Page 1
Colin Horrabin, G3SBI
17 Denbury Ave, Warrington, Stockton Heath, WA4 2BL, United Kingdom; colin.horrabin7@ntlworld.com
The HF7070 HF/LF
Communications Receiver Prototype
A detailed look at high performance receiver design.
The HF7070 receiver is a double conversion superheterodyne with a first IF at
45 MHz and a second IF that is centered
at 44 kHz before going to a 25 bit audio
ADC. The output from the ADC goes to an
advanced 24 bit fixed-point DSP system.
The radio covers dc to 30 MHz and has a
noise figure of 12 dB without the use of a
preamplifier before the first mixer. Out-ofband IP3 at 50 kHz spacing is 45 dBm, giving an SSB IP3 dynamic range of 115 dB.
The IP3 within the 15 kHz bandwidth of the
roofing filters is 19 dBm at 100 Hz spacing
in an SSB bandwidth, which results in an
IP3 dynamic range of 97 dB. This in-band
linearity sets new standards for an up-conversion radio and gives superb high fidelity
reception of FM, AM, SSB and CW signals.
To complement its excellent technical
performance there are all the usual DSP features for the user. These include a sensitive
band scope on the LCD panel of the radio.
The band scope can also be displayed on a
computer connected via a TOSLINK optical cable. With a noise floor of –145 dBm
in a 50 Hz bandwidth, it can display submicrovolt signals.
The receiver analog front end has two
H-Mode mixers using fast bus switches.
The first mixer is terminated by quadrature
hybrid-connected two-pole 45 MHz filters of
15 kHz bandwidth, which is followed by the
first IF amplifier. This amplifier, with a noise
figure of only 1.3 dB and an IP3 at 40 dBm,
drives 4 poles of roofing filter, which is followed by a second amplifier. This amplifier
drives the second H-Mode mixer that gives
an output centered on 44 kHz. There is a
balanced stage of amplification at 44 kHz
before the 25 bit audio ADC. The 6 poles of
roofing filter at 45 MHz gives 115 dB image
rejection at the second H-Mode mixer, so an
image rejection mixer is not required.
The HF7070 was designed by the British
electronic engineer John Thorpe of JTdesign
based in Matlock, England. John also
designed the Lowe receivers and the highly
acclaimed AR7030 HF/LF receiver manufactured by AOR UK.
In terms of its technical performance the
AR7030 represented very good value for
money and was made from 1996 until production ceased in 2007 due to the restriction
of hazardous substance (ROHS) directive.
Over that period some 5000 units were sold,
most of which went for export (even the
French Navy bought a few).
Originally, John designed the HF7070
receiver for AOR UK, but they ceased
trading two years ago. John is a consultant
and had other design commitments. So, I
designed and built an up-conversion front
end for the 7070 a few years ago to help
with the development work. John was able
to re-engineer this for mainly surface mount
components and improve on its technical
performance.
Just after John was presented with the
front end board, I got to know Martein
Bakker, PA3AKE. Martein was keen to
build a holy grail version of the CDG2000
transceiver and had made technical measurements on transformers and fast bus switches
for H-Mode mixers. These measurements
proved quite useful to John. Those readers who are familiar with the CDG2000
Amateur Radio transceiver project will
QEX – July/August 2013 37
Page 2
recognize the similarity of the HF7070 front
end block diagram (see Figure 1) to the
receiver in the CDG2000. Any reader who is
interested in the detailed technical measurements for the HF7070 “proto2” receiver will
find them on PA3AKE’s website at http://
martein.home.xs4all.nl/pa3ake/hmode/.
Fundamental Design Issues in the
Analog Signal Path
Table 1 shows some of the key parameters
of the receiver front end design. It is the job of
the two-pole 45 MHz filter to provide some
protection to the first IF amplifier for strong
signals more than 10 kHz from the selected
frequency.
In an up-conversion radio the third order
intercept (IP3) usually reduces significantly
for off-channel signals within the bandwidth
of its roofing filters. The close-in performance of the HF7070 has surprised quite a
few people and Table 1 shows how this excellent close-in IP3 performance is achieved.
The values in the table of Net NF and Net
IP3 are slightly better than the practical measurements on the “proto 2” receiver. In the
table, the noise figure (NF) of the receiver
at the antenna is 11 dB. This gives a noise
floor of –129 dBm in a 2.4 kHz bandwidth.
Together with a Net IP3 of 24.5 dBm the
result is an in-band dynamic range of 102 dB.
