Kitchen AM Radio 2015.
This page is about how to get best sounding AM radio reception of strong
local radio stations.
Fig 1. Kitchen Radio, 2015.
The plywood box has the tubed AM radio circuit.
There's a 1970s generic Pioneer AM-FM tuner sitting on top of radio.
Its only used to provide mono FM sound. The AM radio has a single EL34 in triode
for 5W which makes fine music for the 80 litre speaker box on floor. The
mechanism uses a sliding pointer along a bar
with a dial cord & wheel on the
capacitor tuning gangs. The tuning meter was recently made in Taiwan in about 1993.

Box was made from material from a 1950s radio-gram.

I built this radio in 1999. I had been fixing old radios and guitar amps since
1992, and I didn't need to try to find enough spare parts, aka old junk to build a
complete radio. The junk seemed to just grow because ppl donated what they didn't
want, or they didn't want to fund the repairs. I first tried to build a synchronous AM
radio similar to the design by Mr D.G.Tucker published in
1947 copy of Wireless
World magazine. This respected UK magazine had numerous articles on radio and
audio and is very good reading for anyone interested in electronics before 1960.
"Synchrodyne" used used mainly pentodes and while I should have
succeeded, I found several aspects too difficult to make or understand, and
I realized a
well done "superhet" circuit would be be easier and better for
reception of strong local stations.

By 1999 I thought a good superhet needed a different approach to what one finds
in nearly every AM set, a bunch parts which have been designed by greedy company
CEOs and their accountants. All the many mass manufactured AM sets I have ever
repaired had disappointing performance including spurious noise, hum, high levels
of THD/IMD, and limited RF bandwidth, and were prone to noise from countless
devices such as compact fluorescent lamps, phone answering machines, and many
other things which slipped past all the regulatory authorities' rules on RFI.

The 80L ported bass reflex speaker has 1953 300mm 
"Rola Deluxe" driver and a
Foster 25mm dome tweeter. The two units produce a flat response with crossover
at 4kHz; the old Rola had sensitivity over 95dB / W and could reach 5kHz without typical
cone break up mode uneven response. So no need for a 3 way speaker system.
The Rola worked well with a reflex box with ports, but bass below 70Hz was down so
the crossover includes a series network with L // 8r0 which reduced all above 70Hz by
4dB and extended bass response to 40Hz. I ended up with sensitivity of 92dB and
response 40Hz to 15kHz. There is plenty of low bass in much music and I wanted to
hear it all. The overall result
sounds natural, with warm midrange and clear treble free
of sibilance.

Controls are all MANUAL but within easy arm reach from kitchen table where I spend
time for meals, thinking and diary writing. This habit of life in a kitchen is a highly
specialized anti-marriage technique, a pity really, but if any sheila spent longer than a
minute in close proximity she'd be thinking of "makeovers" including
banishment of
man +
junk from kitchen. ( Why can't women just sit and listen and think? )

It is only mono channel sound - quite OK for a kitchen. AM comes from the tubed
tuner and FM from the
Pioneer AM/FM tuner. The Pioneer is supposed to give good
FM but in fact I find it sounds dull / honky compared to AM. I can switch between the
same ABC networked broadcast news from AM Radio National 846kHz to Classic FM
and the AM always sounds crisper - cleaner and "less boxed". I have to turn up the
treble FM, although tests revealed FM tuner did have AF response to 16kHz, -3dB.
I found treble boost or cut above 2kHz was essential because it seems the FM or AM
transmitted has much varying levels of treble. But I've found +/= 6dB boost or cut
is enough. The AM source from Pioneer tuner is nowhere near as good as my tubed
AM tuner.

I have reasonably good AGC control for some locals here such as print handicapped
station RPH at 1,125kHz. It is usually a good quality signal but its only
while other stations nearby are 5kW, such as
ABC Radio National 846kHz.

To get away from the normal old radio sound I used :-

A. Ferrite rod antenna with a single tapped winding, Q < 35, which tracks well to
the oscillator tuned LC, and which picks up magnetic part of radio signal, not the
electrostatic part which includes so much noise from compact fluorescent lamps etc.
( The alternative to a ferrite rod antenna is the shielded loop and they are good, but
you need room for them. )
B. 6N8 pentode for
input RF amp with AGC voltage applied for auto volume control.
6AN7  frequency converter, a well known reliable tube, with partial AGC applied.
D. Modified IF transformers made originally in about 1950, for wider pass band,
while retaining "skirt" selectivity.
E. 6BX6
"sharp cut off pentode" for linear IF amp without AGC.
12AU7 cathode follower for linear low Z drive to diode + C + R AF detector
followed by RC ripple filter, with output follower to reduce audio THD / IMD > 0.5% for
85% modulation and negligible for 0% to 30% modulation which has the majority of the
dynamic range broadcast. This assumes IF carrier level applied to detector is between
4Vpk and 8Vpk.
F. Two pairs of L+R RCA terminals for selectable external audio inputs from
FM tuner, radio gram, CD, i-pod, etc.
G. Line level preamp with variable shunt NFB to equalize levels of all sources.
H. Passive treble control to boost or cut F above 2kHz to max of +/- 6dB at 10kHz.
EL34 in triode for SE amp to make 4.5 Watts max, THD < 1%, allowing use
of any available wide bandwidth hi-fi speaker.
H. Heavy hi-fi speaker to support radio set, and which does not color sound,
in a separate cabinet so audio vibrations cannot cause microphony with tubes.
I. PSU with high value electrolytic C and Si diodes, and more bypassing of
B+ and B- rails for total noise free operation.  

SHEET 1. Kitchen AM radio, RF, Mixer, IF.

RF input is generated from a ferrite rod antenna. This produces much lower signal
level than the old fashioned RFT normally fitted to most older sets which needed an
antenna which was at least a 3 meter length of wire to a curtain rood at a window.
The ferrite rod antenna picks up the magnetic portion of the radio wave which has
much less electrostatic noise than when using a wire antenna.

The Q of the ferrite antenna is high, over 50 is possible, so that RF bandwidth at
550kHz is 11kHz, which means audio BW could only be 5.5kHz if the following IF
bandwidth was wider.
I found best operation was with slight Q reduction with 47k R
loading across the ferrite LC
. Resulting LC bandwidth is 20kHz minimum at 550kHz for
a resulting Q = 27.5.

I've often used a ferrite rod antenna to replace the small RF input transformers in old
sets which only suited a wire antenna. I'd buy a bare rod, still available in 2015, either
150mm or 100mm long, and wind a coil with about 50 turns of solid Cu wire from
Cat-5 cable. This 0.5mm dia wire has about 0.25mm polythene insulation, and Q can
easily be over 50 at 550kHz, and coil self C < 5pF. The coil turns depend on amount
of inductance wanted to suit the tuning gang capacitance. Many old sets have tuning
C = 5pF to 440pF.  A rod antenna has its coil usually set at 1/3 the way along the rod
which allows coil to be moved further towards a rod end or to center so allowing
inductance to be varied for best tuning at the low end of AM band with tuning cap
maximally meshed. With that done, tuning at high end of band is done with trimmer C
across tuning C which is fully open.

If C max with tuner C plus a trimmer C = 460pF, the theoretical L in mH needed may
be calculated :-
L = 25.35 / ( MHz squared x CpF )
For 530kHz, and 440pF,
L = 25.35 / ( 0.53 x 0.53 x 440 ) = 0.202mH, ( or 202uH ).
The coil can be on a cardboard former able to slide along rod to change L value about
+/- 30%. This range is seldom needed.
A coil position is found tuning F = 530kHz when tuning C is maximum at say 460pF.

