Origin of terms "Roofing filter" and "Reciprocal mixing"

[John KC0G]

I am curious to know the origins of these terms. The earliest references which I have seen are in an article (1970) and paper (1971) by B.M. Sosin of Marconi Communications Systems Ltd. They were motivated by the introduction of the H2900 series of receivers.

The article was: B.M. Sosin, "A breakthrough in h.f. receiver design", Point-to-Point Communication, January 1970, pp 4-14. On p.11, it states "In the second conversion, the frequency is changed down to 30MHz and a narrow band 'roofing' crystal filter is used. (The term 'roofing filter' is applied to a filter of narrow bandwidth, at an early stage in the receiver, to give protection against adjacent frequency interference)."

Pat Hawker, G3VA, referenced this paper and the term "roofing filter" in his Technical Topics column the RSGB's Radio Communication, February 1970, p. 92. He used a similar description is his June 1970 article in Wireless World on communications receivers.

The paper was: B.M. Sosin, "H.F. Communication Receiver Performance Requirements and Realization", The Radio and Electronic Engineer, Vol. 41, No. 7, July 1971, pp 321-329. This paper was in two main parts the first covering "Performance Considerations", and the second the "H2900 Solid State Receiver" The latter part referenced the earlier article. On p. 323 Sosin stated: "Only recently has a phenomena called 'reciprocal mixing' been appreciated.

This paper was presented at a meeting of the I.E.R.E. Communications Group in London on 7th October 1970. Pat Hawker was present and commented upon it in the RSGB's Radio Communication, January 1971, p 27, stating that it was the first time that he had heard the term (reciprocal mixing).

If any of you know earlier use or references to these terms, I would like to hear of them

Thanks and 73

John

[Andrew2]

Interesting that in that description, the 'roofing filter' comes at the 2nd IF. I've always regarded the first IF the place to put the roofing filter, so as to give max protection to all successive stages.
I wonder what the filtering arrangement after the first mixer was, and why wasn't that filter called the roofing filter?
Or maybe I'm reading it wrong.

[John KC0G]

You are reading it correctly.
The first local oscillator runs from 81.8 to 109.8Mhz in 2MHz steps.
The first IF is at 79.8 to 81.8MHz with a bandpass filter
The second LO runs from 49.8 to 51.8MHz in 1Hz steps
The second IF is at 30MHz with a crystal roofing filter
The third LO is at 32MHz
The third IF is at 2MHz, with LSB, USB and carrier filters for independent sideband reception.

I now see that Sosin also used the term "roofing filter" in his 1971 paper

[Synchrodyne]

Certainly that is a pair of interesting and thought-provoking questions.

“Reciprocal mixing” I imagine reared its head once synthesizers came into use for professional HF receivers, early versions suffering from relatively high phase noise as compared to their conventional oscillator predecessors, and also as compared with the Wadley loop. Given that at the same time, receiver performance was improving, or at least improvements were being sought in other directions, particularly dynamic range, the phase noise problem arrived at an inconvenient time, and could have been a significant limiting factor to realizing the desired overall improvements.

Based upon limited data on hand, nothing was said about phase noise or reciprocal mixing in respect of the Plessey PR155, GEC RC410, Marconi Hydrus or Redifon R550/551, all early examples with partial or full frequency synthesis. Racal may have been mute on the issue until the release of the RA1772, whose good reciprocal mixing performance was strongly featured. But then Racal wouldn’t have wanted to get into a dynamic range debate involving its RA17 and particularly its RA217 models, although it probably could have claimed low phase noise.

So maybe Sosin and his team were the first to “flesh out” the reciprocal mixing concept, and perhaps they even coined the term. Its lack of mention in respect of the Hydrus suggests that it came out of the H2900 development programme.

“Roofing filters” in a functional sense have been a round for a long time. I had a vague notion that the actual term came into use when double conversion, with say 10.7 MHz 1st IF and 455 kHz 2nd IF came into more widespread use for consumer-type HF receivers, and a standard block filter, crystal or ceramic, was used at 10.7 MHz. Possibly though it traces back to dual-conversion VHF RT receivers of the kind where the main selectivity was done at 455 kHz, with a wider filter at 10.7 MHz.

In the HF realm, the examples preciously quoted, Plessey PR155, GEC RC410, and Redifon R550/R551 all had what were roofing filters in their respective 1st IF sections, but I don’t think that any of these makers used the term “roofing filter”. So perhaps Sosin’s team also coined this term.

