Reconciling T_TUNER.EXE and T-Network Tuner of W9CF
Richard
Parameters:
Frequency: 7.059 Mhz
R load: 1485.5 Ohms
X load (jx) 1455.3 Ohms
C1 max: 950 pF
C1 Q : 1500
C2 max : 950 pF
C2 Q : 1500
L Q : 300
Input resistance: 50 Ohms
W9CF's T-Network on automatic tune gives:
C1: 61 pF
C2: 950 pF
L : 9.70 uH
I was just trying to get T_Tuner to read same figures, but I don't seem
to be able to. I change "A", but no joy.
I know, the T network tuner offers such flexibility in settings, and I
guess it's hard to pin down the best settings. But, I just wanted to
try to match the settings with both these programs.
In any event, should I always seek to set C2, capacitor connected to
antenna, to the highest value possible?
TIA. Rich.
Richard
"Richard" <nearlyne...@ntlworld.com> wrote in message
news:b8dup1$8rok4$1...@ID-150848.news.dfncis.de...
> I cannot get the same results with the same parameters.
>
> Parameters:
>
> Frequency: 7.059 Mhz
That should be 7.050 Mhz.
Richard
"Richard" <nearlyne...@ntlworld.com> wrote in message
news:b8dup1$8rok4$1...@ID-150848.news.dfncis.de...
> W9CF's T-Network on automatic tune gives:
>
> C1: 61 pF
> C2: 950 pF
> L : 9.70 uH
> In any event, should I always seek to set C2, capacitor connected to
> antenna, to the highest value possible?
Actually I notice that endfeed.exe for the L Network gives:
\ | /
TX___C1_______________|
|
|
Shunt L
|
|
GND
C1: 60pF
L : 10uH
So, best really to just have an L Network set up on 7.050 Khz.
Richard
Still don't know why I cannnot reconcile T_TUNER.EXE and T-Network Tuner
of W9CF on 7.050 Mhz, but on the issue of L tuner or T tuner, clearly
sometimes the L tuner is the best circuit and the extra capacitor in the
T tuner is just "getting in the way". But, of course, you may have to
use T tuner if you wanted to match a wide frequency range.
Given the results of endfeed.exe it looks like :
C1: 61 pF
C2: 950 pF
L : 9.70 uH
is the best T tuner match, because it comes closest to the optimal
circuit, which is the L tuner with L: 10 uH, C: 60pF. (Me thinks).
JGBOYLES
>antenna, to the highest value possible?
In this instance yes.
>I was just trying to get T_Tuner to read same figures, but I don't seem
>to be able to. I change "A", but no joy.
I just did a quick look on the Smith Chart, and I got a shunt inductance of
7.6uh, and a series capacitance of 60 pf for an L network, and if a T then the
other C would be set at maximum.
Part of the problem is you are trying to match a very high impedance to 50
ohms. On the Smith Chart, the curve for a shunt inductance, starting at the
point of your antenna's Z is almost concentric to the Z=50 ohm circle. It
intersects Z=50 ohms at a very small angle, so it makes the determination of
the shunt inductance difficult. I guess W9CF's and Reg's calculating engines
have similar difficulties.
I have compared the two programs before, with a more reasonable antenna
impedance, and found them to be in agreement.
73 Gary N4AST
Richard
"JGBOYLES" <jgbo...@aol.com> wrote in message
news:20030426164903...@mb-m28.aol.com...
Yep, I think that is partly a nuisance in more ways than one, that my
antenna is pretty much an exact halfwave on 7.050 Mhz.
A T network I suppose can be tricky if there are several settings that
can give the same kind of match. Somewhow W9CF's program zero's in on I
suppose the best, if indeed there is one best.
Like you probably did at the time, I was just trying to get the same
result with both programs, sort of taking the view, sensibly or not,
that W9CF's program was giving the best match of the two programs.
Heck, I don't know what to do with "A" in Reg's program. I thought if I
can get the same match, I'd note what "A" turned out to be and I'd use
it thereafter with T_TUNER.EXE. But, I could not seem to alter "A" to
get the figures W9CF's program gave. All the other parameters were
identical (except I don''t know what input resistance is supposed to be
with W9CF's program, obviously 50 or 75 Ohms).
So, I've still not reconciled the two programs.
I think C2 is max rotational travel in this case. Now there are two
senses of max, a general rule that C2 should always be turned to max
capacitance (end stop of C2), or a general rule that C2 should be at max
possible value, constrained by circuit demands, which could be anywhere
from a few pF to end stop max. In this case before us, C2 should be
right to end stop max, max rotational travel. But in other cases the
max that can be tolerated might be when C2 is mid position (say). But
whatever the case, whether C2 max is limited by physical endstop or
circuit demands, C2 should be at either kind of maximum. (I think).
As I say, in this case it's best if C2 was not there, and we just had an
L network. But, that does not mean the T network is redundant, it has
it's uses or advantages in certain situations.
