Tam/WB2TT wrote:

I will have to try your configuration. I did C - L - C/2 - L -C, with
C=1000PF, L=2.2uH. According to SWCad, the gain is flat above 2 MHz, but
there is 120 degrees phase shift at 4 MHz, relative to the phase at 100 MHz.

The center component should be 2C, or 2000 pF for your experiment, not
C/2. The circuit is simply two pi networks in cascade, each having all
component reactances equal to the "transmission line" Z0. Each pi
section mimics a quarter wave transmission line.

Roy Lewallen, W7EL

"Roy Lewallen" wrote in message
...
Tam/WB2TT wrote:

I will have to try your configuration. I did C - L - C/2 - L -C, with
C=1000PF, L=2.2uH. According to SWCad, the gain is flat above 2 MHz, but
there is 120 degrees phase shift at 4 MHz, relative to the phase at 100
MHz.

The center component should be 2C, or 2000 pF for your experiment, not
C/2. The circuit is simply two pi networks in cascade, each having all
component reactances equal to the "transmission line" Z0. Each pi section
mimics a quarter wave transmission line.

Roy Lewallen, W7EL

Roy,
I wasn't too clear, but I have 2 T networks back/back. That makes the center
cap C/2. I am going to run SWCad on the Pi configuration later, and see what
that does.

Tam/WB2TT

"Tam/WB2TT" wrote in message
...

Meanwhile, I measured the impedance of a Drake 100W dummy load with and
without the HPF. All readings without the HPF are within Drake spec.

FREQ NO HPF With HPF

4 Mhz 47j2 33j13
7MHz 47j2 55j3
14MHz 47j1 50j7
28MHz 48j2 47j3
50MHz 49j2 54j11
144MHz 53j11 74j36

Everything was connected with UHF adapters, and no coax was used. Ignore the
VHF readings,. as the filter was not built that carefully. Capacitors are
mica (actual values 1000, 560, 1000), and inductors appear to be 68-2
(2.2uH).

If I get a chance later today, I will rewire it into the Pi configuration
with the same inductors.

Tam/WB2TT

Tam/WB2TT wrote:
"Tam/WB2TT" wrote in message
...

Meanwhile, I measured the impedance of a Drake 100W dummy load with and
without the HPF. All readings without the HPF are within Drake spec.

FREQ NO HPF With HPF

4 Mhz 47j2 33j13
7MHz 47j2 55j3
14MHz 47j1 50j7
28MHz 48j2 47j3
50MHz 49j2 54j11
144MHz 53j11 74j36

Everything was connected with UHF adapters, and no coax was used. Ignore the
VHF readings,. as the filter was not built that carefully. Capacitors are
mica (actual values 1000, 560, 1000), and inductors appear to be 68-2
(2.2uH).

If I get a chance later today, I will rewire it into the Pi configuration
with the same inductors.

That's about what I'd expect. The increasing X with frequency is
consistent with a small amount of series stray inductance which is
unavoidable in the physical construction. It can be minimized, of
course, by careful construction.

A typical homebrew HF filter will begin becoming poor at VHF and above
due to series self inductance of the capacitors and shunt self
capacitance of the inductors, plus other effects. Component selection
and layout can help a lot, but it might be necessary to cascade a
VHF/UHF filter with the HF filter if very wideband rejection is
necessary. A network analyzer or spectrum analyzer with tracking
generator or noise generator are invaluable in solving those problems.
Of course, the more stuff you put in the path, the more you're likely to
disturb the measurement.

Roy Lewallen, W7EL

Tam/WB2TT wrote:

Roy,
I wasn't too clear, but I have 2 T networks back/back. That makes the center
cap C/2. I am going to run SWCad on the Pi configuration later, and see what
that does.

If you've cascaded two sections, you have two 1000 pF capacitors in
parallel at the center. That makes a total value of 2000 pF at that point.

Roy Lewallen, W7EL

"Roy Lewallen" wrote in message
...
Tam/WB2TT wrote:

Roy,
I wasn't too clear, but I have 2 T networks back/back. That makes the
center cap C/2. I am going to run SWCad on the Pi configuration later,
and see what that does.

If you've cascaded two sections, you have two 1000 pF capacitors in
parallel at the center. That makes a total value of 2000 pF at that point.

Roy Lewallen, W7EL

It's a high pass filter to reject the AM broadcast band. So, the two 1000 PF
caps are in series. Am I missing something?

Tam

Tam/WB2TT wrote:
"Roy Lewallen" wrote in message
...

Tam/WB2TT wrote:

Roy,
I wasn't too clear, but I have 2 T networks back/back. That makes the
center cap C/2. I am going to run SWCad on the Pi configuration later,
and see what that does.