The practical results are NF 12 dB and an IP3
of 19 dBm, giving an in-band IP3 dynamic
range of 97 dB.
It is necessary for the analog front end to
have enough gain so that noise from the analog stages dominates the quantization noise
from the ADC in the digital signal fed to the
DSP. This gives a smooth, audible transition
from noise to signal.
filters and a NF of 17 dB, which yields a
dynamic range of 77 dB. So you can see why
people with knowledge of receiver design
are impressed with the close-in performance
of the HF7070.
Referring to Table 1, in-band IP3 is limited by the 4-pole roofing filter. As the net
IP3 requirement increases, as the signal is
amplified the linearity will be degraded if the
net value exceeds the stage IP3. This does not
happen with the HF7070, but it is interesting
to note that the 44 kHz amplifier has an IP3
output of 60 dBm.
A paradox is that a bit of signal path loss
at the right place is a design virtue. The IP3
dynamic range at a 100 Hz test tone spacing in a 2.4 kHz bandwidth is 97 dB. The
Yaesu FTDX-5000 has an IP3 of –8.5 dBm
within the bandwidth of its 9 MHz roofing
The First H-Mode Mixer
The experimental 7070 front end that I
gave to John used the Fairchild FST3125
fast bus switch in the mixer. An important
result from Martein’s experimental work
was that the more recent Fairchild FSA3157
actually proved to be the best switch for use
in H-Mode mixers. This is a SPDT switch
with a 0.5 ns break-before-make action. This
SPDT switch is ideal for being driven by a
fundamental frequency squarer because it is
not necessary for the drive logic to generate a
complement signal, as was the case with the
Table 1
Key Parameters of the HF7070 Front End
Stage Net Stage Net Stage Net
Gain Gain IP3 IP3 NF NF
dB dB dBm dBm dB dB
Antenna low pass filter –2 –2 45 24.5 2 11.1
First H-Mode mixer –5 –7 45 22.5 5 9.1
2 pole 45 MHz filter –1.5 –8.5 24 16 1.5 4.1
First 45 MHz amp 10 1.5 40 26 1.3 2.6
4 pole 45 MHz filter –2 –0.5 24 24 2 7.7
Second 45 MHz amp 10 9.5 40 34 1.3 5.7
Second H-Mode mixer –5 4.5 45 29 8 14.02
44 kHz amp 18 22.5 60 47 6 6.02
25 Bit ADC 0 22.5 50 47 4 4
Out of Band IP3 at
50 kHz (dBm)
Net Gain (dB)
Stage Gain (dB)
Noise Figure (dB)
QX1305-Horrabin01
Figure 1 — The HF7070 front end block diagram with Out of Band IP3, Net Gain, State Gain and Noise figure indicated for various stages.
45 43 38
-2 -7 -8.5 +1.5
-2
11.1 9.1 4.1 2.6 7.7 5.7 14.02 6.02 4.0
35 MHz
Low Pass
Filter
-5 -1.5 +10
H-Mode
Mixer
LO
45 to 75 MHz
Fundamental
Frequency
Squarer
Double-tank
Oscillator
Voltage Control
2-pole
15 kHz
45 MHz
Filter
10
Analog Signal Path
A
Frequency Synthesizer
-0.5
-2 +10 -5 +18
4-pole
15 kHz
45 MHz
Filter
÷
4
11.25 to
18.75 MHz
V R
Phase
Detector
+9.5
10
DDS
Chip
Control
Set
Frequency
+4.5
45 kHz
H-Mode
Mixer
LO
44.9555 MHz
Fundamental
Frequency
Squarer
TCXO
44.9555 MHZ
A
+22.5
A
ADC and
DSP Clocks
25 Bit
ADC
To DSP
System
38 QEX – July/August 2013
Page 3
FST3125. The choice of Mini-Circuits transformers used in this mixer gives the radio an
out-of-band IP3 of 45 dBm at a 50 kHz test
tone spacing with high sensitivity down to an
input frequency of 10 kHz.
15 kHz Bandwidth 2-Pole 45 MHz
Crystal Filter
Two of these filters are connected via
quadrature hybrids to terminate the mixer.
The use of quadrature hybrids with two identical filters always presents a nominal 50 Ω
termination to the mixer, even when individual filters present a reactive load.
A design goal of the HF7070 was to build
a sensitive receiver without a preamplifier
before the first H-Mode mixer. Because the
HF7070 is a general coverage multimode
receiver, it was always our intention to use
a 15 kHz bandwidth roofing filter system
to accommodate FM and DRM signals.