Owen Duffy
Kevin's 'Autotune' function does not find the lowest loss match for the
values you provided. If you play with the thing, you will get lower loss
with values like 85pF / 6.3uH / 16pF. It seems to me that his 'Autotune'
algorithm will alows set one of the capcitors to maximum and find the
other values for match, but in this case lower loss match is achieved
with both capacitors well below maximum. Whilst the difference is
performance in this case is minor, it does indicate that the
conventional wisdom that you will not get lowest loss unless one
capacitor is at maximum does not apply at least to the simple way in
which the circuit losses are modelled in Kevin's calculator.
Interestingly, Reg's calculator will not use the maximum C in your
example, but does show a quite low loss config with low values of A at
85pF / 6uH / 13pF, and the match and loss figures are pretty consistent
with Kevin's calculator when set manually.
I am embarking on writing a set of library functions for Excel, for T
and pi configurations (which will also include the simpler L
configuration, and a bit bewildered by an analytical solution, will try
an iterative one. I note one of the purveyors of auto tuners claiming
multiple algorithms, I don't know whether this is hype, done for speed
or whether some algorithms don't converge on a solution for some
impedances, where other algorithms may.
I would be interested in any thoughts.
Owen
Richard Clark
wrote:
>
>I am embarking on writing a set of library functions for Excel, for T
>and pi configurations (which will also include the simpler L
>configuration, and a bit bewildered by an analytical solution, will try
>an iterative one. I note one of the purveyors of auto tuners claiming
>multiple algorithms, I don't know whether this is hype, done for speed
>or whether some algorithms don't converge on a solution for some
>impedances, where other algorithms may.
>
>I would be interested in any thoughts.
>
>Owen
Hi Owen,
This sounds like the classic problem of the solution becoming trapped
in a local minima. There are a number of standard tests to break free
(look-aheads and look-behinds) that are described in the Calculus
texts if you wish to stick with conventions. This treatment often
occurs early in the text concerned with derivatives. There are also a
number of genetic-algorithm type approaches that attempt to do the
same thing holistically; however, as far as I am concerned, that
avenue is strictly academic, and having been academic for nearly two
decades suggests it doesn't really work that effectively in the first
place.
In those days when computers ran in the 10's of MHz, it paid to be
smart (but smart was not always achievable). Today with machines
running at 100 times that speed with 1000 times as much memory, brute
force has a satisfying way of dealing with what were "intractable"
problems.
73's
Richard Clark, KB7QHC
Richard
"Owen Duffy" <ow...@oda.com.au> wrote in message
news:3EAB48B9...@oda.com.au...
> Kevin's 'Autotune' function does not find the lowest loss match for
the
> values you provided. If you play with the thing, you will get lower
loss
> with values like 85pF / 6.3uH / 16pF. It seems to me that his
'Autotune'
> algorithm will alows set one of the capcitors to maximum and find the
> other values for match, but in this case lower loss match is achieved
> with both capacitors well below maximum. Whilst the difference is
> performance in this case is minor, it does indicate that the
> conventional wisdom that you will not get lowest loss unless one
> capacitor is at maximum does not apply at least to the simple way in
> which the circuit losses are modelled in Kevin's calculator.
>
> Interestingly, Reg's calculator will not use the maximum C in your
> example, but does show a quite low loss config with low values of A at
> 85pF / 6uH / 13pF, and the match and loss figures are pretty
consistent
> with Kevin's calculator when set manually.
Okay thanks for that information. I did not know what Kevin's program
was doing. I thought maybe calculating *the best" match, but not so.
Perhaps his page should state that the autotune function will find *a*
match, but not necessarily *the best* match. I think folks like me who
don't know a whole lot, will tend to assume autotune is giving the
single best match, and seek to set their tuner accordingly.
Of course, this is fairly academic because the differences in losses
will be insignificant between good matches.
Also, Reg's program, it's more thinkable that it is not necessarily
calculating the best match because you have to mess with "A". Perhaps
that is a bit like turning the controls with Kevin's program, but then
again maybe not. I must look at Reg's program and see how to use it,
I'm not sure how to right now.
It must be said, one wonders how these programs help that much. Take
Reg's program. I mess with "A". Well it seems just an academic
exercise, unless "A" relates to something that is knowable about my
components/system. Of course these programs do help, but I am unsure
exactly how. If I mess with "A", lets say I can get the best match.
Well, that's fine, but the calculated values are meaningless unless the
value of "A" is an actual parameter of my system. IOW, I don't think
either program can give you the values for the best match for your
system, Kevins probably because it sets C2 to max, and Reg's because "A"
is not derived from your components or system. It's just some figure you
mess with. But of course, even capicitor Q is all a guess anyway. So,
it's clear the values you get you should just treat them as in the
ball-park. But I'm not even sure if they can produce an "in the
ball-park" figure for one's system one is contemplating building (tuner
and antenna), because of the unknowable parameter that effects the
computed values. In Reg's program it's "A". You don't know what "A" is
in your system.
