Here are some classic HF receivers from the '40s, '50s, '60s and '70s:
|RACAL RA17L||RACAL RA1772|
|RCA AR88LF||Collins R390A|
The sub-rack at top right houses my homemade TTL computers. Please visit my projects page to read about these. Just visible top left is my homemade magnetic long-wire balun (MLB) which is a trifilar-wound 9:1 ratio broadband transformer. This converts a receiver's 50-ohm input impedance to 450-ohms which is a better match for an end-fed wire. Of course, the antenna impedance is frequency dependent but the transformer seems to help a bit. I don't know why these devices are sold as baluns. Surely it's a magnetic long-wire unun (MLU)?
Starting with the oldest, my AR88 is unusual in having a blue front panel; they're usually gray. It also has an S-meter which is rare. They were supplied to Britain in large quantities during WW2 without S-meters due to scarcity of materials. This is the AR88LF (low frequency) model with coverage down to 75KHz but no medium wave. It has a lovely warm valve sound and I mostly use it for listening to Radio 4 Long Wave.
I've had two problems with this set. Firstly, the audio output transfomer blew. Apparently, this is quite common. I left it in-situ, disconnected, and temporarily fitted a replacement under the chassis. I've since acquired a spare original which I must get around to fitting!
The second problem started with sudden changes in volume level. Eventually, the set died completely. I traced this to R19, the 33K screen grid bias resistor for the 6SA7 mixer, which had gone open circuit. Replacement was stressful because I had to cut out other components to reach it!
My other American set is a Collins R390A built by Capehart Corporation in 1961. Designed in the '50s, classified until the '60s, and manufactured in volume until the '70s, this set has a mechanically driven veeder root counter digital frequency readout and is acknowledged to be one of the best HF receivers of all time.
I knew it was faulty when I bought it because I took a signal generator with me to test it! It was dead from 4 to 8MHz. As it came with full service manuals, I decided to take a chance. The problem was a broken clamp connecting a gear wheel to a cam shaft in the front-end tuning rack. This set has automatic preselector tracking. Permeability tuning slugs, in 6 octave ranges, are driven by the MHz and KHz knobs via a gearbox and differential. Unfortunately, the broken clamp was buried deep inside the clockwork. I had to send the tuning rack to Rick Mish of Miltronix who has re-manufactured so many of these sets that he can fix them with his eyes closed.
My oldest RACAL is an RA17L built in February 1962. Designed in the '50s, the RA17 was the first receiver to use the ingenious Wadley-loop tuning system which divides the HF range into 30 bands, each 1 MHz wide, using a single 1MHz master oscillator, a VHF VFO and some clever mixing and filtering. It's equivalent to 30 crystal controlled converters in front of a conventional 2-3MHz superhetrodyne receiver. The latter has a highly stable VFO, enabling absolute frequency setting precision of better than 1KHz.
Although the set was lively enough when I acquired it, it went deaf within a few weeks of use. I had it fully re-conditioned by Robin Filby in 1999. You can see his 'next service due' sticker in the middle of the front panel. Robin is an expert on RACAL valve equipment and well known amongst RACAL enthusiasts worldwide.
Currently, my set has a minor fault. Occasionally, the 1MHz crystal oscillator fails to start. Eventually, it bursts into life if I leave it on for a minute or two. I must get around to changing the crystal one of these days!
In the 1960s, RACAL embraced transistors with the RA217, RA1217 and, later, the RA1218. Still based on the Wadley-loop principle, these sets are transistorised versions of the RA17. The RA1217 has a mechanically-driven digital frequency readout. Of high-quality, modular construction, this set is beautiful to look at inside. Performance wise, it sounds great. I listen to it through my Hi-Fi as there is no internal speaker.
I've only had one problem with this set: One of the capacitors in the mains input filter went short circuit blowing the mains fuse.
In the 1970s, RACAL left the competition behind by releasing the RA1772 synthesised communications receiver. When other synthesizers could only be programmed with decade switches, the RA1772 had a tuning knob offering a VFO-style search facility. It also had a bomb proof front-end. The dynamic range is such that it doesn't need an attenuator and the preselector can be switched out of circuit most of the time.
