[Sticky] Radio Receiver Intermediate Frequencies
The second loop was quite complex, involving 37.5 and 3.75 MHz feeds from harmonic generators and a VFO.
Actually, the second Wadley Loop involved two VFOs, as shown in this diagram, although the interpolation VFO (200 to 300 kHz) was, strictly speaking, not within the loop proper.
The main loop is highlighted in mauve, the second loop in yellow. The latter, unlike the main loop, does not have any of its legs overlapping with the IF path. One may imagine that with that complex of frequencies, very good screening was required for the so-called synthesizer compartment. Both loops were of the premixer type, with the output of a narrow-range interpolation VFO added to the main drift-cancelling feed.
Another Marconi receiver of the same era as the Hydrus was the N2020 naval unit of 1968. Like the MST models, this was intended to use the same synthesizer as its counterpart transmitters, and had self-tuning of the RF circuits. Thus it also followed the integer IF “rule”. In this case, the required frequency coverage was 240 to 525 kHz and 1.5 to 28 MHz. Here a 2 MHz 1st IF could not be used, given that it would have been in-band. That left 1 MHz. Evidently that was viewed as not being sufficiently high for good image rejection at the higher HF frequencies. So for frequencies above 8 MHz, an initial IF of 4 MHz was used. Thus, above 8 MHz, the receiver was triple conversion, with IFs of 4 MHz, 1 MHz, and 100 kHz. Below 8 MHz, it was double-conversion, with IFs of 1 MHz and 100 kHz. The 100 kHz final IF was, at the time, normal for SSB/ISB HF receivers.
Different again was the Marconi N2050/Mimco Apollo marine main receiver of 1970. This was double conversion in the range 1.4 to 28 MHz, with IFs of 1.1 MHz and 100 kHz, and single-conversion (100 kHz) IF in the range 15 kHz to 1.4 MHz, except for the 65 to 140 kHz range, which was double conversion (1.1 MHz and 100 kHz IFs). The need for continuous “search” manual tuning capability, plus fully tracked, tight selectivity front end tuning likely pointed to the use of a traditional valve-era configuration, with 10 switched bands covering the whole range. The extra high stability required for marine SSB was provided, on eight marine HF bands only, by a simple synthesizer coupled with a relatively narrow-span interpolation VFO. In a way, the 8 HF marine bands were somewhat analogous to bandspread bands on a conventional receiver.
The 1.1 kHz 1st IF was unusual. Previously, where the choice was not otherwise constrained, Marconi had used 1.6 MHz for receivers tuning down to 2.0 MHz, and 1.2 MHz for receivers tuning down to 1.5 MHz. In both cases the 1st IF was 80% of the lowest tuning frequency. Applying the same ratio to a 1.4 MHz tuned frequency points to 1.12 MHz, which could be rounded down to 1.1 MHz. So on that basis, it was a reasonable number. More than that though, the second mixer used a 1 MHz oscillator input to mix down from 1.1 MHz to 100 kHz. That 1 MHz came from the synthesizer, simply divided down from the 8 MHz master oscillator. That would have been another reason to use a 1.1 MHz 1st IF. There was evidently some flexibility in the 1st IF choice, unlike the N2020 case, where 1 MHz was otherwise determined, and so it was necessary to synthesize a 1.1 MHz feed for the 2nd mixer. The synthesizer is highlighted in yellow in this diagram:
Released at the same time as the N2050/Apollo was the Marconi Nebula, more-or-less as a “junior” partner. The Nebula was a rebadged Eddystone EC958/5, the marine version of the EC958. It was also a 10-band, tracking-tuned receiver, covering 10 kHz to 30 MHz. A high stability mode in the range 1.6 to 30 MHz was provided by a Wadley Loop. It was triple-conversion above 1.6 MHz, with IFs of 1335 nominal (range 1335 to 1235 kHz), 250 kHz and 100 kHz. Below 1.6 MHz, it was either double-conversion (250 and 100 kHz IFs) or single-conversion (100 kHz IF).
The 1335 kHz nominal 1st IF appears to have derived from established Eddystone practice. The earlier 830/7 valved receiver had a 1st IF of 1350 kHz, tunable over the range 1250 to 1450 kHz, this providing for incremental tuning on the ranges above 1.5 MHz. In the case of the EC958, it seems likely that the 1335 kHz number, with one-sided adjustment to 1235 kHz, was chosen to suit the specifics of the Wadley Loop. But whereas the 830/7 went direct from 1350 to 100 kHz, the EC958 had an intermediate step of 250 kHz. The Wadley Loop specifics might have pointed to this approach, in terms of generating the feed to the 2nd mixer with minimum clashes. Also, it allowed sideband inversion in the 250-to-100 kHz conversion by having both infradyne and supradyne options. That would have been difficult in say a 1335-to-100 kHz conversion, given the nature of the 2nd mixer feed. 250 kHz was evidently suitable for the purpose, and not only that, it was an established IF, albeit mostly used for final processing in SSB/ISB receivers where very compact filters were desired.
