Craig,
Re, 40Mhz BPF covers.
I have seen a number of Ra17 and Ra117 series, and the MA79 too based on
the same circuitry, but never come across filter covers with mesh. They all
possessed the spring-finger earthing method, and pushed into the chassis
cavity of the filter. These springs are in close proximity to the L/C
components within. The caps are all low pF values, and when tuned for
alignment, I note the tiniest movement twiddling, has a disproportionately
large effect of the BPF filter transfer function shape. In other words...
is exquisitely sensitive to the very smallest capacitance changes. So it
doesn't surprise me at all that those earthing fingers probably have a
large effect on the overall parasitic capacitance within each filter
section cavity; and further: as each section is closely coupled
electrically the effect is cumulative. Hence take cover off - put back on
and it's all slightly different capacitance wise!
Every Racal unit I've seen that has not been "breathed upon" by someone
and/or presumably untouched since leaving the factory, had black plastic
labels on the BPF covers with the warning: "Do not Remove".
It was Bob Finch who pointed this problem out to me in the 1970s when I
first obtained an RA117 and SSB unit and needed to align it (in those days
with a wobbulator and separate sig gen to produce the marker pips). I've
observed many times since he was quite right.
I do not think it is a problem with the quality of earthing of the fingers;
I'm satisfied that is more than reasonable; but is simply the variability
of capacitance with the close proximity of the earthing cover fingers
internally every time the cover is disturbed.
Thank you for the explanation of the "sniffer loop" idea. This introduces
considerable inconvenience for many reasons. It did occur to me thinking
about it earlier that one would have to drill a hole in each cover section
cover to the side of the existing central hole to introduce a fine wire
detector loop - but even _that_ probably will interfere with electrical
capacitance effects and BPF shape. You have now confirmed a sniffer loop
will have to be introduced inside each section.
It seems to me the 40 Mhz BPF is a crucial and highly sensitive part of the
receiver design - highly sensitive to component values and physical shape
in the design. It must have been a real headache when designed.
I think it is an Achlles Heel - a bottleneck - of the whole circuit
design, but imposed on Racal with not much in the way of alternatives
available because of the very stringent requirements of as close as
possible a perfect 'square edged' filter to give _exactly_ a 1Mhz
bandwidth, centred on 40 Mhz - definitely not easy in the 1950s !
Even when adjusted to as near "perfect" as possible it has limitations in
that the required bandwidth of exactly 1Mhz is only _just_ possible to meet
- but at the expense of allowing compromises: a large increase in overall
attenuation across the whole passband (ie. by widening it to the 1 Mhz);
... and rapid decrease in receiver sensitivity at each end of the Khz scale
due to the rapid (square) shape fall-off in passband at each end; and
finally, passband ripple, changing sensitivity of receiever across the Khz
scale.
Aligniment is tricky but doeable and though takes some fiddling is soon
learned.
I'm of the opinion that BPF needs to be isolated as much as possible for
alignment. Which is why I developed my method of using the input and
output valves as buffers to keep it as isolated as possible during
alignment. This method only needs a small cap from the BPF output valve
available for output to the Spec Analyser.
As a final note, the input and output image impedances presented to the
filter also seriously affects the performance (shape) of the filter, which
is why the two end sections are the most sensitive in twiddling to changing
passband shape; also:- even changing either of the E180F high gain valves
at input and output side can change performance of the filter and
performance of the whole receiver. I think it was recommended that
realigment of the BPF be done if these valves were changed.
Regards,
Douglas. G3ZQE.
On Thu, 10 Oct 2024, 00:23 Craig Sawyers via groups.io, <c.sawyers=
[email protected]> wrote:
Douglas
Quizzing David in more detail, it looks like a pair of sacrificial
covers is needed, with additional small holes through which a single
exciting loop from a sig gen and single spectrum analyzer loop to be
introduced.
A very similar approach was shown in a paper by Dishal "Alignment and
Adjustment of Synchronously Tuned Multiple-Resonant-Circuit Filters"
from 1951. Still copyright by the IEEE, and I've failed to find a
bootleg copy.
But here's a thread on this method
and doubtless others regarding Dishal; that link is just the first I
found.
It seems very strange that the 40MHz filter is so exquisitely sensitive
that removing the covers then screwing them back in messes the response
up. Racal cannot have let that one go. I realise that in very late
production they replaced the spring finger plate grounds with metal mesh
edged covers, so perhaps the effect you notice is down to imperfect
grounding of he spring finger plate grounds?