I could be talking nonsense, if I am, I'm sure someone will put me
right.
Seems like there is still an opening for some academic work to be done
on the T network problem regarding an algorithm that will produce the
best match. That is if it's solvable.
Reg Edwards
> on the T network problem regarding an algorithm that will produce the
> best match. That is if it's solvable.
Richard, T-network analysis was taken to its ultimate conclusion long before
Marconi.
But before you start what is YOUR definition of "The best match" ?
---
Reg
Richard
"Reg Edwards" <g4fgq...@btinternet.com> wrote in message
news:b8gepo$j8f$1...@titan.btinternet.com...
> But before you start what is YOUR definition of "The best match" ?
Because it seems that any of the suitable matches are likely to have
insignificant practical differences in terms of strength of signal
recieved at either destination or receiving end, I'm not that hung up
about what might be *the* best match. So, in a sense, we can say to
ourselves don't worry about whether the match is the best loss-wise,
they are all okay from that angle.
However, when it comes to designing your antenna system that is a
different kettle of fish, because what interests you are component
values. Although even here there is no big problem, because you set the
component values and the programs tell you whether you can get a match
or not with the values given. If you cannot get a match then you need
to get a new inductor or capacitor, so eventually you get a match with
the new component values. So that's the usefulness of the programs, just
in telling you whether you can get a match with the components you have,
or they tell you what components you might have to buy. The programs do
work out the match values, and take into consideration the component
values. Lets face it, a program need not take into consideration the
value of your components, it could just work out what the values should
be, but that might not be as useful.
Anyhow you have components to hand at C1 & C2 950pF; L 30 uH. You want
to transmit on 7050 Khz.
You use W9CF's program and it tells you can get a match: Use C1: 61
pF;C2: 950 pF;L : 9.70 uH.
You say, that's great, my components will tune up, I'll look to set my
components to around those values. Now with W9CF's program you initially
think, wow I need to use all 950 pF of C2 capacitance, glad I got 950
pF. But of course you could set C1 and C2 to less than 950 pF and I'm
sure you would get a match.
Then you stumble on Reg's program and it tells you that you can also
match. You set "A" to 1.
to get: C1: 83.8 pF; C2: 13 pF; L : 6.08 uH. Minimum losses in network.
When you see Reg's values you say at first, I don't understand it, the
values are so different. Then after getting to know not to bother about
that, you say, wow I can get by with capacitors under 100 pF. Also you
think, how can I get 13 pF for C2. Also, with "A" being 1, C2 is
probably the maximum; change "A" to 1.5 and C2 goes to 6.3pF. Actually
Reg's values are unrealisable.*
Okay so which is best? Well from received signal strength angle it does
not matter. And perhaps it does not matter heat-wise in the components.
But, W9CF's values are realiseable, whilst Reg's are not.
So a significant difference has resulted from using W9CF's program and
Reg's. At least it looks that way intially. Only because C2 with Reg's
program is not realiaeable. Well, might not be.
* Of course you can seek to find a different match combination manually
in practice, or with W9CF's program. Or alter your antenna length.
The thing is with Reg's program, even though you specfy C1 and C2 as 950
pF the program does not seek to use the max value of C2. It just comes
out with what the best match is. With W9CF's, that program seems to
want to use the full value of C2 in autotune. I think that might lead
to more losses, but again probably is insignificant signal strength-wise
or any other wise. Although you might have to use a value of L that's
just bigger than what you got. :c) But, this is autotune, and perhaps
it's best you should mess manually with the controls.
Summary of sorts:
So, it seems using T_TUNER, it gives a really efficient match, but you
cannot realise C2, unless you change antenna length a tad. That is the
significant difference from using W9CF T Tuner and T_TUNER.EXE.
-------
"A" is variable from 1 to 20. For a given Z match, when A = 1 capacitor
values are largest and power loss is least. As "A" increases so does
coil inductance, capacitor sizes decrease and, most important,
component power losses increase. When the program user sets a
limit to capacitor size the program automatically increases "A"
and the inductance value such that a Z-match is obtainable with
smaller capacitors. But a compromise versus power efficiency
must be made.
As per Reg.
-------
Richard
> Summary of sorts:
>
> So, it seems using T_TUNER, it gives a really efficient match, but
> you cannot realise C2, unless you change antenna length a tad. That is
the
> significant difference from using W9CF T Tuner and T_TUNER.EXE.
I think also that the values that W9CF's program gave in autotune, and
what Reg's program gave are not reconsilable. Because W9CF wants to set
C2 to max.
Wes Stewart
wrote:
[snip]
|
|I am embarking on writing a set of library functions for Excel, for T
|and pi configurations (which will also include the simpler L
|configuration, and a bit bewildered by an analytical solution, will try
|an iterative one. I note one of the purveyors of auto tuners claiming
|multiple algorithms, I don't know whether this is hype, done for speed
|or whether some algorithms don't converge on a solution for some
|impedances, where other algorithms may.
|
|I would be interested in any thoughts.
|
Before reinventing the wheel, you might look at xlzizl.xls found at:
It will find a lower loss solution that Kenvin's.
