"Richard Clark" wrote in message
...
On Mon, 1 Sep 2003 14:12:08 -0400, "Tarmo Tammaru"
wrote:


In general, the addition of negative feedback to change the output
impedance
does not change the dynamic range.


Hi Tam,

H.W. Bode, of Bell Labs, developed a body of work that proves you
completely wrong. Negative feedback enhances every facet of amplifier
design and to claim otherwise is wholly misinforming.

73's
Richard Clark, KB7QHC

Hi Richard,

Everything, except the output dynamic range. Please go through my numbers.
The equation I gave is for the point at which the amplifier goes non linear,
and there actually is no feedback. There is a large error voltage at the
input, but the amp can't do anything about it because it is already in
saturation. For a 5V input/output and a 1K load the output terminal of the
op amp is already at VCC. A more practical example might have been a 13.8
volt power supply regulator running off 16 volts, with a current sensing
resistor in series with the output. Anyway, it was a rainy afternoon kind
of thing.

Tam/WB2TT

Richard,

Let me be the lurker on for the day. Actually, you have a good example. I
see the parallel equivalent resistance for the device at 1.5 MHz is about 2
Ohms. At 12.5V with a Vcesat of an optimistic .5V, the power that can be
delivered at conjugate match is 12^2/(2 * 2) = 36W. Not the rated 100.

The 421 is a fairly inefficient transistor. The 50W MRF450 has a *series*
equivalent output impedance at 30 MHz of 174 - j.5. The parallel resistance
will be about 175. Matched power will be 12^2/(2 * 175) = 0.4W. There is
nothing wrong with that.

One of their ap notes states that on newer devices they specify the
conjugate of the desired load, rather than the intrinsic output impedance.
On some devices you simply can't tell which they mean. On the very popular
MRF150, it is footnoted as being "conjugate of optimum load". It may be that
the FET resistance is too high to be meaningful.

Tam/WB2TT
"Richard Clark" wrote in message
...
On Mon, 1 Sep 2003 10:19:39 +0100, "Ian White, G3SEK"
The MRF421 is used in two of my HF rigs and you have yet to offer what
you have. You didn't look did you? Have you ever looked? Have you
ever repaired a Finals' deck? Ever design one that is comparable?

From that specification sheet(s):
Figure 7 - Series Equivalent Impedance, Zin
shown with both Smith Chart and tabular results over the range of 2
MHz to 30 MHz:
Freq. Zin
30 0.7 - j.5
15 1.39 - j1.1
7.5 2.8 - j1.9
2.0 5.35 - j2.2

Richard Clark wrote:
On Mon, 1 Sep 2003 10:19:39 +0100, "Ian White, G3SEK"
wrote:

Richard Clark wrote:
On Sun, 31 Aug 2003 19:42:42 +0100, "Ian White, G3SEK"

Motorola publishes the equivalent series (or parallel) resistance for
the MRF-xxx (pick your own that corresponds to the actual device used
for the finals in your own transmitter).

Sorry, that horse won't run. RF power transistor data sheets specify the
load impedance that needs to be presented *to* the transistor from the
outside world, in order for the device to function as specified.

Clearly you did not look at such a sheet, and certainly not from
Motorola.

The MRF421 is used in two of my HF rigs and you have yet to offer what
you have. You didn't look did you? Have you ever looked?

Of course I have, over several years: looked at data sheets (Motorola
more than any others), at numerous application notes, and at Dye and
Granberg's textbook.

Have you
ever repaired a Finals' deck?

If you mean, have I worked in the mobile radio industry, then the answer
is no.

Ever design one that is comparable?

Oh yes: designed, built, used - and understood.


From that specification sheet(s):
Figure 7 - Series Equivalent Impedance, Zin
[snip]
Yes, Zin is the true input impedance of the device - but "Zout" is not.

Figure 9 - Output Resistance versus Frequency, Zout
slightly less than 2 Ohms at 1.5 MHz to 1.0 Ohms at 30 MHz shown in a
clear and unambiguous chart over the entire range of frequency and
resistance described as Rout, Parallel Equivalent Output Resistance,
(Ohms).

"Described" is the operative word. What these terms really mean is a
different matter - see the quotes from Motorola below.


Sometimes they state the conjugate of the required load impedance, and
sometimes they don't say which.

