An Introduction to o f impeda nce.

G enerally the crystal resistance can be neg-

lected since the cflcctivc Q of the crystal is high and so it is
usual to plot reactance- frequency diagrams rather than
impedance graphs when con sidering a filter response. The
Crystal Filters graph of reactance against frequency for a single crystal is
shown in F ig. 2 a nd it is seen that as the frequency is increased
a series resonance occurs (zero reactance) followed by a
PART I parallel resonance (infinite reactance). These frequencies
arc respecti vely the resona nt frequency (fr) and the anti-
By HAMISH V. BELL, B.Sc. (G3MAZ)* resonant frequency (fa) of the crystal unit and are most
important in the consideration of crystal filters.
A nother way o f expressing these significant po ints is to
W lTH the ever increasing popularity of the single sideband
(A3a) mode of transm ission , attention is being focused
on what, for the amateur, are two unusual typ es of networks.
term points of zero reacta nce (such as.fi·) as ZEROS and points

One is the resistor-capacitor network which will act as a REACTANCE

90° phase shifti.ng device over a range of audio frequencies, I
and the other is the crystal filter. Jt is the intention of this
article to review the latter in a simple manner, and so
possibly elucidate some of its mysteries. INDUCTIVE
t I
I
I
I
I

It is well known that a quartz plate when mounted between : 10 tr~que:ncy

two metal electrodes and excited electrically will vibrate 0 r-------~£------L~--~------~

!/
mechanically at a certain fixed frequency. This frequency
is dependent upon the size and shape of the plate, and the CAPACITIVE
position in the parent crystal from which it was cut. Although
this is a mechanical vibration the effect can be represented t I
ncar resonance by an equivalent electrical circuit as show11
in Fig. 1. Fia. 2. Graph of reactance aa:ainst frequency for crystal .
The capacitance Co is the capacitance of the metal
electrodes with a quartz dielectric, an d includes the etTect of in finite reactance (such as fa) as POLES. The pole-zero
of any air gap between the electrodes and the crystal. The plot of the crystal equivalent circuit wi ll then look like Fig. 3.
components L, C and R arc governed by the cut and the It can be shown that the pole-zero spacing, S, is related to
dimensions of the quartz plate. Typical values of these . Co b 1
lt 1e ratio . fr C .. .
components for an 80 kc/s crystal might be as fol b ws: C y t 1c expreSSIOn S = 2 Eo· From thiS 11 1s
L = 39·7 H R = 900 ohms obvious that any capacity added in p ara llel or series with
C = O·JpF Co = l7pF
Q= '~L = 22,000 Fig. 3. Po le-zero plot of crystal.
This va lue of Q is much higher than any tha t can be o b-
tained u si ng inductors and capacitors with the rc.sull that
the crystal will decrease the spacing between the resonant
and tl1e anti-resonant frequencies.
If inductance is placed in series or parallel with a crystal
the opposite eflcct occurs. For the series case (Fig. 4), it is
Fig. I . Equivalent electrical seen that the c rystal pole remains fixed while its zero is
circuit of crystal near re!ion· lowered in frequency. A further zero or series resonance is
a_n ce. Effect of spur ious
frequencies shown by dashed also introduced higher in frequency than the c rystal anti-
lines. resonant frequency. In Fig. 5 the parallel case is illustrated
and here the c rystal zero re maills fixed while the pole is
moved higher than ils o riginal frequency. As in the series
case a further significant frequency is introduced, a pole,
lower in frequency than the c rystal zero.
Besides the main responses at fr and fa, other responses
filters employing crystals are capable of high attenuations may be found. These spurious responses arc higher in
outside the passband with rapid transitions from passband frequency than the parallel resonant frequency and are due
to stopband. to mechanical coupling within the crystal. They may be
. Co. I I' . . represented electricall y by further series circuits (shown by
Th c rat1o C 1mposes a fundamcnta 1m1tat1o n upon the
characteristics of the crysta l and thus of any filter in which
it is used . This ratio depends upon various electrical and
mechanical constants for the quartz and has a theoretica l
minimum value of 140 with stray capacity increasing this
figure. in practice. however, for low frequency crystals it
is found that a smaller va lue is rea lizable, usually around
125 [1].

:
The elect rical circuit o f Fig. I wi ll have two s ignificant CRYSTAL tr lo
frequencies. At these frequencies, the combination wi ll ALONE 0
appear either as a series tuned circuit or as a parallel tuned II !2
circuit , and at other freq uencies will have intermediate values WITH Lo 0 0
Fig. 4. Effect of induct~ncc in serie.s with crystal.

