FET Questions

[G8HQP Dave]

Another issue with dual-gate MOSFETs for HF front-ends is that they may have too much gain. Fine for VHF, where noise figure is more important, but at HF you want as little gain as necessary before the mixer - often none at all.

[Synchrodyne]

Originally Posted by G8HQP Dave

Another issue with dual-gate MOSFETs for HF front-ends is that they may have too much gain. Fine for VHF, where noise figure is more important, but at HF you want as little gain as necessary before the mixer - often none at all.

I suppose that what is desired ahead of the first mixer depends upon the mixer characteristics. In some cases gain might be desirable to offset mixer noise, but perhaps then only at the higher end of the HF band. In other cases, a reasonably amount of RF selectivity could be required, to protect a mixer that is not all that linear and also for single-conversion receivers with low IFs to obtain good image rejection.

This was discussed in a valve receiver context in the thread:

https://www.vintage-radio.net/forum/...ad.php?t=98269,

and it is worth quoting G8HQP Dave’s comment:

Originally Posted by G8HQP Dave

The frequency at which a low noise RF amp is needed depends on both the antenna and the mixer noise. With a poor antenna an RF amp might even be useful on MW. Given a reasonable antenna and an ECH81 (or similar) mixer my guess is that you need an RF amp from a few MHz up, and a low noise RF amp from perhaps 20MHz up. Even then, over much of the HF range the main source of receiver-generated noise is thermal noise in the first tuned circuit. At higher HF there is a transition to valve shot noise then grid noise at VHF.

Too much RF gain could reduce receiver performance more effectively than too little RF gain. Note that the ECH81 used in European designs is actually quite a good mixer: not too noisy, and reasonable strong signal handling.

From there one may transpose to a receiver that followed a typical valve receiver signal path but used fets instead of pentodes and heptodes, etc. Real-world examples might not be so common. The Drake SPR-4 was broadly similar to the R-4. And the Eddystone EC958, 1830 and 1000 all had signal paths that were not totally unlike those of earlier valve models.

So with say a dual gate mosfet rf amplifier and either a dual-gate mosfet or push-pull jfet mixer, a question is would mixer noise be such that at some point in the HF band it would be preferable to use an RF amplifier? I assume that a fet mixer is less noisy than a heptode or even a pentode, but perhaps not as quiet as would be desired at say above 20 MHz. But perhaps the fet mixer would have less “headroom” than a heptode or pentode? That said, the data available for the US VHF TV tuner case indicates that at VHF anyway, mosfets in the RF and mixer positions matched or bettered valve performance in terms of cross-modulation.

But these fet mixers are evidently not linear enough to work without the protection of RF selectivity, and good selectivity might be easier to obtain around a gain stage. And the gain stage itself, unless extremely linear, probably needs the protection of input selectivity. Some valve HF receivers had two RF stages, ostensibly to obtain even greater pre-mixer selectivity and image rejection rather than for extra gain. But I don’t think that practice was carried over to the fet era, perhaps because the additional rf gain would have been counterproductive. For example, the Eddystone 880 had two RF stages (ECC189 and 6BA6) ahead of its 6AK5 1st mixer, whereas the EC958 had a single RF stage, albeit with a bandpass input, using a jfet-mosfet cascode, ahead of its dual-gate mosfet mixer. That may have been simply the outcome of different choices, as the 958 was not exactly a solid-state clone of the 880. Anyway, it does seem that where, mosfet or jfet 1st mixers were used, then they were usually preceded by just one RF stage. The previously mentioned JRC NRD525 might also be illustrative, in that the RF amplifier remained in circuit even if the front-end tuning was switched out. One could infer that the mixer was a tad noisy at the upper end of the HF band to work without some RF amplification.

On the other hand, good four-diode DBMs seem to be capable of working adequately well through to 30 MHz with nothing in front of them but a low-pass filter. That is assuming a sufficiently high 1st IF that all images are well out-of-band.

Originally Posted by Radio Wrangler

In the late '80s early '90s the 3SK125 was the RF amplifier FET of choice, and also used in push-pull mixers. THe NRD525 and family as well as the IC765 by my elbow used them. Their performance was quite good, but better was about to happen. The Japanese designers had a terrible tendency to say 'Well that's the signal through a crystal filter, we can forget about linearity, now' and so the closer-in intermod performance degraded terribly. THe audio distortion was bad too.

