Reg wrote
They who attempt to grasp support by stating the manufacturer's type
number
of the instruments used are most in need of the self-confidence it falsly
generates.

Hi Reg. What exactly are you talking about? I had a few minutes in
between
Hurricane Ivans wrath to get the Emergency generator cranked up and had a
chance to read this. Lucky you don't have these things in the UK.
73 Gary N4AST
============================

Gary, we have heard the news over here about the devastating Hurricane Ivan.
We get them here at about 1/2 strength of yours only once every very few
years. And even then it's only over a relatively small area.

If you can't understand what I am wittering about then its due either to the
storm stress you are under or because you are one of those suffering from
delusions of accuracy. In your case I prefer the former excuse.

I hope your generator started up OK and that you and your family suffer the
bare minimum of danger and damage. At least communications between us are
still intact. My best wishes.
---
Reg, G4FGQ

"JGBOYLES" wrote in message
...

All radio people suffer from delusions of measuring accuracy.

RF power measurements are the most inaccurate of all.

The accuracy of measurements are a function of the instrument user.


They who attempt to grasp support by stating the manufacturer's type
number
of the instruments used are most in need of the self-confidence it falsly
generates.

Hi Reg. What exactly are you talking about? I had a few minutes in
between
Hurricane Ivans wrath to get the Emergency generator cranked up and had a
chance to read this. Lucky you don't have these things in the UK.
73 Gary N4AST

Gary,

I saw an interesting curve at a resistor manufacturer's web site. It plotted
resistor error as a function of F(MHz) x R(Meg) for 1/4 W carbon resistors.
To make a long story short, the resistor error will be about 20% where the
Megahertz x MegaOhms = 1. That means the resistor value will be 1/
Frequency. So, at 30 MHz, the resistor will be in error by 20% if it is
bigger than 1/30 =.033 Meg, or 33K. That, I believe ignores capacitive
effects. Personally, I have never tried to put RF through a resistor bigger
than a few hundred Ohms.

It occurs to me that you can ignore capacitive effects if you make all
resistors identical. For instance, if you want a 3:1 divider make the series
resistor 10K, and the shunt resistor two 10K resistors in parallel. Of
course, you will need a high impedance load on it. Let's see if anybody
shoots this down.

Tam

On Thu, 16 Sep 2004 21:26:48 -0400, "Tam/WB2TT"
wrote:

To make a long story short, the resistor error will be about 20% where the
Megahertz x MegaOhms = 1.

And not so curiously Trc = 1 MOhm · 1 pF = 10^-6
F = 1 / T
F = 1 MHz
perhaps the product rule should be:
Megahertz x MegaOhms x picoFarads = 1

The 20% error is, of course, simply the rolloff response at the RC
inflection point described by 1/Trc.

Let's see if anybody
shoots this down.

Hi All,

I think chipping at the clay feet of saints is more appropriate
metaphor.

What is the saint? The RF response of the resistor. It should be
suspect right out the gate. Being suspect, you employ the
conventional techniques already evidenced even by the cheapest Power
Meter builder (MJF) by swamping the stray capacitance with series
capacitors (paralleling the resistors). One capacitor is either
variable, or further paralleled with a trimmer. The saint is also the
unspecified requirement: is this divider BEFORE OR AFTER the detector?

If before, and thus subject to RF, the simple RC compensated divider
has served for eons. If after, and thus subject to only DC - who
cares? The one clay foot of the discussion.

The other clay foot of the discussion is that for placement before OR
after the detector, ALL ratios are post-hoc determinations (in other
words, design with variable components fully expecting you WILL be
wrong). Further, ALL descriptions to this point have been of
normalized levels. With the RC compensated divider, you are throwing
the knee if rolloff into lower frequencies so that ALL frequencies of
interest reside on the same slope. Hence the common "calibration"
procedure has you adjust the resistors for the low frequency readout,
and the capacitors at the high frequency readout. This "calibration"
is simply distributing the error so that it doesn't accumulate
outrageously.

The greater challenge is how do you know how much power you are
setting your meter to read? Compounding errors are common in RF.