The 2nd mixer feed was generated either by a 935 kHz crystal oscillator mixed with the output of a narrow-range VFO, or, in the high-stability mode, by a 935 kHz signal from the Wadley Loop, similarly mixed. In that mode, the 1st VFO was “locked” in 100 kHz increments. This was a conventional Wadley Loop in that one of its legs overlapped the 1st IF path. It was also of the pre-mixer type, like both in the Hydrus.
The loop is highlighted in yellow. A subsidiary AFC function acting on the 1st VFO, highlighted in green, kept the 935 kHz signal on-frequency, needed because the Wadley Loop was narrow-band.
The AFC system was a true feedback loop. The Wadley Loop I should imagine was so-described because its key components do form a loop. But the signal flow within that loop is not unidirectional, so it is not a true loop in the narrower definition of the term.
The four Marconi receivers mentioned, the Hydrus, the N2020, the Apollo and the Nebula all had a 100 kHz final IF, at the time the customary number for final processing in commercial/industrial SSB/ISB receivers. But their earlier IFs were largely circumstantial, dictated in part by choices made for the overall circuit configurations, and to some extent by established maker practice. The available information gives some pointers to the reasons for their selection. Otherwise, it is necessary to derive plausible (but not necessarily correct) explanations by working backwards, as it were.
There was a departure from the 100 kHz final IF in the Marconi H2900 receiver of 1970, which chronologically was in the same bracket as the above-mentioned 100 kHz cluster. The H2900 was described by Marconi as a “breakthrough” receiver, with a synthesizer that tuned in 1.0 Hz steps. Basically it was a one-box, dual diversity point-to-point ISB receiver, with a tuning range of 1.25 to 30 MHz. It was of the triple-conversion, upconversion type, with a fully-tuned front-end. The 1st IF was 79.8 to 81.8 MHz, or 80.8 ± 1.0 MHz. The 2nd was 30 MHz, and the 3rd and final IF was 2 MHz, an unusual number at which to do the sideband and carrier filtering.
With upconversion HF receivers, the 1st IF appears to have been a matter of individual maker/designer choice, inter alia related to synthesizer design. For example, the higher the IF, the smaller the synthesizer swing ratio. Given that the 1st IF would be above 30 MHz by a lesser or greater amount, then if the final IF were to be 100 kHz, direct conversion to that low number might have given rise to image issues, indicating that an intermediate, 2nd IF was desirable. In that case, the choice of the 2nd IF also was also an individual maker/designer issue. In the case of Wadley Loop receivers, the possibilities were predetermined by the loop characteristics. And in some cases, the 1st and 2nd IFs were related by the need to obtain from the synthesizer a convenient oscillator feed to the 2nd conversion mixer.
In the H2900 case, one might suppose that direct conversion from 80.8 to 2 MHz was probably feasible. However, it would seem that its triple-conversion nature was an integral part of the tuning process, in that the 1st IF was tunable over a ±1 MHz range. That the 2nd IF was a relatively high number, 30 MHz, as 2nd IFs go, may well have been required by synthesizer considerations. The oscillator feed to the 1st mixer was stepped in 2 MHz intervals, the smaller intervals down to 1.0 kHz being covered by the oscillator feed to the 2nd mixer. Conversion from 30 MHz direct to 100 kHz may not have been considered favourably, indicating that a higher final IF was required. At the time that the H2900 was designed, the later standard 1.4 MHz number was embryonic, so was probably not an obvious choice, so Marconi evidently chose its own number to suit, namely 2 MHz. This may also have been related to synthesizer design, bearing in mind that the oscillator feed to the 2nd mixer also came from the synthesizer, as did the 2 MHz carrier insertion oscillator feed for the SSB demodulators. Given the one-box nature of this receiver, a frequency for which compact selectivity filters could be used was also likely to have been a desideratum.
The H2900 though did not set a precedent when it came to following Marconi receivers. It may be noted though that the 2 MHz final IF of the H2900 had a precedent in the 2 MHz 1st IF of the MST receivers. And the 30 MHz 2nd IF later turned up as the 1st IF of the McKay Dymek HF receivers, starting with the DR22 circa 1977. These were triple conversion, upconversion, with 2nd and 3rd IFs of 10.7 MHz and 455 kHz, respectively.
In 1976 the H2540 arrived, being a one-box MF/HF General Purpose and point-to-point ISB receiver, covering the range 15 kHz to 30 MHz, and intended for remote as well as local control. This was of the double conversion, upconversion type, with a 1st IF of 68.6 MHz and a 2nd IF of 1.4 MHz, the latter by then the new norm for HF SSB/ISB receivers. The exact number for the 1st IF, 68.6 MHz, derived simply from the choice to use an existing synthesizer, namely that used in the contemporary H1540 transmitter drive. In the H2540 case, the synthesized oscillator input to the 1st mixer moved in 1 kHz increments, so there was no need for an incrementally tuned 1st IF.
