Noise Figure data for VHF TV and FM front ends

[Radio Wrangler]

We've all got somewhat different viewpoints, so in discussions like these, we all keep learning something new from each other. We are all basically in agreement, so it's a nice, friendly discussion. If a non-combatant is enjoying it, then that's brilliant.

David

[G6Tanuki]

I think there was a bit of a difference between the UK and US TV/FM front-end design-mentality.

In the UK there were historically few stations on either VHF TV [bands I and III] or FM [Band- II] - so intermodulation effects were less of a problem. Typically you would have 1 BBC station on Band I and one ITv station on Band III [joy and blessings and lots-of-friends fell upon those who could teceive both ATV and Granada on Band III, specially when their episodes of TV-series were broadcast on different days with a shift-of-episode!]. Band-II meant you enjoyed/suffered three BBC National FM stations and nothing-else unless there was a 'lift' on.

In the US, at least in urban areas, pretty much every TV channel was occupied, and tuning across the FM dial got you dozens of stations even with a set-top 'rabbit-ears' antenna, whereas in US-Rural areas people were putting-up 7-element Channel-2 beams and Channel-Master rotators to struggle to get more than a single TV station; even then it could be a case of "We can only get the DuMont network on thursdays-when-it's-not-raining".

So UK TV designers had it kinda easy - little crossmod/intermod problems to contend with, though in some areas signal-strength was low leading to 'Fringe' models with an extra RF-amp and/or flywheel-sync to try and drag a signal out of the noise.

US manufacturers needed both low-noise for the rural viewer/listener and the ability to watch/listen-without-birdies when there's a 50Kw station on the next-channel-up transmitting from a tower five miles away.

Somewhere I've got a schematic for the front-end of a late-60s Fisher FM-tuner that uses no RF amps, two loosely-coupled signal-frequency tuned-circuits, and a quad of hot-carrier-diodes as the mixer. That was quite innovative at the time, but clearly fulfilled a need.

"Noise Figure" is relatively simple to measure in a single-frequency situation; it becomes somewhat meaningless when there's a vast slew of strong signals either side of the frequency at which you're measuring the noise-figure.

[SteveCG]

G6Tanuki summarises the TV/FM tuner manufacturer's dilemma, which also manifests itself in the choice of aerials for the receiving different environments. If you are in the Urban setting with high field strength from multiple transmitters then a wide-band aerial design would easily suffice - and in addition money could be saved on materials (boom and element metal) by not really covering the lower part of the band well (eg in the USA coverage of channels A2 and A3) since the signal strengths were well above the receiver's noise figure (and downlead losses) let alone the Galactic noise floor ! As G6Tanuki comments, a rather different situation to those in remote rural locations where out and out S/N was the determiner.

[Synchrodyne]

Even with just two VHF TV channels used in a given area, cross-modulation could be a problem with bipolar VHF TV tuners. See my post #6 above, 8th paragraph.

Thus I developed an aversion to receiving equipment (other than portables) that used bipolar devices (at least the consumer small-signal types) in their RF and mixer stages. The lack of signal-handling capability also existed at MF, for example with wideband AM tuners. The AWA AM3 (bipolar) overloaded and got into cross-modulation quite easily, whereas I never succeeded in coaxing the Quad AMII (valves) into bad behaviour, even at a location about 2 km from the regional transmitters and with 30 metres or so of longwire aerial, signal strength such that the EM84 tuning indicator bands were overlapping.

It would appear that the use of low-noise RF devices, materially lower than the circuit target, allowed more design flexibility in some cases. The following is taken from the overview description of the Marconi H2001 Hydrus solid-state point-to-point HF SSB/ISB receiver of 1968:

“The degree of pre-selection, and hence the number of tuned circuits which may be inserted between the antenna and the signal-frequency amplifier, is a compromise between sensitivity and selectivity. It is generally accepted that a noise figure of 6 dB to 8 dB may be usefully employed at the upper-frequency limit of the h.f band ( 30 MHz). Provided that the noise figure of the active devices used in the signal-frequency amplifier can be made sufficiently low, say 2 dB, then greater scope is offered to the designer in providing antenna pre-selection. During the past two years, field effect transistors (f.e.t.s) have entered the scene and are now available at economic prices. These devices offer attractive advantages over the more conventional bipolar transistors, resulting in improved signal-handling capabilities at lower noise figures. Depending upon circuit configuration, improvements in crossmodulation performance of up to 20 dB have been measured. Coincident with better signal handling, device noise figures of 2 dB or less are realizable.

“In the Hydrus equipment it has been possible to introduce a band-pass coupled circuit in the antenna input, which produces 30 dB to 40 dB attenuation at 10% off tune. This circuit feeds the signal-frequency amplifier, which consists of two junction f.e.t's connected in cascode. Noise figures of between 4 dB and 7 dB have been achieved. The cascode arrangement is preferred as it gives good isolation between the input and output of the amplifier and it also provides a convenient terminal for the application of a.g.c.”

Thus the Hydrus had a four-gang, single amplifier stage RF amplifier with bandpass input and single-tuned interstage. In the valve era, Marconi typically used four gangs with two RF stages, with single-tuned input and interstages. Thus the use of a low noise device allowed similar or better RF selectivity using only a single stage, and probably with much better selectivity before the RF amplifier. Other Marconi group receivers of the same era, such as the Eddystone EC958 and Marconi N2050 Apollo, also used single RF stages (dual-gate mosfets in those cases) preceded by a bandpass input, whereas their valve predecessors had two RF stages with single-tuned circuits.