Within the 50 turns, a tap for grid is at about 13 turns away from the earthy end of coil.
The wires from tap to grid and from live end of coil to tuning gang must be less than
70mm so they do not act as wire antennas. The grid tap has low output impedance and
the tube input capacitance does not react with coil to limit HF tuning.

The rod+coil should be able to be tuned across a range of 531kHz to 1.720MHz.
The ratio of HF : LF = 3.24 :1. The theoretical C minimum to get to 1.72Mhz is 43pF,
comprising the minimum C of tuning gang + trimmer C + coil self C + any
stray C between wires and chassis.
Not all AM radios achieve this range of F because the total minimum C across the coil
is much more than I say is possible here. I found many older sets could only manage
550kHz to 1.5MHz. This didn't bother many ppl because most AM stations are within
this reduced band. But if you lived where your favorite was at 1.6MHz, then you would
be disappointed.

The LC Q reduces as the F increases for the high end of the AM band. This is
OK because the purpose of the LC input coil is to give MODERATE selectivity so that
a 5kW station does not overload a mixer causing "cross modulation" interference when
listening to a 300Watt station only 45kHz away. 
There could be more than one ferrite rod+coil arranged to point in different directions
to favor one wanted station, but the operation becomes difficult for anyone to get right
and benefits are few.
Ferrite rods are directional, and like AM loop antennas, should be able to rotated in a
horizontal plane at least 90 degrees. Usually a position can be found to favor most

The low RF signal from my ferrite rod LC means a following RF amp is needed.
I chose a mini 9 pin tube 6N8 remote cut off pentode with AGC applied to vary its gain
for automatic volume control. The 6N8 boosts the RF to be applied to a triode-heptode

6AN7 frequency converter (aka mixer ). 6AN7 anode output feeds IFT1 primary, and IFT1
sec feeds a 6BX6/EF80 sharp cut off pentode IF amp which feeds IFT2 primary.
IFT2 sec feeds the detector circuit on SHEET 3.

There should be a future supply of these 3 tube types which were made in huge numbers
until about 1970. The 6N8 was chosen because I have a few, and is suitable for gain control
with AGC voltage. Its two diodes are normally used in an IF amp stage to generate AGC Vdc,
but are not used here. A good alternative to 6N8 is 7pin 6BA6 remote cut off pentode.

V1 6N8 anode is fed through an untuned choke L1 which is a spare old input RFT
normally used for for a wire antenna, arranged with pri and sec in series to make
an "informal choke" of about 0.85mH, and with low C and high enough self resonant Fo.
Such a choke has reactance of 2k9 at 550kHz, rising to a peak at about 1.4MHz
due to its self resonance. The R3 2k2 R4 22k across L1 allow viewing V1 performance
on a CRO with low input C probe without affecting V1 performance.

With no tuned LC in V1 anode circuit there is no further RF bandwidth reduction and
all following selectivity is due to the two IFTs. The gain of V1 is quite low because
it spends most of its time with AGC about -3Vdc with strong local stations which keep
gm low for low gain. V1 also has unbypassed Rk 470r. The overall effect is
that very low power local stations of 300Watts generate low AGC to get V1 to have
maximum gain, so their sound level is only slightly lower than strong stations, needing
a small increase of volume control.

V2 6AN7 triode-hexode is my favorite 1960s frequency converter. The cathode is
common for both tube sections and connected to 0V. The triode acts with an oscillator
transformer with tuned grid LC which generates a huge
40Vrms with negative Vdc bias
generated by C11+R10 which stabilizes the oscillator level. The triode grid feeds grid 3
in hexode by internal connection. The triode anode powers the grid coil via an untuned
anode coil with less turns than grid coil.

The hexode is thus being turned on and off by the oscillator signal. The hexode
operates as a non linear amp with dual grid grid inputs, and interaction due to
capacitance between the two two inputs is avoided by having two screen grids.

Any RF input to hexode g1 pin 2 produces a number of intermodulation harmonic
products in the heptode anode current. The anode Ra is fairly high. An oscilloscope
will show hexode anode voltage at IFT1 primary to be mainly the oscillator frequency
in addition to a blur of other frequencies. The IFT1 LC has its highest tuned Z at 455kHz,
and all F away from 455kHz are shunted by L or C. The magnetic field changes
mostly at the tuned 455kHz, so output from the IFT1 secondary shows mainly only 455kHz.

The oscillator is arranged to always run at 455kHz above the tuned F for the input
RFT with rod+coil and tuning gang. So if RF input range is 530kHz to 1,730kHz,
oscillator moves from 985kHz to 2,185kHz, a ratio of 2.22, a lower F range change
which should be easy to get by using just the right value of fixed C in series with
a tuning gang identical to that for the input RF LC.
In my radio, I have a fairly miniature Kriesler brand two gang tuning C made in about
1965 to suit the last of the tube sets and beginning of SS circuits.
The oscillator gang has smaller moving plates to produce a smaller C change
and to give oscillator F which is more likely to always be 455kHz above the
RF input. Kreisler saved the 10c cost of 1 series C. Often the sets with identical C
gangs had a series C which changed its C slightly with temperature to compensate
for tuning drift as the whole set warmed up from say 15C to say 40C inside the box.
Such a C is not essential.
The action of having oscillator F always being 455kHz above input RF is called
"tracking" and in practice it is impossible to get perfect tracking when using
identical tuning gangs and a fixed C in series with oscillator gang. RDH4
explains it all better than I have room for here. Kreisler probably used the non identical
tuning gangs to get better tracking, but I have never been sure they achieved that goal.

Alignment of this type of radio, calibration of dial.....
Gear needed :-
Oscilloscope, RF signal gene with Ro 50r and for 455kHz to 1.75MHz
at least and with adjustable AM modulation between 0% and 100%, with envelope
THD < 1% at 90% mod.
Digital frequency meter. Voltmeters. Screw driver with plastic or wood shaft with
small metal end bit for adjusting IFT screws to move ferrite cores. Screwdrivers must
have no effect on C or L of the LC coils.

External PSU with Ro < 1k0, and with adjustable Vo for -10Vdc.
Several 300mm leads of insulated wire with alligator clip at each end.
Leads to bring test signals to points on RF and IF circuits.

1. Turn on radio, wait until set is warm after at least 15 minutes. Spend this time
checking Vdc and Idc around the circuitry.

Connect DMM to read Vdc at
TP2. The aim is to adjust 2 tuned LC in each IFT to
get wanted band pass shape.
Connect short lead with alligator clips across 47k at ferrite rod input.
Connect oscilloscope to output of AF detector at V4b cathode. Turn volume control low.

Connect low level 455kHz signal without mod to anode of V1 6N8 via 5pF capacitor.
Increase 455kHz until -Vdc at TP2 begins to become more negative than when first
measured without any 455kHz.
tuning each L in this order to achieve maximum increase of Vdc at TP2 :-
IFT2 sec, IFT2 pri, IFT1 sec, IFT1 pri. Repeat the process 3 times, and if Vdc at TP2
exceeds -10Vdc, reduce the 455kHz input.
Use 1kHz modulation to sig gene and watch detected AF wave on CRO. At 95%
modulation, there should be no clipping of detected 1kHz. There should be no change to
Vdc at TP2. If all appears well, initial IFT alignment is OK.

3. IFT1 now much have primary tuned to 460kHz, and secondary tuned to 450kHz.
This will widen the overall pass band. The best way to check the shape of the pass
band is to apply 455kHz which is frequency modulated to give +/- 40kHz swing.
The shape of the pass band is seen on CRO at V4a cathode and there may be 2
peaks each side of 455kHz, but exceeding 1dB above center 455kHz.
careful adjustment of sig gene F and with no FM or AM mod will allow -3dB poles
to be recorded, at about 8kHz above and below 455kHz. The slope of attenuation
further away from 455khz should have no slight peaks.