Another point; the component Q's given in the example are a bit
optimistic in my mind.
Mark Keith
> senses of max, a general rule that C2 should always be turned to max
> capacitance (end stop of C2), or a general rule that C2 should be at max
> possible value, constrained by circuit demands, which could be anywhere
> from a few pF to end stop max. In this case before us, C2 should be
> right to end stop max, max rotational travel. But in other cases the
> max that can be tolerated might be when C2 is mid position (say). But
> whatever the case, whether C2 max is limited by physical endstop or
> circuit demands, C2 should be at either kind of maximum. (I think).
>
> As I say, in this case it's best if C2 was not there, and we just had an
> L network. But, that does not mean the T network is redundant, it has
> it's uses or advantages in certain situations.
There are cases where I prefer the L network, but the T network is not
too bad as long as you are careful tuning. In real world power
measurements, I found using the least inductance seemed to be the most
important for the least loss. In about 95% of cases where least coil
is used, C2 will be at maximum. So they tend to go hand in hand. But I
see a few cases , probably with severe hi Z mismatches where C2 will
not be all the way at max. Pretty close though. And you are right, a
T net tuner will tune a certain load in about 29 positions. But here,
I found carefully using the least inductance needed is a pretty sure
way to have the least amount of loss.
Using settings that didn't use least inductance, I noticed losses of
up to about 20 %. When using least inductance, and "usually" max C2,
the losses were nearly unreadable on a garden variety watt meter. In
this case, C1 can end up about anywhere...Usually about in the middle
of the range. MK
Ian White, G3SEK
>There are cases where I prefer the L network, but the T network is not
>too bad as long as you are careful tuning. In real world power
>measurements, I found using the least inductance seemed to be the most
>important for the least loss.
That will always be true for the T-network. Always start from lowest
inductance, and increase L in small steps until you can find a match.
Then stop.
>In about 95% of cases where least coil is used, C2 will be at maximum.
>So they tend to go hand in hand.
Again, true. If you can use an L-network, it will always have lower loss
than a T (made from the same components). On the other hand, the T has a
wider matching range.
This points to an interesting solution devised by G3LNP: he modified an
existing T tuner by mounting a simple cam switch on each capacitor
shaft, which would short out the capacitor when turned to maximum. That
gives the options of a normal T-network, plus two different
configurations of L-network by shorting either C1 or C2.
The operating instructions are: use one of the L configurations if you
can, and if neither will match, switch to the T and use the lowest
possible inductance.
The only slight disadvantage is that the L-network requires a precise
value of L, as well as a precise value of C, so it's more suited to a
roller coil than to a switched inductor. On the other hand, the
switchable arrangement offers lots of options for homebrewers. If you
don't want to use cam switches, you could short out C1 and C2 by
separate switches. Also, you could plan a switched inductor that is more
finely adjustable than is normally necessary in a T-network.
--
73 from Ian G3SEK 'In Practice' columnist for RadCom (RSGB)
Editor, 'The VHF/UHF DX Book'
http://www.ifwtech.co.uk/g3sek
Richard Clark
<G3...@ifwtech.co.uk> wrote:
>
>The only slight disadvantage is that the L-network requires a precise
>value of L, as well as a precise value of C, so it's more suited to a
>roller coil than to a switched inductor. On the other hand, the
>switchable arrangement offers lots of options for homebrewers. If you
>don't want to use cam switches, you could short out C1 and C2 by
>separate switches. Also, you could plan a switched inductor that is more
>finely adjustable than is normally necessary in a T-network.
>
>--
>73 from Ian G3SEK
Hi Ian;
Sorry for posting through you, but it seems this discussion about
tuners is for - uh - tuners.
Didn't this all start out as a fixed match for a dedicated radiator?
I see a lot of hand wringing over minutia that certainly might plague
an operator in the shack, but really, this needs only to be done once,
dipped in Resin and painted black (who really needs a box for this?).
It seems the grief introduced to the topic is in an exponential
relationship to the amount of software introduced.
However, I do like the self shorting caps. Imagine finding that "just
enough" capacitance for the perfect match in the last few micrometers
separation before the contacts close. Imagine trying to find it
again. ;-)
Reg Edwards
> Kevin's 'Autotune' function does not find the lowest loss match for the
> values you provided. If you play with the thing, you will get lower loss
> with values like 85pF / 6.3uH / 16pF. It seems to me that his 'Autotune'
> algorithm will alows set one of the capcitors to maximum and find the
> other values for match, but in this case lower loss match is achieved
> with both capacitors well below maximum. Whilst the difference is
> performance in this case is minor, it does indicate that the
> conventional wisdom that you will not get lowest loss unless one
> capacitor is at maximum does not apply at least to the simple way in
> which the circuit losses are modelled in Kevin's calculator.