Another reason for this is that narrower
bandwidth crystal filters have lower design
impedances, which would give a greater
insertion loss and therefore require a preamplifier before the first mixer. This filter and
its surrounding circuitry introduce a loss of
only 1.5 dB.
The First 45 MHz Amplifier
This amplifier drives the second roofing filter and it is the input IP3 of this filter
(26 dBm) that limits the in-band performance
of the radio.
The amplifier is a 45 MHz version of the
4 × J310 amplifier designed by Bill Carver,
W7AAZ, as used in the CDG2000 transceiver. The noise figure of this amplifier is
1.3 dB. Its output IP3 of 40 dBm makes it a
particularly important building block in the
HF7070 to satisfy the requirement of having
a sensitive, linear receiver without a preamplifier before the first H-Mode mixer. This
amplifier uses source gate feedback, which
gives excellent reverse isolation so that its
input impedance is not affected by its output
driving a crystal filter, which in its transition
region can present a reactive load.
15 kHz Bandwidth 4-Pole 45 MHz
Crystal Filter
This filter has an in-band IP3 at its input
of 26 dBm and this ultimately controls the
in-band IP3 of the radio. Some Amateur
Radio equipment manufacturers offer narrower bandwidth VHF roofing filters for their
radios. You don’t want this for two reasons:
(1) the insertion loss increases and (2) the
in-band IP3 for a narrower design bandwidth
with the same quality of quartz crystals gets
worse at roughly 6 dB per octave. This effect
was discovered by PA3AKE when he was
designing his 9 MHz roofing filters with the
fabulously linear crystals supplied by the
German firm Quarztechnik.
So what you really need in an up-conversion radio is a wide roofing filter, but
the linearity of following circuitry must not
degrade the in-band IP3 of the crystal filter.
This seems to be a problem area with many
transceiver designs.
The second 45 MHz amplifier is similar
to the first amplifier and uses the 4 × J310.
Its output IP3 is 6 dB higher than the net IP3
at that point.
The second H-Mode mixer gives 115 dB
rejection of the image because of the stop
band of 6 poles of roofing filter, so an image
rejection mixer is not required. This mixer
gives a push-pull output centered on 44 kHz.
The 44 kHz Amplifier
John designed this amplifier with a noise
figure of 6 dB, a gain of 18 dB and an output
IP3 at 60 dBm. A point worth noting is that
it is easier to get low-noise, high-IP3 gain
at 45 MHz than at audio frequencies. This
was an important consideration for the gain
distribution of the front end. There is also a
potential problem for the unwary that may
affect other designs: many operational amplifiers have poor IP3 characteristics.
You get good IP3 results at high signal
levels, but the IP3 tones do not fall away as
the signal amplitude is reduced; they remain
at 60 dB down. There were a few operational
amplifiers that gave the correct IP3 behavior.
The reason for this problem may be due
to crossover distortion in the output stage of
most operational amplifiers. John was aware
of this problem when he designed the 44 kHz
amplifier. It is interesting to note that the new
ICOM IC-7410 with its 36 kHz IF appears to
use a version of the TL074 that is known to
have the IP3 problem.
Apart from the front panel board, the
radio has an analog board and a digital board.
After some discussion the 25 bit ADC went
on the digital board. The second H-Mode
mixer makes the transition from 45 MHz to
44 kHz and it gives a push-pull balanced output. This is followed by a balanced amplifier
at 44 kHz located on the analog board whose
outputs are connected by a short length of
strip cable from the analog board to the balanced inputs of the 25 bit ADC located on the
digital board. Whatever common mode noise
is present due to the strip cable connection
between the two boards is well within the
common mode rejection of the ADC.
The 25 Bit ADC
The radio uses a top-of-the-range, stereo,
24 bit audio ADC with both channels driven
and the signals digitally added giving a theoretical 25 bit performance, It is worth remembering that the ADC is effectively the main
gain block. An estimate of its noise figure
Further Developments
The manufacture of the ten HF7070
production prototypes has been funded by
a British company owned by a radio amateur. He was happy to learn that this article
was appearing in QEX, but he asked that
neither he nor his company be mentioned.
In view of this it seems unlikely there will
be a significant production run because he
is concerned that there would be too much
competition from SDRs.