Craig
------ Original Message ------
From "Douglas Denny" <ddennyoptom@...>
To [email protected]
Date 09/10/2024 23:55:36
Subject Re: [RacalRA17forum] Marconi TF1168
Craig,
This Zverev reference is very interesting indeed. A completely new angle
on filters for me. I must look more into it.
I'm still dubious about the practical application for alignment with the
Racal 40 Mhz BPF however.
As I have mentioned before: just taking off the cover and replacing again
can drastically disturb the BPF characteristics and shape; so this
procedure would require the covers to be remaining in place. How is this
procedure of shorting either side to be done without removing the covers?
and how will one couple the Spec Analyser to the single resonator when
there is just a small hole in the top with the cap trimmer visible (which
is at HT incidentally so presumably to be done power off!).
Sorry, - just don't see it.
On Wed, 9 Oct 2024, 08:03 Craig Sawyers via groups.io, <c.sawyers=
[email protected]> wrote:
Douglas
This is a compedium of the mails on this topic from UK Vintage Radio (
), David is a list admin,
with the handle "Radio Wrangler"
Craig
"There is a dirty trick for aligning coupled resonator filters (whether
the coupling is spatial, L or C, top or bottom)
Each resonator should resonate at the geometric centre frequency when
the other resonators to each side of it are disabled (shorted)
You need to use very very light coupling from your tracking generator
and your analyser to do this, but it works. Spectrum analysers have
tons
of sensitivity, which can be used for very light coupling. With the
floating resonators of the 40MHz filter a coupling loop on each of the
tracker and the analyser is needed.
There isn't just a mathematical basis to this method, it is actually
involved in the design of such filters.
Trying to just twiddle the trimmers on a high order filter is an
unstable process, and even with a display of the shape in front of you,
settings tend to diverge, not converge.
Racal used preset jigs and all sorts of custom hardware to make setting
easy, but the 3dB k and q method only really got going after Zverev in
1967. The older m-derived and image parameter methods were much more
tedious to work out and didn't give you an adjustment method as a
freebie. There was a bit of a quiet revolution in filter design methods
in the mid sixties that came out of Bell labs and Westinghouse. Filter
design was a big issue for telecomms firms of the day.
==================================
David,
Thank you for the swift reply, and for your "dirty trick"... Apologies
in advance for the tidal wave of questions...
I know very little about coupled resonators - does the 40MC/s filter
contain mutual-inductance coupled resonators?
When you say geometric centre frequency - do you mean sqrt(39.5 *
40.5)?
Also, I presume that since the filter bandwidth is relatively wide at
~1MHz, each of the resonators needs to be set for a different
frequency.
Should I evenly space the resonators? Geometrically space them? Do they
need to be aligned in increasing or decreasing resonant frequency
order?
Do you have a specification at all?
When you say "the other resonators to each side of it are disabled
(shorted)", do you mean ALL of the other resonators should be shorted?
Or just the immediate two resonators either side?
When you say "light coupling", I presume that a single turn for the
tracking generator and spec analyser input should be adequate.
How would you recommend achieving this, given that the proximity of the
shielding can will probably affect the resonant frequency of each
resonator?
I am not familiar with the work of Zverev - I will look for some online
reference material...
Many thanks again for your advice,
====================================================
The 40MHz filter is a constructed as a string of LC resonators relying
on magnetic field coupling between adjacent ones though there will be a
little E field as well.
You have the geometric centre right.
Each resonator is set in isolation to the geometric centre, and the
coupling to its neighbours pulls it off of the centre. In a neat feat
of
mathematics the coupling factors not only move the resonances to the
right places, they also get the right Q for each resonator to complete
the pattern which synthesises the required shape.
The source and load impedances applied to the ends of the filter set
the
Qs of the end resonators, and the coupling factors pass some of the
damping along to the inner resonators. The coupling just happens to get
the pattern of the Qs right. Handy Eh?
Zverev wrote the book which set out "Modern Filter Theory" it turned
the
decision of where to place the roots of the differential equation
defining the transfer function of a filter into a science. It's an
approximation for most idealised filter shapes, but you can decide how
complex a filter you want to make and this defines how close an
approximation you'll get.
