"K7ITM" wrote in message
...
On Jan 29, 6:01 pm, Owen Duffy wrote:
"Suzy" not@valid wrote :

...

Thanks for that. I really want to emulate a throughline (like the Bird
43 I once had) and am thinking more in the terms of a stripline on PCB
(using BNC), with parallel lines to sniff the RF and diodes to convert
for the meters. But what I'm usure of is the sort of dimensions I
should use for 70 cms, and wther these are critical. How long should

If you make the coupling line very short wrt wavelength, you can analyse
it with a lumped constant approximation.

You can think of the coupled line as located in the electric field of the
main line, and it will have a voltage difference to ground.

Similarly, the coupled line will be cut by magnetic flux due the the
current in the main line, and so a voltage will be induced end to end in
the coupled line.

By adjusting the resistor at one end of the coupled line, you adjust the
contibution of these current and voltage derived samples, and can adjust
them so balance each other (ie no meter deflection) when V/I on the main
line is 50.

When the coupled line is short, the characteristic impedance of the
coupled line is not very critical.

You should be able to achieve sufficient sensitivity for 5W pwr on 70cm
with 20mm of coupled line.

The Zo of the though line is more important, it is the main determinant
of the insertion VSWR of the instrument, so you need to strive to achieve
close to the desired Zo, presumably 50 ohms.

Next, don't put the coupled line to close as it will load the main line
and degrade the insertion VSWR.

Then adjust the R at the end of the coupled line to null the DC output on
a 50 ohm load.

Repeat for the other coupled line (if you use one).

Check for symmetry, ie that reversing the instrument gives exactly the
same readings on the dummy load.

Tom suggested Schottky diodes. Dick Smith has 1N5711 Shottky diodes (at
exhorbitant prices), or you could get 1N34 germainium diodes from Jaycar.

BTW, you did mention the Bird 43, the above is not frequency compensated
like the Bird slugs, deflection will be frequency dependent.

Owen

PS: I have only met one VK YL ham named Susan, I sometimes wonder what
happened to her, haven't heard her on air in decades.

Owen's comments reminded me that I always used to think of coupled-
line hybrids as sampling the magnetic and electric fields at a point,
and of course you can adjust the ratio by adjusting the load as he
suggests. But then something began nagging me: his example of a 20mm
(2cm) line isn't exactly short compared with a "full length" quarter
wave coupled line. Because of the relatively slow propagation in
stripline over FR4 PC board material, a quarter wave is only about 10
cm for 450MHz signals. But thinking of it in terms of a distributed
line system, you're just terminating the coupled line in the proper
impedance to not get reflections off that end; so this same thing
works even with full 1/4 wave lines to adjust the directivity.

However, if you have one through line and one coupled line, and you've
made them very symmetric so you can swap the two and not see a
difference, then if you have to terminate the coupled line in other
than 50 ohms to get perfect directivity (no reflection from that
port), it also says that the through line is not 50 ohms. As Owen
points out, you'd really like that through line to look like 50 ohms
(or other system Z0 if you wish), so it doesn't disturb the system
it's installed in. There's the incentive to make the coupler
symmetrical and tuned so the coupled line terminates properly in 50
ohms, to get best directivity.

I'll readily admit that the details of coupled lines from a fields
perspective is a bit beyond my full understanding, so there may be an
error in my thinking about this, but I believe it's pretty accurate.
The thing that comes first to my mind is that for stripline, the
propagation velocity for even and odd modes is different, but still, I
think if the lines are lightly coupled, the paragraphs I wrote above
are a valid way to look at the situation.

Geez, Schottky diodes should be dirt cheap these days. It's germanium
that are hard to find around here, unless they are old stock.

Cheers,
Tom

I'd like to thank all you kind fellows for your assistance, but
unfortunately a lot of it is rather above my head. I need to be pointed at
an article that spells out exactly what dimensions, how to build etc. (eg is
there one in the ARRL handbook?) And Owen, no in this case Suzy if not
short for Susan, so I'm not the person you knew.

On Jan 30, 9:57 am, "Suzy" not@valid wrote:
....