Collaboration between British and
American electronic engineers during
World War II resulted in some great design
work. The thing to remember about the
HF7070 is that although it has been
designed by a British electronic engineer,
all the components that really matter,
apart from the 45 MHz monolithic filters
with their excellent IP3s, were designed in
the USA. At least in that sense, history is
repeating itself.
It would have been more cost effective
to use an MMIC in the 45 MHz IF strip, but
nothing was available that met the technical
requirements. That has changed recently,
however. The American firm Mini-Circuits
has introduced the PHA-1+ and the dualmatched version known as the PHA-11+.
Although they were designed for microwave
applications, the noise figure (NF) and
IP3 out is at its best between 45 MHz
and 100 MHz, making them useful in IF
amplifiers for up-conversion radios. Typical
performance at 45 MHz is: NF 2.2 dB: gain
18 dB: and IP3 out 42 dBm.
The vector network analyzer designed
by Paul Kiciak, N2PK, together with the
Windows software for the VNA written
by Dave Robert, G8KBB, was used to
measure the PHA-1 input impedance at
45 MHz, which is 80 Ω in parallel with
25 pF. The best way to use this part in
the existing HF7070 IF strip is to use two
in parallel with an output attenuator to
replace the first 4 × J310 amplifier, and
one to replace the second amplifier, again
with an output attenuator.
A change to the HF7070 front end
architecture by having all 6 poles of roofing filter connected via quadrature hybrids
immediately after the first H-Mode mixer
could, in principle, further increase closein receiver dynamic range. Obviously,
careful shielding around the filters is
required. All the mechanical and electronic
components for this experiment have been
obtained and I hope to report the results
in a future QEX article. From a practical
printed circuit point of view, rather than
use the matched pair the thermal affect
on amplifier noise figure would have been
reduced if two individual PHA-1s could
have been used. Unfortunately, this did
not fit well with the circuit board design.
Another version of the PHA-1 where the
input and output pins are transposed
could be useful. Perhaps Mini-Circuits will
consider this.
QEX – July/August 2013 39
Page 4
(NF) is 4 dB, but if it was 12 dB the effect on
receiver noise figure would be to increase it
by only 0.1 dB.
+V
QX1305-Horrabin02
Overall Front End Performance of
the Signal Path
The HF7070 has excellent technical specifications and this is reflected in its outstanding on-air performance. This didn’t happen
by chance; a lot of thought has gone into the
receiver front end architecture.
It is unusual to have a sensitive up-conversion receiver that does not use a preamplifier
before the first mixer. In many ways the front
end architecture makes use of the principles
established by Bill Squires, W2PUL, in the
1960s in his SS-1R receiver. His receiver
used the 7360 beam-switching tube in the
mixers and, although it was a tunable IF
design, gain distribution was carefully controlled before the main crystal filter. It also
lacked a preamplifier before its first mixer.
Just like PA3AKE’s holy grail receiver
with a 9 MHz IF, the HF7070 is only as good
as the in-band and out-of-band linearity of
its crystal roofing filters. Martein has made
IP3 measurements on a number of different
manufacturers’ 15 kHz bandwidth, 45 MHz
fundamental-mode monolithic crystal filters.
Only one makes the grade in terms of IP3
and its 3rd order law compliance with input
signal level.
RFC
L L2
C1
1 nF
220
39 nF
220
The Double-Tank Oscillator
Out 1 Out 2
82 k 82 k
C2
L
+5J310
Output
J310
Buffer
+108 V
A Low Phase Noise Local Oscillator
for the HF7070
The local oscillator in the HF7070 is a
DDS/PLL design using a double-tank VCO.
This was designed by John Thorpe and it
is similar to the one used in the AR7030
receiver, but with improved performance.
The basic double-tank oscillator is shown in
Figure 2. The principle involved is that this
circuit can only oscillate if the cold end of the
active tank (the one with the J310) is a low
impedance to ground. This can only occur if
the dummy tank is series resonant at the same
frequency as the active tank. This means that
as you move away from the carrier two resonators are active, which increases the rate of
phase noise fall-off with offset frequency.
This circuit has never been analyzed from a
phase noise point of view by a mathematician, but measurements show phase noise
falls off at 30 dB/decade compared to 20 dB/
decade in a single resonator oscillator.