Zverev's Handbook of Filter Synthesis covers the theory and then gives
catalogues of achievable shapes, than gives tables of normalised values
for making these shapes.
ust about every filter book written since 1967 cites Zverev. Many books
have been written on how to understand Zverev, but they don't put it so
directly
Zverev scares a lot of people off. He writes equations of attenuation
when equations of transfer function are now the norm so his equations
are upside down. His poles are our zeroes, his zeroes our poles.
It's not a book to go into alone. You need a guide to show you what it
can do. Otherwise it will seem weird and you may not notice the
valuable
bits.
A small single turn loop for each of tracker and analyser is right.
Ideally you want them on opposite sides of the resonator
David
==============================
Ah, I must learn to type faster and think faster a few posts happened
while I was writing.
Use the one resonator at a time process. You do NOT stagger them at
all.
You tune them all to the geometric centre frequency. One stage at a
time
with the neighbours each time shorted.
When you take the shorts off, the coupling moves them all onto the
right
frequencies.
Staggering is automatic. it does it for you itself.
When the full filter is running you will see Chebyshev bumps in it, but
if you twiddle the trimmers you'll find that they all interact. It
isn't
a case of trimmer X moves bump Y. You don't get anything so easy.
That's
why most trimming attempts of people unfamiliar with a design go astray
even with wobbulators spectrum analysers etc.
Cryptography relies on mathematical one-way streets, functions which
are
easy to do in one direction, but much harder to do in the reverse
direction. Filter twiddling is like that. You have to learn to cheat.
David
=======================================
The source and termination resistors set the Qs of the end stages of
the
filter, and via the coupling factors set the Qs of all the other
stages.
If the resistance at either end goes high, the Qs go too high for the
designed response and the ripple gets big.
Lower the resistances and the Qs fall too low for the designed shape
and
the filter rolls off.
If you're filter-hacking on a simulator like Spice, you can vary width
by scaling the coupling factors, then fiddle with the terminating R at
each end to adjust to get the flatness/ripple right.
David
===================
There's a danger of confusion here with talk of zeroes and Zverev.
Zverev drew equations of attenuation but now we tend to use equations
of
transfer function. These are reciprocals of each other, so no big deal.
But the zeroes of transfer function are the poles of attenuation and
vice versa. So Zverev and current usage have the words zreoes and poles
swapped.
The tune one stage at a time method works for "All-pole" filters (poles
of transfer function) realised as coupled resonators. The schematic
posted earlier of Racal filters as simple L-C tanks spatially separated
is such a filter... they have no notches.
This technique does NOT put the poles on the required frequencies. It
puts them all on the same frequency. Then it relies on the coupling
factors to pull the poles apart to shift them onto the required
frequencies. It is rather neat that the same coupling factor does the
right frequency pull and sets the right Q both at the same time.
The passband pole positions show up as blunt ripples in the passband
top, making them very hard to adjust if you try this way. Better, they
appear as much sharper nulls in the return loss. So a tracker and
return
loss bridge on a spectrum analyser can be used this way.
Filters with real-axis zeroes give stopband nulls... EG Elliptic ==
'Complete Cauer' designs. these are usually designed by the transformed
lowpass technique and suffer from its limitations.
I've played around with narrow bandpass filters with notches achieved
by
basing things on pseudo-crystals.
David "
------ Original Message ------
From "Douglas Denny" <ddennyoptom@...>
To [email protected]
Date 09/10/2024 00:56:03
Subject Re: [RacalRA17forum] Marconi TF1168
>Craig,
>
>Thanks for your interesting reply.
>The method you described shorting of the 40Mhz resonators in turn
surprises
>me that it is reported to work, but admit I fail to understand it at
all.
>If it does work - great !.. a new relatively simple new method to
try.
>They are relatively closely coupled however, and the image impedances
>looking in and out at each resonator stage with the close coupling
would
be
>badly affected if shorted I would have thought - which in turn would
badly
>affect the bandwidth shape I would expect.
>It is true the two end resonators are the most sensitive to bandwidth
shape
>changing characteristics when adjusted, which is another indication
of the
>importance and sensitivity of the input and output impedence seen at
the
>filter by the pre and post circuitry. Interesting comments though...
thank
>you for that.
>
>I was always impressed at how acutely sensitive is the 40Mhz BPF for
>alignment, especialy so are the caps when adjusted - for obtaining the
>accuracy needed for a good passband shape. So much so that just
removing
>the cover and replacing it will change the passband shape - sometimes
>drastically.