>
> Interestingly, Reg's calculator will not use the maximum C in your
> example, but does show a quite low loss config with low values of A at
> 85pF / 6uH / 13pF, and the match and loss figures are pretty consistent
> with Kevin's calculator when set manually.
>
> I am embarking on writing a set of library functions for Excel, for T
> and pi configurations (which will also include the simpler L
> configuration, and a bit bewildered by an analytical solution, will try
> an iterative one. I note one of the purveyors of auto tuners claiming
> multiple algorithms, I don't know whether this is hype, done for speed
> or whether some algorithms don't converge on a solution for some
> impedances, where other algorithms may.
>
> I would be interested in any thoughts.
Owen and Richard,
Okay, so I'm thinking -
I was unaware of the existence of W9CF's T-net calculator until Richard
introduced it to this thread. I like it. Lot of work went into it.
Apart from arbitrary options there are other reasons why the two calculators
do not exactly agree. (Disagreement is magnified by Richards unfortunate
choice of an almost exact 1/2-wavelength antenna.)
The arbitrary "A" input parameter does in fact vary the network operating Q.
I deliberately labelled it "A" because Q for a T-network is not well defined
and the various opinions would have led to the usual nitpicking, protracted,
distracting, confusing, arguments.
There's only one coil in the network and so "A" (approx=Q) actually selects
the value of the inductance which then sets the network's working Q.
A fundamental property of an impedance matching network is that the smaller
the circuit working Q the greater its power efficiency. So a sensible
initial guess for "A" is 1.
But another property of the circuit, for an impedance match to be possible,
working Q MUST BE GREATER than a value determined by the ratio of the
terminating impedances. So when the ratio of the terminating impedances is
very high a high Q and low efficiency are inevitable.
But we cannot continue decreasing circuit Q and the inductance value for
ever - the high-voltage capacitor values C1 and C2 would become too large
to fit into the tuner box or the bank-balance would go into the red. So the
program operator is allowed the option of setting the maximum allowable
capacitor values to the largest values he can manage to salvage from his
junk box.
If program T_TUNER, with a working Q of unity (a small inductance), needs a
capacitor value greater than the highest allowable value for C1 and C2 then
the program is obliged to increase the number of turns on the coil in order
to obtain an impedance match. This is exactly what the tuner-builder would
do in his shack.
So T_TUNER arrives at the greatest possible tuner efficiency allowed by the
restriction on capacitor size. Everything can be reduced to a problem in
economics and statistics.
Actual calculated efficiency requires intrinsic Q's of coil and capacitors
to be entered in the program. And this complicates the circuit by
associating a loss resistor of calculable value with each L and C. This
changes the input and output impedances of the network. So the network no
longer exactly matches its terminations. In effect, when internal losses are
taken into account, it is the terminations which change relative to the
network so the program has to do more calculations before presenting its
results. These more exact values will differ from other programs and will
also differ from tables of L and C component tables which appear in the
handbooks.
In practice, of course, discrepances of 15 or 20 percent are of no practical
significance when a tuner is used in conjunction with the uncertainties of
HF antennas, feedlines, baluns and SWR meters. Much Ado About Nothing!
==========================
I once had a tuner program which worked in reverse. For given network L and
C values program calculated what the input impedance of transmission line or
antenna must be when SWR=1. In the absence of a true bridge, in conjunction
with a tuner of which the component values were known, the program plus
tuner was used to 'measure' experimental antenna and counterpoise
impedances.
==========================
Automatic tuners do not work by measuring antenna impedance, calculating
T-network L and C values, and then picking the actual components from a box
of all possible values.
They work in the same way as a methodical impedance bridge operator who
knows what he's looking for - minimum magnitude and phase of the bridge
output voltage. The bridge in the tuner is EXACTLY the same as in an SWR
meter. Bridge output is split between magnitude and phase detectors.
Electric motors or relays vary, or vary in steps, the L and C values.
Computing algorithms needed for motor and relay control must be dedicated in
every detail to the particular tuner. Logical rather than mathematical
operations apply. In the case of relays to reduce hopping-about-time there
could possible be some pseudo-random logic.
---
Reg, G4FGQ
OC_CAM's Razor
> I see a lot of hand wringing over minutia that certainly might plague
> an operator in the shack, but really, this needs only to be done once,
> dipped in Resin and painted black (who really needs a box for this?).
So a vertical in Bismark, ND has the same impedance in the middle of winter
as it does in the middle of summer??? Methinks you have been living in
good weather for too long. :-) Bismark winters Vs Bismark summers are
definitely not "minutia". Try "roughing it" in Bismark in February and
get back to us (if you survive). :-)
--
73, Cecil http://www.qsl.net/w5dxp
-----= Posted via Newsfeeds.Com, Uncensored Usenet News =-----
http://www.newsfeeds.com - The #1 Newsgroup Service in the World!