In a superhet receiver local oscillator,
sideband noise causes reciprocal mixing
and, as a result the dynamic range associated
with good close-in IP3, can be limited by the
sideband noise of the local oscillator. The
frequency reference used in the HF7070 frequency synthesizer is a 44.9555 MHz TCXO
that has a good phase noise profile. The
TCXO is used for the local oscillator feed to
10 pF
100 k 100 k
Basic Oscillator Circuit of the G3PDM Receiver
Figure 2 — The HF7070 double-tank oscillator (top) compared to the tube-based oscillator
used in the G3PDM/W1 receiver designed by Peter Martin in 1970.
the second H-Mode mixer and it also clocks
an AD9951 DDS chip. Other divided outputs
from the TCXO are used to provide clocks
for the 25 bit ADC and the DSP system.
Operation of the local oscillator frequency synthesizer is as follows. The output
frequency from the double-tank VCO is
divided by 4, which is then phase locked to
a DDS-generated reference frequency in the
range of 11.25 to 18.75 MHz. The result is
that the VCO tunes 45 to 75 MHz in small
steps. The VCO is buffered and goes to a
fundamental frequency squarer whose output drives the FSA3157 switches in the first
10 pF
E88CCE88CC
H-Mode mixer. The loop bandwidth of the
PLL system is about 1 kHz, within which
VCO oscillator phase noise is reduced by the
action of the loop.
A measured value of phase noise for
the HF7070 at 200 Hz to 1 kHz offset on
the 7 MHz band is a –116 dBc/Hz plateau
falling to –128 dBc/Hz at 7 kHz. Six SSB
signals could be present within the 15 kHz
bandwidth of the roofing filter and would be
separated in at least an 82 dB dynamic range
because of reciprocal mixing
For CW reception with a 250 Hz DSP, the
bandwidth dynamic range would be at least
40 QEX – July/August 2013
Page 5
Table 2
Phase Noise Comparison
Frequency Offset (kHz) 3 5 10 20 30 50 100 200
AR7030 dBc/Hz –113 –118 –128 –137 –142 –147 –155 –160
HF7070 dBc/Hz –116 –122 –133 –142 –145 –150 –158 –162
HF7070 at 7 MHz –120 –126 –138 –147 –150 –154 –159 –162
92 dB. If the DSP bandwidth was reduced
to 50 Hz you could resolve CW signals in at
least a 99 dB range.
Outside the PLL bandwidth the phase
noise profile is that of the double-tank oscillator itself. The HF7070 is not phase noise
limited beyond 7 kHz from the carrier,
reaching –150 dBc/Hz at 30 kHz from the
carrier on the 7 MHz band. According to
Leeson, phase noise in a single-tank oscillator reduces beyond a corner frequency at
20 dB per decade as you move away from
the carrier. In the double-tank oscillator our
measurements have shown that phase noise
decreases by about 30 dB per decade.
By examining Table 2 you’ll see that
phase noise falls at around 30 dB per decade
of frequency offset in both the AR7030 and
the HF7070. The results for the AR7030 are
calculated from the Peter Hart review of the
AR7030 receiver in the July 1996 RadCom.
The results characterize the basic phasenoise profile of the AR7030s double-tank
VCO, which is the result of the very narrow
PLL bandwidth used in the AR7030 receiver.
The HF7070 is added for comparison and
adjusted to compare with the AR7030 measurements at 21 MHz.
I developed the double-tank circuit
in 1994 to complement my design of the
H-Mode mixer and John developed it further
for use in the AR7030 receiver. The owner’s
manual for the AR7030 includes a complete
circuit of the radio, so it is surprising that
John’s version of the double-tank oscillator
has not been copied for use in other commercial developments (such as other Amateur
Radio equipment). In fact, the real origin of
the double-tank circuit goes back a bit further than 1994, all the way to the G3PDM
receiver designed by Peter Martin in 1970.
His receiver used a 7360 as the mixer and he
had a push-pull double-triode local oscillator
VCO to drive it. There was only one resonator, each end of which went to the plate of a
triode. Each triode plate was connected to
the high-voltage power supply through an
82 kΩ resistor. You can’t do that with J310s,
but if you compare the two circuits they do a
similar job. So, it is fair to say that Peter got
there first.
Semiconductor designers appear to have
a love affair with band-gap voltage references in what they consider to be low noise
systems. These usually have a voltage of
Figure 3 — This photo shows the RF/IF circuit board in the prototype HF7070 receiver. The
signals enter at the top right, and the white blocks in that section of the board are relays for an
input attenuator. Below the relays are the low pass filter inductors and then the Mini-Circuits
transformers of the first H-mode mixer. To the left is the first toroid of the quadrature hybrid, two
poles of filtering and the second quadrature hybrid. To the left of center is the 4 × J310 amplifier.