RACAL packed a lot of features in the 4U high case. Some versions have an internal FSK demodulator. My set has AFC (automatic frequency control) which phase-locks the 2nd conversion 34 MHz oscillator to the received carrier for synchronous AM reception. LSB, USB or AM demodulation may be selected with AFC. My set also has the ISB (independent sideband) adapter but I still haven't managed to find an ISB transmission to try it out on!
RACAL achieved the exceptionally high dynamic range using four hard-driven BSV81 switching MOSFETs in the first mixer. R P Rafuse described this type of mixer in his 1968 paper "Symmetric MOSFET Mixers of High Dynamic Range." Here's a close-up. Top left is the 35.4 MHz crystal filter. At the right you can see the 2N3866 local oscillator drive transistors:
The synthesizer is a cascade of 3 multiplying loops and 2 summing loops, implemented in discrete 74 series TTL. They don't make them like that any more! Left to right we have the lower loop (F4), lower transfer loop (F3), upper loop (F2), upper transfer loop (F1) and HF loop (F0):
Each VCO covers much less than a 2 to 1 range. Three separate switched VCOs are used to generate the final output frequency f0. The transfer loop oscillators are nice and close to 1 MHz making it easy to separate the desired difference product from the unwanted sum by filtering. The same principle, with fewer loops, was used by Marconi in the excellent 2018 / 2019 series signal generators.
I have one criticism of the RA1772: The printed circuit boards are difficult to remove because the wiring harness is soldered directly to them. Repairs can be awkward:
* If you need to buy a spare 5082-7300, be sure to get the same luminosity (denoted by a letter code on the back in the top right corner).
My RACAL RA1792 was just the right size to fill the empty space in my racks. I want to replace the Telford Electronics front panel with an original:
It's interesting to compare this set with the RA1772: Looking at the front-end, the RF amplifiers and first mixers are similar - though not identical; also, common-gate JFET IF amplifiers follow the post-mixer crystal filters in both sets; however, the RA1792 had a new single-loop fractional-N synthesizer:
The RA1792 synthesizer was disclosed in this RACAL patent from 1978. I also found this IFR (formerly Marconi) application note on fractional-N synthesis with references to patents by Dana, RACAL, Marconi and Hewlett Packard.
Patents can be downloaded from the European Patent Office and Google Patents.
RA1792 schematics are available from Premium-Rx.
My experimental Fractional-N synthesizers are here.
The pulse to voltage converter is shown above as it was disclosed in the original patent, but this is actually somewhat misleading. Close inspection of the schematics reveals a second output controlling PFD gain by varying the charge-pump current. In the steady state, the DAC reference is fixed. It's actually the full-scale (2π radians) PFD output which varies. Same difference, you might say. More on this later.
The best way to understand this synthesizer is to first accept that it works. When the N.F command is 500.0100, the VCO frequency is a rock-solid 50.001 MHz. The first accumulator counts 0100, 0200, 0300 ... 9900 and overflows once every 100 reference cycles. The VCO is divided by 500 for 99 reference periods and 501 when the accumulator overflows. Every millisecond, 100 reference cycles elapse, and the VCO divider counts a total of 99*500+501 = 50,001 VCO cycles.
The accumulators count fractional VCO cycles, or VCO phase if you like. An overflow is worth 1.0000 cycles or 2π radians. The first accumulator tracks VCO phase as measured at the PFD output. When it overflows, an extra VCO cycle is swallowed making the VCO appear 2π radians behind; but the accumulator also lost 1.0000 by overflowing, so they remain in agreement. We don't need a register to count overflows because the number of whole cycles accumulated is notionally always zero.
The second accumulator integrates the first. Overflows from the second accumulator retard and advance the measured VCO phase, by swallowing and always thereafter inserting an extra VCO cycle, in such a way as to have no long term effect. The temporary retardation is not reflected in the contents of the first accumulator; but the integral in the second is 1.0000 short, as if there had been a 1.0000 deficit at the integrator input. So the second accumulator accurately predicts the integral of the PFD output.