Sometimes? You don't seem to sure, and you offer nothing specific.

I am very sure. Some manufacturers state the conjugate, others don't.
Even Motorola have changed their terminology over the years: sometimes
it's "output impedance" or "Zout", sometimes it's "Z(subscript OL)",
sometimes it's "Z(superscript *)(subscript OL)".

Motorola's AN1526 was written in the 1990s to clear up this mess. This
is the key paragraph [my comments in square brackets]:

"Almost every RF power device in Motorola's RF Device Data Book has a
section identifying the device's large-signal series equivalent input
and output impedances. Most often, the device output impedance is
referred to as 'the complex conjugate of the optimum load impedance into
which the device operates at a given output power, voltage and
frequency.' That is certainly a statement requiring some careful
thought, especially since the term 'output impedance' is somewhat
misleading [so even Motorola admit that]. ... as described in [AN282]
it is the conjugate of the LOAD impedance at the fundamental operating
frequency which allowed the transistor to 'function properly' [when the
load impedance was varied in a test jig]."

AN1526 goes on to explain this in much more detail. All of it is
consistent with my earlier statement that the parameter sometimes
labelled "output resistance" is actually the LOAD that needs to be
presented *to* the device from the outside world, for optimum
performance.



That is not the output impedance *of* the transistor itself (except in
certain special cases). However, the data sheets sometimes do
ambiguously call it the "output impedance". It's a confusing mess.

The data sheets are quite ordinary to the designers however. I've
never sat through a design seminar where any RF design engineer has
made such a statement as yours (much less the outright howler of
circularity below). In fact their queries related to exactly these
specifications. How is it that your experience is different?

Maybe it's because I haven't swallowed someone else's half-digested
information, but have done my own thinking and not given up till it
really made sense.


However, I'd hoped by now that we were all agreed about this particular
dead horse. It's buried somewhere in the Google archives of this
newsgroup.

The rest of the argument about transforming the impedance from the
device collector/drain anode to the output socket is correct, but it's
being applied to the wrong impedance. The answer you'll find at the
output socket will always be 50 ohms, because all you've done is lead
your horse around in a very tight circle.

Odd that in a chain of signal flow, that you see it being circular.

I never mentioned signal flow - I'm talking about your circular logic!

The output network is designed to transform a 50-ohm load at the output
into the complex load impedance that needs to be presented to the
collector. But you incorrectly claim that the data sheet gives the
"output impedance" of the device itself. You then claim that by
transforming that "device output impedance" through the output network,
you can calculate the output impedance of the transmitter. Since you are
simply back-tracking through the network design calculations, the answer
you get will always be 50 ohms! It's circular logic in the tightest
possible loop.


For those lurkers who get short-shrifted of quality in these
half-debates I offer another Motorola reference: "RF Device Data
Volume II." Observe in AN2821, "Systemizing RF Power Amplifier
Design," in the sub-section "Amplifier Design":

That's actually AN282A.

"After selection of a transistor with the required
performance capabilities, the next step in the design
of a power amplifier is to determine the large-signal
input and output impedance of the transistor."
...
What you cut out here was the key reference to "collector LOAD
resistance", presumably because you missed the significance of "load".

"Having determined the large-signal impedances, the
designer selects a suitable network configuration
and proceeds with his network synthesis."

ALL may note this is exactly what has been described by me. ALL may
note nothing of equal scope and depth is offered in rebuttal

AN282, from which you quote, was first published in 1968. The truth
about load impedance is in there to be seen, but I'd be the first to
agree that it's not stated clearly. Motorola then confused the issue by
continuing to talk ambiguously about "output impedance" for at least
another 20 years.

AN1526, from which I'm quoting above, supersedes AN282. It was written
about 25 years later in an attempt to clear up that inherited mess of
loose definitions... but apparently with limited success.


--
73 from Ian G3SEK 'In Practice' columnist for RadCom (RSGB)
Editor, 'The VHF/UHF DX Book'
http://www.ifwtech.co.uk/g3sek

On Mon, 1 Sep 2003 17:52:43 -0400, "Tarmo Tammaru"
wrote:
Hi Richard,

Everything, except the output dynamic range. Please go through my numbers.
The equation I gave is for the point at which the amplifier goes non linear,
and there actually is no feedback. There is a large error voltage at the
input, but the amp can't do anything about it because it is already in
saturation. For a 5V input/output

What exactly does this mean? That you expect 5V in will track with 5V
out? Say so. Shortcuts in specification lead to disaster. If this
is not what you mean (and you did not specify an input in the
original) your ambiguity allows for any interpretation.

and a 1K load the output terminal of the
op amp is already at VCC. A more practical example might have been a 13.8
volt power supply regulator running off 16 volts, with a current sensing
resistor in series with the output. Anyway, it was a rainy afternoon kind
of thing.