The H1540 transmitter drive does provide some insights to the H2540 IF choices. In this case, the SSB/ISB modulation was done at 1.4 kHz, indicating that the “new” receiver final IF was also used at the transmitting end. This was then translated to 68.6 MHz, and from there to an output in the 1.5 to 30 MHz range. The Marconi commentary in respect of the 1.4 MHz generation frequency was:
“The choice of 1.4MHz for the sideband filters is complex, and is governed in part by the following considerations:
“The frequency of the filters should be outside the h.f band, and as high as possible, consistent with obtaining an adequate carrier rejection. Generally the higher the frequency, the easier is the design of the first mixer and i.f amplifier in the frequency translator. It is also important that the heterodyne frequency for the modulator is easily derived from the master oscillator. The chosen frequency should be well within the range of the crystal filter manufacturers' current techniques, so that high-quality filters with an adequate passband phase characteristic can be manufactured for high-speed data transmissions.”
For many practical purposes, the HF band began at 1.5 MHz (all though formally it began at 3 MHz), so 1.4 MHz was about as high as one could go.
Then in respect of the 68.6 MHz number, Marconi said:
“It will be seen that an intermediate frequency of 68.6 MHz has been chosen. This frequency is dictated primarily by the needs of the second mixer which has an output frequency in the range 1.5-30 MHz. The choice of input frequency is limited by a number of factors. The upper frequency limitation is that of the synthesized local oscillator which has been taken as 99.999 MHz in order to avoid the complication of introducing a 100 MHz decade into the synthesizer control loops. The lower frequency limit was chosen so that the unwanted third order mixing products appearing at the output would be greater than about 38MHz which is well outside the HF band.
“A further constraint is that the intermediate frequency should be low enough to permit the use of high selectivity bandpass filter having good attenuation characteristics either side of its centre frequency. This is to ensure that noise generated in circuits before the filtering, does not contribute significantly to the wide-band noise content of the drive output.
“The first mixer has a signal input frequency of 1.4 MHz, as discussed earlier in the section dealing with the modulator. It is convenient if the heterodyne frequency is a multiple of the 5MHz standard frequency to which the whole drive system is locked and in fact, 70 MHz has been chosen. The output frequency is, of course, the intermediate frequency referred to above.”
This does show clearly how the seemingly strange number of 68.6 MHz came about. And essentially the same arguments may be marshalled in reverse in respect of the H2540 receiver.
It may be added that the 70 MHz frequency used for the heterodyne frequency also easily divided down to provide the 1.4 MHz number required for SSB carrier reinsertion at demodulation.
An observation that could be made about the 1.4 MHz final IF is that for dual-conversion receiver purposes at least, it did not necessarily have to be outside of the HF band. Final IFs that were in-band were not that unusual, and for example the H2900 had a 2 MHz final IF. But 1.4 MHz was also used for single-conversion spot-frequency receivers, in which case it did have to be outside the HF band. As discussed upthread, whilst 1.4 MHz did have a prior history of use, its arrival as a de facto standard appears to have coincided with the impending adoption of SSB for marine communications purposes, with Redifon being an early user, and with the filter makers offering a full range for that frequency. The use of the same number for transmitter drives, and single-conversion and double-conversion receivers was certainly quite rational.
Receivers in VHF land mobiles (i.e. taxi, police) in the 1960s commonly used a double conversion design with a 1st IF of 10.7MHz followed by a 2nd IF on 455KHz. Later, with the availability of crystal filters and silicon transistors, they became single conversion to 10.7MHz followed by untuned IF gain. These radios worked extremely well, pulling signals in the 0.2uV range out of the noise. I used them for many years. Most VHF air radios still use this design.
Thanks for that!
10.7 MHz (dating from 1945) and 455 kHz (standardized in 1937, but in use earlier than that) were probably the most widely used broadcast receiver standard IFs, and were frequently appropriated for non-broadcast applications.
I suspect too that the 10.7 MHz, 455 kHz pairing was probably the most common IF combination for double-conversion receivers of all types. Even when regular AM receivers drifted from 455 kHz et al to 450 kHz, the double-conversion combination mostly stayed at 10.7 MHz, 455 kHz.
Regarding land-mobile equipment, the Pye R401 (AM) and R402 (FM) receivers were interesting cases, in that both had PLL synchronous demodulators. They were described as single-conversion, with 10.7 MHz IF. But perhaps one could also describe them as double-conversion, with a 10.7 MHz 1st IF and a zero Hz 2nd IF!