I'd like to thank all you kind fellows for your assistance, but
unfortunately a lot of it is rather above my head. I need to be pointed at
an article that spells out exactly what dimensions, how to build etc. (eg is
there one in the ARRL handbook?) And Owen, no in this case Suzy if not
short for Susan, so I'm not the person you knew.

OK, back to the first posting I made here. How about if I build one,
test it, and post the design and results somewhere? As I noted in one
of my postings, if I were to make one, I'd first consider how much
power I wanted to read, full scale, so if my offer is appealing, let
me know how much power you want to measure, max. Don't tell me a
kilowatt if you really are going to use it at 10 watts, because if I
design for a kilowatt, 10 watts will be low enough that you won't be
able to read it very well. In fact, I'd propose 10 watts as a
reasonable full scale for a lot of ham uses.

Cheers,
Tom

On Jan 29, 4:26*pm, K7ITM wrote:
...
To a pretty close approximation, the coupling will go as the sine of
the electrical length, so for example if you use the design for a 20dB
coupler but make it only 1/8 wave long, the response will be about
sin(45 degrees) or .707 times as much, or 3dB lower: *it will be about
a 23dB coupler at that length. *

Actually, the sine approximation isn't all that good. Since it's not
been easy for me to find on the web and others may have similar
trouble finding it, I'll reproduce a better equation he

coupling = (Zeven-Zodd)*sin(theta) / [2*Z0*cos(theta) + j*(Zeven
+Zodd)*sin(theta)]

whe
Zeven and Zodd are the even-mode and odd-mode impedances of the
coupled line pair:
Zeven is the impedance seen when testing one line of the pair,
when both lines are driven by the same voltage.
Zodd is the impedance seen when testing one line of the pair,
when the two lines are driven by equal-amplitude but 180
degree out of phase voltages.
Z0 = sqrt(Zeven*Zodd)
theta = electrical length of the coupled line section

To find the coupling in dB, just take 20*log10(coupling).

You can get the even and odd mode impedances from a program like altc,
available at http://altc.sourceforge.net/. You may also be able to
find formulas that give them to you for common structures like
stripline and microstrip just from the dimensions and relative
dielectric constant of the substrate.

In the hopes this may be valuable info for one or more of the
lurkers...

Cheers,
Tom

"K7ITM" wrote in message
...
On Jan 30, 9:57 am, "Suzy" not@valid wrote:
...

I'd like to thank all you kind fellows for your assistance, but
unfortunately a lot of it is rather above my head. I need to be pointed
at
an article that spells out exactly what dimensions, how to build etc. (eg
is
there one in the ARRL handbook?) And Owen, no in this case Suzy if not
short for Susan, so I'm not the person you knew.

OK, back to the first posting I made here. How about if I build one,
test it, and post the design and results somewhere? As I noted in one
of my postings, if I were to make one, I'd first consider how much
power I wanted to read, full scale, so if my offer is appealing, let
me know how much power you want to measure, max. Don't tell me a
kilowatt if you really are going to use it at 10 watts, because if I
design for a kilowatt, 10 watts will be low enough that you won't be
able to read it very well. In fact, I'd propose 10 watts as a
reasonable full scale for a lot of ham uses.

Cheers,
Tom

Hi Tom

That's a kind offer, but you needn't go to all the trouble of building it.
Just a pointer at the practical design will do (but no complex theory --
over my head!).

I want to investigate various 70 cm antennas (central frequency in Australia
435 MHz). TX is switchable 5 10 20 watts. I want to standardise on BNC, and
have readouts on analogue meters (probably 1 mA movements)

"Suzy" not@valid wrote in message ...

"K7ITM" wrote in message
...
On Jan 30, 9:57 am, "Suzy" not@valid wrote:
...

I'd like to thank all you kind fellows for your assistance, but
unfortunately a lot of it is rather above my head. I need to be pointed
at
an article that spells out exactly what dimensions, how to build etc.
(eg is
there one in the ARRL handbook?) And Owen, no in this case Suzy if not
short for Susan, so I'm not the person you knew.