73's
Richard Clark, KB7QHC

"Richard Clark" wrote in message
...
On Thu, 16 Sep 2004 21:26:48 -0400, "Tam/WB2TT"
wrote:

To make a long story short, the resistor error will be about 20% where the
Megahertz x MegaOhms = 1.

And not so curiously Trc = 1 MOhm · 1 pF = 10^-6

I think the curve ignores C, and is based on skin effect only. There is no
explanation for the data.


F = 1 / T
F = 1 MHz
perhaps the product rule should be:
Megahertz x MegaOhms x picoFarads = 1

Go to http://www.xicon-passive.com/resistor.html and click on CC. There is
also info on resistor performance vs frequency in the W6SAI book. He shows
curves for 5 different carbon resistors vs frequency without identifying the
resistor values. As a gross average, they show about 50% error at 15 MHz.

The 20% error is, of course, simply the rolloff response at the RC
inflection point described by 1/Trc.


The curve goes from 0 - 100. I arbitrarily picked 20 % as being a point
where there is apreciable error.

Tam/WB2TT

Let's see if anybody
shoots this down.

Hi All,

I think chipping at the clay feet of saints is more appropriate
metaphor.

What is the saint? The RF response of the resistor. It should be
suspect right out the gate. Being suspect, you employ the
conventional techniques already evidenced even by the cheapest Power
Meter builder (MJF) by swamping the stray capacitance with series
capacitors (paralleling the resistors). One capacitor is either
variable, or further paralleled with a trimmer. The saint is also the
unspecified requirement: is this divider BEFORE OR AFTER the detector?

If before, and thus subject to RF, the simple RC compensated divider
has served for eons. If after, and thus subject to only DC - who
cares? The one clay foot of the discussion.

The other clay foot of the discussion is that for placement before OR
after the detector, ALL ratios are post-hoc determinations (in other
words, design with variable components fully expecting you WILL be
wrong).

So true. I notice the series C in the Kenwood meter is variable.

Further, ALL descriptions to this point have been of
normalized levels. With the RC compensated divider, you are throwing
the knee if rolloff into lower frequencies so that ALL frequencies of
interest reside on the same slope. Hence the common "calibration"
procedure has you adjust the resistors for the low frequency readout,
and the capacitors at the high frequency readout. This "calibration"
is simply distributing the error so that it doesn't accumulate
outrageously.

The greater challenge is how do you know how much power you are
setting your meter to read? Compounding errors are common in RF.

73's
Richard Clark, KB7QHC

Oh, yea! I forgot the bit about Peak...silly me.
73, Steve K9DCI

"JGBOYLES" wrote in message
...
E = Root (P*R)

Scaled V is V / 273 but you'll have to go further to stay in the
dynamic
range of the multiplier. V^2 should be around 10V or whatever the mult
can
output.

Thanks for checking my calcs. Steve. I had to do what your spreadsheet
did by
hand. I should note that since I have to convert the voltage to the
multiplier
to DC, at 1500 watts we are working with 273*SQRT 2 or 386 volts. I size
the
divider so that 386 this gives 10.0 volts to the multiplier. With a
dual
polarity 15VDC supply, the multiplier has enough dynamic range.
73 Gary N4AST

On Fri, 17 Sep 2004 10:35:25 -0400, "Tam/WB2TT"
wrote:
I think the curve ignores C, and is based on skin effect only. There is no
explanation for the data.

Skin effect would tend to increase resistance which contradicts the
trend. As for explanation:

From "Electronic Components and Measurements," Wedlock and Roberge,
1969:
"At high frequencies the performance of a resistor will depart
from Ohm's law because of stray capacitance and lead inductance."
[pg. 77]
There is an identical curve to your reference shown in Figure 7.4,
same page:
"Change in resistance of a ½ Watt carbon-composition resistor as a
function of frequency. Frequency in MHz times resistance in
Megohms"
In Chapter 18 "RF Impedance Measurements":
"Such behavior is often termed stray capacitance or stray
inductance. Because these effects are usually undesirable and
serve to limit the high frequency performance of components, they
are also called parasitic effects." [pg. 276]

However, my expression of this being rolloff was too simplistic as the
curve does not follow the typical 10dB/Decade characteristic. Rather,
it shows a 6dB/Decade+. Some of this may be accounted for in lead
reactance, but at the Megohm scale this is inconsequential for
conventional leads.