Further to my recent post on the Marconi H2900 and H2540 receiver IF systems, here are their respective block schematics:
Returning to the Eddystone EC958/Marconi Nebula, I suspect that one reason for its triple conversion layout was that its sub-HF coverage extended down to 10 kHz. Thus the desired final IF of 100 kHz would be in-band for the tuning range that included 100 kHz, so that an alternative number would be required for single-conversion in that tuning range at least. The 1335 kHz 1st IF was associated with the incremental tuning facility and the Wadley loop available for the HF band 1.6 to 30 MHz, so was probably not convenient to use for one of the LF bands. That indicated the need for a middle IF, between 1335 and 100 kHz, and adequately above the top of the tuning range containing 100 kHz, namely 53 to 126 kHz. 250 kHz fitted the bill. Concomitantly, the range containing 250 kHz, namely 125 to 295 kHz, was converted directly to 100 kHz.
Also, the use of a middle IF kept the Wadley loop clear of the final IF, and as previously mentioned, allowed the final conversion to be switched between infradyne and supradyne for sideband inversion. For the drift correction to work in this case, the second conversion needed to be supradyne. The second conversion oscillator feed was in the range 1485 to 1585 kHz, converting a 1st IF range of 1235 to 1335 kHz to 250 kHz. One may see that the 250 kHz choice also ensured that the second oscillator was kept (just) below the 1.6 MHz bottom of the HF range. In fact it looks as if Eddystone’s established 1350 kHz IF number was shifted downwards slightly to provide some margin under 1.6 MHz. The one-sided incremental tuning, downwards from 1335 kHz, rather than the customary two-sided approach may also have been part of keeping the second oscillator feed below 1.6 MHz.
I think that some validation of the foregoing reasoning is found by looking at the following Eddystone 1830 model, a simpler receiver using some of the EC958 modules, and which was the direct replacement for the 830. It was double conversion above 1.5 MHz with IFs of 1350 ± 50 kHz and 100 kHz, with infradyne second conversion. From 1.5 MHz down to its lowest tunable frequency of 120 kHz, it was single-conversion, with 100 kHz IF. Thus, absent the complication of the high-stability Wadley loop, Eddystone had returned to the pattern used with the 830, which was double-conversion above 1.5 MHz, with 1350 ± 100 kHz 1st IF followed by infradyne conversion to 100 kHz, and single-conversion, to 100 kHz for the range 300 to 1500 kHz.
Before the 830 was the 910/1 marine receiver, which was double conversion with IFs of 1400 ± 50 kHz and 85 kHz, the latter also an established final IF for communications work, although one which faded out in favour of 100 kHz quite early on. The 910/1 tuned 1.5 to 30 MHz and 375 to 525 kHz. It looks as if the 1400 kHz nominal 1st IF was moved to 1350 kHz nominal for the 830 in order to allow for a ±100 kHz incremental range instead of ±50 kHz, whilst still keeping the range safely below 1.5 MHz. As mentioned a long way upthread, the 1400 kHz 1st IF had been used previously for the Pye/Rees Mace CAT marine receiver.
The 1350 kHz/100 kHz combination was used again by Eddystone in the 1837/1838 receivers of 1977. In this case the incremental tuning range was quite small, 1350 ± 10 kHz. One could say that there was an inertial effect here, in which 1350 kHz or thereabouts was carried over through several receiver generations. Thus it is sometimes desirable to look at the history in any retroactive “outside-in” analysis of receiver IF choices.
As mentioned previously, one area in which IF numbers appeared to be somewhat random was in respect of the 1st IF of upconversion HF receivers, where they usually varied by make and sometimes amongst different models from the same makers. In the early days of synthesized receivers, say the later 1960s and early 1970s, the available synthesizer technology favoured keeping the 1st IF as low as reasonably possible, albeit adequately clear of the 30 MHz upper end of the HF band. Thus numbers around 35 MHz or so were found. With improving synthesizer technology, the numbers tended to move upwards. Perhaps the fact that a higher 1st IF meant a lower 1st oscillator range, in geometric terms, was considered advantageous. With these higher numbers, the reasons for the choice is probably not easily deducible by inspection; designer comment might be necessary. With the early synthesizer designs, though, a certain amount of “outside-in” analysis is possible.
One early example is the Plessey PR155 group, dating from 1966, and including the later PR155n variants. These were all triple conversion, with IFs of 37.3 MHz, 10.7 MHz, and 100 kHz. They probably just predated the advent of the 1.4 MHz final IF for SSB processing, hence the use of the well-established 100 kHz number. That being the case, the step from somewhere around 35 MHz to 100 kHz was probably too large, so a middle IF was needed. 10.7 MHz was the obvious choice, being a widely used standard number with a very wide range of filters available, including for HF applications. I suspect that there would have to have been a very good reason to do otherwise.