That makes me wonder if, in the transition from bipolar transistors to fets for the VHF TV case, there was in fact lower device noise, this providing a margin that allowed for tighter input selectivity at the same overall noise factor as the bipolar case. Nonetheless, VHF TV tuners generally retained the four-gang form with a single-tuned input and bandpass interstage, although in some cases, such as the 1968 Zenith unit, the desired RF selectivity was obtained in a three-gang (three-coil in American terminology) circuit that nevertheless matched prior valve four-gang performance.

In the case of UHF TV tuners, the classic American design with a diode mixer had a bandpass input. Presumably, that was unavoidable given that was the only point at which RF selectivity could be provided. Early European UHF tuners using say the PC88, PC86 valve combination had an aperiodic input (for lowest noise, I think) and a bandpass interstage. For the UK case, a single-tuned input, in addition to the bandpass interstage, was needed as the easiest way to provide the much higher image rejection that was required by the decision to use a 39.5 MHz IF rather than the European standard 38.9 MHz. That pattern was generally carried over to the bipolar era, although I think that there was at least one case (Thorn?) where, for lowest noise, an aperiodic input was used with a triple bandpass interstage. The American varactor-era UHF tuners typically used a single-tuned input and bandpass interstage, effectively following the UK pattern. That was true for the RCA KRK-160 (bipolar RF amplifier) and KRK-226 (mosfet RF amplifier). That the latter had a lower noise factor than the former suggests in this case, the device noise benefit was taken more-or-less in full, with little or none used to offset to allow higher input selectivity.


Cheers,

[Radio Wrangler]

We've come around almost full circle, nowadays. We can do better amplifiers with bipolar transistors using 'noiseless feedback' type circuits... See David Norton's and Qubit's patents. Chris Trask also applied the same ideas to mixers.

These are RF amplifiers linearised by plenty of feedback, but the feedback division ratio is done by an RF transformer, offering less noise than a feedback resistor would contribute. Cake and eat it! It does come off sometimes.

What Marconi was saying was that with a low noise active device, you can afford some loss ahead of it. This loss could be a resistive attenuator, and it will have a favourable result on big signal handling, intermod, crossmod etc.

OR you could use the same loss budget to afford a tighter preselector. This, if narrow enough, can give you even better reduction of unwanted signals. So if your budget allows you to have, say, 3dB of loss before your low noise amp, it's better to spend it in a filter than in a pad.

There is a sort of figure of merit for a preselector. How much can it attenuate your unwanted stuff, compared to how much it attenuates your wanted stuff.

Seymour B Cohn wrote a paper on this. Designing best rejection for the amount of in-band insertion loss. It turned out his optimised filters were not classic Butterworth, Chebyshev, Gaussian etc shapes, they had rather erratic ripples at the the edges of their passbands, but they did have the best bang per buck ratios.

He happened to use shared inductance bottom-coupling in coupled resonator filters. The amateur radio world completely missed the point of least insertion loss for given rejection, and looked at the bottom coupled topology and promptly named it 'The Cohn Filter' subsequent articles (and Pat Hawker's column) cemented this name in people's memories.
Cohn did an awful lot of filter work, not just that bit. There is no one Cohn filter. Also the uniqueness of that work lies in the component value calculation, not in the exact circuit topology used in his example of how to implement it.

David

[SteveCG]

A further aspect to add to the mix...

What improvement possibilities did a lower noise front end, or increased aerial gain for that matter, give to the viewer of Analogue TV in the VHF part of the spectrum?

We know that the picture quality as perceived by a typical observer, decreased with a poorer signal to noise ratio, Here is a six point scale used in the USA:

Unusable -- 10 dB S/N
Inferior -- 19 dB
Marginal -- 24 dB
Passable -- 29 dB
Fine -- 34 dB
Excellent -- 42 dB

Data: "Antenna Application Reference Guide", Page 6-9, Fig 6-2.

The BBC had a similar six point scale, with the CCIR having a five point scale.

Studying this scale shows that for the picture quality to perceptibly improve for people not in the Excellent class an improvement of about 5 dB is needed in the S/N.

Now typically a single channel, 3 element TV yagi has a gain of about 6 dB with respect to a dipole, with a 6 element version having a gain of about 9 dB wrt a dipole. If we consider the physical stress limits that a Band I aerial imposes upon any chimney stack mounting, then it is clear that a bigger aerial can only bump your picture quality by about 1 scale point - unless the viewer invests in separate mounting tower. So all that is left to the viewer is to reduce the effective noise figure of his receiving system. To do this he can either buy a TV set with a lower noise figure, or if possible, install a mast head low noise pre-amplifier (assuming that there are not strong unwanted signals present to overload the pre-amp). Either way he might only improve his viewing by another scale point.

From all this we conclude that if you are not getting a very good picture with a simple receiving set-up then your scope for improvement is limited. Of course what is acceptable to one viewer may differ from one to another. Somebody may be happy to just get something whereas another may expect the impossible in perfection ( he's seen all the TV set makers adverts in the showroom! ).

This all goes to show that broadcast reception is really a 'N dimensional' problem...

( And we've not considered Digital here at all, whether Radio or TV )