332 R.S.G.B. BULLETIN JANUARY, 1962

 L R C crystal arm is equal and opposite in polarity to the voltage

GJ
due to the capacitor arm, there will be a balance and .no
output will be obtained. This infers thai at some frequency
0 0 the reactance of the crystal must be capacitive and equal to
the reactance of C2. At th is frequency on the allcnuation-
frcqucncy curve there will appear a point of infinite allenua-
Lo tion. Lf on the other hand, the crystal reactance is inductive
and numerically equal to the reactance of the capacitor,
to then ihe output voltage will be a maximum since the voltages

::
CRYSTAL tr
X of the two arms will add together. At other points, the out-
ALONE
puts will vary between these extremes.
II 12 The reactance curves for the crystal and for the capacitor
WITH L 0 X X are shown in F ig. 7. Also included in this diagram are the
Fia. S. Effect of inductance in parallel with crystaL
appropriate attenuation-frequency curves for values of C2
greater than and less than Co, the crystal capacitance.
dashed lines in Fig. I) placed in parallel with the main It is seen that the passband extends from a frequency [1,
circuit, and their eA'eet will be referred to later. higher than the crystal resonant frequency to /2, a frequency
Having reviewed ihe electrical characteristics of the crystal just less than its anti-resonant frequency, i.e. it lies between
unit and seen how these characteristics may be modified by the highest and lowest significant frequencies. The passband
suitably connecting inductors and capacitors, the use of these is only a matter of a few hundred cycles in width and will
clements in filters can now be considered. not be flat. The frequency of infinite attenuation, [ aJ, is
below the passband when C2 is greater than Co and higher
Filter Configurations
The basic filter configuration may be considered to b~ the
lattice, or bridge circuit, since other configurations may be
transformed into the lattice form by means of mathematical
processes.
Crystal G ate
The simplest form of lattice filter and the one which will
be used to introduce the method of finding the passband and

Fir. 8. Four•crystal lattice filter.

OUTPUT

INPU~ than the passband when C2 is less than Co. The capacitor
C2 is the familiar phasing control which can be used to
reject an interfering carrier near to the desired signal by
adjusting the frequency of infinite attenuation to coincide
with the interfering signal.
Fia:. 6. Crystal gate or one-crystal lattice filter. The main disadvantages of this simple form of filter are
that the passband is not very wide, and also the frequency
frequencies of infinite attenuation, is the so-called .. crystal of infinite allenuation can only be placed on one side of
gate." This form of filter has been used for many years in the passband at one time. This results in the skirt on the
communication receivers as a means of increasing the i.f. other side sloping rather badly.
selectivity for c.w. and also to a certain extent for telephon y
as well. Four-Crystal Lattice
The circuit is shown in Fig. 6 and is. to all intents and The form of lattice usually favoured in commercial prac-
purposes, a bridge circuit. When the voltage due to the tice employs fo ur crystals as shown in Fig. 8. The coils which
terminate the filter can be shown mathematically to be
R~octonc~ Rcoctonc~
effectively in parallel with the crystals, thus spreading the
..
~
Jl crystal pole and zero frequencies and introducing a further
pole as was indicated in Fig. 5.
u
"'c:
'0
If the poles and zeros produced by crystal X I are coincident
with those of crystal XJ and the poles and zeros of X2 are
coincident with those of X4 then only the reactance diagram
~ of X I and X2 need be considered to obtain the attenuation
·v
0
characteristic. In the diagram of Fig. 9, the reactance of
a. the series crystal, X I, is represented by a solid line and the
0
v reactance of X2 by a dashed line. Crystal X2 has a resonant
.
A ttqnuot ton ·
I frequency higher than X I by the amount of the pole-zero
separation. Jn this way one of the poles of X2 coincides
with the zero of X I and one of the poles of X I coincides
with the zero o f X2. A passband then exists between the
lowest and highest significant frequencies, actually lying just
within these frequencies. The bandwidth can be made about
3 kc/s at frequencies up to 500 kc/s with tbis circuit. Two
frequencies of intlnitc auenuation arc produced. and these
can be controlled to some extent by a small capacitor across
one of the higher frequency crystals. as a lso can the flatness
Fie. 7. Reac.tance..frequency and ;attenuation-frequency graphs for o f the passband. However, a ll is not as simple as it seems
crystal -gate filter. with this fllt..:r. One o f the main requirements is thnt the