The designers liked to give customers variable bandwidth IF filters, and did it by sliding two filters at 8 point something MHz and 455kHz against each other. So you wound up with 16 poles of crystal filters and bad ringing and group delay unflat enough to even become audible. Ugh!

The NRD525 had a relatively straightforward signal path, lacking say the variable passband multiple-conversion complexity of the ICOM R70 and R71. But it still had not-very-good audio, although reputedly not as bad as that from the R70/R71. One may allow that the inbuilt amplifier and speaker were of a “courtesy” nature only, but even using the line output to feed an external amplifier and speaker system, it was not stellar. But I doubt that the audio quality was anything much to do with the choice of rf and if devices; rather it was lack of attention to detail. One found similarly quite poor performance with AM (MF) sections of many Japanese hi-fi tuners, in contrast to their FM capabilities, although there were exceptions here and there, notably some Sansui models. The ICOM R7000 VHF-UHF receiver also had relatively poor audio even via its line output. Quite telling was that audio from the TV receive adaptor accessory line output was much better than that from the main receiver line output when one tuned into to a TV transmission. The R7000 by the way had a 3SK121 dual-gate fet (mos or GaAs I don’t know) rf amplifier (a separate one for each of the 4 tuning bands) feeding a DBM.

Originally Posted by Radio Wrangler

I have grave doubts about the usefulness of varactor tuned preselector filters given the dynamic ranges on HF. Marconi published an interesting paper in their review studying levels found across the HF region and receiver overload causing loss of reception. The figures were fair, but it was famous for B M Sosin trying to push his wonderbox, the H2900 receiver by taking a poke at the competitor's performance. Famously Marconi backpedalled and sent a new expurgated edition out and asked recipients to destroy the old copies. Yeah, right.

My (relatively simplistic) understanding of the Marconi paper was that Sosin’s position was that HF receivers with relatively sharp and good front end selectivity were better than those with more “open” front ends. Thus his test procedure favoured such receivers, including the Marconi Apollo and Eddystone EC958 and 1830, all of which had tuned fet RF stages, with bandpass inputs in the Eddystone cases. That said, I have no knowledge of the circuitry of the Marconi H2900 and Hydrus models (other than that the latter was upconversion and used fets), both of which were said to have come out well. Still, would it be fair to say that the overall concept embodied in the RA1772 won the day?

Originally Posted by Herald1360

The RA217 didn't have many fans, even at Racal! Strong signal performance was considerably worse than its predecessor RA17 series kit. It was solid state, though....

So even in the commercial world in the late 1960s, “solid state” was a sales feature and possibly a customer desideratum as well, even if performance was noticeably lower than could be achieved with valves. In the consumer world, that was more understandable, but it did for example result in the production of bipolar germanium or silicon FM tuners of mediocre performance. In Stephen Spicer’s excellent book on the history of the Leak company it is evident that the press to release a solid-state FM tuner was running ahead of device capability until suitable (and one presumes suitably priced) mosfets came along, the result in this case being the “Stereofetic” model. A sidebar question here is why did the Stereofetic use a single-gate mosfet RF amplifier, with agc, when a dual-gate type was the norm. Designer whim or based upon sound reasoning?

One may also ask whether the RA217 would have been at least somewhat better had it used fets instead of bipolar transistors in the RF and IF signal path?


An earlier and allegedly not-so-good solid-state exercise was the Eddystone 960, which was notionally the bipolar counterpart to the valved 940. Whereas the latter was a four-gang, 2RF, 2 IF design, the 960 was three-gang, 1 RF, 3 IF. I guess that two RF amplifiers ahead of a bipolar transistor mixer would have been just too much for it, hence the difference. But one may wonder why the 960 was not four-gang with a bandpass input.

Was the Plessey PR155 based upon bipolar transistors, or did it have some fets? I suspect that it might have been released on the cusp of the fet era.

Originally Posted by G6Tanuki

It's always interesting to compare designs: my RA217 has a manually-tuned preselector which then feeds two simple bipolar transistors as a RF amp ahead of the first balanced-diode mixer. There is a "wideband" position on the preselector but that's unusable when you've got a 100-foot longwire attached!

I wonder how well960s/1970s FET mixers compared on HF to the approach used in the Atlas 210/215 transceivers - no RF gain, preselector straight into a 4-diode double-balanced-mixer with lots of l.O. drive, then into the first IF filter. It always seemed to make sense to me to put the signal into a nice narrow filter as soon as possible.