OK, back to the first posting I made here. How about if I build one,
test it, and post the design and results somewhere? As I noted in one
of my postings, if I were to make one, I'd first consider how much
power I wanted to read, full scale, so if my offer is appealing, let
me know how much power you want to measure, max. Don't tell me a
kilowatt if you really are going to use it at 10 watts, because if I
design for a kilowatt, 10 watts will be low enough that you won't be
able to read it very well. In fact, I'd propose 10 watts as a
reasonable full scale for a lot of ham uses.

Cheers,
Tom

Hi Tom

That's a kind offer, but you needn't go to all the trouble of building it.
Just a pointer at the practical design will do (but no complex theory --
over my head!).

I want to investigate various 70 cm antennas (central frequency in
Australia 435 MHz). TX is switchable 5 10 20 watts. I want to standardise
on BNC, and have readouts on analogue meters (probably 1 mA movements)
Oh, and no surface mount components. Eyes not good enough!

On Jan 30, 2:08 pm, "Suzy" not@valid wrote:
"K7ITM" wrote in message

...

On Jan 30, 9:57 am, "Suzy" not@valid wrote:
...

I'd like to thank all you kind fellows for your assistance, but
unfortunately a lot of it is rather above my head. I need to be pointed
at
an article that spells out exactly what dimensions, how to build etc. (eg
is
there one in the ARRL handbook?) And Owen, no in this case Suzy if not
short for Susan, so I'm not the person you knew.

OK, back to the first posting I made here. How about if I build one,
test it, and post the design and results somewhere? As I noted in one
of my postings, if I were to make one, I'd first consider how much
power I wanted to read, full scale, so if my offer is appealing, let
me know how much power you want to measure, max. Don't tell me a
kilowatt if you really are going to use it at 10 watts, because if I
design for a kilowatt, 10 watts will be low enough that you won't be
able to read it very well. In fact, I'd propose 10 watts as a
reasonable full scale for a lot of ham uses.

Cheers,
Tom

Hi Tom

That's a kind offer, but you needn't go to all the trouble of building it.
Just a pointer at the practical design will do (but no complex theory --
over my head!).

I want to investigate various 70 cm antennas (central frequency in Australia
435 MHz). TX is switchable 5 10 20 watts. I want to standardise on BNC, and
have readouts on analogue meters (probably 1 mA movements)

OK, I gave this some thought last night. I see a couple problems...

Though you could use 1mA meter movements, that puts you at a detected
power level high enough that the meter scale won't be linear in power,
assuming Schottky or germanium diode detectors. To me, having a
linear power scale is a big advantage, because then you can reasonably
accurately figure SWR without having to worry about temperature
compensation of the detectors. There's a reason that Bird power
meters use a sensitive microammeter movement. (I think I've heard
30uA full scale, but I'm not sure about that.) Anyway, that's why I
suggested using a DVM for readout.

The second problem is, if you want to implement a microstrip design,
how do you get the trace width right? If you're afraid of surface
mount parts, how will you control the trace width to +/- a fraction of
a millimeter? On 1.6mm thick PC board, assuming FR4 with a relative
dielectric constant of 4.75, you'd like to have a trace width about
2.78mm to get a 50 ohm line. If your trace is 3.5mm wide, you get a
bit under 44 ohms, and if your trace comes out 2.0mm wide, you get a
line that's almost 60 ohms. If you can do the PC board
photographically and have confidence that you can control the trace
width to within 0.1mm, that would work. If you're doing it by
scribing the copper and pulling up unwanted copper, I think you'll
have to be working under a pretty good microscope to get to much
closer than a mm of the desired width-- or maybe cut it on a milling
machine.

I suppose there's still the possibility of cutting the trace a bit
narrow on purpose and adjusting the impedance by adding a grounded
plate above the board. It could be spaced an adjustable distance away
by mounting it with threaded rods (long screws), and adjusted to make
the traces 50 ohms. But there's still the problem of making the two
(or three) all the same width. Not knowing how you might be able to
do this, I'm rather discouraged about how this would come out. Maybe
there's a better way to make the coupled lines that's easier for a
typical ham with minimal shop facilities to handle. Or maybe if there
was enough interest, someone could make some boards with guaranteed
performance.