We started with an assumption that the VCO frequency was rock-solid. This implies a clean VCO control voltage. We have a complex waveform at the PFD output; but we also have an accurate prediction of its integral in the second accumulator. The two most significant digits of the integral are converted to a voltage by the DAC, differentiated by a capacitor, and summed with the PFD output in the correct phase and proportion to cancel it.
Finally, we return to the purpose of the pulse to voltage converter. The full-scale DAC output must cancel exactly 2π VCO radians of phase error at the PFD output. But the full-scale PFD output is 2π radians of the reference frequency, which is anything from 404 to 704 times as many VCO radians. Fortunately, the ratio of VCO period to reference period also varies over the same 404 to 704 range. The pulse to voltage converter exploits this, adjusting PFD gain to keep full-scale PFD and DAC outputs in exactly the right proportion.
Earlier designs used one accumulator (and a DAC). So why bother with the second accumulator? It overflows more regularly than the first, reducing the peak phase error which must be cancelled by the DAC; it also exchanges low frequency energy at the PFD output for higher frequencies more easily removed by the loop filter. With a single accumulator, the PFD output never exceeds 2π radians; but its integral does several times over. With two accumulators, the integral is never allowed to exceed 2π radians.
You'll find more RACAL info at Keith's Vintage RACAL Enthusiasts Site.
Wireless Set No. 38 Mk II* was a WW2 man-pack AM transceiver covering 7.3 to 9 MHz with a range of up to 2 miles. It was simple to operate, having only a send/receive switch and a tuning control. The antenna could be a vertical rod or a wire laid on the ground. The operator wore the set at his front with the battery pack on his back. A throat-microphone kept his hands free.
These sets required 150V at 16mA plus 3V at 480mA on transmit falling to 9mA and 240mA on receive. I lashed-up a crude mains power supply to test them. The receivers were still very lively. The fixed volume level can actually be uncomfortably loud in the headphones. They were designed to pick out weak signals at a time when the HF bands were much quieter than they are today.
Using 4 HP hot carrier diodes in the first mixer, and triple conversion - including up conversion to a high first IF of 37.3 MHz, this was a very advanced receiver for its time. Absolute accuracy is aided by a 70" long KHz film scale. Stability is excellent, thanks to a permeability-tuned interpolating VFO in the ingenious synthesizer.
Quadrature terms, generated by RC phase-shift networks, are combined in an image-rejecting mixer to suppress the (A-B) difference frequency. The wanted sum frequency (A+B) is cleaned-up by a phase-locked loop. To overcome the limited capture range of the analogue phase detector, a slow (~ 1 Hz) ramp is injected into the loop to sweep the VCO onto frequency. Once locked, the loop acts to correct this disturbance.
Another PR155 owner, Bill Cooper, who also owns a Marconi 2382, sent me spectra of his interpolation VFO showing similar 50 Hz sidebands.
He attenuated them by inserting steel and ferrite objects between the VFO and the transformer.
Received more spectra from Bill demonstrating how wrapping Mu-metal foil around the VFO dramatically attenuates the sidebands.
April 2017: Duncan Richardson, who worked at Plessey in the 1960s, salvaged a reject transformer from the factory, and used it for a Hi-Fi project. His amp suffered hum problems which were cured by replacing the transformer. He concluded the original was faulty, hence why it had been discarded, perhaps having badly-assembled iron stampings leading to large leakage flux.
Fractional-N synthesis was invented in the late 1960's by Noel Braymer, founder of Dana Laboratories Inc. of Irvine, California. The Digiphase synthesizer was the first indirect frequency synthesizer with steps less than the reference frequency, low output spurious and low random phase noise. Spurious was below -80 dBc and close-in phase noise was -110 dBc/Hz.