The block schematic for the PR1553 is shown, as it happened to be on hand:
In the PR155 group, the 1st, 2nd, and 3rd mixer oscillator frequencies were synthesized from a 1 MHz harmonic comb derived from a 1 MHz master oscillator. In the case of the 1st oscillator, interpolation between the 1 MHz points was done by a nominally 1 MHz range VFO, whose output was mixed with the selected 1 MHz harmonic and then used to control the 1st VCO via a PLL.
The harmonic range used was 35 to 64 MHz. Given that the harmonic comb as generated had to be filtered to suppress those members within the HF band (30 MHz and lower), I think we may suppose that 35 MHz was the lowest usable harmonic.
Then in the interests of stability, the interpolating VFO would have to be kept to a reasonable range, say around 1.5:1. A reference point is the Racal Wadley loop receivers, where the interpolating oscillator covered 2.1 to 3.1 MHz. So 2 to 3 MHz is a reasonable approximation. The VFO output could be added to or subtracted from the 1 MHz harmonics, but for the 35 MHz harmonic, subtraction probably moved the output uncomfortably close to 30 MHz. That pointed to addition, suggesting that the 1st IF was going to be somewhere around 37 MHz.
Conversion from the 1st IF to the 2nd IF of 10.7 MHz using a 1 MHz harmonic as the 2nd oscillator frequency implied that the 1st IF was going to be non-integer. Infradyne conversion was outruled, since this would have required the 27 MHz harmonic, which was necessarily suppressed by the harmonic comb filter. Thus supradyne conversion was necessary, and using the 48 MHz harmonic produced the 37.3 MHz 1st IF for a 2nd IF of 10.7 MHz.
In turn that established the required range for the interpolation VFO, nominally 2.3 to 3.3 MHz, but with some “overscan” at 2.2 to 3.4 MHz, giving a tuning ratio of 1.55:1. Thus was provided a 1st VCO range of (35 + 2.2) = 37.2 MHz to (64 + 3.4) = 67.4 MHz).
The 3rd conversion, 10.7 MHz to 100 kHz, was done by dividing down from an appropriate 1 MHz harmonic. For the original PR155, it could be switched between infradyne and supradyne in order to effect sideband inversion, using 10.6 and 10.8 MHz oscillator feeds respectively divided down from the 53 and 54 MHz harmonics. The PR1553 looks to have been slightly different.
One could say that given the synthesis scheme employed and the choices of the 2nd and 3rd IFs, the 37.3 MHz 1st IF was the lowest that reasonably could be employed.
The following Plessey upconversion HF receiver, the PR2250 of 1979, was double conversion, with IFs of 65 and 1.4 MHz. It was also fully synthesized. One may deduce that by then, there was no great difficulty in synthesizing a 65 to 95 MHz 1st VCO output. Possibly that range was selected because it was a little under the 100 MHz point, and so avoided the need to move from one “hectade” as it were, to another. Also, synthesizing a non-integer 2nd oscillator output, in this case 63.6 MHz, was evidently not a problem, either.
Somewhat more difficult to analyze than the Plessey PR155 was the Redifon R550/R551 HF receiver of 1969. However, a recent discussion on UKVRR has provided some helpful background.
The R550/R551 was an upconversion receiver with double conversion. The 1st IF was 38.0 MHz, and the 2nd IF was 1.4 MHz. It was probably one of the first receivers to use the 1.4 MHz final IF for SSB processing.
Being a little later than the Plessey PR155 receiver, it had a more complex synthesizer. The oscillator feed to the 1st mixer was synthesized in 10 MHz, 1 MHz, and 100 kHz steps, with a VFO interpolating between the 100 kHz points. The reference for the synthesizer was a 1.4 MHz master oscillator that also served as the carrier insertion oscillator.
Apparently in the interests of easing the difficulties of synthesizer design, the 1st VCO output was not fed directly into the main PLL, but was first mixed down to a much lower frequency, in this case 3 to 32 MHz. That range, just over a decade, might well have been the lowest reasonably possible.
One may see that the 1st VCO and the mixed down feed to the PLL would both occupy 30 MHz blocks of spectrum. The mix-down would need to be infradyne in order to preserve the directional sense, so there would need to be room between the two blocks for the mix-down signal. Thus the mix-down signal was going to be adequately above 32 MHz, and in turn the 1st IF adequately above that.
Although the mix-down signal could come from an independent crystal oscillator, by using the same crystal oscillator for the second conversion, its output could be included in a drift-correcting loop. That though required that the second conversion be supradyne, in turn implying that the oscillator needed to operate at 1.4 MHz above the 1st IF and so also 1.4 MHz above the lowest 1st VCO frequency. That made it too high in its raw state for the mix-down function, where it needed to be lower than the lowest 1st VCO frequency. Thus it needed some mixing down for this purpose, but done in a way that did not introduce another opportunity for drift.
That was done by separating the synthesizer into two parts, one doing the 10 and 1 MHz steps, and the other doing the 100 kHz steps and the interpolation between them. The output of the latter, covering a nominally 1 MHz range was mixed with the crystal oscillator output, their difference being used for the mix-down.