I guess that we need to be careful here not to stray too far from the heading topic, but comparison of fets and the circuit layouts in which they were usually used with other devices and their typical circuit layouts is I think within bounds.

The only qualitative comparison I have in this regard is from 20 or so years ago and relates narrowly to shortwave program listening when such is done for extended periods, and not to HF reception generally. My main receiver was a Liniplex F2 with the outboard OSC-1 synthesizer. The F2 was designed expressly for SW program listening, and made no concessions to general coverage requirements. The RF went straight into a DBM, with only a low-pass filter in the way. The DBM output then went direct into a 10 kHz bandwidth 35.4 MHz 1st IF filter. The second conversion and 2nd IF strip were bipolar (RCA CA3026 type, I think), and fed a sophisticated tracking PLL demodulator with selectable sideband facility. Not a fet in sight. My backup receiver was the NRD525 (all-fet signal path and with a tuned RF stage) from which a 455 kHz IF output was taken to a Sherwood SE3 Mk II outboard PLL demodulator, also with a selectable sideband facility although that needed to be accessed by offsetting. In both cases the audio line outputs fed into the stereo system. The antenna was a trapped dipole in the attic, maybe 9 metres high at its peak. For listening to the BBC World Service, the Liniplex was always better than the NRD525/SE3 combination. It was useful to have the latter set up to the incoming frequency just ahead of a frequency change time so as to maintain program continuity, and otherwise set to an alternate frequency, but it would have been a rare occasion if, after having switched to the NRD525, I did not quickly retune the Liniplex and switch back to it. Oh, and also, my OSC-1 was an early version that went up to 22 MHz only, so the NRD525 was also used during the periods when the BBC was broadcasting on 25 750 kHz, usually a very steady albeit lowish signal.

I suspect that the differences here stem from the different design objectives for each receiver rather than because of their different signal paths and active device technologies. And I was not “swamped” with HF signal, either. My recollection is that during the evenings, the BBC on 5975 and 6175 kHz ran at a bit above 40 dB (100µV) on the Liniplex meter.

Cheers,

[Radio Wrangler]

Hmmm....

The input impedance of Mosfets is famously high. Their gates look like small capacitors, with either a small series resistance component, or else a very large parallel resistance component, depending on how you prefer to model them. If we think of available power gain, then we have to include the effects of input and output matching circuits. Particularly at lower frequencies, we can make input matches which have a very high impedance step-up ratio while resonating-out the capacitive component. These have very large voltage step-up ratios and so the gain comes out a lot higher than you'd expect from just looking at gm and load resistance figures. Additionally, at low frequencies, the feedback Miller capacitance from drain to gate becomes negligible.

So applied in matched circuitry, fets give increasing gain at lower frequencies. I design with GHz mosfets in the hundreds of watts league. Their increasing gain with reducing frequency makes achieving stability at low frequencies rather difficult. The matching networks rarely terminate ports so many octaves from the design band. Use any of the usual microwave/RF CAD packages and their Linvill/Rollets stability analyses focus on the wanted band and fail to understand just how wide the band is in which singing its own song is undesirable. THis is really a failure in imagination of the people responsible for the software and a tick box mentality "Does it do stability analysis... TICK" But there wasn't a box for whether the stability analysis did the necessary job.

Ordinary small signal dual gate mosfets are very unsuitable for MF/HF front ends.

Everything depends on band conditions. If you want a contact on 21MHz when the band is closed, it is like any VHF/UHF band, you want very low noise indeed. When the band is open and raging with the traditional calls of "DX ONLY!" and "You are 5&9, please repeat your callsign" then you need something bullet proof at high levels with very little gain, usually none.

I don't know the Icom 7000, but I do have an R9000 kicking around the place. It's audio amp and speaker aren't bad. They are not the limiting factors. The SSB and FM demodulators are OK, the AM demodulator is poor. The intermod of in-channel components down the IF signal path is poor. It uses diode ring mixers at VHF/UHF and sticks block converters in front for the higher ranges to 2GHz. On HF and below it has a separate front end with push-pull jfets.