73's
Richard Clark, KB7QHC

The other clay foot of the discussion is that for placement before OR
after the detector, ALL ratios are post-hoc determinations (in other
words, design with variable components fully expecting you WILL be
wrong). Further, ALL descriptions to this point have been of normalized
levels.

Hi Richard, I haven't been able to keep up with this like I wished because of
that pesky Hurricane. If you put the detector circuit before the voltage
divider, then the resistors see DC which they are a lot happier with. The
detector diode will have to be 700VDC PRV rating, and the filter cap. will have
to be sized properly.
I guess the diode will have some frequency dependent properties, but as long
as it still acts like a diode, and the forward bias drop is around .6V it ought
to work. This looks like good alternative to frequency dependent resistors.
What say you?


73 Gary N4AST

FYI
I have used either 470 or 510 ohm carbon comp resistors way back when as
20dB probes for a sampling scope and Spectrum analyzer work. If I recall,
they were quite comparable to the scope's regular probes for pretty fast
digital signals.
I would, however, get nervous with the high values stated / needed here.
At first I was thinking I'd series-up 500 or 1K resistors if I had to do
this, but then strays start to become significant as has been well
discussed.

If you have the equipment, I like the frequency response / sweep idea. This
way you may even be able to adjust the compensating caps for best high freq
(10M) response. You could also use the calibrator on most good scopes to
start out....
73,

--
Steve N, K,9;d, c. i My email has no u's.

"Richard Clark" wrote in message
...
On Fri, 17 Sep 2004 10:35:25 -0400, "Tam/WB2TT"
wrote:
I think the curve ignores C, and is based on skin effect only. There is
no
explanation for the data.

Skin effect would tend to increase resistance which contradicts the
trend. As for explanation:

From "Electronic Components and Measurements," Wedlock and Roberge,
1969:
"At high frequencies the performance of a resistor will depart
from Ohm's law because of stray capacitance and lead inductance."
[pg. 77]
There is an identical curve to your reference shown in Figure 7.4,
same page:
"Change in resistance of a ½ Watt carbon-composition resistor as a
function of frequency. Frequency in MHz times resistance in
Megohms"
In Chapter 18 "RF Impedance Measurements":
"Such behavior is often termed stray capacitance or stray
inductance. Because these effects are usually undesirable and
serve to limit the high frequency performance of components, they
are also called parasitic effects." [pg. 276]

However, my expression of this being rolloff was too simplistic as the
curve does not follow the typical 10dB/Decade characteristic. Rather,
it shows a 6dB/Decade+. Some of this may be accounted for in lead
reactance, but at the Megohm scale this is inconsequential for
conventional leads.

73's
Richard Clark, KB7QHC

On 17 Sep 2004 20:36:21 GMT, (JGBOYLES) wrote:

What say you?

Hi Gary,

For a diode, the 1N4007 is rated at 1000V. However, you have to
design with Peak voltage in mind, and then add 50% safety factor for
good design. Further, feeding a capacitive filter requires you DOUBLE
the PIV rating. All in all, this suggests your design choice should
tend toward the divider before the detector. Then the question
becomes, do you want a peak reading meter, or an averaging meter? A
peak reading detector (AKA Clamp) will lightly load the divider
whereas the averaging will load it more (and ruin any fixed divider
ratio - which returns us to the variable component to be designed in).

73's
Richard Clark, KB7QHC

On Fri, 17 Sep 2004 16:25:56 -0500, "Steve Nosko"
wrote:

At first I was thinking I'd series-up 500 or 1K resistors if I had to do
this, but then strays start to become significant as has been well
discussed.

Hi Steve,

This would probably improve the parasitic capacitance while increasing
the parasitic inductance. Off hand, I think the inductance would
probably be tolerable in the HF.

73's
Richard Clark, KB7QHC