4. With 50% AM modulation of 455kHz, the audio response at V4b cathode should give
a flat line between 20Hz to 7kHz where signal begins to reduce, with -3dB at about 8kHz,
and -6dB at 10kHz. If these results are obtained, the IF amp is aligned OK.

5. Remove
V2 6AN7. Connect CRO to TP1 R3 to monitor V1 anode output.

6. Remove link across 47k at ferrite rod. Connect sign gen with no mod to V1 grid input
5pF and low level input.

7. Close tuning cap to maximum C. Adjust ferrite rod coil position for max signal at TP1
for 530kHz. Calibrate the 530kHz dial position on a cardboard template fixed to dial plate.
Prevent coil moving with masking tape. 

Open tuning cap to minimum C, and adjust trimmer C for mid position.
Adjust sig gene for higher F to find the highest F for the LC. If you get 1,600kHz,
open trimmer C, and if you then get 1,900kHz, all is well and you may set gene
to 1,730kHz and adjust trimmer C for peaking at this F.
Repeat steps 7 & 8, and plot the card for 530 and 1,730.
If the HF cannot reach 1,730 with trimmer opened, it indicates coil self C is too high
and turns must be reduced. So remove 10% of turns from rod coil at tuning cap end
and and repeat 7 and 8. The coil will need to be moved slightly towards centre
of rod for 530 and the HF should reach 1,730 with trimmer mid-set. This may need to
be repeated 3 times to get a coil which will give the wanted 530 to 1,730 range,
HF:LF ratio 3.26:1.
With this done, calibrate dial card in pencil above 530 to 1,730 in 100kHz steps,
600, 700, 800, etc.

9. Plug in V2 6AN7.

10. Connect fixed -10Vdc to TP2, so that AGC action during tests is avoided.

11. Apply 600kHz RF, no mod, at V1 grid input and tune the set so dial pointer
is at 600kHz.
Connect F meter to V4a cathode.

12. Adjust oscillator coil inductance by turning screw of its ferrite slug so that
IF frequency = 455kHz, confirmed by frequency meter.

13. Apply 1,500kHz RF input and tune the set so pointer is at 1,500kHz.
Adjust the trimmer C across the oscillator tuning gang so IF = 455kHz.

14. Reset RF input to 650kHz and tune the set back down to 650kHz and
on the same dial position as step 11. The IF should be 455kHz. If not,
then it means the change to C of oscillator LC at 1,500kHz has also changed
the Fo at 650kHz, so repeat steps 11 & 14.

15. If you cannot get 455kHz IF at the two dial positions, and you are 10kHz
away, then you have a problem with range of oscillator F and the oscillator
tuning gang. In my set I have dissimilar RF and Osc F gangs, and fortunately
I got the initial tracking to be OK for step 14.
If there are identical gangs, there will be a C in series with Osc gang. This
C may have to measured, and a smaller value connected and with a shunting
trimmer C of say 10-40pF

16. Step 14 is repeated with series C adjusted until the dial can be set to 600
or 1,500 and IF = 455kHz, +/- 2kHz.

17. Set dial pointer to 1,000 and apply 1,000kHz RF. If you are very lucky, the IF
= 455kHz, +/-2kHz.
While you may eventually get the IF to be 455khz at two dial positions, at
other positions the IF is 5kHz away from 455kHz. This can be unavoidable, and
what you have is the best anyone could ever do. Fortunately, the small error in
tracking will not ruin the overall performance of the radio.

18. Allow several days to get this right, with lots of swearing, cursing, and mistakes.
But after doing 30 radios, you'll get the hang.

V3 6BX6 IF amp is a mini 9 pin high gm "sharp cut off" pentode which is not
designed to have AGC applied. Its anode loading is the high tuned impedance
of IFT2 primary LC input. Thus anode Vac can be up to 100Vrms but Ia change
is only a few mA, so distortion is less than 5%. Normal operating Va will be much
lower than 100Vrms, and THD is low. I have an unbypassed R16 = 180r which gives
local current FB to further reduce distortion of the IF envelope shape.

The 6BX6 anode signal creates a negative AGC Vdc via C16 33pF and IN914.
AF is filtered low by RCRC networks with R13+R17 680k, and C9+C14 47n.
Each RC has pole at 5Hz, so the audio LF or hum cannot modulate the signals in V1
and V2.

A fixed -1.45Vdc is applied from bottom of R23 to ensure there is always
a slightly negative bias to grids of 6N8 and 6AN7, and prevent gain becoming
too high which causes grid current, excessive Iadc, instability and distortions
in V1, V2.

ABC Radio National at 846khz broadcasts AM with 9kHz of audio BW which
sounds excellent if your radio has a total of 16kHz of combined RF and IF
bandwidth before the IF signal is applied to the audio detector. Almost no
easily available AM radios have IF bandwidth this wide.

One problem with wide IF bandwidth is that there's no way of easily knowing 
when the set is tuned correctly. The only time the set is correctly tuned is when
IF = 455kHz, which is the preset aligned Fo for the IFT LC. With wide band IF,
there is little change to audio volume during tuning. Tuning eye tubes or tuning
meters usually work on the changing levels of -Vdc AGC, and the maximum AGC
may not be when the IF = 455kHz. Reliance on -Vdc of AGC is only useful for
sets which have a narrow IF band with one peak at 455kHz, easily heard while

So I built a second additional IF amp
with bjts and 3 x IFTs found in transistor radios
to make a band pass filter with band about 1.2kHz to power a tuning meter.

Sheet 5. Tuning indication.
If tuning indicators were ever used in old radios, a magic eye tube or a
meter was powered by the -Vdc AGC level. But for most old sets you just twiddled
the tuning knob for the best sound with least HF sibilance. There was only one peak
of IF at 455kHz, and that was where audio volume was maximum.
Domestic radios were always tuned by ear for long distance or short wave when meters
or other indicators didn't show any reaction.

For hi-fi AM radio, the IF pass band is flat so AGC hardly changes for IF
change of +/- 6kHz.  The additional IF amp ( on a small board at rear of case )
has overall Q > 400 so that the meter only moves from far left to right when
IF is between about 447kHz and 462kHz. When I peak the meter with a tiny
movement of tuning knob I can get IF between 454.5kHz and 455.5kHz, ie,
within +/- 500Hz of wanted F.

Googling anything like this drew a blank; nobody else seems to gone to the
trouble I have, and it appears I have invented this indicator circuit method.
There are schematics using 455kHz ceramic filters or crystals but nothing
was able to do what I have here, with a high enough Q and simple circuit above.
The 3 tiny IFTs used in old AM radios are still easily available. Unlike IFTs for
tube radios, these have only one tuned primary winding with a tap at 33% and a
fixed C across all turns. Idc feeds the tap from +12Vdc, and bottom of coil
is driven by collector. There is a second winding with no C and less turns than
tuned winding. So there are 6 connections, 5 for the coils and 1 to ground
the transformer can. The secondary coil with less turns suits the low base input
resistance of a following stage. All this must be well understood before working
with such junk.
With such IFTs, there is no flat topped pass band to widen the AF response.
I am exploiting the limited bandwidth to make an indicator. The high Q of these
LC depends on not having R or a bjt loading all primary turns. Most cheap
transistor radios and AM sections of AM-FM tuners achieved IF pass band
of 3kHz, allowing only 1.5kHz of audio bandwidth, so Q of 2 IFTs is about
150. With 3, Q is over 400.