In the top left quarter of the board there is a four pole filter, the second 4 × J310 amplifier, the
second H-mode mixer and the 44 kHz push-pull amplifier. The narrow cable strip carries the
amplified signal to the analog-to-digital converter on the digital circuit board. The bottom section
of the circuit board includes the frequency synthesizer and low-noise voltage regulators.
QEX – July/August 2013 41
Page 6
1.2 V and an RMS noise of about 10 mV in
a 10 kHz bandwidth. They have a characteristic noise profile with a noise plateau that
extends out to about 300 kHz before it falls
sharply to the thermal noise floor. Most chip
designs that make use of band-gap references
fail to provide a pin to which a large capacitor can be connected to reduce low frequency
noise from the reference.
There is suspicion that close-in phase
noise within the PLL bandwidth on the HF
7070 is limited by band-gap noise on the
phase detector chip made by Philips: the
74HCT9046. To improve this situation it is
necessary to know how band-gap noise on
the chip may be turned into noise that would
affect the charge pump. This turned out to be
easier said than done.
In the old days manufacturers of integrated circuits would show internal circuits
on their data sheets. Philips was contacted to
see if they would provide a circuit of the chip
so it could be established exactly what effect
band-gap noise could have on the charge
pump, but a circuit was not forthcoming.
They were again contacted, but finally told
us they were not interested in pursuing the
issue further because no other ‘9046 users
had asked these questions before and the
chip was selling well. As far as Philips was
concerned, this was a non-issue.
In truth, these issues have only recently
arisen because state of the art up-conversion HF receivers are approaching 120 dB
dynamic range, and in the case of PA3AKE’s
holy grail down-conversion receiver, exceed-
ing 120 dB. Semiconductor designers really
need to take notice of the shortcomings of
band-gap references in ultra low-noise systems and provide a pin to deal with low frequency noise from the reference.
On-Air Performance of the HF7070
Receiver
The technical performance of this radio is
what Pat Hawker, MBE, G3VA, (SK) would
call “superlinear.” Not only does the radio
have an SSB IP3 dynamic range of 115 dB
at 50 kHz, its design answers the problem
of how to get good IP3 dynamic range in an
up-conversion radio for signals within the
bandwidth of its 45 MHz roofing filter (effectively achieving a dynamic range of 97 dB in
an SSB bandwidth). What was unknown was
how a radio with this built-in in-band linearity would actually sound on the air.
The answer is “amazing.” The fidelity is
remarkable and what is particularly noticeable is the way selected signals seem to
stand out above the noise floor and adjacent
signals. The first time you hear this radio you
know you are listening to something special.
Unless you hear it yourself, the performance
is impossible to describe.
130
Reciprocal Mixing Dynamic Range
(2.4 kHz Bandwidth)
120
110
100
Dynamic range (dB)
90
80
0 10 20 30 40 50 60 70 80 90 100
QX1307-Horrabin04
Figure 4 — This graph shows the measured dynamic range (DR) of the HF7070 prototype
receiver. The reciprocal mixing dynamic range was measured with a 2.4 kHz bandwidth. Notice
that the 2nd order IMD DR is completely flat across the entire measured range. The 3rd order IMD
DR increases rapidly above a signal spacing of about 5 kHz and then becomes quite flat at a
signal spacing of 20 kHz. Ideally, the IMD DR should not change as the signal spacing of the
input signals changes. This radio has a 3rd order IMD DR that is better than 110 dB at 20 kHz
spacing, and levels off at 115 dB, which is quite remarkable. This graph was copied from Martein
(PA3AKE) Bakker’s website: http://martein.home.xs4all.nl/pa3ake/hmode/ar7070-proto2.html.
Signal separation (kHz)
3rd Order IMD Dynamic Range
2nd Order IMD Dynamic Range
How I Came to be Involved in the Design of the HF7070
In 1995 I had suggested to John Thorpe that he consider the double-tank oscillator for use in the AR7030 receiver. We had spoken many times on the phone, but we
didn’t meet until 1998.
In 2000 Dave Roberts, G8KBB, George Fare, G3OGQ, and I started to design the
CDG2000 transceiver that used an H-Mode mixer with the FST3125 fast-bus switch.
As an analog radio, it made use of the 4 × J310 amplifier and the 9 MHz IF amplifier
design that Bill Carver, W7AAZ, presented in QST.