R.S.G.B. BULLETIN JANUARY, 1962 333

crystal frcquencit!S shall be precisely adjusted so that the a relatively high current demand when one crystal is series
zeros of the pair of series crystals a rc coincident with each resonant. This will tend to unbalance the input circuit which
other and with the poles of the shunt crystals. If the crystals is high impedance and so balanced for voltage only.
are not c losely matched, the passband will be unt vcn and Using surplus Ff241 type crystals. the input impedance of
also stopbands may appear within the passband. T hese this filter configuration will be about 250 K ohms for a one
factors all add up to a circuit which is not the easiesr to per cent bandwidth at the frequencies employed, around 300
to 500 kc/s. This frequency range allows most normal i.f.
Reoctonco transformers to be used as i11put matching devices and pole-
zero spre:tding inductors when modified as in Fig. 11. The
lnductiv<t comments in the next section on dissipation should be borne
in mind when assessi11g the suitability of any particular
transformer, since several factors inHucnce the choice.
The crysta ls employed should be measured and adjusted
so that the relevant poles and zeros arc on the same frequency.
within a few cycles. Care should also be taken to ensure
Copocitivo that a ll the crysta ls show approximately the same activity.
T his is a n indication of their Q and since the frequencies of
the crystals in a h<tlf-lattice filter difTer by only 1·5 kc/s or so,
it is also an indication of their series resistance. lf desired.
the series resistance can be measured directly as described
Att<tnuotion later in the measurements section. The series resistances and
hence the Q va lues arc only important in that a large differ-
ence between crystals will affect the transmission through
the fi lter and, to a lesser extent, the slope of the skirts of
the passband. However these are not very critica l parameters
and differences in series resistance of up to 30 per cent should
provide ncceptable results.
Fir. 9. Four-crystal lattice filter r eactance-frequency and a ttentua- The half-lattice filter has been very popular for some time,
tfon-frequency craphs.
since the use of i.f. transformers for input :md output coupling
devices nnd the availability of suitable crystals simplifies
adjust. In commercial practio.:c some of the difficulties of construction. It is also quite tolerant of circuit constants so
matching arc overcome by plating two sets of electrodes on that by appropriate juggling of the transformer tuning slugs
a single quartz plate: the amateur has to turn to other circuit
configurations to achieve the same ends. L2

Two-Crystal Lattice
A suitable circuit is shown in Fig. 10 and is generally called
the hnlf lattice lilter, although in fact it is the unbalanced
equivalent of the four-crystal la !lice. The crystals X I a11d
X2 are the same as the series and shunt crystals in Fig. 8,
and have the same frequency relationship. The reactance
and attenuation diagrams arc the same as for the four-crystal
lattice. A capacitor C2 is placed across the higher frequency
crystal to enable its pole-zero spacing to be altered slightly.
In this way the crystals can be matched to give a flat passband Fia . II . Two-crystJ.IIattice w ith capacitive centre-tap.
with steep sides and a frequency of infinite attenuation above
and below the passband. The circuit Ll Cl is tuned to the and the capacitor in shunt with the high frequency crystal
mid-band frequency. as also is L2 C3. For best results the it is possible to achieve attenuations greater than 35db over
input circuit should be as shown, i.e. with the inductor Ll the stopbands with very little dip in the passband.
physically centre-tapped. but this is generally difficult to Where a higher attenuation in the stopband is required.
arrange because of the relatively low frequencies at which two or more sections may be placed in cascade so that their
these fi lters arc used. The alternative (Fig. II ) is to art ificially effects wi ll be additive. This procedure allows less critical
centrc-wp by means of two capacitors in series, connected adjusllncnt of the filter, a lthough even so it must be laid
in p lace of C l . These capacitors arc each twice the v;1 lue of out cmefully if capacitive coupling round the fi lter is to be
CJ and should be high quality silver mica types in order avoided and high a ttenuations achieved.
thnt the fi lter characteristic wi ll no t be greatly a ffected by
tcmper:llurc. Dissipation and Termination
T his method is usually satisfactory but could be contri- Since the properties of these filters depend upon the
butory to some attcntuation in the passband, since there is extremely high Q factors of quartz crystals it may be won-
dered what etfcct the Q of the inducto rs used in the filter has
on the overall characteristics. Owing to their high Q factors.
crystals can be considered dissipation-less compared with
coils. and so the actual filter without terminations does not
introduce loss within the passband. Any inductors in series
or para llel with a crystal can be transformed mathemat ically
to become part of the terminations; thus this concept of a
dissipation less filter is true for any filter. This trea tment leads
to th e view that the attenuatio n outside the passband is
mainly dependent upon the filter and any loss within the
passband is due to the terminations.
The lo~~ within the passband can be of two types. The
Fir~ 10. H :llf-lattice or two•crystal latticf' filt~r. (Crmtiuurd on paf(P 342}