The 455kHz signal from low resistance source at V4 cathode has
bandwidth = 17kHz, so it does not affect the Q of the bandpass filter
for meter.

Brighter tech heads reading here may laugh at what I am doing and insist
I should use a phase locked loop with 455kHz crystal controlled oscillator
to provide a +/- Vdc signal to drive a meter or magic eye.
I won't hold my breath waiting for them to post up a suitable simple schematic
which they have tried and tested.

Or they may like me to use a 455kHz crystal controlled oscillator with PLL and
apply +/- Vdc to alter bias of vari-cap diodes to change the C of the tuned
oscillator LC network. This means once a station is tuned, there is Automatic
Frequency Control, AFC, which locks the variable oscillator F to whatever F is
required to give a 455kHz difference between station RF and the oscillator RF.
If the station is at 846kHz, oscillator should be at 1,301kHz. The PLL should try to
keep the oscillator at 1,301kHz if the RF of input LC set is tuned away up to +/-
10kHz from 846kHz. Once tuned far enough away from 846,
say to 864kHz, there is insufficient IF signal generated to be compared to
crystal 455kHz oscillator, so the "lock" releases to let the variable tuned to
vary when tuning for another station. This way a set is either tuned, or it isn't,
but the RF still needs to be tuned to give maximum RF input to the mixer tube
so a meter driven by AVC Vdc is needed. There can be LF oscillations if the
PLL "hunts for a lock". Ordinary varicap diodes cannot operate in a tubed oscillator
which generates such high signal the diodes conduct. However, medium voltage
zener diodes of say 68V, 5W rated can have up to 50Vdc applied and they don't
conduct, so being better for tubes than low voltage rated varicap diodes. I have
used 5 in parallel to make a splendid vari-cap to change an RF gene for 455kHz
to give FM, 455kHz +/- 40kHz, to plot response of the IF amps, similar to using a
"sweep", or "wobbulator".

If anyone wants to go to all this trouble, they are welcome. An AM radio can easily
have far more complexity than it really needs to have, and if you want to explore 
complexity for better performance then I suggest synchronous reception and AF detection.

I welcome anyone to give me a SIMPLER schematic involving a crystal controlled
455kHz oscillator and a ratio detector or IC to compare radio IF to the oscillator
F and thus be able to produce a similar easy-to-built-and-use outcome.
Ratio detectors with 6AU6 in FM tube radios became very popular but needed a
special 10.7MHz IFT. Those are fairly easy to make with solid wire, but for
the AM band it may be more difficult.

Much more about AM radio may be read in books written before 1960.
RDH4 is quite good. I hope my approach with tubes performs better than
anything from old books which seem to show the same basic stuff full of
shortcomings so carefully included by penny pinching company accountants.

I cannot advise anyone to make their own IFT transformers. They are
most unlikely to get the performance available from the many old and usable
IFTs from the past which used litz wire and just the right type of ferrite cores
with screw adjustment and the right size of shielding can.
The larger cans about 100mm high and 40mm x 40mm seem to be best
to get good skirt selectivity and Q. The can size relates to coil size, and
the coils used in old tubed radios had to be easily made by women working
at benches while chatting about recipes and their dim witted husbands.
When IFTs were miniaturized to suit tiny SS radios arriving from Japan in
1960, I guess automation replaced the highly skilled ladies, or the Japanese
girls all strained their eyes.

I have not tried use of 2MHz IFTs which would need an oscillator F
from 2,531kHz to 3,701kHz. Almost no maker I know ever made sets
with IF = 2MHz, perhaps because there was a chance broadcast band
sets could interfere with other communications sets because of the high
oscillator F. Maybe regulatory authorities banned 2MHz.
If the LC for 455kHz has Q = 45.5, then pass band = 10kHz but if the
IF = 2MHz, the same Q of 45.5 gives pass band 39.9kHz, and audio
BW of 20kHz, somewhat excessive for AM. In practice, such higher
Q as F rises may be unlikely and a lower Q gives pass band is too wide.
So you'd need 3 IFTs. There are a couple of old schematics online,
one by Miller, but I've not had time to try them.

SHEET 2. AM detector Sheet 2.
This arrangement of triodes cannot be found in any AM radio sold by any
radio manufacturers before today.
The IF signal from IFT2 secondary is a fairly high Z source, and feeds very
high Z input to grid of V4, 1/2 12AU7. V4 is in cathode follower mode and
its input C < 20pF so it has little loading effect on IFT2 LC.

V4 cathode has low Rout = 1/Gm = 1/0.0017 = 588r, and powers the 1N914
diode to charge C2 with half cycles of 455kHz IF waves. C2 value is chosen to
have reactance Xc of more than 2 x Rout at the carrier F.
Carrier F = 455kHz, and with C2 = 270pF, Xc = 1,294r. C could be less,
but R value becomes too high and the circuit becomes prone to noise.
If C is higher, the peak charging current in 12AU7 may become too high and cause
grid current which limits the IF input signal.

Fig 2. Basic Waves.
Curve A shows a 14kHz AF wave used to amplitude modulate a 
wave seen Curve B.

Amplitude of peaks of 455kHz follow same shape as the applied AF signal
assuming the de-modulation process is linear.
The amount of modulation shown is the maximum possible, ie, 100%.
The AM wave shown may be detected with diodes + C + R to give two opposite
phases of the modulation signal.

If AF is increased beyond what is needed for 100% mod, the "envelope shape"
begins to have asymmetric clipping on one side of the modulation wave. The clipping
is detected and recovered audio modulation has rapid increase of THD.
Curve B is the possible IF wave after an IFT in a radio set. It is usually a high
Z signal source which I think is best converted to low Z with a cathode follower
or darlington pair emitter follower.
The followers give lower THD than without them when driving a following passive
diode+C+R network.

In practice, although some modulation from radio stations reaches 100%, most of
the AF modulation is done at an average level of 25%. This is easily confirmed
when viewing output of a radio station at V4 cathode. Radio stations often use audio
compression to reduce dynamic range and have higher average mod levels and therefore
a "louder sound" But many locals here show most modulation is under 30%.
When using a test AM signal with 100% modulation with a single AF, say 400Hz,
many receivers will have less than 100% mod in their IF signal because of phase
shifts in the mixing process before IFTs, so max modulation depth at receiver
detector may be 90%. Nothing is perfect, But this aspect is not a problem.

Fig 3. Detecting AM waves.
Fig 3 explains the waves in my kitchen radio AF detector.
any 455kHz signal the Vdc at C2 = 6.5Vdc, about 0.5Vdc below cathode
idle Vdc. Biasing of V4a grid = 0Vdc, and about 3mAdc flows in V4 and V5.
I show the 455kHz non modulated wave at V4a grid as 8Vpk. From this, diode and
tube turn on resistance produces Vdc at C2 which is +6Vdc above the idle Vdc of
+6.5Vdc. So with carrier shown, expect to see Vdc at C2 = +12.5Vdc.
To view the AF waveform after C2, best point for CRO connection s cathode of V5,
a low Rout port, and 455kHz ripple will be very low after LPF R3+C3. 
Without modulation you will see a flat line which I show in red here. The Vdc level
should not change when modulation between 0% and 100% is present.

The 455kHz ripple voltage should have fairly constant shape and amplitude with
low frequency AF modulation up to about 95%. This is due to C2 being discharged
by R2 connected to -320Vdc rail. The C2 discharge current remains a virtually
constant 0.24mA. The forward voltage drop across diode is reduced because
it has idle current even where 455kHz swings towards 0.0Vac during 100%
modulation. Thus distortion caused by change of ripple voltage amplitude during
detection is much reduced. This detector works better than if you have R2
of much lower value connected to 0V instead of -320V.