Tam/WB2TT

Hi Tam,

Let's review as you asked:

Consider the following. An ideal DC op amp operating from a 10V supply with
0 output impedance, and infinite gain.

1. Put a 1K resistor in series with the output.

Hence the "ideal" Op Amp is crippled from the start, any expectations
of superlative performance have already been abandoned. Further, the
"ideal" Op Amp is not rail limited, nor is it convention to discuss Op
Amp designs with single ended supplies. These are more short cuts
that could have been expressed with very little greater length to
conventional usage.


2. You now have an op amp with 1K output impedance.

#2 in fact adds absolutely nothing but repetition.


3. Connect the inverting input of the op amp to the other end of the 1K
resistor. Call that the output terminal. and connect a load resistor Rl from
there to ground. ( The gain from the non inverting input to Rl =1 in the
linear range)

That being the new output of the amplifier, R1 as you call it, now
loads that amp to ground. (By the way, if you are going to have two
resistors, convention would label them R1 R2; not R and R1.)

The gain is NOT 1 by sheer, obvious placement of the resistors you
describe. You elsewhere supply a gain that does not agree with this
#3. What you imply are the mu and beta gains, but you do not really
go into that distinction, nor do you perform the math that bears on
their usage.


4. Because of the infinite feedback,

Infinite feedback? Poor specification where I have to presume you are
in error and meant infinite gain (for the previously "ideal" amplifier
- which it is not now). If the output were strapped back to the
inverting input, that "might" qualify as infinite, but your load is
the gain determinant of the amplifier. If you meant that the Op Amp
output is impressed upon the inverting input, that goes without saying
for all linear applications doesn't it? If by this your statement of
a gain of 1 above was along the same lines, it suffers by similar
degree.

the output impedance is now 0 again,

No, it is not, you have ascribed (by description) a 1000 Ohm output
impedance (resistance). The amplifier is not the Op Amp component, it
is the assembly of components presented to the load and is modified by
the gain that you incorrectly ascribe above.

but all of the load current still flows through the 1K series resistor.

Which confirms my statement and directly follows from the addition you
originally offer.

It
will not, for instance, deliver 5V into a 910 Ohm resistor because the
amplifier will have saturated before that point. For a given RL, the maximum
voltage you can get out is (10 x Rl)/(1000 + Rl) with or without feedback.


With OR without feedback? Which is it? Do you have feedback or don't
you? The additional baggage of your statement both adds nothing, and
is self conflicting. Do I now have the choice to express it has no
feed back?

Dynamic range is not the same thing as rail limited and rail limiting
is certainly not within the canon of "ideal" amplifiers. Dynamic
range is dimensionless and generally described in dB and is a function
of noise. Feed back has a direct correlation to the amount of noise
added by the amplifier and thus impacts Dynamic Range directly.

It would have taken a whole lot less to simply use the conventional
741 Op Amp as an example, warts and all, to express the same issue
which merely points out that a poor design works poorly. You even
anticipate this poor aspect through the modifications after the fact:
A more practical example might have been a 13.8
volt power supply regulator running off 16 volts, with a current sensing
resistor in series with the output.
which illuminates how a design engineer builds from known limitations
toward known loads. In other words, if the engineer faces a 10V rail
limitation, he could have as easily added a DC-DC up converter to
solve it. Anyone can trap another through crafted specifications.
I've got several many squirrels up a tree right now.

73's
Richard Clark, KB7QHC

On Mon, 1 Sep 2003 19:17:40 -0400, "Tarmo Tammaru"
wrote:

Richard,

Let me be the lurker on for the day. Actually, you have a good example. I
see the parallel equivalent resistance for the device at 1.5 MHz is about 2
Ohms. At 12.5V with a Vcesat of an optimistic .5V, the power that can be
delivered at conjugate match is 12^2/(2 * 2) = 36W. Not the rated 100.