334 R.S.G . B. BULLETIN JANUARY, 1962

Two iurther amateur band ranges arc available with crystals
providing coverage for the remaining portions of the 28 to An Introduction to Crystal Filters
29·7 Mc{s band, whilst five universal ranges of 600 kc{s each (Continued from p;oae lJ.4)
may be selected between 3 and 30 M c{s by the choice of
suitable crystals. This provides a total of twelve 600 kc{s first is frequency sensitive and gives rise to an uneven pass-
bands which may be selected from the fro nt panel. Internal band; the second provides a constant loss and can be con-
spurious responses, always a problem in this type of re- sidered as a purely resistive effect. This latter loss is of little
ceiver, are held to a level which is less than an equivalent consc.quence! since it can be made good in any amplifier
I p.V s ignal. The dial calibration is marked at intervals of assocmted w1th the filter, by merely increasing its gain.
10 kc/s, with I kc/s vernier dia l divisions on the rim of the The former type, however, is a lillie more difficu lt to
tuning knob, both of these being adjustable for calibration eliminate as it usually results in a deep dip in the middle of
purposes. The S meter, which is operative on aU modes, the passband. This effect can be due to two causes the first
forms part of a bridge circuit in the anodes of the r.f. and i.f. being that the pole of the l.f. crystal does not coin~ide with
amplifiers. However, after the S meter had been calibrated the zero of the h. f. crystal. Any dip due to this can only be
in accordance with the instruction manual it was found fui.Jy ~liminated by ensuring that these two frequencies do
that only a local signal would produce an S9 reading. cotncJde. The second cause arises when the dynamic
impedance of the input tuned circuit at resonance is com-
Operation parable with the input impedance of the filter, thus lowering
The receiver has been used for the reception of a ll modes the effective impedance across which the signal is developed.
of transmission and naturally the 3·6 kc{s bandwidth posi- If the dip becomes more than about 6db below the max_imum
tion provides the most faithful reproduction of a.m. signals, response of the filter, the distortion of signals passed through
whilst the 2·1 kc/s position is optimum for s.s.b. The 500 the filter is most noticeable. By using inductors having
c{s bandwidth enables c.w. signals to be read through heavy r~latively high Q fac~ors it is possible to adjust the termina-
QRM, and when the Q Multiplier is used this can provide tions so that the dtp m the passband is well below this figure
true single signal reception. At any position of the band- and thus a reasonable flat response may be obtained. A
widtll switch the passband tun ing control allows selection of co.il of 1·8mH having a Q value of 50 o r greater would enable
upper or lower sideband without readjustment of the main th1s. slate of affairs to be achieved at 465 kcfs with the half-
tuning control. Sideband signals are as easy to tune as their latllce filter described. However, since the dynamic impedance
a.m. counterparts, and in the event of heavy QRM the 500 of the coil depends upon the frequency and the inductance
c/s bandwidth is usable although obviously speech quality as well as the Q factor, it is not possible to generalise. A far
suffers. The adjacent channel selectivity is steep sided and beuer guide to the values required in any particular case
approaches that provided by mechanical filters whilst the can be obtained by considering the dynamic impedance
passbarids are continuously movable above and below the of the tuned circuit and comparing it with the filter input
fixed b.f.o. frequency of 50 kc/s. The bandwidth at 60db impedance in the passband.
down on the three selectivity positions is 2·75 kc/s, 7·5 kc/s, lfZd isthedynamicimpcdancc, t hen Zd = L/CR whereL, C
and 10·5 kc/s respectively.
It is desirable that a receiver capable of high selectivity
and R relatetothetunedcircuit,and Q = ~w HenceZd =
should also have good frequency stability and in this respect 2:rfofJL where fo is the ~es~nant frequency of the tuned
the Model 2-B is second to none. The manufacturers claim c1rcu1t. Now for least d1p m the passband the dynamic
a warm-up drift of less than 400 c/s, and a shift of Jess than impedance should be several times greater than the fi lter
JOO cycles after this initial period; experience has given no i!lP~t impe~ance, but if t~is is not possible, an absolute
grounds for argument with these figures, which are the hm.1t for sausfactory operauon can be considered to be when
result of the use of a crystal controlled first oscillator and a zd equals the filter input impedance.
carefully designed low frequency variable esciUator.
A review of this nature can only deal briefly with the
salient points of the equipment used and many details must A Three Band Cryst.a l Controlled Co nverter
be left unmentioned, but it should be said that the feel of (Continued from paee llS)
the main tuning control, which is coupled by a cord drive to amo':'nt of oscillator output on any band may be made by
the dial mechanism, left something to be desired and a altcnng the value of h.t. dropping resistor to the coil con-
geared flywheel arrangement would be preferred. The cerned.
mains transformer in the receiver is designed for 60 c/s Once the crystal oscillator is functioning satisfactor y, LJ
operation which probably accounts for a large proportion of and L2 should be adjusted so that signals on all three bands
the heat radiated from this component. However, during an may be peaked with the 100 pF tuning capacito rs. The
extended period of operation the Model 2-B gave every 14 Mc/s band should resonate ncar the maximum capacity of
satisfaction and is highly recommended as a communications these capacitors and 28 Mc/s near minimum capacity. Once
receiver of outstanding performance. The latter, in the set up for a given band no further adjustment should be
experience of the writer, has been excelled only by a receiver necessary, all tuning being carried out on the main receiver.
the cost of which was roughly three times that of the Drake
product. Results
The Medel 2-B receiver reviewed was supplied by K.W. The convener has been in regular use by the writer for
Electronics Ltd. of Vanguard Works, Heath Street. Darlford. five years working into a C R 100. The combination gives
Kent, and costs £138, including the crystal calibrator and cxce!l~~~ results from the stability point of view and the
loudspeaker/ Q Multiplier. sensJIJVJty appears to be very good. The converter has also
been used with other receivers during recent National Field
Days with complete reliability and has earned the approval
of all who have used it.
Since this article was written it has been pointed out that
The G4Z U FB-5 A e rial un~er certain conditions harmonics from the crystal
H E type of bead required for the G4ZU FB-5 aerial
T described in Technical Topics on page 283 of the Dec-
cember issue of the BuLLETIN is Mullard B4 pattern FXI308
oscillator may cause TV!. Reducing the HT voltage to the
cir<?uit concerned by increasing the value of the dropping
res1stance together with further decoupling of the HT line
and not FXJ300 as mentioned. will effect a cure.