So, the net answer is that I am not aware of any earlier use than by Sosin of the “reciprocal mixing” and “roofing filter” terms; it does not turn up in a relatively superficial search amongst some of the usual suspects.


Cheers,

[Radio Wrangler]

The H2900 receiver was quite remarkable in a number of ways, thedesigners were obviously aware of several limiting factors that weren't fully appreciated elsewhere for a nuber of years.

I thoroughly enjoyed taking that prototype apart to look at how they'd done various things and to measure individual sections. The 2N4391 JFET, I'd used a lot an an analogue switch, but they showed it made a decent high dynamic range amplifier and its wide area made for more than just low RDSon. The mixers were a fair bit plainer than the diode rings developed by Hugh and Guy, but the Marconi tea were working several years earlier. I had to make a synthesiser which could keep up with the mixer, and without the luxury of a bank of crystal oscillators. In the and, I'd learned all I could from 'Sosin's Folly' and I didn't fancy using it with thumb wheel decade switch tuning and about a second to lock up. So I passd it on so someone else could have some fun exploring it.

Cerainly the FDM telecomms test industry understood phase noise and reciprocal mixing in the early seventies.

By the way, phase noise is a misleading simplification. The idea of a carrier sitting amidst a pedestal of noise is wrong. The illusion of a carrier is caused by finite resolution analysis. Look more carefully and it's noise all the way in, there is no carrier, just climbing noise. And this fits with the mechanisms of oscillators.

So it isn't a measurement of how much noise there is, a wideband power meter will tell you that, it's amatter of how steeply the muck-heap is stacked.

(I;m reminded of a bit by Terry Pratchet, when one character asked another about the flat Earth on the backs of four giant elephants standing on a giant turtle, and just what was the turtle standing on? The reply was "I've got news for you IT'S TURTLES ALL THE WAY DOWN )

David

David

[G8BBZ]

Hi All,
With regard to the Redifon R550/551, the synthesizer phase noise issue was recognised in the design phase of this equipment in the late 60's, though I am not sure if the term "reciprocal mixing" was current at that time. The receivers used an analogue VFO for the last 100 kHz to ensure a clean LO signal to the 1st mixer. An ancillary unit - the ARU11 - was available which can be switched in replacement of the VFO to provide synthesized steps down to 100Hz, though this did affect the RF performance of the receiver if the 10's kHz, 1's kHz or 100's Hz steps were not set to zero. There is a note in the User manual recommending that these decades be set to zero unless really necessary, though the degree of impairment of the receiver performance was never specified.
The companion Transmit Drive - the GK203 - which is essentially an R550 in reverse, used a different 2nd loop synthesizer module which had steps down to 100Hz as free tuning transmit drives were not popular and presumably the phase noise issue was deemed less critical.
The 38MHz filter in the receiver 1st IF was certainly referred to as a "roofing filter" by the time the receiver went into production (1970), but I do not know where the term originated.
cheers
Peter G8BBZ

[Radio Wrangler]

I used a GK203 with an RA117 as my HF station for about 10 years. I never noticed any issues with TX phase noise, but then you don't - it's a problem for other people. But it looked quite decent on the spectrum analyser.

This thread is stirring memories. George, G3RJV used to have the later model Redifon SD1 sitting in the vicarage study

David

"Roofing Filter" the one at the top, for frequency anyway.

[Synchrodyne]

In ‘Technical Topics’ for 1966 July, Pat Hawker included the comment:

“Although it is not our intention to draw paper comparisons, it may be noted that the direct single conversion to a fixed first i.f. used in the PR155 and the RC410R allows a preliminary crystal filter to be inserted immediately following the first mixer which in the PR155 is of the balanced ring type.

This was part of an item looking at transistor HF receivers generally, and the GEC RC410R, Plessey PR155 and Racal RA217 in particular.

His use of the term “preliminary crystal filter” for what were in fact roofing filters suggests that at the time, the author was not aware of the term “roofing filter”, and given how widely he read the available technical literature, one may infer that if it were in use at all back then, it still had limited circulation.