When we were discussing the frequency synthesizer, I suggested that we ask
John and AOR UK if they would let us use John’s VCO design. They did. Since the
CDG2000 receiver section uses a 9 MHz IF, the output of John’s VCO was divided by
at least two, giving –150 dBc/Hz at 25 kHz on 20 meters and –150 dBc/Hz at 12.5 kHz
on 40 meters.
The divider was a 74AC74 bistable, so it produced a square wave output, giving
both not-Q and Q signals to drive the mixer, and required no adjustments to give a
receiver IP3 of 40 dBm.
On one of my visits to John I stopped by AOR UK to thank them for their help. While
I was there, Mark Sumner told me that John was investigating a DSP based receiver
then known as the AR7070, but progress had been slow. Originally John intended the
signal path to be 45 MHz/455 kHz/20 kHz to the 24 bit audio ADC.
Mark supplied me with some of the Hertz Technology 15 kHz bandwidth, 45 MHz
fundamental mode crystal filters to test. Their transmission characteristics were very
good with low insertion loss and a good stop band. I realized that if 6 poles of filter
were used you could go straight down from 45 MHz to 44 kHz, simplifying John’s
original design and reducing the second H-Mode mixer image by 110 dB. Also the very
good in-band IP3 of these filters offered the promise of hitherto unobtainable close-in
dynamic range. I designed and built the prototype up-conversion front end and gave
the design information to John along with the prototype board.
One of my hallmarks at Daresbury Laboratory was the tendency to push any
given technology to the limit. That is why I am now experimenting with the use of the
Mini-Circuits PHA-1 MMIC, which in principle could put 10 dB on the in-band IP3 and
give 45 dBm at 20 kHz in a front end like that used in the HF7070. I don’t really see
John making use of this development because it is unlikely he will design another upconversion receiver. Together with a local oscillator using the AD9910 DDS chip with a
1 GHz low-phase-noise clock, however, this development would represent the ultimate
up-conversion receiver technology.
42 QEX – July/August 2013
Page 7
Figure 5 — This photo shows the first prototype receiver front end. It proved the author’s design before it was adapted for the HF7070
receiver. This photo is courtesy of Martein Bakker, PA3AKE: http://martein.home.xs4all.nl/pa3ake/hmode/g3sbi_intro.html.
In the 15 kHz wide roofing filter, the signal you are listening to is in the center of the
pass band where its group delay characteristics are at their best. This is one factor in the
exemplary performance of the radio.
John knows I am a CW enthusiast so my
personal agenda was to make sure that the
HF7070 was a good CW receiver, and it is.
When I first had the opportunity to listen
to the HF7070 proto1 receiver, John had a
lap top computer connected to display the
bandscope. Apart from the larger display, it
updates much faster than the LCD display on
the radio front panel and I could read CW off
the screen. I was watching a DX station on
15 meters working split, and I could see other
stations replying. This is remarkable stuff !
Acknowledgements
Because the first part of this article concerns hardware, this is probably the right
place to thank all the people who helped
design and build the prototype up-conversion
front end.
A chance conversation with Pat Hawker,
MBE, G3VA, in 1993 about an item in
RadCom’s “Technical Topics” introduced me to the pioneering work of Jacob
Makhinson, N6NWP, in receiver front-end
design. That set in motion a chain of events
that led to my development of the H-Mode
Useful Reading Material
• “Super-Linear HF Receiver Front Ends,” (Colin Horrabin, G3SBI, evaluates
the N6NWP high performance mixer), RadCom, “Technical Topics,” Sep 1993, pp
54-56.
• “G3SBI’s High Performance Mixer,” (H-Mode mixer), RadCom, “Technical
Topics,” Oct 1993, pp 55-56.
• “Crystal Filters for High Performance Mixers,” (H-Mode mixer termination with
crystal filters using quadrature hybrids), RadCom, “Technical Topics,” Jan 1994,
37-39.
• Reed Fisher, W2CQH, “Twisted-Wire Quadrature Hybrid Directional Couplers,”
QST Jan 1978, pp 21 to 23. More consistent results can be obtained by using
bonded bifilar wire to build your own quadrature hybrids. Such wire is available
from the Scientific Wire Company (www.wires.co.uk/acatalog/bb_wire.html).
• The CDG2000 transceiver description at the Warrington Amateur Radio Club
website at www.warc.org.uk/proj_cdg2000.php.
• The G3PDM receiver, as described in the Radio Communication Handbook
Fifth Edition, 1982, pp 4.50-4.57.