342 R.S.G.B. BULLETIN JANUARY, 1962

 series inductor and capacitor in parallel with the components
An Introduction to giving the main resonance. .
The crystals used for the low frequency filters a re fortun-
ately free from spurious responses, due to the particular cut
Crystal Filters used. and so this factor is not important. This is not the
case with crystals designed for higher frequencies and indeed
some of the spurious responses can be as little as 3db below
PART 2-H.F. Filters the main response. The spurious frequencies may be moved
within certain limits during manufacture but even in a well-
By HAMISH V. BELL, B.Sc. (GJMAZ)* designed crystal cannot be eliminated entirely. The final
plac ing is dependent upon the cut and the physical properties
ItypeN low-frequency
I
Part the discussion related to crystal filters in the
was
I
spectrum below Mc/s since these are the
most generally used and about which the most is known.
of the crystal. Methods used to reduce spurious responses
include specia l grinding tec hniques and also the plating
of only part of the crystal on each face, the latter having
However, it has been shown [2 1 that the selectivity of a the secondary effect of reducing the parallel capacitance.
simple bandpass filter is dependent only upon the pass- Such measures. o f course. are not generally available to the
band width and the position of the frequency of infmite amateur but nevertheless it is interesting to consider com-
attentuation relative to the cut-off frequency. In other words, mercial practice.
it is not, in theory, dependent upon the centre frequency of
the filter. Suitable Crystals for H . F. Filte r s
" In theory," there of course lies the difficulty! There is a Coming now to crystals which are availnble cheaply to the
limitation in practice in the value of Q factor which is obtain- amateur, the obvious choice is the FT243. It is not , however,
able and in the stability of the crystal unit. Nevertheless,
there is a great deal of incentive to devise a workable high DESIRED S IGNAL DESIRED SIGNAL
frequency filter, since its use would be advantageous both 2 8 to 28·003 Mcfs +
INTER FE RING SIGNAL
~
in receivers and s.s.b. transmitters. Fewer frequency con-
versions would be required and high selectivity could be
achieved so that the receiver, or transmitter, would be less
j-~~-l~F~IL:T~E~R~------~
liable to generate spurious signals.
Assuming there is no limitation on the mid-frequency T+-----t._~_j BANDWIDTH
INTERFERING 5 to 5·003 Mc/s
of a filter. any configuration used at the lower frequencies SIGNAL SPURIOUS
can be used as a high frequency filter. However the more 28·010 M c/s RESONANCE
stringent requirements as regards stabil ity and Q factor 5·010 Mcfs
usually resull in designs which arc difficull to achieve practic- LOCAL OSCILLATOR
ally. 23 Mcfs
To make a four-crystal lattice filter for frequencies of the
order of 5 Mc/s with access only to surplus crystals would Ftc. l l. EHe:cr: o f Jpurious responses in niter u sed for reception.
be virtually an impossibility and so some other approach
just a ma!lcr of selecting a crystal marked with the required
must be found. frequency and hoping that it will fit the bill. Possibly the
One such approach is that suggested by W3TLN (QST.
figures in Table I, relating to one crystal of this type, will
January. 1959) based upon an article by Kosowsky in the
help to impress the point. Measurements were made on
several c rystals, using uccuratc frequency determining
equipment. and these figures are representative of the others.
It is seen that there arc three spurio us responses of which
the first is the one most likely to cause trouble.
In order to reduce the effect of these responses, several
filter sections should be used in cascade since it is unlikely
that two spurious responses will occur at the same frequency.
The result will be a lower value of attentuation at the spurious
resonance frequencies but no actual passband will be formed.
For filters used for transmitting. this is quite satisfactory,
Fie. 12 ... Back to back'' connection for hiah frequency filter •ua·
since the audio response can be shaped to cut otT sharply at
c e.sted by WJTLN . 300 and 3300 c/s. There is thus very littl~ s ignal passed to the
filter at the spurious resonance frequencies. For reception,
Proc.I.R.£. (February, 1958). This employs a pai r of two- TABLE I
crystal lattices connected "back to back" as in Fig. 12. M easurements on FT14l Crystal
The coil Ll serves as a load for each section and according
to W3TLN is apparently not critical as regards inductance, Marked frequency 6850 kc/s
although in common with any inductor associated with a Measured rc,:onanc: frequency 68~8·299 kc/s
filter, it should have as high a Q as is possible. For details Measured ~ ntircsonlnt frequency 6849·935 kc/s
Poh~·t.ero spJcinc 1·636 kc/s
regarding the construction of a filter of this type the article
by W3TLN should be consulted; instructions for shifting Spurious Responses
the crysta l frequencies are a lso given. Level relative
Higher th;:an
anti resonant frequency to m3in response
Spurious Responses
At this point. a short digressio n is permissible to consider 5·02~ kc/s - 12db
12·658 kc/s - 5db
the efl'cct of spurious crystal responses on the filter character- 13·759 kc/ s - II db
istic. It was mentioned earlier that spurious responses
could occur in crystals and could be represented by another Equlv:alent series rc.sinance OJt reson:ance =::::: 150 ohms
Effective p>nllel cop•city ~ 11 ·1 pF
• 151 Kings~·uy. 111/tllf~l. l.t.mc:s.