I have chosen -Vdc rail = -320Vdc because it was so easy to do with a couple
of power diodes and C from a PT with a suitable HT winding. But -120Vdc would
also be fine providing the R1, R2, R4 values are reduced to give same idle Idc.
The circuit works best where Vdc between top C2 and a Vdc rail is at least
10 x expected maximum carrier Vpk.

I show V4 and V5 = my favored 1/2 12AU7 but other triodes could be used, 6DJ8,
12AT7, 6CG7, 6SN7, or trioded 6BA6, 6AU6. Never use 12AX7, 12AY7 which have
such low gm they fail to work properly at all. J-fets could be used but these cannot
produce such a high output audio level and Vdc rails must not exceed +/- 24Vdc.
The humble 12AU7 can make up to 25Vrms as shown, so overload distortion
is not a worry.
A darlington pair follower with bjts can be made using 2N2222 + BF469 which should
work very well with +/- 50Vdc rails and slightly higher C2, 470pF and different R values
Fig 4. AF detector time constants
Fig 4 shows an audio sine wave = 26.65kHz, with a wave time = 37.5uS,
and with amplitude of 6Vpk, 4.2Vrms.

The C = 270pF, and R = 1.36M. There is an idle Idc flow = 332V / 1,360k
= 0.244mAdc.

The time constant for C and R is calculated :-
Time in uS = CpF x R ohms / 1,000,000 = 270 x 1,360,000 / 1,000,000 = 367uS.

This means that if current input to diode is switched off, the Vdc would fall by
0.63 x Vdc across R, so from 332V to 122.8Vdc in 367uS, and then to 45.4Vdc after
734uS and so on until the V across R = 0V.

The initial current I in R = 0.244mA, but it declines as V across 1.36M reduces so
the rate of discharge from C slows so the complete discharge of C takes a long time.

But if could keep the discharge current from C constant, we would find the time
taken to V across R to = 0V is equal to the time constant.

If the voltage change across R is a small fraction of the total possible V change,
then the rate of V change is very nearly constant and we can draw the line
of this as a straight line on a graph. It makes the investigation of the behavior of our
circuit much easier to draw, and contemplate.

The rate of V change across C determines the maximum audio level and the
highest audio frequency
without slew distortion.

For a small fall in voltage of say 6Vpk, the current is virtually constant so rate of V
change is calculated :-
Rate of V change = max Vdc / TC = 332V = 367uS =  0.904V/uS.

Now the time between peaks of carrier 455kHz = 2.2uS, therefore Vpk-pk
fall in Vripple = carrier wave time x rate of change = 2.2 x 0.904 = 1.988V.
So say 2Vpk-pk, and for a saw tooth wave it is 0.57Vrms = 13.6% of the audio
voltage = 4.2Vrms.

If the AF level is doubled to 8.4Vrms, then ripple voltage remains the same and is
6.8% of AF. The fastest rate of V change at zero crossing with 26.65kHz doubles,
so there will be slew distortion. Therefore the fastest sine wave modulation can
be 13.33kHz.imit within fastest

The analysis here tells us we can get good audio HF performance while keeping
Ripple voltage low, while not overloading the 12AU7 and causing grid current flow.
If C is increased, overloading could occur, and grid current could affect the input
wave and load the IF circuit. If C is reduced Vripple increases, recovered Vm
reduces, and R noise increases.

In practice, if the modulation below 5kHz is 4.2Vrms maximum the R could be 2M7 to
discharge C 270pF. The detector would pass 13kHz with no slew distortion for 4.2Vrms.
However all audio speech and music signals have most of the energy between 80Hz
and 1kHz, and above 5kHz the the signal voltage falls at about 6dB/octave so there
is little need for the detector to ever have to cope with more than 1.4Vrms at 13kHz
so there is no slew distortion. The higher R value halves the ripple voltage.

Other detectors may use a class B SS amp as I show at AM modulation.

Those SS fetishists can dive much deeper into AM as described at

While some radios use the -Vdc after detector, I found it better to generate
-Vdc for AVC using a diode and low value C fed off IF amp anode and equal
to what I show in SHEET 1 above with IN914.

V4 cathode is directly coupled to V5 grid via 1N914, and R7 100k.
I have C5 68pF to 0V so that output from C2 has low pass filter with -3dB
cut off at 23kHz. This means virtually no attenuation occurs to audio up
to 9kHz, the maximum we need to worry about for AM radios.
But attenuation for 455kHz is from 0.55V ripple to 0.03Vrms which cannot
affect V6 preamp or following power amp stages.

SHEET 3. Kitchen AM radio Sheet 3.
The AM radio power amp uses paralleled 12AX7 input tube driving EL34
in triode mode. This amp is state of the art SET amp, with the usual features
to get true hi-fi performance.
The EL34 triode idles with Pda + Pdg2 = 17.5 Watts, well short of maximum
Pda rating = 28 Watts. The EL34 will last a very long time if idle Pda =
0.63 x max Pda rating. Other output tubes strapped in triode could be
KT66, 6L6GC, 5881, 6550, KT88, KT90. Adjust R11 cathode resistor so idle
Pda < 0.63 x rated max Pda. Least THD would be with KT88 in triode where
idle Pda = 26.4Watts, Po max for 5k0 anode load = 5.5Watts at 4% THD,
if Ea
= +350V, grid bias = -39V,
Iadc = 75mA, or you have B+ = +390V, Ek = +39V,
R11 = 520r.
Load for max Po with KT88 = 2k9, and Po max = 7.9Watts at 7%,
Normal loud listening levels with 5k0 load will rarely exceed average of 1Watt when
THD = 1.7%. There is some 2H cancellation between input driver V7 12AX7
and output tube. If GNFB is 12dB then THD > 0.5% at 5.5W for 5k0 load.
EL34 will give similar THD figures for same 12dB GNFB at 4.5Watts with 5k0
Notice the R&C networks for HF and LF stability. High quality 8W rated OPTs for
5k0:4r,8r,16r should be used for 6550/KT88. But for EL34 triode my OPT is 6W
rated and was formerly used with EL84 in SE pentode mode in a non-repairable
1965 tape recorder where OPT was slightly better than those used in old radios.

You should find the sound quality is just magical.

SHEET 4. Kitchen radio SHEET 4, PSU.
This PSU provides :-
B+ = +385V at 88mAdc with CLC filter to feed all anode supplies.
B- = -320V at 6mAdc for cathode load resistance current.
6.3Vac with CT for all heater filaments.
+/- 12Vdc for the solid state 455kHz frequency detector for tuning meter.
I repaired a large number of tubed and some solid state AM - FM radios
over 19 years between 1993 and 2012 when I retired.
Not one equals or betters the performance I get with above schematics.
I did re-engineer a number of large floor stander sets plus a few
"mantelpiece" sets and obtained a wealth of experience of how to get old
junk to sound far better than when they sold to the public 60 years ago.
Back then the public rarely heard any hi-fi anywhere, so their expectations
were very easy to satisfy.
Canberra's AM radio stations.
6665kWABC Canberra (2CN) - Canberra ACT news/talk 
84610kWABC Radio National (2FM) - Canberra ACT
1008300WSky Racing Network (2KY) - Canberra ACT sports/racing 
10535kWForever Classic (2CA) - Canberra ACT classic hits 
11252kWRPH Print Radio (1RPH) - Canberra ACT reading service 
12065kWTalking Canberra (2CC) - Michell ACT news/talk 
14402kWSBS Radio 2 (1SBS)|rep - Canberra ACT public/multi-langua service
1629-Rete Italia|rep - Tuggeranong ACT Italian 
1638-Rete Italia|rep - Canberra ACT Italian 
1647400W2ME Radio Arabic (1ME)|rep - Canberra ACT community,cl. hits,pop,narrowcast Arabic