The desired frequency is dialled-in on the front panel or remotely controlled (using TTL level 8-4-2-1 BCD code) via a 52-way Amphenol connector at the rear. Frequency being the rate of change of phase, the requested frequency is digitally integrated to determine what the VCO phase should be at any moment. The VCO is disciplined to have exactly this required phase by a combination of analogue and digital wizardry.
In "Analog Circuit Design - Art, Science and Personalities" edited by Jim Williams', Garry Gillette, the engineer tasked with turning Noel Braymer's idea into a working product, gives a fascinating account of their 3-year struggle to make Digiphase work. An early prototype revealed the immense challenges they faced using the technology of the day. In the end, they nearly failed to achieve the expected spurious rejection until a lucky "unplanned discovery" was made quite accidentally. I won't spoil it for you.
The digital circuitry is a mixture of original 7400 series TTL (no LS in those days) and Motorola MC1000 series MECL for fast edges and low jitter where speed was critical. The analogue sections feature hot carrier diodes for mixing and switching; JFETs; dual-gate MOSFETs; UHF PNP transistors; and discrete transistor DC amplifiers. Monolithic dual transistors are used; but there is only one integrated circuit (ML709) operational amplifier in the 7010. There are several 0.05% tolerance resistors and the calibration interval was 6 months.
According to Garry Gillette, only a small number of Digiphase synthesizers were ever made. Some were sold to NASA, who needed precise control of frequency and phase to compensate for Doppler shift when tracking extremely weak signals from distant space probes. Another application was controlled chirp for coherent radar. The series comprised 7010, 7020 and 7030 models. The 7020 and 7030 have "search" controls and external FM modulation. The 7010 has no "search" and only PM. The 7020 has 0 - 90dB output attenuation whilst the 7030 has an F/100 output covering 0 - 110 KHz.
Having pioneered the introduction of high-precision digital multimeters (DMMs) in the 1960's, and despite the remarkable achievement of Digiphase, Dana hit upon hard times in the mid-1970's and the company was eventually sold to RACAL, becoming RACAL DANA in 1979.
The 7020 has an additional card providing "SEARCH" and external FM modulation. This applies an analogue-derived offset to the digitally-controlled frequency. Full-scale deviation can be selected in decade ranges from ±1 Hz to ±100 KHz.
VCO zero-crossings (whole cycles) are counted and compared to the most significant digits in the phase accumulator by a digital phase detector. Current pulses controlled by the fractional decades of the phase accumulator are also fed into the loop filter integrator summing node. The widths of these pulses scale with VCO period so they accurately represent fractions of the VCO cycle. By this combination of digital and analogue means, the net current integrated is proportional to the difference between actual and required VCO phase to a high precision.
The digital phase detector output and residual cancellation pulses occur at different times within the 10µs cycle. Following the loop filter integrator, a sample-and-hold gate which closes once per cycle ensures the VCO control line is not subjected to transients and also acts as a notch filter for rejecting the 100 KHz reference frequency and its harmonics. This allows loop bandwidth to be quite wide without compromising reference spur attenuation.
The VCO tunes from 40 to 51 MHz in 1 Hz steps. The 0 - 11 MHz output is generated by mixing with an internal 40 MHz reference. By dialling-in 0.000000 or using SEARCH, the output can be adjusted down to DC. The DC output voltage may be held steady at any point on the sine curve by zeroing the requested frequency. SEARCH also permits tuning below zero into negative frequencies!
External FM is added to the SEARCH control wiper voltage and integrated, converting FM to PM which is injected into the loop integrator summing node via the front-panel decade range attenuator. When it exceeds full scale, a precise charge is removed from the SEARCH integrator and the associated phase accumulator decade is incremented or decremented.
With ECL and TTL devices date-coded 69, 70 and 71, some in ceramic packages, this 7030 is older than my 7010 and 7020. Nevertheless, it too arrived in working order, despite the collector of a series-pass transistor being shorted to chassis, taking the +35V rail to +44V unregulated. I checked the power supplies first with all other modules disconnected.
|Copyright © Andrew Holme, 2003-2009.|