R.S.G .B. BULLETIN FEBRUARY, 1962 381

 Cl 2CI 2CI the case of surplus crystals, these
obviously a re not known.

:Fl:
Measurements can be made if
suitable apparatus is available,
and a method of carrying out
these is described later. With

:I I~
a knowledge of the approx-
imate values of the various
components it is, however,
possible to adjust the filter
experimentally.
(a)

Fie. 14. Buic sections of ladd e r filters (a) Lowpau, (b) Hiehpau, (c) Bandpass. Lowpass Filter
After some initial calculations
to determine approximate
however, the situation is somewhat different, since a signal component values, the experimental approach was adopted
at the input can beat with the local oscillator to produce the with a single section of a lowpass filter, using an Ff243 type
spurious resonance frequency, and will be passed to the crystal, whose resonant frequency was 8 150 kc/s. The
detector a long with the desired signaL This is illustrated in parallel capacitance of these units is high, in this case 13 pF
Fig. 13. GcneraUy the attentuation will be greater for the and so no additional parallel capacity was required.
interfering signal than for the desired s ignal, but their relative SmaU trimmers were used for C/ and C2, and the circuit
strengths may coumeract this. arranged as shown in Fig. 15. The impedance of the filter
depends upon the bandwidth and the crystal series capacity,
Ladder Filters but in general will be low, between a few hundred ohms and
A form of fi lter not generally employed at low frequencies 5K ohms. The resistor T-network is to enable the signal
is the ladder type, three forms of which are illustrated in generator to be isolated from the input of the filter. The
Fig. 14. The reason for its lack of use is the low percentage output was fed via a pi-network to match the impedance to
bandwith ava.ilablc since the filters are comprised of crystals a receiver fitted with an S-meter. The voltage divider action
and capacitors only. At high frequencies, however, although reduced U1e input to the receiver and in some cases it may be
the percentage bandwidth is the same, the actual bandwidth desirable to use a cathode-follower instead of the resistive
in cycles per second is sufficient for art i.f. amplifier. Will\ network. The filter would then be terminated by a 1·2K
the c rystals available, this configuration does look as though ohms resistor.
it may be a possible approach to high frequency crystal From this single section an attenuation of about I0-15db
filters. was obta ined in the stopband, but minor passbands appca.red
Owing to the large number of spurious responses, it is due to the spurious responses and here the attenuation
much better to employ only one crystal in each section and dropped to virtually zero. The rejection notch immediately
put a number of section s in series. For s.s.b. transmitters h. f. of the passband was very deep and sharp. As a result it
therefore it is only necessary to connect either a number of 8150 kc/s
lowpass sections or a number of highpass sections in series 3· 3K.O. 560.0. I·BK.O.
to select the required sideband, the audio response being
tailored to provide the necessary 3 kc/s bandwidth. For a
receiver i.f. amplifier it would be necessary to have a band-
560.0. 3-30 ·2·2 470
pass characteristic formed by putting a number of highpass pF Kn .n ~
filter sections after a number of lowpass sections. Alter-
natively a bandpass section could be used but this would be
more critical to adjust and the choice of rejection OSCILLATOR INPUT TO RECEIVER
frequencies would be limited. The suppression of spurious -
frequencies would also be reduced. Fir. IS. Experimental sin1le section of lowpass tadder filter usine
Taking the circuits of Fig. 14, it is seen that no d.c. path FTl43 crystal.
exists between input and output and so the response wi ll fall
off at low frequencies. The bandwidth depends upon the would provide a high degree of suppression for any signal
ratio C/ ~Co where C/ is any capacity added in parallel placed at that frequency such as a carrier. The disadvantage,
as was expected, was the na.rrow bandwidth available due to
with the crystal. The placing of the frequency of infinite the h igh parallel capacity of the crystal. The bandwidth was
. C2 of the order of 1·5 kc/s and so would not be suitable for the
attenuation will depend upon the ratio CJ + Co. These generation of s.s.b.
ratios bold in each of the three cases, the element values Highpass Filter
being modified as shown on the diagrams. A similar highpass filter was built using the same crystal,
If the maximum control over bandwidth and the placing but resu lts were not so good as for the lowpass version. The
of the attenuation frequency is to be obtained, the ratio ~o input and output impedances will be generally slightly higher
and the maximum attenuation is very dependent upon the
should be low, a value of 250 having been fow1d feasible for series resistance of the crystal at resonance. The passband
the range 5 to 8 Mc/s. However with FT243 types, it is was very narrow with the spurious responses providing
likely to have a much higher val ue than this, since the crystals alternate stop and passbands higher in frequency.
were designed for good frequency control cha.racteristics
rather than h.f. filter applications. Conclus ions on Ladder Filters
The formulae relating to the design of these fi lters are It wou ld seem that the highpass filter is likely to be more
quite simple, and can be easily applied, but they require a critical than the lowpass version as rega.rds requiring a low
knowledge of the values of the elements of tbe crystals. In series resistance crystal, inferring of course a high Q.factor.