Canberra's FM radio stations.
87.65WRaw FM|rep - Canberra ACT dance/urban
87.81WUCFM - Canberra ACT college (Univ.-Canberra) 
88.01WRadio Austral|rep - Canberra ACT community espanol 
88.7100WACT TAB (1TAB) - Tuggeranong ACT sports (racing) 
89.5100WValley FM (1VFM) - Tuggeranong ACT community 
90.3100WArtSound FM (1ART/T)|rep - Tuggeranong ACT community/jazz/blues/world... 
91.120kWCMS Radio (1CMS) - Canberra ACT community multi-languages 
91.920kWOne Way FM (1WAY) - Canberra ACT religious 
92.720kWArtSound FM (1ART) - Canberra ACT community/jazz/blues/world... 
94.3100WOne Way FM (1WAY/T)|rep - Tuggeranong ACT religious 
96.7500WQBN-FM (2QBN) - Queanbeyan NSW community 
97.5300WHot Country Canberra - Queanbeyan NSW country,narrowcast 
98.320kW2XX FM (1XXR) - Canberra ACT community 
99.9100WABC News Radio (2PB/T)|rep - Tuggeranong ACT news
100.7200W104.7FM (2ROC/T)|rep - Tuggeranong ACT CHR-pop 
101.580kWABC Triple J (2JJJ/T)|rep - Canberra ACT
102.320kWMix 102.3 (1CBR) - Canberra ACT hot ac 
103.980kWABC News Radio (2PB/T)|rep - Canberra ACT news
104.720kW104.7FM (2ROC) - Canberra ACT CHR-pop 
105.580kWSBS Radio 1 (2SBSFM)|rep - Canberra ACT public/multi-langua service
106.380kW1ABCFM - Canberra ACT classical 
107.1100WMix 102.3 (1CBR/T)|rep - Tuggeranong ACT hot ac

Canberra's digital radio stations.

10B   104.7FM - Canberra ACT CHR-pop 
    ABC Canberra (AM666) news/talk 
    ABC Jazz jazz 
    ABC Radio Grandstand sports 
    ABC Radio National
    Chemist Warehouse Remix hot ac
    Double J alternative 
    Forever Classic (2CA,10523AM) classic hits 
    Hot Country Canberra (97.5FM) country,narrowcast 
    Mix 102.3 hot ac 
    Mix '80s CHR-80s
    Mix '90s CHR-90s
    My Canberra Digital hot ac
    SBS Chill electronic,chillout
    SBS Pop Asia pop-Asian
    SBS Radio 1 (105.5FM) public/multi-langua service
    SBS Radio 2 (1440AM) public/multi-langua service
    Talking Canberra (2CC,1206AM) news/talk 
    The Edge Digital CHR-rhythmic 

 Triple J Unearthed local/unsigned artists

To Index page

To radio re-engineering

Additional general notes.

AM radio broadcasts in Australia are officially from 531kHz to 1,602kHz. Some
stations are allowed to operate conditionally to 1701kHz. The RF carrier of each
station is controlled by a crystal or equivalent and is at a frequency exactly divisible
by 9.0, ie, all AM stations at on 9.000kHz spacing across the band, 531, 540, 549
etc. Audio modulation frequencies are allowed up to 9kHz with rapid attenuation above.

Canberra's ABC Radio National has carrier = 846kHz. The RF amplitude
determines the volume of the sound heard, while the rate of change of amplitude
determines the frequencies heard from audio between 20Hz and 9kHz.
This is linear AM modulation. The 846kHz transmitted radio signal has a spectrum
of "modulation products" between 837kHz and 855kHz.

Amplitude modulation of an RF signal 'F1' by a second audio signal 'F2' produces
a radio signal with a fixed F1 signal, and
"modulation products" above and below
F1, = (F1 + F2) and (F1 - F2). If carrier = 846kHz and modulation = 5kHz, 3 RF
signals are present,
the 846kHz carrier plus modulation products at 841kHz, and
851kHz. The "products" are called lower and upper sidebands. The carrier is at
constant amplitude, with sidebands each able to vary between zero and 1/2 the
carrier amplitude for modulation levels between 0% and 100%.

An oscilloscope will show the AM wave as a single F sine wave at 846kHz
but its amplitude is varied at a rate of 5kHz. There are really 3 frequencies
interacting, and a spectrum analyzer with very narrow band filters will show
the single 846kz carrier and two sideband peaks each side of carrier, 841kHz
lower sideband and 851kHz upper sideband.

If the modulation process is not linear some additional sideband F are created.
If the carrier Vc = constant 2Vrms, then each upper and lower sideband peak can
rise to a maximum of 1Vrms. This is 100% modulation depth. If more AF is
applied to modulator, the envelope seen on the oscilloscope will have very
rapidly increasing THD, and the detected audio signal will appear to clipped
on positive or negative wave peaks.
The oscilloscope shows that if carrier wave without modulation = 2Vrms, then
the maximum envelope wave voltages above and below 0V are each also 2Vrms.
Two audio wave signals of 2Vrms are able to be detected, each AF wave having
opposite phase. 99% of all AM radios detect only one phase of audio, and often use
a single diode plus C + R which works the same as a 50Hz 1/2 wave power rectifier.
Most radios generate a negative going audio wave, ie, the lower 1/2 of the AM wave
so that the -Vdc of carrier also detected can be used in sets for AGC.
All old tube radios had 1/2 wave detectors producing up to maybe -25Vdc and
possibly 25pk volts of audio. Full wave rectifiers and voltage doublers all were
not needed, and were found to be difficult to optimize.

It is fairly easy to detect both phases of audio modulation to produced a balanced
AF signal but not one maker has ever seen any benefit because the single ended
AF source was adequate, even when a PP audio amp was used.

THD/IMD and poor audio bandwidth of old radios is due to several factors:-
1. Non linear IF amp distorts the envelope shape, adding several % THD to the AF detected.
2. Diode peak detector is not linear, adding several % THD....
3. Detected AF fed to cap coupled volume pot which causes cut off distortion which isn't
understood by anyone, but it happens.......
4. Audio amp has minimal stages, usually high ยต triode feeding 6V6 tetrode in tetrode
mode but without any GNFB, adding several % THD.
5. PSU is poorly filtered. You know the radio is turned on by the hum you hear.
6. Very poor speakers in the same box as radio, adding several % of THD/IMD at
all levels above a whisper.
7. Noise pick up by RF input stage using small radio frequency transformer and 3 meter
long wire antenna.
The 7 factors may produce harmonic artifacts which may exceed the wanted HF
audio content which is excluded by the sets poor AF bandwidth of say 3kHz.
So the missing wanted HF is replaced by bad sounding sonic garbage.
Audio bandwidth less than 3kHz is determined by the overall RF and IF bandwidth
of tuned LC networks in the set being less than 6kHz. Most AM superhets have
5 tuned LC circuits, one RF LC, and 4 x IF LC. The bandwidth of each LC may be
8kHz to allow 5kHz of audio. But when 5 x LC are cascaded, the overall RF+IF BW
can be 4kHz, allowing only 2kHz of audio BW. These  radios have good "skirt selectivity",
ie, a 5kW local station cannot be heard when tuned to a low power local station only
45kHz away; the attenuation is over -60dB.
The skirt selectivity is good for listening to distant stations, but I never need to hear
distant stations because my ABC favorites are networked so a distant ABC RN
station has the same program as the local RN. Many commercial and regional radio
are stations are networked and give boring pop music or torkbak with long adds
so none are worth a listen.