Hi Tam,

rated 100 WHAT?

Sheesh.

73's
Richard Clark, KB7QHC

On Tue, 2 Sep 2003 00:28:12 +0100, "Ian White, G3SEK"
wrote:


AN1526, from which I'm quoting above, supersedes AN282. It was written
about 25 years later in an attempt to clear up that inherited mess of
loose definitions... but apparently with limited success.


Hi Ian,

25 years after the 60's, and yet AN282 is still published by Motorola
in 1991 by my copy. Further, it appears to survive through to 91
without any appearance of AN1526 which superseded it. This would
suggest that this application note, if published in the 90's
represents quite a bulk of material (nearly half again the total AN's
by number following 1991) published in 10 Years? 1968 + 25 = 1993
which suggests rather a revolution in thought over two years and
clearly not within the scope of credibility. These publishing date
games are too inspecific with your reference clearly in front of you.
Don't you know to surer accuracy?

I still see nothing of substance, merely suggestion:
That is certainly a statement requiring some careful
thought, especially since the term 'output impedance' is somewhat
misleading [so even Motorola admit that]. ... as described in [AN282]
it is the conjugate of the LOAD impedance at the fundamental operating
frequency which allowed the transistor to 'function properly' [when the
load impedance was varied in a test jig]."
Does not say what it is, but what it was by testing - hardly
revolutionary nor upsetting.

Certainly you could come up with a smoking gun couldn't you? Complete
with an actual, demonstrable specification for the item I offered
(seeing as you still lack any concrete example). What does your new
and updated resource say about the MRF 421. Does it abandon that
discussion entirely to this new-age era of all being unknowable?

This is still nay-saying and cut-and-paste philosophy without any
correlatives to actual equipment in the field. Why is it so simple to
correlate for me, and for you to dispute through t'ain't so with no
further elaboration?

For example you countered my query as to "did you build your own RF
deck" and the paucity of specifics could create a vacuum. What
impedes your own counter examples of actual implementation? What
prevents your discussion of design issues considered that are germane
to this side-bar? Where is the scope and depth? What service do you
offer those mythical "lurkers" or do we agree that they are simply an
egoists device?

If you think for yourself, be prepared to stand and deliver.
Otherwise this commentary is more posture than content.

73's
Richard Clark, KB7QHC

Why of course that's the explanation! But we're sure thankful we've got
you to tell us what the *real* device characteristics are, and how
things really work! Or at least to give us programs that give us the
for-sure right answers when the underlying principles are too hard for
us to understand. Thanks!

Roy Lewallen, W7EL

Reg Edwards wrote:
Evidently data sheets and books are written by
marketing and sales departments.

Hi Richard,

Thanks for your comments. The next time I do anything like this I will make
the thing more real world. Should have done the power supply thing. I have
built enough of those. As I said, this was a rainy afternoon thought. I
sprinked in some commenys below.

Tam
"Richard Clark" wrote in message
...
On Mon, 1 Sep 2003 17:52:43 -0400, "Tarmo Tammaru"
wrote:
Hi Richard,

Everything, except the output dynamic range. Please go through my
numbers.
The equation I gave is for the point at which the amplifier goes non
linear,
and there actually is no feedback. There is a large error voltage at the
input, but the amp can't do anything about it because it is already in
saturation. For a 5V input/output

What exactly does this mean? That you expect 5V in will track with 5V
out? Say so. Shortcuts in specification lead to disaster. If this
is not what you mean (and you did not specify an input in the
original) your ambiguity allows for any interpretation.

and a 1K load the output terminal of the
op amp is already at VCC. A more practical example might have been a 13.8
volt power supply regulator running off 16 volts, with a current sensing
resistor in series with the output. Anyway, it was a rainy afternoon
kind
of thing.

Tam/WB2TT

Hi Tam,

Let's review as you asked:

Consider the following. An ideal DC op amp operating from a 10V supply
with
0 output impedance, and infinite gain.