The mix-down signal was chosen to be 35 to 36 MHz. Presumably this was the lowest reasonable range given that the mixed-down output was 3 to 32 MHz. This determined the 1st VCO range, which was (3 + 35) = 38 MHz) to (32 + 36) = 68 MHz. Thus the 1st IF was 38.0 MHz. And the 2nd conversion crystal oscillator frequency was (38.0 +1.4) = 39.4 MHz.
From this the required output from the 100 kHz step synthesizer and VCO could be calculated as being between (39.4 – 36.0) = 3.4 MHz to (39.4 – 35.0) = 4.4 MHz. That was achieved by having the 100 kHz step synthesizer operate between 4.1 and 5.0 MHz, and the interpolating VFO operate between 600 and 700 kHz, and mixing to form their difference, (4.1 - 0.7) = 3.4 MHz to (5.0 – 0.6) = 4.4 MHz. Presumably these ranges were chosen as providing the best trade-off when it came to stability and the avoidance of potentially harmful spurs.
That I think shows that given the choice of a mix-down synthesizer and the 1.4 MHz 2nd IF, then 38.0 MHz was about the lowest 1st IF reasonably possible.
The following Redifon upconversion HF receiver was the R1000 of 1978. This was fully synthesized, but retained the same 38.0 and 1.4 MHz IFs as the R550/R551. Give that by then, other makers of similar-class receivers were using – apparently without difficulty – 1st IFs in the 60 MHz range, Redifon’s choice to stay with 38 MHz does look as if it contained an element of inertia. In this case, the 2nd conversion was infradyne, using a synthesized 36.6 MHz oscillator feed.
By the time that the Racal RA1772 HF receiver arrived on the scene, c.1972, synthesizer design had evidently advanced to the point where the “frequency gymnastics” seen in the Plessey PR155 and Redifon R550/551 were no longer needed, and the 1st VCO output could be fully synthesized. The RA1772 was double conversion, with IFs of 35.4 and 1.4 MHz. The second conversion was infradyne, with a synthesized oscillator feed of 34.0 MHz.
Nonetheless, one may infer that at that time, it was still desirable, from the cost and “doability” perspectives, to keep the 1st IF as low as reasonably possible, and that it was still preferable to generate an integer 2nd oscillator second feed, meaning that the 1st IF would be non-integer. If we take 35 MHz as being about the lowest 1st IF that was adequately clear of 30 MHz, then the RA1772 was close to it, offset slightly upwards to accommodate the need for the integer 2nd oscillator. That it got closer to 35 MHz than say the Plessey PR155 and Redifon R550/R551 could have been because it did not require a larger margin for the “gymnastics”.
The following Racal receiver, the RA1792 “Anglo-American” of 1979, had IFs of 40.455 MHz and 455 kHz. The latter was probably to suit American requirements, given that the American market was the primary target. That the 1st IF showed only a moderate upward movement might have been because this was something of a seriously cut-cost receiver. Apparently, it had quite a different synthesizer design as compared with the RA1772, but there might have still been some cost advantage in not going too high for the 1st IF. It also looks as if offsetting the 1st IF to allow for an integer 2nd oscillator output (40 MHz) might have provided some cost advantage, at a time when others did not see that as necessary.
Eddystone was a late entrant to the professional upconversion HF receiver field, with its 1650 series of 1984. That lateness may have had something to do with the success of its EC958 series. The 1650 was double conversion, with IFs of 46.205 and 1.4 MHz. That 46.205 MHz number does look rather odd. A closer inspection shows that the main synthesizer had operated with 2 kHz steps, and that the interpolation down to 5 Hz steps was done by the second oscillator, which was thus adjustable over a 2 kHz range. Possibly it was done this way to facilitate the use of a single PLL in the 1st oscillator synthesizer.
But by itself that does not explain the oddity of the 46.205 MHz 1st IF number. For that, a closer look at the second oscillator synthesizer is needed. The VCO output was mixed down with a reference signal to provide a 4 to 6 kHz input to the controlling PLL. Presumably this was the best way to provide 5 Hz steps over a 2 kHz range. The reference signal used was the 8th harmonic, 44.8 MHz, of the 5.6 MHz synthesizer reference oscillator. The latter choice appears to have been made on the basis that simple division by 4 resulted in the required 1.4 MHz carrier insertion frequency.
Returning to the 2nd oscillator mixdown, for the 4 to 6 kHz output with a 44.8 MHz reference frequency, the VCO output must range from 44.804 to 44.806 MHz, with a mean of 44.805 MHz. Given the 1.4 MHz 2nd IF, this in turn implied a nominal 46.205 MHz 1st IF for infradyne conversion, which was used. (Supradyne conversion would have resulted in a 1st IF of 43.205 MHz.