Audio bandwidth is determined mainly by the receiver IF bandwidth.
Fig 5.
Fig 5 shows 2 measured responses for typical well made IFTs for HMV radios
made in 1950s using litz wire coils cam wound in 3 pyes. BLACK and BLUE curves
are for IFT1 and IFT2, and their summed response is shown as the RED curve.
Notice the curves have asymmetrical shape
above the center F = 455kHz. This is due to
slightly higher stray coupling between
input test signal and output leads at higher F.
The curves show the coils of both IFT or at least one IFT could be moved closer for
more IF BW if wider AF is wanted. Notice the phasing of connections for primary
and secondary coils. The response will measure very poorly if the incorrect phasing
is used, and the result with best attenuation away from 455kHz is the correct
connection. Many who making their own IFTs now will need to realize the importance
of coil distance and connection phasing. Some ppl will use fixed RF coils with L =
477uH, ( 0.47mH ) and C = 260pF using 220pF fixed and parallel to 10-50pF trim cap.

The overall
RF and IF bandwidth is mainly determined by the two IFTs in most sets,
each with two magnetically coupled tuned LC. The Q of the litz wire coils produce
a high Q over 40. The amount of magnetic coupling of pri and sec IFT LC is determined
by the distance between the pair of IFT coils, and this determines pass band response
Coils say 20mm apart, magnetic coupling is low, insertion loss is high, maybe -20dB
but selectivity and Q is high, which best suits long distance reception. with low AF bandwidth
and low noise.
Coils say 15mm apart, magnetic coupling is medium, insertion loss -12dB, and selectivity
slightly less but OK, and Q is lower, permitting 4kHz audio bandwidth.
Coils say 12mm apart, magnetic coupling high, insertion loss -10dB, selectivity OK for local
stations at least 45khz apart, Q is low, and AF is widened to a maximum of about 9kHz.
Fig 5.
Fig 5 shows the response of a typical 455kHz IFT from 1950s.4
The AF bandwidth is always 1/2 the IF bandwidth.

This all assumes all the IF LC are tuned to 455kHz. However, when 2 tuned LC are brought
close enough, the IF pass bandwidth cannot be widened to more than maybe 16kHz before
the shape of the pass band becomes twin peaked. It becomes almost impossible to easily
adjust the tuning of each LC to 455kHz unless one has a signal gene with F meter plus a sweep
or wobbulator or FM gene display the response of the IF so symmetrical twin peaks are

The twin peaks of IFT1 can be both reduced if the pri and sec are tuned slightly above and
below 455kHz. This makes the pass band flatter hence flatter audio response.

There is often good reason to slightly reduce distance between any 2 LC in a given old
IFT to be able to get more IF BW and thus get wider AF BW. But to the inexperienced
DIYer, the mod is difficult, can ruin the IFT, and it needs experience, a good signal gene
and F meter and a CRO + probe with 10:1 function.

Should anyone want to wind their own IFTs, I suggest the page at

Synchronous reception.
For a good synchronous design you need an RF oscillator with its sine waves  exactly
synchronized with those of a station to which you are tuned. Before solid state chips
were invented, phase locked loop ( PLL ) techniques were rarely used; I have never
seen a tubed PLL. So a "locked oscillator" was used so a small sample of wanted station
RF was amplified, then limited, then applied to an oscillator so that it "pulled" the oscillator
Fo to being the same as the wanted station. This is OK for FM radio where a constant
level 19kHz pilot tone can be used to lock a circuit producing 38kHz sub-carrier for stereo.
But for AM, I found the locking didn't want to work so well because if the modulation reached
100% the locking signal ceased so oscillator would jump away from lock, and the better
way would always be PLL with stand alone oscillator controlled by Vdc applied to vari-cap
I did try once to make a "synchrodyne" set using a Mr D.G.Tucker's circuit from 1947.
I needed special trifilar RF transformers and 4 x Ge diodes for a ring diode mixer network.
I found no available info about such transformers in 1999, and I easily concluded the
superhet was a whole lot easier.

In 2015, I investigated how a Gilbert Cell works, usually with bjts inside IC, as in MC1496.
I found it could be done with 3 twin triode 6DJ8/12AT7, but MC1496 has 6 bjts for the
basic pair of cross coupled differential amps. I found discrete bjts or triodes were all
so fiddly and complex I gave up.

To produce an audio signal in a receiver which is equal to the audio used to modulate
the the transmitted carrier and without any interference from any other station,
there are two basic requirements :-
1. The receiver must have RF pass band of twice the highest audio
BW used for modulation measured at the input to the AF detector circuit.
If the highest AF = 9kHz, ideally the RF pass band at AF detector input is 18kHz.
Beyond this pass band, there must be a high rate of RF attenuation so a strong
local local a station 45kHz away from the wanted station has a level at least
-60dB below the wanted station.
2. Radio stations with the same carrier F as the wanted station must be located
very far away so their signal strength at receiver is so low it cannot be heard.
From searching listed stations, I found one at Cairns on 846kHz, same as a
Canberra station. Distance is over 2,000kM. Local stations are not closer
than 45kHz apart.

AFAIK, the highest audio F used for AM modulation in Australia is 9kHz, and without
any emphasis of high of low AF. But many stations use audio compression to make
everything sound more lively. AM radio stations across Australia are 9kHz apart,
so when listening to distant stations, it may be possible two stations are 9, 18, or
27kHz apart so that both are heard if the selectivity is poor. Where they are
9kHz apart, you can get a 9kHz whistle because the carrier of one station 9kHz
away from a wanted station is detected as a 9kHz tone. The other side bands
of the other station are also detected and cause "monkey chatter". For listening
to distant stations, or "DXing" a wider the RF and IF bandwidth in a receiver makes
 it less possible to just hear one station. Older AM radios were often made with far
less RF and IF bandwidth, perhaps only 4kHz, which allows only 2kHz of audio.
While these may be OK for speech it will never be hi-fi. Having a directional
antenna can improve DX listening, but such things must be large, and interest in
AM is now much less than in 1955. Short wave reception also depends on good
selectivity and low total IF and RF bandwidth.

Solid state receivers often have only 3 tuned circuits, one input ferrite rod plus
2 tunable IFTs with a single LC. These have high Q, and allow about 2kHz of audio,
but away from the tuned F the "skirt selectivity" is much less than in tube sets with
5 LC with medium Q, The RF input LC and the 4 LC in the pair of IFTs, which
produce a flattened pass band.

If you want to tune to say 1,008kHz here, the SS radio will appear at first to be
able receive it OK but the AF BW is often only 1.5kHz. Attenuation of an unwanted
station say 45kHz away 1008 may only be 12dB with one tuned LC even if it has
a high Q. 90kHz away from 1008 may be give 18dB, so that 3 tuned LC give
54dB, not so good if the unwanted station has signal strength 20dB higher than
wanted. But with 5 LC of a typical tube radio, 60dB attenuation is achieved easily
at 45kHz away from  wanted.