-----== Over 80,000 Newsgroups - 16 Different Servers! =-----
Richard
> > So, it seems using T_TUNER, it gives a really efficient match, but
> > you cannot realise C2, unless you change antenna length a tad. That
> > is the significant difference from using W9CF T Tuner and
> > T_TUNER.EXE.
>
> I think also that the values that W9CF's program gave in autotune, and
> what Reg's program gave are not reconsilable. Because W9CF wants to
> set C2 to max.
You have to understand that in the beginning I saw two programs, each
that would calculate *the* values required. They did not calculate the
same *the* values. Could not understand that at first. So seeks to
figure out if both programs can be reconciled, ie, made to obtain the
same values, by keeping Kevin's autotune figures and adjusting something
in T-Tuner. Finds that I cannot reconcile the two programs, the
values, because there is nothing in T_TUNER.EXE that I can alter apart
from "A" and adjusting "A" does not alter the values to match Kevin's
autotune values.
Why is that? Remember Kevin's program set C2 to maximum. You cannot
get Reg's program to set C2 at max. But why? That is pure academia.
Now if we do things the other way we probably would get joy. That is,
take Reg's values and set the values manually in Kevin's program. Yep,
then you will see the match situation that Reg's program came out with.
Now we have some sort of reconciliation, that is, we have the same L and
C values in both programs and there is a match. Reconciling is mostly
about figuring out the programs, what is making them give different
figures.
It probably does not matter much which set of values you use, whether
from Kevin's or Reg's program, when you consider merely efficiency. But
it might* make some difference in other ways, like whether some value is
realiseable (ie C2). But with a T tuner you can adjust things, so, you
obviously use the calculated values as something to aim for, you may try
to keep close to it. So, if you don't want to adjust your antenna
length, so a value becomes realiseable, you just alter L a bit and C1.
* Probably unlikely because of flexibility of T net.
A value that is unrealiseable is something like C2 being 3pF.
(Just thinking out loud mostly).
Richard Clark
<Cecil....@ieee.org> wrote:
>Richard Clark wrote:
>> I see a lot of hand wringing over minutia that certainly might plague
>> an operator in the shack, but really, this needs only to be done once,
>> dipped in Resin and painted black (who really needs a box for this?).
>
>So a vertical in Bismark, ND has the same impedance in the middle of winter
>as it does in the middle of summer??? Methinks you have been living in
>good weather for too long. :-) Bismark winters Vs Bismark summers are
>definitely not "minutia". Try "roughing it" in Bismark in February and
>get back to us (if you survive). :-)
So you have a new old wive's tale that tuned antennas don't work in
Bismark, hmm?
More hand wringing. On reflection, you DO act like that continental
fop with an attitude problem with the pope - you both fold.
OC_CAM's Razor
> So you have a new old wive's tale that tuned antennas don't work in
> Bismark, hmm?
Depending on the bandwidth, some tuned antennas don't work year-round
in Bismark. Ice-coated window-line comes to mind. Please feel free to
move up there, spend a year proving me wrong, and then get back to us.
Hint: Roy reports that window-line doesn't work all that well even
when the H2O is not solid and not four inches thick.
Roy Lewallen
Stewart who measured the window line and came to about the same conclusion.
Actually, ice is quite a good dielectric. Pure water is lossy at HF due
to movement of its polar molecules in their attempts to align with the
field. Frozen in the crystalline structure of ice, that loss mechanism
goes away. So I wouldn't expect ice, even thick ice, to cause as much
loss as even a thin coating of water. The dielectric constant of the ice
(about 3 -- compared to 80 for water) would cause some detuning, however.
Roy Lewallen, W7EL
Richard
Of course the reason why you cannot get Reg's program to give the same
values in this instance, is as explained: That you cannot alter a
parameter (like L) sufficient to acheive the match in values. That's
all there is to it really. So, you have to rely on Reg's values and
alter parameters in Kevin's program if you want to match values, you
cannot do it the other way round.
If Reg's and Kevins programs did the calculation the same way, you would
always get them giving out pretty much the same values I'm sure. But as
mentioned, Kevin's seems to want to set C2 at max and finds a match
after taking that into account. Reg's does not set C2 at max and get's
on with it''s thing to find another match point. So, you end up with
two diffeing values for matches.
Does one program do a better job than the other at autotuning? Maybe,
but I suspect in practice you could go for either worked out value and
be satisfied.
Okay, that's my final comments on this. Just me wanting to get
something clear in my head.
ave such an inpact in C and C2 that cit's pretty simple
Owen Duffy
Richard wrote:
> always get them giving out pretty much the same values I'm sure. But
> as mentioned, Kevin's seems to want to set C2 at max and finds a
> match after taking that into account. Reg's does not set C2 at max
> and get's
Kevin's algorithm tests whether it can find a solution, and if it can,
it ALWAYS sets one of the capacitors to maximum, then finds the
appropriate value for the other two components. In my experience, the
values it selects are always a very good match, just they might not be
the lowest loss match.
