the A
probe is sampling the main transmission line, the sample is of the
resultant of the algebraic sum of the forward and reflected waves
rather than a sample of the forward wave alone (well, nearly alone) as
you would get with a directional coupler. My suggestion of placing the
A probe on the source side of the attenuator is to reduce the
contribution of the reflected wave to the A probe measurement. The
attenuator was proposed mainly for isolation of the forward wave
component for measurement, rather than a source matching issue...
which also exists.

Have I got this wrong?

All comments on practical limits of RL from pads / attenuators noted,
and understood.

Owen



--

Thanks,

I went back through the calibration procedure.

1. The meter appears to be performing as it is supposed to. The 'static readings
work repeatably and predictably. With a balanced configuration shorts and
opens read correctly.

2. The pad I was using appears to have a frequency shift. With the 'legs
balanced for length' the measurement would change with frequency. This really
confused the measurement process. I put it aside.

3. I have some 23 Ohm couplers. These of course confuse readings when used as
normal couplers. However if they are used in a truly balanced configuration -
that is the legs are configured as close to identical as possible then they
appear to work.

I am going to shelve this until I find or make a 50 Ohm bi-directional coupler.

Thanks again - Dan


Owen Duffy wrote:
On Mon, 02 Jan 2006 16:10:02 -0800, dansawyeror
wrote:


The coupler measures
about -14 dbm.


What does this mean? Are you trying to tell us that the power on the
coupler port is 14dB less than the through power? What has dBm got to
do with it?

You didn't report the power in the coupler port with a s/c and / or
o/c at the measurement plane. Did you perform this cal?


(The phase angle and reflected ware very touchy. It was almost impossible to
adjust by changing frequency. It was easier to 'adjust' it by sitting very still
and moving my arm.)


That is understandable. How much coax between the A probe and the
load, and the B probe and the load... how many degress does this total
electrical length change for a 1% change in frequency? Does that
explain some of the phase sensitivity?


The antenna is a copy from the ARRL handbook. It is a 4 inch segment, a 1 inch
long by 3/4 inch diameter 5 turn coil, and a 4 inch tip. It is mounted over a 2
foot square aluminum plate. This antenna should have an input impedance less
then 20 Ohms.


Don't you need to measure some "known" loads. Why not try a 50 ohm
load tee'd to a s/c stub (quarter wave at a known frequency) and see
if you get the predictable results at different frequencies around
resonance. Then try two 50 ohm loads in parallel with the stub. (339mm
of RG58C/U should have a Z of around 6000+j0 ohms at around 146MHz, at
half that frequency it should be 0.85+j50, etc...


How can it measure very close to 50 Ohms? Is there something wrong with this
analysis?


See if you can trust your measurements on known loads before wondering
why the unknown load isn't what you expect when using unknown
measurement technology... too many unknowns.

Owen
--

On Tue, 03 Jan 2006 08:25:48 -0700, Wes Stewart
wrote:


Owen

Let me pick a nit or two.

....

Firstly, thanks for posting AN 77-3, it is a long time since I read
it, and have little recollection of the recommended low frequency test
setup.

The HP 11549A is described as a power splitter, so I am guessing that
it is some kind of hybrid (ie as in hybrid transformer) that in that
role, whilst splitting the power to the "output" ports, will prevent
power flow between "output" ports, so isolating the A probe to some
extent from the reflected wave on the unknown load side.

Additionally, the 8491 attenuator in the load path will improve the
return loss at the B side 11549A port, so that combination seems to be
stabilising the loads presented to the splitter (which if it is a
hybrid, improves is cross port isolation), and improving the RL by the
action of the attenuator and splitter.

Without knowing the loss in the 8491 (I know they were available in
10dB, but I think there were -3, -6, -10 and -20s), or the isolation
across the splitter, it is hard to quantify the total isolation of
reflected wave from the A probe.

It may be that Dan should consider constructing a hybrid or Return
Loss Bridge, whatever you want to call it, it will be cheaper and have
less loss that a dual directional coupler for HF measurements. IIRC,
the ARRL has some simple designs.