1. Put a 1K resistor in series with the output.

Hence the "ideal" Op Amp is crippled from the start, any expectations
of superlative performance have already been abandoned. Further, the
"ideal" Op Amp is not rail limited, nor is it convention to discuss Op
Amp designs with single ended supplies. These are more short cuts
that could have been expressed with very little greater length to
conventional usage.
**********************************
I am making a not quite perfect amplifier from a perfect one. I used that so
I would not have to get into details about 80 db gain, .5V dropout voltage,
10 meg input impedance, 20 Ohm output impedance, etc. Single rail
amplifiers, including these that have a common mode range that includes
ground are quite common.
***********************************

2. You now have an op amp with 1K output impedance.

#2 in fact adds absolutely nothing but repetition.
************************************
right, but I want the reader to consider the 1K as part of the amplifier,
and I have placed it at a point where its effect is obvious
*************************************

3. Connect the inverting input of the op amp to the other end of the 1K
resistor. Call that the output terminal. and connect a load resistor Rl
from
there to ground. ( The gain from the non inverting input to Rl =1 in the
linear range)

That being the new output of the amplifier, R1 as you call it, now
loads that amp to ground. (By the way, if you are going to have two
resistors, convention would label them R1 R2; not R and R1.)

The gain is NOT 1 by sheer, obvious placement of the resistors you
describe.
***************************************
The gain is precisely 1 at the point that I call the amplifier output. The
gain from input to the output terminal of the original amplifier is
variable, and depends on the load.
***************************************
You elsewhere supply a gain that does not agree with this
#3. What you imply are the mu and beta gains, but you do not really
go into that distinction, nor do you perform the math that bears on
their usage.

************************
Precisely. People might read this who have no idea what mu and beta are, but
they know Ohm's law.
************************

4. Because of the infinite feedback,

Infinite feedback? Poor specification where I have to presume you are
in error and meant infinite gain (for the previously "ideal" amplifier
- which it is not now). If the output were strapped back to the
inverting input, that "might" qualify as infinite, but your load is
the gain determinant of the amplifier. If you meant that the Op Amp
output is impressed upon the inverting input, that goes without saying
for all linear applications doesn't it? If by this your statement of
a gain of 1 above was along the same lines, it suffers by similar
degree.
***************************
I have reduced the open loop gain A of the amplifier by the ratio
Rload/(Rload + 1K), but it is still infinite. I may be using terms that are
not universally used. When feedback is used to reduce gain by say 20 db, it
is common to refer to this as 20 db of feedback (May be an audio term). I
have a unity gain amplifier, as I have defined it, and reduced the open loop
gain by infinity.
***************************8

the output impedance is now 0 again,

No, it is not, you have ascribed (by description) a 1000 Ohm output
impedance (resistance). The amplifier is not the Op Amp component, it
is the assembly of components presented to the load and is modified by
the gain that you incorrectly ascribe above.
********************************
I am defining the 1K as part of the amplifier, but I placed it at a point
where you could see it. As defined, if you placed +6V on the noninverting
input, and a load of 10K on the output, the output voltage would be 6V. If
you changed the load to 2K, the output voltage would still be 6V. The
impedance is 0. A person might assume he could easily obtain 11 ma from such
a low impedance source. He can not. Not even at .01V.
**********************************
but all of the load current still flows through the 1K series resistor.

Which confirms my statement and directly follows from the addition you
originally offer.

It
will not, for instance, deliver 5V into a 910 Ohm resistor because the
amplifier will have saturated before that point. For a given RL, the
maximum
voltage you can get out is (10 x Rl)/(1000 + Rl) with or without
feedback.


With OR without feedback? Which is it? Do you have feedback or don't
you?
***********************
I realize this sounds unclear. Consider the node that drives the 1K
resistor. It can not go higher that 10V; the short circuit current is 10 ma.
This is true both for
a) loop is closed. That is the voltage across the load ic connected to the
non inverting input.

b) loop is open with inverting input at ground and noninverting input at
+10V
************************

The additional baggage of your statement both adds nothing, and
is self conflicting. Do I now have the choice to express it has no
feed back?

Dynamic range is not the same thing as rail limited and rail limiting
is certainly not within the canon of "ideal" amplifiers. Dynamic
range is dimensionless and generally described in dB and is a function
of noise. Feed back has a direct correlation to the amount of noise
added by the amplifier and thus impacts Dynamic Range directly.