Thus the apparent oddity of the 46.205 MHz 1st IF is accounted for, or, in other words, why it was that and not say 46.0 MHz exactly. But its actual placement is not explained. The use of different 5.6 MHz harmonics for the 2nd oscillator mix-down would, for the infradyne case, have allowed a range of different 1st IF numbers, for example from 35.005 MHz for the 6th harmonic to 68.605 MHz for the 12th harmonic. One may infer that there were other reasons, not readily apparent, for the actual choice.
This rather small sample survey of synthesized upconversion professional HF receivers has shown the following range of IFs:
Plessey PR155 et al: 37.3 MHz 10.7 MHz 100 kHz
Redifon R550, R551: 38.0 MHz 1.4 MHz
Racal RA1772 35.4 MHz 1.4 MHz
Marconi H2540 68.6 MHz 1.4 MHz
Redifon R1000 38.0 MHz 1.4 MHz
Plessey PR2250 65.0 MHz 1.4 MHz
Racal RA1792 40.455 MHz 455 kHz
Eddystone 1650 46.205 MHz 1.4 MHz
The 1st IFs vary by make and by model, QED. The 2nd (and where used, 3rd) IFs are all standard numbers.
The Marconi H2900 point-to-point HF communications receiver was mentioned upthread. I have since found an article (*) which provided some detail about the choice of the triple-conversion approach and the IF values, something not always included in HF receiver descriptions. It is worth including some of that in this thread, , and it is simplest to simply quote from the pertinent part of the Marconi article, with comments inserted after each paragraph.
“Choice of Intermediate Frequencies
“The choice of intermediate frequencies needs justification where one must consider the advantages and disadvantages of single, double and triple superheterodyne. Advances in the design of crystal filters allows the construction of sideband filters working at 2 MHz. Choosing the final intermediate frequency (IF3) of 2 MHz eliminates the necessity of additional conversion to the region of 100 kHz.”
That the need to for a final IF of 100 kHz (or some other similarly low frequency) to facilitate sideband filtering was disappearing was demonstrated by the widespread adoption of the 1.4 MHz final IF, with Redifon having been one of the first so to do. At the time that the H2900 was in design, it may not have been clear that 1.4 MHz would be the future norm. Be that as it may, Marconi did not give a rationale for the 2 MHz number as such. Perhaps it represented the upper limit of what was possible with the newer sideband filters. Or perhaps it was because 2 MHz was already in the Marconi lexicon, for example as the 1st IF for the MST point-to-point receivers. Marconi did though choose 1.4 MHz for the later H2540 model.
“In a receiver with a single i.f. the radio frequency is converted directly to a relatively low intermediate frequency. Even if the latter is as high as 2 MHz, which will inhibit reception round about 2 MHz, in order to avoid image reception the requirement on the input filtering is impossible. This scheme must be rejected outright.”
That position does seem to be a bit severe. For example, the established MST receivers had a 2 MHz 1st IF. The Marconi N2050 (Apollo), which was more-or-less a contemporary of the H2900, had a 1.1 MHz 1st IF. And the Eddystone EC958, which just preceded it, had a 1.335 MHz (nominal) 1st IF. In the latter case, Eddystone had considered whether its design objectives could be met with a sub-1.6 MHz 1st IF before proceeding on that basis. Perhaps with the H2900, Marconi was looking for even better image rejection than any of those provided. On the other hand, the marine requirements for front-end selectivity, met by both the N2050 and the EC958/5, were regarded as being more severe than most of those applicable to HF receivers.
“The next, less simple arrangement would be to use a high first i.f. above the highest receiver frequency, followed by a relatively low second i.f. The situation is considerably improved in this case but closer examination indicates that the amount of isolation between the first and second frequency changer is again almost impossible to obtain in practice. Lack of this isolation will cause image receptions at twice the second i.f. Reception of two input signals spaced at the second i.f. will also be present. This arrangement also requires the heterodyne oscillator to be continuously variable over the 30 MHz band from a frequency synthesizer. Although it is possible to combine the two high-frequency signals generated in the synthesizer into one continuously variable signal, the mixing process involved would be detrimental to the reciprocal-mixing performance of the receiver. This is another reason for avoiding a dual frequency-conversion system.”
Dual-conversion of the type rejected became the norm, and in fact was adopted by Marconi for the H2540. Thus it looks as if Marconi was being very careful about isolation between the 1st and 2nd IFs. That the H2900 was doing in one box what had previously required a 7 ft rack may have been a factor, also that the H2900 was a dual-diversity receiver, whereas the other single-box units of the period were simply single receivers. Synthesizer performance was a valid concern at the time. Redifon had used a quite complex process in its R550/R551 upconversion receiver, and that was probably lower down the overall performance ladder. But that changed quite quickly, as evidenced by the Racal RA1772 of the early 1970s.