>
>Douglas Denny G3ZQE
>
>On Mon, 7 Oct 2024, 16:41 Craig Sawyers via groups.io, <c.sawyers=
>[email protected]> wrote:
>
>> I'm lucky enough to have a R&S Swob II that I bought from Rob Filby
>> donkeys years ago (fully functional with all the probes) and a
CT501
>> that I bought for a bewilderingly low ?40 maybe 10 years ago -
clearly
>> no one knew what it was, and I was the only bid on eBay.
>>
>> BTW, although I have never tried it, apparently a sneaky trick
with
the
>> 40MHz filter is this. Short circuit every resonator except one.
Very
>> loosely couple a SA to the remaining resonator, and set that to to
>> 40MHz. Do that for each resonator in turn. Every one in isolation
set
to
>> 40MHz. When all the shorts are taken off, the resonator couplings
pull
>> each other to precisely the right frequencies to give the required
>> bandshape. Apparently the only ones that need care are the two end
>> resonators, because they have to be loaded correctly.
>>
>> This is from a guy whose career was designing RF stuff (like
spectrum
>> analysers) for HP. He tends to know every trick in the book
regarding
RF
>> filter design and alignment. He is also an RA17 fan.
>>
>> Yes the TF1186 was a strange and wonderful thing; I suspect that
not
>> many were made.
>>
>> I've just hauled my Tek 7L5 SA out; It is a while since I used it,
and
>> quite forgot that it has a tracking generator option. 10Hz to 5MHz,
with
>> a resolution bandwidth down to 10Hz. That I believe is more is more
than
>> capable of aligning the IF chain filters ;-)
>>
>> Incidentally Douglas, I was just reading an old post of yours
describing
>> the order of bandwidth filter alignment - IOW start with the 100Hz
and
>> work up in frequency.
>>
>> Craig
>>
>>
>> ------ Original Message ------
>> From "Douglas Denny" <ddennyoptom@...>
>> To [email protected]
>> Date 07/10/2024 13:16:52
>> Subject Re: [RacalRA17forum] Marconi TF1168
>>
>> >I still have two TF114H sig gens stored somewhere in the attic,
and
have
>> >aligned RA117s with them in the 1970s (along with an early
wobbulator
for
>> >the 40MHz bpf).
>> >They are a beautiful example of precision engineering in test
equipment of
>> >the time; if left on continuously were rock stable and very low
jitter and
>> >spectral purity compared to some modern sythesised equipments. I
think one
>> >amateur professed to use one as a vfo for vhf transmitter on 2
metres
>> >around that era.
>> >I have no knowledge about the TF1168.
>> >
>> >Douglas Denny. G3ZQE.
>> >
>> >
>> >
>> >On Mon, 7 Oct 2024, 11:58 Craig Sawyers via groups.io,
<c.sawyers=
>> >[email protected]> wrote:
>> >
>> >> This is pretty much a historical curiosity. This was the "book"
method
>> >> of aligning the crystal and other bandpass filters on the IF
strip, and
>> >> IF transformers in the RA17 series of receivers.
>> >>
>> >> Basically described as a "High Discrimination Oscillator", it
was
>> >> precision set to 100kHz against an internal crystal calibrator,
and had
>> >> three large tuning knobs covering 10kHz, 1kHz and 100Hz total
range,
>> >> selected by a switch.
>> >>
>> >> Nothing much on the web about this boatanchor (a hefty
40kg/88lbs), no
>> >> manuals anywhere. There are two references - one an old Marconi
>> >> catalogue that describes its characteristics, and a similarly
old
>> >> Marconi price list that says it was ?328 in 1961 (around ?10k
in
>> today's
>> >> money). However, as an example of the ravages of time and
value,
the
>> >> TF144 signal generator was even more expensive - but last year
I
was
>> >> given two. Both of which work bang on spec BTW, half a century
on.
>> >>
>> >> Of course now a spectrum analyser is the way to go. However,
for
>> >> curiosity's sake more than anything, does anyone know anything
about
>> >> this Marconi TF1168 beast?
>> >>
>> >> Craig
>> >>
>> >>
>> >>
>> >>
>> >>
>> >>
>> >
>> >
>> >
>> >
>> >
>>
>>
>>
>>
>>
>>
>
>
>
>
>
Douglas Denny