It would have taken a whole lot less to simply use the conventional
741 Op Amp as an example, warts and all, to express the same issue
which merely points out that a poor design works poorly. You even
anticipate this poor aspect through the modifications after the fact:
**************************
I didn't want to do that. Instead of a 741 I would use an LM358.
**************************
A more practical example might have been a 13.8
volt power supply regulator running off 16 volts, with a current sensing
resistor in series with the output.
which illuminates how a design engineer builds from known limitations
toward known loads. In other words, if the engineer faces a 10V rail
limitation, he could have as easily added a DC-DC up converter to
solve it.
***************************************
I didn't want to build a good circuit, I wanted to build a bad circuit, and
show that all the feedback in the world was not going to make it a good
circuit.
***************************************

Anyone can trap another through crafted specifications.
I've got several many squirrels up a tree right now.

73's
Richard Clark, KB7QHC

"Richard Clark" wrote in message
...
On Mon, 1 Sep 2003 19:17:40 -0400, "Tarmo Tammaru"
wrote:

Richard,

Let me be the lurker on for the day. Actually, you have a good example. I
see the parallel equivalent resistance for the device at 1.5 MHz is about
2
Ohms. At 12.5V with a Vcesat of an optimistic .5V, the power that can be
delivered at conjugate match is 12^2/(2 * 2) = 36W. Not the rated 100.


Hi Tam,

rated 100 WHAT?

Not the rated 100 Watts.
This is just like the 6V6 thing we did earlier.

Tam

Sheesh.

73's
Richard Clark, KB7QHC

Richard Clark wrote:
On Tue, 2 Sep 2003 00:28:12 +0100, "Ian White, G3SEK"
wrote:


AN1526, from which I'm quoting above, supersedes AN282. It was written
about 25 years later in an attempt to clear up that inherited mess of
loose definitions... but apparently with limited success.


Hi Ian,

25 years after the 60's, and yet AN282 is still published by Motorola
in 1991 by my copy. Further, it appears to survive through to 91
without any appearance of AN1526 which superseded it. This would
suggest that this application note, if published in the 90's
represents quite a bulk of material (nearly half again the total AN's
by number following 1991) published in 10 Years? 1968 + 25 = 1993
which suggests rather a revolution in thought over two years and
clearly not within the scope of credibility. These publishing date
games are too inspecific with your reference clearly in front of you.
Don't you know to surer accuracy?

I only know what Motorola say in HB215/D, 'RF Application Reports'. As
you know, Motorola don't go back and re-write old application notes -
they only publish newer ones, leaving users to sort out which aspects
have been superseded and which are still valid.

AN282 is still included in HB215/D and older compilations because it
contains valuable information on other topics.

The application notes themselves are not dated. The list of references
in AN1526 states that AN282(A?) was published in 1968. The publication
date for AN1526 can be bracketed to the early 1990s (after the latest
dated reference that it quotes, and before 1995 when it was re-published
in HB215/D).

That is sufficient to establish my point: that the thinking in AN1526 is
based on about 25 years further experience after AN282.



I still see nothing of substance, merely suggestion:
That is certainly a statement requiring some careful
thought, especially since the term 'output impedance' is somewhat
misleading [so even Motorola admit that]. ... as described in [AN282]
it is the conjugate of the LOAD impedance at the fundamental operating
frequency which allowed the transistor to 'function properly' [when the
load impedance was varied in a test jig]."
Does not say what it is, but what it was by testing - hardly
revolutionary nor upsetting.

Certainly you could come up with a smoking gun couldn't you?

If you can't get it from the key paragraph I quoted, then read all 15
pages of AN1526. If you still can't see that your notion about "device
output impedance" is shot clear through, then neither Motorola and I can
help.

Complete
with an actual, demonstrable specification for the item I offered
(seeing as you still lack any concrete example). What does your new
and updated resource say about the MRF 421.

You know perfectly well that AN1526 won't say anything about your
specific pet device, so what was the point of asking that question?

Does it abandon that
discussion entirely to this new-age era of all being unknowable?

And that is an even worse travesty of what Motorola and I are saying.

If you want to measure the *true* output impedance of an MRF 421 - as
distinct from the load impedance given in the data sheet - then go ahead
and do it. After all, you're the one who claims it is an important
design parameter.

I'm the one who says it is (a) not what you think it is; and (b) not
important anyway.

Now it's up to other people to judge the technical truth of the matter.



--
73 from Ian G3SEK 'In Practice' columnist for RadCom (RSGB)
Editor, 'The VHF/UHF DX Book'
http://www.ifwtech.co.uk/g3sek