On the positive side, I did find BNC jacks that edge-mount on PC
boards, so that part of it becomes easy at least.

Cheers,
Tom

Hello Tom

My responses**

"K7ITM" wrote in message
...
On Jan 30, 2:08 pm, "Suzy" not@valid wrote:
"K7ITM" wrote in message

...

On Jan 30, 9:57 am, "Suzy" not@valid wrote:
...

I'd like to thank all you kind fellows for your assistance, but
unfortunately a lot of it is rather above my head. I need to be
pointed
at
an article that spells out exactly what dimensions, how to build etc.
(eg
is
there one in the ARRL handbook?) And Owen, no in this case Suzy if
not
short for Susan, so I'm not the person you knew.

OK, back to the first posting I made here. How about if I build one,
test it, and post the design and results somewhere? As I noted in one
of my postings, if I were to make one, I'd first consider how much
power I wanted to read, full scale, so if my offer is appealing, let
me know how much power you want to measure, max. Don't tell me a
kilowatt if you really are going to use it at 10 watts, because if I
design for a kilowatt, 10 watts will be low enough that you won't be
able to read it very well. In fact, I'd propose 10 watts as a
reasonable full scale for a lot of ham uses.

Cheers,
Tom

Hi Tom

That's a kind offer, but you needn't go to all the trouble of building
it.
Just a pointer at the practical design will do (but no complex theory --
over my head!).

I want to investigate various 70 cm antennas (central frequency in
Australia
435 MHz). TX is switchable 5 10 20 watts. I want to standardise on BNC,
and
have readouts on analogue meters (probably 1 mA movements)

OK, I gave this some thought last night. I see a couple problems...

Though you could use 1mA meter movements, that puts you at a detected
power level high enough that the meter scale won't be linear in power,
assuming Schottky or germanium diode detectors.

**Pardon my ignorance, but isn't it just a case of using an op amp or
whatever to suit whatever meter movement I have? ANyway, I have now sourced
a 100 microamp meter (MU65) with a 3.9K resistance (sounds strange as the 1
mA one has a 210 ohm resistance.

BTW, I'n not bothered about the linear issue. I will be having two meters to
show forward and relected powers simultaneously. I don't want to clculate
actual SWR.

To me, having a
linear power scale is a big advantage, because then you can reasonably
accurately figure SWR without having to worry about temperature
compensation of the detectors. There's a reason that Bird power
meters use a sensitive microammeter movement. (I think I've heard
30uA full scale, but I'm not sure about that.) Anyway, that's why I
suggested using a DVM for readout.

The second problem is, if you want to implement a microstrip design,
how do you get the trace width right? If you're afraid of surface
mount parts, how will you control the trace width to +/- a fraction of
a millimeter? On 1.6mm thick PC board, assuming FR4 with a relative
dielectric constant of 4.75, you'd like to have a trace width about
2.78mm to get a 50 ohm line. If your trace is 3.5mm wide, you get a
bit under 44 ohms, and if your trace comes out 2.0mm wide, you get a
line that's almost 60 ohms. If you can do the PC board
photographically and have confidence that you can control the trace
width to within 0.1mm, that would work. If you're doing it by
scribing the copper and pulling up unwanted copper, I think you'll
have to be working under a pretty good microscope to get to much
closer than a mm of the desired width-- or maybe cut it on a milling
machine.

I suppose there's still the possibility of cutting the trace a bit
narrow on purpose and adjusting the impedance by adding a grounded
plate above the board. It could be spaced an adjustable distance away
by mounting it with threaded rods (long screws), and adjusted to make
the traces 50 ohms.

**But how do you check that in a workshop with no test gear?

But there's still the problem of making the two
(or three) all the same width. Not knowing how you might be able to
do this, I'm rather discouraged about how this would come out. Maybe
there's a better way to make the coupled lines that's easier for a
typical ham with minimal shop facilities to handle. Or maybe if there
was enough interest, someone could make some boards with guaranteed
performance.