Owen
--

On Wed, 04 Jan 2006 21:35:40 GMT, Owen Duffy wrote:

On Tue, 03 Jan 2006 08:25:48 -0700, Wes Stewart
wrote:


Owen

Let me pick a nit or two.

...

Firstly, thanks for posting AN 77-3, it is a long time since I read
it, and have little recollection of the recommended low frequency test
setup.

The HP 11549A is described as a power splitter, so I am guessing that
it is some kind of hybrid (ie as in hybrid transformer) that in that
role, whilst splitting the power to the "output" ports, will prevent
power flow between "output" ports, so isolating the A probe to some
extent from the reflected wave on the unknown load side.

Nothing so exotic

www.k6mhe.com/n7ws/HP-11549.pdf


Additionally, the 8491 attenuator in the load path will improve the
return loss at the B side 11549A port, so that combination seems to be
stabilising the loads presented to the splitter (which if it is a
hybrid, improves is cross port isolation), and improving the RL by the
action of the attenuator and splitter.

Without knowing the loss in the 8491 (I know they were available in
10dB, but I think there were -3, -6, -10 and -20s), or the isolation
across the splitter, it is hard to quantify the total isolation of
reflected wave from the A probe.

It may be that Dan should consider constructing a hybrid or Return
Loss Bridge, whatever you want to call it, it will be cheaper and have
less loss that a dual directional coupler for HF measurements. IIRC,
the ARRL has some simple designs.

Maybe, but I think Dan is trying some VHF measurements and the coupler
approach is really the way to go.


Wes

I now have use of a b-directional coupler for HF and would like to perform the
following 'tests' however I do not understand them. Can you explain the last a
different way. I don't understand "two 50 Ohm loads in parallel with the stub".

Tests of stubs are now clearly reading correct for various lengths across
frequencies. That is once the 1/4 wave is determined the next odd and even
is very predicable.

Thanks,
Dan


Don't you need to measure some "known" loads. Why not try a 50 ohm
load tee'd to a s/c stub (quarter wave at a known frequency) and see
if you get the predictable results at different frequencies around
resonance. Then try two 50 ohm loads in parallel with the stub. (339mm
of RG58C/U should have a Z of around 6000+j0 ohms at around 146MHz, at
half that frequency it should be 0.85+j50, etc...

On Thu, 05 Jan 2006 22:24:54 -0800, dansawyeror
wrote:

I now have use of a b-directional coupler for HF and would like to perform the
following 'tests' however I do not understand them. Can you explain the last a
different way. I don't understand "two 50 Ohm loads in parallel with the stub".

Parallel two 50 ohm loads with a tee piece, then use another tee to
put the stub in parallel. The tees will be imperfect, but at HF, the
impact should not be major.


Tests of stubs are now clearly reading correct for various lengths across
frequencies. That is once the 1/4 wave is determined the next odd and even
is very predicable.

Do you get the correct answers for 1/8, 3/8 wave stubs in parallel
with a 50 ohm load? (1/8 wave s/c stub has an impedance close to
0+jRo, and you need to put it in parallel with the dummy load, so
series equivalent is 25+j25. 25 ohms in parallel with 1.8 wave s/c
stub should be around 20+j10. Put a 25 ohm load at the end of a metre
of coax and check its transformation at different frequencies
(equivalent to quarter wave, eight wave etc) against your Smith Chart
prog.

They are some examples, work some out for what you have at hand.

Prove that your measurement system works on predictable loads.

Owen


Thanks,
Dan


Don't you need to measure some "known" loads. Why not try a 50 ohm
load tee'd to a s/c stub (quarter wave at a known frequency) and see
if you get the predictable results at different frequencies around
resonance. Then try two 50 ohm loads in parallel with the stub. (339mm
of RG58C/U should have a Z of around 6000+j0 ohms at around 146MHz, at
half that frequency it should be 0.85+j50, etc...


--