There is a common belief that the lowest loss match will be achieved
with the larger capacitor set to maximum, though if you drive Kevin's
simulator by hand on the example data, you will find that a slightly
lower loss match is achieved with quite lower values of C. Perhaps the
simple statement applies to a lossless network, but not necessarily to a
practical network.
I suspect that under certain conditions, using the maximum C does find
the best position, but that outside of those conditions it may not, and
that your example is extreme and might fall outside those conditions. It
is a pity that the web page that accompanies the model doesn't explain
its limitations.
Owen
Richard Clark
wrote:
>Perhaps the
>simple statement applies to a lossless network, but not necessarily to a
>practical network.
This statement goes to the heart of the matter. Imagine the sheer
volume of computation involved if another variable, loss, were tossed
into the mix of either package. [The thorn of becoming trapped in
less than optimal solutions will still follow even with the inclusion
of this factor. And it could only get worse.]
>
>I suspect that under certain conditions, using the maximum C does find
>the best position, but that outside of those conditions it may not,
"Certain conditions" indeed. That is part of the blind-siding of
"undocumented" software. Call it SMD or Software of Mass
Disappointment - well, perhaps too dramatic. Most users are happily
unaware of being disappointed (they really don't have any alternative
but to operate their free programs like a slot machine - waiting for
the three cherries to line up).
>and
>that your example is extreme and might fall outside those conditions. It
>is a pity that the web page that accompanies the model doesn't explain
>its limitations.
This is more common than most would give credit (or blame) for. The
greater problem is when the writer has the arrogance to believe his
work has no limitations (or worse, these are YOUR limitations, a
personal failure as evidenced by his exemplary software performance).
Richard has taken up an examination by comparison to chase out the
differences. However, this is like elevating one crook over the other
because one is less dishonest. All systems require testing (even what
you buy) and that testing should run along the lines of applying known
conditions that result in known results and comparing THEM. Given
several rounds show disturbing results (wouldn't disturb me however,
given the toy status we started with); it would seem obvious from the
angst generated that both should be abandoned and personal
responsibility should kick in. That is, design your own software, or
fall back on traditional methods that include paper, pencil, (or
charcoal on a flat rock) and your favorite form of calculation.
I can hear shrieks building in the galleries full of guru-wannabees as
they come out of the shock of withdrawal. I could anticipate and fill
this posting with their dope-deprived wails of anguish, but really,
that would be extremely boring. Stuff like that from the capitol
fills the headlines every day.
Reg Edwards
I was about to say something - but with you it's a waste of bandwidth.
Richard Clark
g4fgq,
a symptom of withdrawal. ;-)
88's
Richard Clark, KB7QHC
Owen Duffy
Wes Stewart wrote:
> Before reinventing the wheel, you might look at xlzizl.xls found at:
>
> www.qsl.net/ac6la
>
> It will find a lower loss solution that Kenvin's.
>
> Another point; the component Q's given in the example are a bit
> optimistic in my mind.
Thanks, I was aware of Dan's work.
I wanted to develop a function that would return the component values
for an optimal match given the frequency, a range for the variable
capacitors and inductors, the steps by which the inductors or capacitors
can be varied, and a simple Q value for each of the components.
Dan's NETS function tries to do that for idealised components and
depends on the Excel Solver add in to refine the solution using another
function that solves the network for practical components. The guess
returned by NETS is often a long way from optimal. Solver is not really
suitable to a table that for instance might try to calculate the optimal
match at 300 frequencies.
As I suggested in an earlier post, an analytical solution is a bit
bewildering, so I think that the practical thing to do (although a bit
compute intensive) is to write a function that implements an iterative
algorithm much like an auto tuner.
It has already been pointed out that Kevin took a different approach to
an analytical solution to the optimal settings for practical components,
but it is not correct under some situations. Nevertheless, it is
interesting as an estimator of the loss that an operator will probably
encounter if she were to follow the advice in the manuals for the tuners
which commonly suggest that when tuning, try to ensure that one of the
caps is at max.
Owen
Kevin Schmidt
Richard wrote:
> I cannot get the same results with the same parameters.
>
> Parameters:
>
> Frequency: 7.059 Mhz
> R load: 1485.5 Ohms
> X load (jx) 1455.3 Ohms
>
> C1 max: 950 pF
> C1 Q : 1500
> C2 max : 950 pF
> C2 Q : 1500
> L Q : 300
>
>
> Input resistance: 50 Ohms
>
> W9CF's T-Network on automatic tune gives:
>
> C1: 61 pF
> C2: 950 pF
> L : 9.70 uH
>
> I was just trying to get T_Tuner to read same figures, but I don't seem
> to be able to. I change "A", but no joy.
>
> I know, the T network tuner offers such flexibility in settings, and I
> guess it's hard to pin down the best settings. But, I just wanted to
> try to match the settings with both these programs.
>
> In any event, should I always seek to set C2, capacitor connected to
> antenna, to the highest value possible?
>
> TIA. Rich.