If you've cascaded two sections, you have two 1000 pF capacitors in
parallel at the center. That makes a total value of 2000 pF at that point.

Roy Lewallen, W7EL


It's a high pass filter to reject the AM broadcast band. So, the two 1000 PF
caps are in series. Am I missing something?

Sorry, I missed that you had made a T network rather than pi.

In general, a tee network substituted for a pi will have the same
characteristics only at one frequency, but will have different transfer
and/or impedance characteristics at other frequencies. So the
substitution should be done with care if characteristics are important
at more than one frequency.

In this case, though, if you make a tee network which has the same "half
wave" characteristic as the pi at the design frequency, it'll have
identical transfer characteristics (it's got the same filter response)
and complementary impedance characteristics. That is, at frequencies
where one network has an input impedance greater than 50 ohms, the other
will have an impedance that's less, and the phase angles are the
negatives of each other. And, luckily, the transformation is simple for
this particular special case -- the T network reactances are also all
the same and also equal to the Z0 of the "transmission line". So one is
just as good as the other.

The HPF equivalent doesn't of course simulate a transmission line,
although the impedance transformation though the filter is unity at the
design frequency. Otherwise, it works in pretty much an opposite way
from the LPF.

I need to correct and clarify a couple of points I made in my earlier
posting.

The "half wave" lowpass filter simulates a half wavelength transmission
line only at and near the design frequency (where the reactances are all
the same). It doesn't do a very good job either above or below that
frequency. For a better general simulation of a *short* transmission
line, reduce the end pi or T network components to half their values.
This model improves -- in theory at least -- as more sections are added.
In practice, imperfection in the components limits the quality of the
approximation. But I don't think this is of particular interest in
making analyzer measurements. The 7 MHz example terminated with 50 ohms
will show an input impedance within 2 ohms magnitude and 2 degrees phase
of 50 ohms between about 6.2 and 7.4 MHz, so it's good for the entire 40
meter band. But it will disturb measurements on lower bands. You should
construct one for each band and, preferably, one for each general
impedance level you expect to measure. A single one won't do for
multiple bands as I implied.

Roy Lewallen, W7EL

Tam/WB2TT wrote:

It's a high pass filter to reject the AM broadcast band. So, the two 1000 PF
caps are in series. Am I missing something?

Sorry, I missed that you had made a T network rather than pi.

In general, a tee network substituted for a pi will have the same
characteristics only at one frequency, but will have different transfer
and/or impedance characteristics at other frequencies. So the
substitution should be done with care if characteristics are important
at more than one frequency.

In this case, though, if you make a tee network which has the same "half
wave" characteristic as the pi at the design frequency, it'll have
identical transfer characteristics (it's got the same filter response)
and complementary impedance characteristics. That is, at frequencies
where one network has an input impedance greater than 50 ohms, the other
will have an impedance that's less, and the phase angles are the
negatives of each other. And, luckily, the transformation is simple for
this particular special case -- the T network reactances are also all
the same and also equal to the Z0 of the "transmission line". So one is
just as good as the other.

The HPF equivalent doesn't of course simulate a transmission line,
although the impedance transformation though the filter is unity at the
design frequency. Otherwise, it works in pretty much an opposite way
from the LPF.

I need to correct and clarify a couple of points I made in my earlier
posting.

The "half wave" lowpass filter simulates a half wavelength transmission
line only at and near the design frequency (where the reactances are all
the same). It doesn't do a very good job either above or below that
frequency(*). The 7 MHz example terminated with 50 ohms will show an
input impedance within 2 ohms magnitude and 2 degrees phase of 50 ohms
between about 6.2 and 7.4 MHz, so it's good for the entire 40 meter
band. But it will disturb measurements on lower bands. You should
construct one for each band and, preferably, one for each general
impedance level you expect to measure. A single one won't do for
multiple bands as I implied.

(*)For a better general simulation of a *short* transmission line, use a
ladder network with all the reactances equal to Z0 except the end
components. For the end components, make the series L or shunt C half
the value of the rest. (For example, the LPF I showed would have input
and output shunt capacitors with reactance = 100 ohms, and remaining
components with reactance = 50 ohms. A five-component tee type network
would have input and output series inductors with reactance = 25 ohms,
and the remaining components with reactance = 50 ohms.) This model
improves -- in theory at least -- as more sections are added, being able
to imitate longer and longer lines. In practice, imperfection in the
components limits the quality of the approximation. But I don't think
this is of particular interest in making analyzer measurements. The
model I proposed is better for simulating a half wavelength line while
providing filtering.

Roy Lewallen, W7EL