Before about 1930, many AM radios were "TRF" types using at least 3 simple
RF triode amps with LC in series, each with a separate tuning knob. All 3 knobs
had to be each adjusted to get a single station, and no doubt many domestic
arguments were caused by those unable to handle the radio set properly.
Each tuning knob pointed to a numbered dial so people would write down a list
of numbers for each station, so to tune a radio was like having to know the
numbers of a lock on a safe, or a six digit telephone number.
Oh what fun.
Some TRF sets had a 4 gang tuning capacitor so four LC could be tuned with
one knob without 4 confusing interactive knobs. The multi gang cap was installed
in the very expensive sets for the rich, but audio bandwidth was always poor.
Each LC had to be very carefully set up to ensure the tuned F of each was
the same regardless of the knob position. If the Q for each was 50, then at
1,008 kHz the RF BW = 20.16kHz, allowing 10kHz of AF BW. But with 4 LC,
the effective Q increases to 142 allowing 3.5kHz of AF. At 45kHz away from
1008kHz, attenuation is about -48dB, and less than the usual better -60dB or
more with 5 LC in a superhet.
Development of the "super heterodyne" type of radio in 1920s also meant
only one tuning knob was needed. The "superhet" could have 5 to 7 tuned
LC networks with only 2 or 3 requiring tuning by means of a two or three
gang variable capacitor. (One of these tuning gangs was always for a tuned
local oscillator.) Some communications radios had a 4 gang cap to tune 3 RF
LC at radio input with 1 gang for oscillator but they may have had been 6 to
10 fixed F LC for between 3 and 5 IFTs.
For 50 years after invention of the simple TRF AM radio, many claimed the
TRF offered the best hi-fi because less mucking about was done to  the RF
signal before detecting the AF. A few ppl claimed a crystal set tuner with no
active RF amp was best. Well, now we have a crowded AM band with many
stations and lots of noise so beliefs about TRFs & crystal sets can safely be
ignored. By about 1935, nearly all AM radios were superhets,
easily giving the
one wanted station without another in the background. Simple sets allowed
serious short wave reception with a good antenna, and in 1935 you may have
listened to Hitler ranting and raving, marginally less tolerable than the Voice
of America, specialist stations guided by religion, socialism could be heard,
plus a plethora of loud buzz sounds which were radio jammers to drown out
the "opposition"

For strong local stations the minimum superhet needs a single tunable RF
input LC and four LC in two IFTs, so 5 LC in total, and with the pass band
having a flat topped profile 18kHz wide, -3dB.

The tubed superhet has at least a 2 gang tuning cap usually with identical
gangs giving C range 5pF to 480pF, typical. There is usually a 4-40pF trim
C across each gang to get wanted range of RF and oscillator F.
A 10 fold change from say 480pF to 48pF gives a theoretical Fo change from
say 531kHz and 1,701kHz from RF radio stations. The 3.2 ratio is close
to the square root of C change.
The oscillator F range is from say 986kHz to 2,156kHz, 2.18 ratio, with IF =
455kHz. To obtain this, a cap of about 440pF is in series with the tuning
gang fixed plates. This gives maybe 230pF to 48pF, 4.79:1 ratio, and F ratio
is sq.rt. 4.79 = 2.18.

The difference between the RF and oscillator F is the IF = 455kHz in most sets.
Some older sets have lower IF, and some older communications receivers may
have IF at 100kHz to exploit the higher Q available with lower IF, and thus narrow
the overall pass band and exclude noise when tuned to a low power station.

Typical "mixer tubes" aka frequency converters are 6AN7 hexodes or
6BE6 heptodes. These have enough electrodes for operation as a triode
RF oscillator which feeds its output to an available input grid sandwiched between
screen grids so that effectively the tube acts as a non linear amplifier with TWO
input ports. These special tubes are designed to prevent much varying levels of
RF signals from altering the oscillator tuning F, and vice versa, and ensure
operation without production of parasitic F. Their range of F can be from audio
signals to 50MHz. 
The oscillator signal in a mixer tube is a much higher signal than RF input signal.
So the mixer anode current is switched on and off at oscillator F rate.
RF input between microvolts and volts produces intermodulation product F which
are osc Fo + RF, and osc Fo - RF.
The output anode current is fed to an IF transformer which is tuned to the IF,
ie, say 455kHz. The oscillator is set up to always be 455kHz above the tuned
LC of the RF input circuit. If the RF input signal = 846kHz, and oscillator Fo
= 1,301kHz, then the difference in F = 455kHz. The mixer tube produces a large
oscillator voltage signal at IFT input, and when looking with oscilloscope one
cannot see the wanted IF. But the IF is there, at a low level, and IF LC resonance
means that ALL F except 455kHz are excluded by L or shunted by the C.
If the IFT secondary LC signal is examined, we find far less oscillator Fo
and presence of IF is seen easily. The following IF amplifier pentode, say 6BA6
raises the low IF signal to apply to a second IFT with two LC and the output
has virtually no other signal present except the IF. And the IF signal modulation
envelope looks exactly like the RF input signal envelope, ie, the audio information
has been transferred to a new carrier F, ie, the IF.
This process of "mixing" is more accurately known as frequency conversion
and it was called "super heterodyne" reception.

IFTs were typically wound with litz wire
to make coils about 0.48mH. This L was
bypassed with a silver mica 250pF capacitor. The F is varied by adjust screws to
change a ferrite slug position inside the coil former tube. Many older pre-WW2 IFTs
had variable capacitors, and used solid wire, not litz wire, and IF was lower, 220kHz.
IFT plus a an Al shield could be 125mm high and 70mm dia. The
late 1950s IFTs
used fixed C and variable ferrite slug tuning and were typically 75mm high and Al
cans 22mm x 22mm plan size. These later IFTs with litz wire produced the highest
Q and the two LC could be arranged for a fairly flat topped pass band with good
"skirt selectivity" so that only 2 IFTs were needed in most AM receivers.
The thousands of ladies working in electronics factories were required to make
items which gradually became smaller and more difficult.

Most typical
early 1950 IFTs had primary and secondary litz wire coils
wound on a common phenolic former tube about 10mm oa dia and 100mm
long fitted inside an aluminium can 110mm high x 50mm dia, or square section
40mm x 40mm, with coils about 15mm apart. These gave good Q, and I've
found them to be the easiest to change the distance between the two LC
to slightly widen the pass band width.

Some older communication receivers had 3 IFTs, with an RF stage ahead of
the mixer. Helicrafter had a knob to turn a shaft to change IFT coil distance
on 3 IFTs to dramatically reduce the IF pass band to maybe less than 1kHz
which is better to detect a weak Morse code signal among other strong
short wave stations and noise. There was an RF input pentode and 2 IF amp
pentodes, and having enough gain was not a problem. Alignment was easy.
With IF coils closest, The AF pass band was about 4kHz, enough to suit
communication speech and amateur radio operators liked these sets.

RDH4 offers a solution to increase IF bandwidth by switching in a few turns
of a tertiary IFT winding to increase primary and secondary coupling to give
a slightly twin peaked response. See page 455, RDH4, 1955 Ed.
Trio and Kenwood and early Quad AM receivers had this feature. First you
tuned in narrow band mode for highest IF signal at a meter or "magic eye".
Then you switched to
"hi-fi" to add in the tertiary and a
udio bandwidth could be
increased magically from say 4kHz to 8kHz. But you needed to repeat this
for each station which is torture for radio surfers changing stations 20 times
a night. Quad had The Best way of doing it.

I used recycled IFTs from 1950s with both IFT coils closer together and no
switched tertiary. Because I have good test gear, I was able to tune each LC in
IFT1 to 450kHz and 460kHz, and thus get wide flat IF bandwidth with both LC in
IFT2 tuned to 455kHz.
To be able to tune the set properly I have a tuning meter powered by second
IF amp with 3 tiny IFTs from a transistor radio for a high Q and meter tell me
when IF is +/- 500hz away from 455kHz. The sound is then better than all other
AM sets I've owned, or repaired, and better than FM where the signal is from
the same networked source for AM and FM, as it is with news on ABC Classic
FM and AM Radio National.

Keep soldering, and don't electrocute yourself.