382 R.S.G . B, BULLETIN FEBRUARY, 1962

[n either case, a lower value of parallel capacity than that stant over a range of ± 10 kc/s about the approximate
provided by the FT243 crystals would be most desirable. frequency of the crystal to be rneasured.
The variation of the parallel capacity between crystals of The valve millivoltmeter is only required to indicate the
different frequencies is quite wide and docs not show any changes in voltage as the oscillator is varied through the
definite trend. As a result it may be found that crystals of resonant and antiresonant frequencies of the crystal, and so
the same frequency have difrercnt values of parallel capaci- can be of quite simple design. A broadly tuned r.f. stage in
tance and if so, those with low values would be selected for front of a normal valve voltmeter should be sufficient to give
use in filters. a workable deflection. Alternatively a receiver having a
It should be possible to overcome this disadvantage of broad i.f. bandwidth and fitted with an S-meter could also
narrow bandwidths by placing an inductor Jn parallel with be used.
each crystal. This reduces the eRective parallel capacitance Since quite small frequency changes must be measured,
and so increases the pole-zero spacing. The value of induc- the best method is as illustrated in Fig. 16. The signal from
tance required depends upon the initial parallel capacitance the oscillator is fed to the crystal test circuit and also into a
of the crystal and the factor by which it must be reduced to mixer. The other input to the mixer is from a crystal, XI,
enable the desired passband width to be achieved. T he whose frequency differs fiom that of the crystal under test
effective circuit below the resonant frequency of the crystal (X2) by something of the order of 250 kc/s to l Mc/s. The
can be considered to consist of only the crystal parallel output of the mixer is broadly tunable over tllis range and so
capacitance C. and the added inductance L .. Consequently, theditlerence frequency can be measured by comparison with
a BC221 or s imilar frequency meter. The oscillation
the susceptance of this combination is lr.[C. [ 1- ; •] frequency of XI is measured in the normal manner, as
where fco is the parallel resonant frequency of Lo and c•. accurately as possible, although within limits this measure-
The effective value of c. can now be reduced to any selected ment does not govern the accuracy of the final results.
figure by the appropriate choice of /co and hence L.. The The oscillator is now tuned to the resonant frequency of
inductor should, if possible, be made variable to enable the X2 (where a maximum is indicated on the valve voltmeter)
passband to be adjusted easily after the filter has been built. and the beat frequency from the mixer is measured, using
This was not tried with the two sections described above, but the frequency meter. lf the two frequencies XI and X2 only
should ena ble the FT243 type of crystal to provide sufficient differ by a small amount, a calibrated audio frequency
bandwidth for s.s.b. transmitters and receivers. oscillator can be used in place of the frequency meter. The
The type of crystal which is cut especially for overtone anti-resonant frequency is found in a similar manner, where
operation will generally have a low parallel capacity at its the valve voltmeter indicates the first minimum higher in
fundamental frequency to enable the effective parallel cap- frequency than the series resouant frequency. A small
acity at the overtone frequency to be kept low. As capacitor whose value is known (say JO to 20 pF), is placed in
a result, these crystals will make good filter crystals on parallel with the crystal and the new anti resonant frequency
their fundamental frequencies, but of course the overtone is detern1ined. The actual frequency in each case is the
frequency may not be an exact multiple of the fundamental oscillating frequency of XI plus the difference as measured by
frequency. This is because the overtone is due to a mech- the frequency meter. This assumes Xl is lower in frequency
anical rather than an electrical harmonic and so the effective than X2. If the reverse is true, then the small difrerence
element values may produce a slightly different output frequency should be subtracted from the oscillating frequency
frequency to the electrical harmonic. It is therefore necessary of XI.
to measure the fundan'lental frequency carefully. lf the resonant frequency is fr, the antiresonant frequency
Further information on ladder filters, their design and fa, and the antiresonant frequency when the crystal is
matching can be obtained from the monograph mentioned shunted by capacitor Cl is f~> then C 0 can be found as
in reference [2]. follows :
f 1Z - fr~]
Measurements Co = Cl [ fa 2 - f 12
Measurements of the inductance, capacitance and resis- where Co and Cl a re in picofarads, and the frequencies are
tance of a crystal cannot be made directly, but have to be in megacycles per second. This value of Co includes any
deduced from measurements of the significant frequencies. stray capacitance introduced by the measuring circuit,
The instruments required for this purpose a rc an oscillator a lthough the resistive network in Fig. 16 is designed to
covering the desired range and fitted with a fine tuning reduce stray capacitance due to the measuring instruments.
control, a valve millivoltmeter and an accurate frequency The equivalent series capacity is next found using the
measuring device. following relationship :-
The oscillator should produce several volts output at a /az _ frZ]
low impedance and this output should be reasonably con- C = Co [ jrl
where C and Co are in picofarads and the frequencies in
DIFFERENCE megacycles per second.
FREQUENCY The approximate formula,
MEASURED BY C ~ 2.Co.S
FREQUENCY
METER fr. 106
can be used, the pole-zero spacing S being in cycles per
second while fr and the capacitances are in the same units as
before.
From tl1is result and the normal equation for a resonant
circuit, the inductance in henries of the crystal (L} can be
determined:
1 .
L = 4 .,., fr 2 C lrenrtes
where fr is in megacycles per second and C is in picofarads.
Fia. 16. Circuit used for crystal meaaurements. (conduded 011 page 392)