Cheers,

[John KC0G]

Thank you all for your comments, which prompted me to look further. I had previously looked through some of the articles from Point-to-Point Communication (a Marconi in-house journal) and Technical Topics for several years. That included the July 1966 piece, where I overlooked term “preliminary crystal filter”. :-(

I went back to the two-part paper by G.J. NcDonald and C.S. Burnham,
"Review of progress in mercantile-marine radiocommunication, Part 1. Systems and facilities", Proc. IEE, Vol. 118, No. 11, November 1969, pp 1807-1811
"Review of progress in mercantile-marine radiocommunication, Part 2. Technique and Design", Proc. IEE, Vol. 118, No. 12, November 1970, pp 1812-1820
There is no mention of the terms "roofing filter" or "reciprocal mixing". McDonald was with Marconi International Marine Co., and Burnham was with Marconi Communications Systems Ltd. This adds further credence to the theory that the Sosin was the first person from Marconi to publish these terms. The H2900 was an expensive receiver, and some evidence that it solved known technical problems would have been useful.

In Gosling's book "Radio receivers". Sosin wote the chapter on "Receiver performance and Specifications", and took a certain amount from his 1971 paper.

Pat Hawker referred to the potential effects of phase noise in his Wireless World article in the June 1970 issue, but as stated above did not use the term "reciprocal mixing". The effects of phase noise were discussed in a paper at the IEE Conference on frequency generation and control, by two people from Rohde & Schwarz. It was mentioned by Pat Hawker in Technical Topics in May 1968.

Somewhere I have a book (from the 1970's ?) on noise testing of communications channels. At the moment it is conspicuous by its absence.

I am left wondering where else to look. The first IERE conference on Radio Receivers and Associated Systems was held in 1972 The first IEE conference on HF Communication system was held in 1979. The IEE Radiocommunicaton convention was held in 1963, which is too early for this discussion, but the proceedings are fascinating. That leaves professional journals and in-house journals. The latter from the UK are hard to find here in the U.S.A.

[Radio Wrangler]

The book on noise testing will be "The White Noise Book" by M J Tant of Marconi Instruments.

It's based on the Marconi noise power ratio measurement instruments. Scientific Atlanta were also active in the field. HP were a bit too late into the market with a compendious monster (in three stacked boxes) called the Baseband Analyser (3724A, 3725A, 3726A) and a much younger me synthesisers in it...

NPR measurement really did show up things like phase noise, mixer linearity and intermods in crystal filters.

Essentially, you hit a broadband FDM system (60kHz to 18.6MHz baseband) with full band flat white noise at high levels. The noise had narrow slots filtered into it (about 90dB deep) representing a few quiet channels amidst a full load of loud ones. At the far end of the system , you look at the output baseband and try to measure the depth of the notches. The test gear had to be able to measure more than 80dB depth with source connected straight to measuring receiver, and that meant having individual crystal preselector filters for each of the preferred slot frequencies ahead of a very high dynamic range receiver.

The tale of the HP instrument eventually went in the HP Journal of April 1982, with a photo of the monster on the cover. By this time PCM systems were rolling out to replace FDM systems, and the baseband analyser was simply too late to sell more than about a hundred units. However, a lot was learned in developing it, and that knowledge went into much more mainstream and successful RF test equipment.

David

[G8BBZ]

Hi All,
Radio Wrangler's mention of the Marconi White Noise test sets brought memories flooding back of many long hours as a junior engineer spent at (often remote) baseband sites on 1200 channel N+1 FDM/FM microwave systems sending noise down the line so that the more senior guys could find the source of some noise component that was putting the system out of spec. I carried a copy of "The White Noise Book" with me for a long time and still have it in my bookcase. Eventually I graduated to being the one trying to track down the source of the noise - but then the world went digital.
Happy days
cheers
Peter G8BBZ

[G0HZU_JMR]

Phase noise in oscillators was investigated in the mid 1960s by Dave Leeson and he came up with his classic phase noise equation some time around 1966 I believe. This allowed the noise of an oscillator to be modelled wrt flicker noise at 9dB/octave and thermal noise at 6dB/octave and the noise floor was a function of the loaded Q and frequency etc.

I think he was asked to do this work because the phenomenon of reciprocal mixing had recently been appreciated in those days although I don't know when it was first called reciprocal mixing. But clearly there was some demand for this research work to be completed (in a hurry) back in those days.

I've used Leeson's equation numerous times when designing oscillators from HF through to many GHz and it always seems to give good results. The only thing I struggle to predict accurately up at many GHz is the flicker corner frequency. But I usually model the far out phase noise at 100kHz and 1MHz offsets within 3dB of the 'real circuit when tested, even for oscillators up at several GHz

I also have a copy of the "White Noise Book" here and I've done a fair bit of NPR testing at work although this was 20 years ago...