>
I thought I would make a few comments about my tuner applet. I only
became aware of this thread yesterday since I now read this newsgroup
only occasionally.
I wrote the T-network applet as a amusement and most of the work went
into coding the knobs and meter to look reasonably well and to make it
give a feeling similar to that of a real tuner. The auto tune function
was patched on as an afterthought. When I wrote it, I thought that the
minimum loss occured with one of the capacitors set at maximum, but I
now see that that is incorrect. That rule is true for the lossiest case
of a low impedance load, but is not true in general. The reason the
web page did not explain that better is because I was unaware of it.
As an aside, the complete source of the applet is downloadable from the
web page, so it is hard for me to understand how anyone can say it is
undocumented. Perhaps its not documented the way you would like, but
since the whole package is licensed under the GNU GPL, you are allowed
to make any changes you like and distribute the changes. A quick look
at the source will show that the original autotune algorithm consists
of first trying the output capacitor at maximum and if that doesn't work
trying the input capacitor at maximum. It was that simple.
I believe minimum inductor rule should be correct if the inductor losses
dominate, so I have changed the algorithm to find the minimum inductor
value. Note that since our tuners don't have a loss meter on the front,
we need a simple rule like set one of the capacitors to max, or look
for a match that minimizes the inductance. The web page now has, I hope,
a better explanation of the autotune function.
73 Kevin w9cf
Kevin Schmidt
After this discussion about T-network tuners and my little applet in
particular, I thought I had better go back and see if I could understand
the adjustment for low loss better. For the lossiest low impedance
case, the tuner performs best at one of its endpoint settings with one
of the capacitors is at its maximum which also corresponds to a minimum
inductance.
However, the discussion here concerned the other extreme of a high
impedance load where many settings give reasonably low loss. Several
people mentioned that tuning for smallest inductance was the optimum;
however, that cannot be the case in general. For example for matching
50 ohms to 50 ohms, infinitely large capacitors and and infinitely large
inductor will give a match and has zero loss for any Q. This indicates
that a larger inductance gives a smaller loss at least in this contrived
case.
I assumed an input resistance of R0=50 ohms, and that the load
resistance RL is greater than R0. For lossless capacitors and a low
loss inductor, I solved the network equations and then solved for the
minimum inductor
reactance, which is sqrt(RL*R0). The value for lowest loss turns out to
be an inductor reactance of [(R0+RL)/(RL-R0)]sqrt(RL*R0), assuming the
other components have enough range to allow these settings. Note that
the load reactance XL does not enter into this formula, but it generally
needs to be substantially inductive to bring the output capacitor into
an allowed range.
The analytic solution in the limit of very high inductor Q says that
if we can adjust for minimum inductor we get a fractional loss of
sqrt(RL*R0)*(RL+R0)/(R0*RL*Q) while for the optimal size inductor,
the fractional loss is sqrt(RL*RO)*(RL-R0)/(R0*RL*Q).
To test this I put the values Z = 200+j200 at 7MHz into the Tuner
applet with the default component values. The new applet version will
autotune for lowest inductance and gives Cin = 230.3 pF, L = 2.3 uH,
and Cout = 75.5 pF, with a loss of 2.6 percent. From the equation
above, the smallest inductance should be at sqrt(200*50)/(2*pi*f)
= 2.27 uH, so that agrees, and the fractional loss for Q=100, is
sqrt(50*200)*(200+50)/(50*200*100) = 0.025 which also is in good
agreement. The applet also contains capacitor losses and the inductor
Q is not going to infinity so perfect agreement is not expected.
If I now readjust to increase the inductance by a factor of
(200-50)/(200+50) = 5/3, or to 3.8 uH, and then adjust the capacitors
by hand, I get Cin = 227.7 pF, L = 3.8 uH, and Cout = 234.2 pF, with
a loss of 1.6 percent. The fractional loss from my equation above is
sqrt(50*200)*(200-50)/(50*200*100) = 0.015 again in good agreement.
So minimum inductance is definitely not the lowest loss setting in
general. Of course both 2.6 and 1.6 percent are excellent figures so in
reality it doesn't matter much, but tune for minimum inductance is not
the optimal rule. If I blindly apply the minimum inductance loss formula,
it gives for RL=50 a fractional loss of 2/Q. So the minimum inductor
case gives only 2 percent loss with a Q of 100. The optimal case is
the one mentioned above where the output capacitor resonates with the
inductive reactance of the load, the inductor is infinitely large and
therefore gives 0 loss, and the input capacitor is infinitely large. For
higher load resistances, the difference in the loss is given by the factor
(R0-RL)/(R0+RL). So by the time you get to something like RL = 500 ohms,
the minimum inductance formula gives for Q=100, 3.5 percent loss, while
with the optimal inductance the loss is 2.8 percent.
So my conclusion is that while tuning for minimum inductance is a good
rule, it does not always minimize the loss.
73 Kevin w9...@ptolemy.la.asu.edu, http://fermi.la.asu.edu/w9cf