R.S.G.B. BULLETIN FEBRUARY, 1962 383

On lookers at the Opening Ceremony
which was broadcast in sound by
t he B.B.C. and filmed (or television.
(Photo by johrl Miles, Falmouth)

beam loaned by J-Beam Aerials ltd. and a Marconi-built G31GV, G3MFW. G3MSU. G30GT, G30HM, GJOJN,
Army mobile s.s.b. station. G30CB lent his home-built G30YW, G3PEP and GSA W. The overall opera tion was
receiver and G3BHC a transmiuer. under the leadership of J ohn Watson (G3AET), Chairman
Considerable publicity was given to the Memorial Station of the C lub.
by B. B.C. Television and. Westward Television and the Thanks are due to Mr. Boaz, Falmouth Depot Manager
opening ceremo ny was broadcast in sound by the B.B.C .'s of Marconi's Wireless Telegraph Co. Ltd. for his willing
home and ove(seas services. co-operation, the War Office, the Directors and Manageress
The operation reflects great credit on all members ofthe ( Mrs . Shoesmith) oftht Poldhu Hotel and the National Trust.
Cornish Rad ioandTelvision C lub. Behind the scenes workers A film of the event was made by G30UZ and together
included G2BHW, G2DDR, G3NKE. GJNUJ, G3NVJ, with a tape recorded commentary wi ll be available on loan
G6LY and many others. The operators wercG 2JL, G2BHW, to clubs and groups in return for a donation to the C heshire
G3XC, G3AET, G 3CZZ. G3DCJ, G3DLH, G3HES,G3HZV, Homes.

Bibliog r a phy
An Introd uction to C rystal Filters (Continued from p age JB3) Kosowsky. q. r., " The Synthesis a nd Realization of
Crystal Filters'' (Technical Report 298, June 1955,
If desired. the equivalent series resistance o f the crysta l Research Labo ra tory of Electro nics, M.I.T.)
can be found quite easily by noting the reading on the valve Mason. W. P.. " Electro-mechanical Transducers and
voltmeter at the resonant frequency and then replacing the Wave Filters " (D. Van Nostra nd).
crystal by known values of resistance until the same reading "Single Sideband for the Radio Amateur," A.R.R .L.
is obtained. The value of resistance giving the same reading
is the value of the equivalent series resistance. Values
ranging from a few hundred ohms to several ki lohms will
be found with normal crystals. In general, the lower the
series resistance then the higher is the Q and so the more Mo dernizi ng the R.C.A.-AR88 Receive r
(Continued from page 390)
active is the crystal.
The frequency and amplitude of the spurious responses can
be measured using the same method, and hence a full picture (ix) Re-check all adjustments so that flat-topped response
of each crystal can be o btained. This enables each crysta l to curves are obtained o n positions 2, 3 and 4.
be placed in the position where it will be most useful. It should be noted that the response curve of a half-
lattice filter is affected by the Q of the associated coils,
Conclusion especially the loading coils (T4 a nd Tb) . The original T4
The subject of crystal filters is highly complex and should be about right but it may be necessary to experiment
obviously cannot be fully covered in an article of this nature. with Tb in order to get the best resu lts. If there is a deep
Further p ractical difficulties are encountered in the lack of trough in the middle of the response curve on position 2.
suitable crystals and this undoubtedly hampers experiments which cannot be removed by careful trimming, then the L/C
in the field of high frequency filters. It is hoped, howeve r. ratio of Tb is too low and a difl'erent one should be tried. It
that this simplified. treatment will serve as a n introduction to is. in fact, desirable to select Tb to tune with 100 pF or less,
the subject as a whole and so possibly provide a foundation so that the response curve o n position 2 is slightly rounded.
on which further knowledge a nd experiment can be based. Steeper sides to the response curve may be o btained by
using a little neutralization across each h.f. crystal. This
Refe rences device, however, has the disadvantage o f producing side
[1) Starr, A. T .. "Elect ric ~ircuits and Wave Filters" lobes to the response curve (caused by the unbalancing
App. IX p. 451 (Pitman): cfl'e ct on the circui t) and is, therefore, not recommended
[2) Sykes, R. A., " A New approach to the design of especially as ihe sides o f the response with a ll three filters
high frequency crystal filters " switched in are already so steep that the slight improvement
(Bell System Monograph 3180). obtainable by using neutralization would scarcely be noticed.

392 R.S.G .B. BULLETIN FEBRUARY, 1962