On the positive side, I did find BNC jacks that edge-mount on PC
boards, so that part of it becomes easy at least.

Cheers,
Tom

Cheers

On Fri, 1 Feb 2008 06:12:14 +1100, "Suzy" not@valid wrote:

I suppose there's still the possibility of cutting the trace a bit
narrow on purpose and adjusting the impedance by adding a grounded
plate above the board. It could be spaced an adjustable distance away
by mounting it with threaded rods (long screws), and adjusted to make
the traces 50 ohms.

**But how do you check that in a workshop with no test gear?

This is called residual SWR in a reflectometer. You load it with a
known good load, and what SWR you find (or what is exhibited by the
two meters) inhabits the reflectometer itself. Then you flip it over
and apply your source into the goesoutta with the known load on the
comesinna. You then proceed to reduce the residual SWR in both
directions.

Finding a good load is another matter, and I reported one (a precision
RF resistor) with specific characteristics here last week. Consult
the thread "RF Power Resistors from Caddock."

All of $10-$20 to accomplish.

73's
Richard Clark, KB7QHC

On Jan 31, 11:12 am, "Suzy" not@valid wrote:
Hello Tom

My responses**

"K7ITM" wrote in message
....
OK, I gave this some thought last night. I see a couple problems...

Though you could use 1mA meter movements, that puts you at a detected
power level high enough that the meter scale won't be linear in power,
assuming Schottky or germanium diode detectors.

**Pardon my ignorance, but isn't it just a case of using an op amp or
whatever to suit whatever meter movement I have? ANyway, I have now sourced
a 100 microamp meter (MU65) with a 3.9K resistance (sounds strange as the 1
mA one has a 210 ohm resistance.

Ah, OK. I had assumed you wanted to use just the meters, with no
amplifier. With amplifiers, the 1mA meters will be fine. But we need
op amps that have very low offset voltage and drift--I would prefer to
set the meter full scale to correspond to around a millivolt or two of
detected DC. I suppose if you have a way to zero the offset and it
doesn't drift, that'll be OK. Then we need to make sure the amplifier
is reasonably immune to 450MHz signals floating around... This is all
"do-able" but there are some details you'll have to pay attention to.
I'd suggest using a couple of the RF power detector ICs available from
Analog Devices or Linear Technology, but we're back to surface mount
stuff again at that point.


BTW, I'n not bothered about the linear issue. I will be having two meters to
show forward and relected powers simultaneously. I don't want to clculate
actual SWR.

Well, yes, but wouldn't you want to know whether "0.1" on the meter
represented 1/10 the power of "1.0" on the meter, rather than 1/100 of
the power?? If you don't pay a little attention to the level of RF
the detector is actually detecting, you're liable to have that
problem.


....
I suppose there's still the possibility of cutting the trace a bit
narrow on purpose and adjusting the impedance by adding a grounded
plate above the board. It could be spaced an adjustable distance away
by mounting it with threaded rods (long screws), and adjusted to make
the traces 50 ohms.

**But how do you check that in a workshop with no test gear?

What do you mean "no" test gear? You'll have the directional coupler
with meters itself, and a power source. The only other thing you need
is a 50 ohm load to put on the coupler output. Is it not worth having
at least a load you can trust? With a known good load, you feed some
power through the coupler and adjust for zero indicated return; turn
the coupler around and make sure the other port also reads zero. With
an open or short load you should get equal readings on the forward and
reverse meters. (I suppose you need two couplers and a good load to
insure that the through line of the coupler is also the same impedance
as the load...) I have a good network analyzer on my bench at work,
but without a calibration load to test and calibrate it with, I don't
know how good its readings really are.

Cheers,
Tom

K7ITM wrote:
To me, having a
linear power scale is a big advantage, because then you can reasonably
accurately figure SWR without having to worry about temperature
compensation of the detectors.

Can you define what you mean by linear? Straight line?
Since we can only measure voltage and current, in order
to obtain a linear power scale from a linear meter, it
is necessary to supply some pre-display computing
ability (microcomputer).
--
73, Cecil http://www.w5dxp.com