• Wes Hayward, W7ZOI, Rick Campbell, KK7B, and Bob Larkin, W7PUA,
Experimental Methods in RF Design ARRL, 2009. ARRL Order No. 9239; $49.95.
ARRL publications are available from your local ARRL dealer, or from the ARRL
Bookstore. Telephone toll free in the US 888-277-5289 or call 860-594-0355, fax
860-594-0303; www.arrl.org/shop; pubsales@arrl.org.
• Wes Hayward, W7ZOI, and Doug DeMaw, W1FB, (SK), Solid State Design
for the Radio Amateur, ARRL, 1986. This book is out of print, but it occasionally
shows up at Hamfests, on eBay and other used book sites.
• William Sabin and Edgar Schoenike, Single Sideband Systems and Circuits,
2nd Ed, McGraw Hill, 1995.
• Peter Hart, G3SJX, “Review — The AR7030 VLF to HF Receiver,” RadCom,
July 1996, pp 44-46.
mixer and the double-tank oscillator. This
QEX – July/August 2013 43
Page 8
has led to the high performance receivers
found in PA3AKE’s holy grail version of
the CDG2000 and now the up-conversion
HF7070 receiver designed by John Thorpe.
“Technical Topics” also brought together
Bill Carver, W7AAZ, Harold Johnson,
W4ZCB and Gian Moda, I7SWX. Their
contributions to radio technology have been
used in the design of the HF7070 as well.
The influence of Wes Hayward, W7ZOI,
expressed in his various books and articles
in QST and QEX have contributed significantly to the detailed design of the signal
path. Research work by Martein Bakker,
PA3AKE, presented on his website has been
important for the technical status of our
hobby and has helped John Thorpe to firm
up the choices of some critical components.
I retired in 2000 and it would not have
been possible to research and build the
prototype up-conversion front end were it
not for the good will of former colleagues
at Daresbury Laboratory. George Fare,
G3OGQ, was recruited to help with circuit board design work and Mark Sumner,
G7KNY, formerly of AOR UK (now at
MWS Technical Services), provided components needed for experimental work associated with the front end design.
When Martein Bakker, PA3AKE, became
involved in testing various H-Mode mixer
configurations (transformers and switches),
Mark supplied all the bits. It was a real
shame that AOR UK ceased trading before
the project reached the production stage. The
purpose of my experimental front end design
was to establish the circuit techniques that
John Thorpe was then able to improve upon
for his commercial design.
Colin Horrabin, G3SBI, was born in 1941.
His father provided him with a World War II
BC348 radio receiver for his 12th birthday, followed by a copy of the ARRL Handbook for
Christmas. After years building various projects using government surplus equipment, he
obtained his Amateur Radio license in 1963.
He has a degree in electrical engineering and
a degree equivalent qualification in mechanical engineering. Following an apprenticeship
with the British Aircraft Corporation in the
early 1960s, he spent over 30 years working at
Daresbury Laboratory as an electronic engineer. Colin is interested in small DX antennas
for the LF bands, and intends to do some work
on small multi turn spiral wound loops that are
self resonant containing ¼ wavelengths of wire,
which are suitable for transmitting.
A Forum for Communications Experimenters
QEX features technical articles, columns, and other items of interest to radio amateurs and communications
professionals. Virtually every part of the magazine is devoted to useful information for the technically savvy.
Subscribe Today: Toll free 1-888-277-5289 • On Line www.arrl.org/QEX
Subscription Rates: 1 year (six issues)
ARRL MEMBER: for ARRL Membership rates and benefits go to www.arrl.org/join
F US $24.00 F US via First Class $37.00 F Intl. & Canada by air mail $31.00
NON MEMBER:
F US $36.00 F US via First Class $49.00 F Intl. & Canada by air mail $43.00
F Renewal F New Subscription
Name: _________________________________________ Call Sign: ____________________________
Address: ____________________________________________________________________________
City: ______________________________ State: _________ ZIP: ____________ Country: __________
F Check F Money Order F Credit Card
Charge to: FFFF
Account #: ______________________________ Exp. Date: __________
Signature: __________________________________________________
Monies must be in US funds and checks drawn on a US Bank
Subscription Order Card
Published by:
ARRL, 225 Main St,
Newington, CT 06111-1494 USA
Contact circulation@arrl.org
with any questions or go to
www.arrl.org
Web Code: QEC
Project #350
44 QEX – July/August 2013
Loading...