“Triple-frequency conversion is thus necessary. Restraints on the if. and heterodyne frequencies can now be placed in the usual manner taking into account harmonics, particularly harmonics of the heterodyne signals to the high orders. The analysis led to the five possible selections (Table 2).”
“The selected schemes were then investigated in greater detail. I.p.s up to the 15th order were plotted and weighted according to their order and fractional separation relative to the intermediate frequencies. They were also checked by a computer. No particular scheme is outstandingly advantageous but generally of the similar pairs 2 is preferred to 1, and 4 to 5 as the lowest-order spurious frequencies are further off-tune. On the remaining possibilities 2, 3, 4, scheme 3 is an inverting mode which has some slight disadvantages for a drive application and 2 requires the highest frequency for the variable oscillator. Scheme 4 is therefore used as the best compromise. This scheme also has the possible advantage that all the intermediate and oscillator frequencies employed are either outside or near the extreme limits of the h.f. band.”
The second IF, at 30 MHz, was at the upper edge of the HF band, although probably most band activity was below around 28 MHz. The 2 MHz 3rd IF was in-band, given the 1.25 to 30 MHz tuning range. So would have been the emerging 1.4 MHz de facto standard. In general, an in-band final IF does not seem to have been an issue for professional HF receivers of that generation. The use of a crystal-controlled first oscillator appears to have dispended with the issue of its having a frequency range that crossed the 100 MHz point, something evidently preferably avoided with some synthesized approaches.
(*) B.M. Sosin; H.F. Communication Receiver Performance Requirements and Realization; The Radio & Electronic Engineer; 1971 July.
Thanks for the feedback. It was fortunate that Marconi did publish its rationale for the H2900 receiver IF choices. That one would have been very difficult to work out by “reverse engineering”.
Very early on in this sequence, in fact before ‘Radio IFs’ was split off from the ‘TV IFs’ thread, “Anonymous” mentioned the use of a 6.75 MHz FM IF for some French and German receivers (valved) of the early 1960s. (See: https://www.radios-tv.co.uk/community/black-white-tvs/television-receiver-intermediate-frequencies/#post-41483.)
I have recently found another, earlier reference to FM IFs in the 6 MHz range. This was in a 1957 March British IRE paper “Principles of Design of Battery Operated Frequency Modulation Receivers”, by R.A. Lampitt and J.F. Hannifan, both Ever Ready staffers. With the available battery valves, getting enough IF gain on FM was something of a challenge, particularly for portable receivers where keeping the valve count down was paramount. One option shown was double conversion for FM, with a 10.7 MHz 1st IF, and a 2nd IF of around 6 MHz, where more gain was possible. In the example circuit, the actual 2nd IF was 6.5 MHz.
Thus, the “6 MHz” IF was chosen to facilitate higher gain with a given valve complement, presumably with the exact chosen number to be above the top end of the relatively 49-metre SW broadcasting band. A logical question is “why double conversion”. One possible reason is that with the simple front end, lacking an RF amplifier, it may have been thought better to stay with the standard 10.7 MHz number for the 1st IF, which ensured that all images were not just out-of-band, but well out-of-band in respect of the then UK FM band of 87.5 to 100 MHz. Another reason is that the DK92 valve, included as the AM frequency changer, also served as the 2nd frequency changer on FM. In other FM-AM implementations shown in the paper, a DK96 was used as the AM frequency changer, and bypassed on FM. Thus double conversion resulted in better valve utilization.
In the French and German receivers mentioned above were both valved, and both were single-conversion. Presumably 6.75 MHz was chosen over 10.7 MHz for IF gain reasons, and evidently their front ends, almost certainly including RF amplifiers, were considered to be robust enough on the selectivity front for the lower IF not to cause any undue problems.
As previously shown, by the late 1970s, a popular format for professional HF receivers was initial upconversion to a 1st IF of individual choice, followed by downconversion to a 2nd IF of 1.4 MHz for SSB, ISB, etc., processing. Bucking that trend was the Redifon R500 of 1981, which reverted to single-conversion with a 1.4 MHz IF. Redifon argued that good front end selectivity was an essentially element of overall satisfactory performance, and that with it, a single-conversion receiver was not only possible but also more economic. The details were covered in a paper by R.A. Barrs, “A Reappraisal of H.F. Receiver Selectivity”, published in Radio & Electronic Engineer, 1982 July, p.315ff.
Of course, Redifon had previously used single-conversion with a 1.4 MHz IF for its R499 receiver, but this was crystal-controlled, with plug-in fixed-tuned front end selectivity units. In the case of the R500, RF tuning was done by a motor-driven four-gang variable capacitor, controlled by a microprocessor, which kept it on-track with the synthesized oscillator.
Four gangs, in this case a bandpass input and a bandpass interstage, ahead of the mixer, was more than had typically been used for earlier HF receivers with 1st IFs in the 1 to 2 MHz range. Apparently, the cost-benefit trade-off precluded the use of another gang.