I used to have a Marconi H2900 which was one of the prototypes... the whole cabinet one block of aluminium, milled from the solid! Someone had nicked the switch mode PSU transistors, and the control chip was goosed. I fixed this and got it going, but ran into problems with bad connections in the synthesiser. With thumbwheel switch tuning it was really only useful as a point-to-point rig. no good for exploration. Diode mixers were used. RF/IF amps were based on high current jfets intended for signal switch use, but they work at RF well. 2N4391. I lost interest in it and sold it, though it was probably a significant historical artefact

A lot of 60's 70's design was mosfet based because mosfets were seen as valve like and designers were used to valves.

As said, other firms rushed into transistors and stuck on "Solid state" badges like military decorations. It took a while to find out that this was a huge step backwards and longer before design techniques allowed the little devices to be used without a net penalty.

THe RA217 I know well, because I have its steam-rollered brother the 1217 rackmount version. It isn't a wonderful performer, and it certainly needs to have the preselector in just about all the time.

Probably the best HF receivers available new today are to be found only in transceivers because the market for pure receivers is seen as 'less sophisticated' or maybe less likely to pay as much.

David

[Herald1360]

Originally Posted by Radio Wrangler

THe RA217 I know well, because I have its steam-rollered brother the 1217 rackmount version.

Like the concept.

[G0HZU_JMR]

.....

[G0HZU_JMR]

Thus I was surprised when looking closely at an American TV VHF tuner schematic (attached) to find that the mixer had the local oscillator injected at gate 1 along with the signal, and not at gate 2.

Further checking across other sources suggested that this practice, if not universal, was quite widespread.

Looking at the circuit you posted up I think they used gate1 for both RF and (VHF) LO because this provided part of a cheap system that allowed it to switch between VHF and UHF tuning.

i.e. I think the 'mixer' Q2 can also be used as a plain IF amplifier in UHF mode. To me it looks like (in UHF mode) the UHF IF input signal will arrive via CR10 along with a bias voltage that will help to turn off the oscillator Q3 (Q3 only used in VHF mode?)

You can see that UHF mode will reverse bias the oscillator Q3 Vbe via CR10 and help to disable the Q3 LO.

So I think the mixer at Q2 isn't always a mixer. Maybe this is why they feed the LO to gate 1 because this allows the 'mixer' to behave as an IF amplifier when in UHF mode. The circuit design will be driven by cost rather than performance so they found a cheap way to combine 'LO switching', and 'UHF/VHF mode selection' and also make a mixer behave like a plain IF amplifier in UHF mode.

I should point out that I know next to nothing about TV tuners and I'm just looking at the circuit provided. I'm assuming that the UHF mode turns off the Q3 LO because that is how the circuit appears to be designed.

[G0HZU_JMR]

In case anyone suggests that they could still have used gate 2 for the LO with this system then there are a few things to consider.

In VHF mode where the DG MOSFET Q2 needs to act as a mixer then Q2 would usually be biased (lightly) as a 'mixer' at gate 2 and fed a large LO waveform at gate 2 to provide the mixing action. Sometimes this low voltage bias is taken from the source via a large resistor (eg 100k). Sometimes it is biased differently and there is 0V bias at gate 2 using a simple shunt resistor to ground and this requires a lot of LO voltage swing.

So this type of bias would be a bit weak for normal IF amplifier service in UHF mode. So this wouldn't work well in the UHF mode where Q2 needs to behave as an IF amplifier as the gain would be a bit low.

When the LO wasn't needed (on UHF mode) this also produces a problem at Gate 2. Ideally gate 2 would need to be ac decoupled to provide a constant and clean/stiff bias here when acting as an IF amplifier as it aids amplifier stability for one thing. But this clashes with the need to feed an LO here in VHF mode where gate 2 needs to be high impedance to allow lots of LO voltage swing. So the ac decoupling on gate 2 would get in the way when in VHF mode if they fed the LO to gate 2.

So it looks like they opted for a cheapo low performance mixer with both RF and the LO (in VHF mode) fed to gate 1. Gate 2 is biased and decoupled to suit the (performance compromised) VHF and UHF modes of operation for Q2.

At least that's how I see the circuit

[Synchrodyne]

Yes, it could simply be a case of low-cost design.

Meanwhile, I have unearthed some additional information.

RCA seems to have been a major player in the introduction of dual-gate mosfets, this following its activity in single-gate mosfets. From the start it stressed the functionality of gate 2 for both agc in amplifiers and local oscillator injection in mixers. To support this it provided two worked examples using a pre-production dual-gate mosfet TA7151. One was as an AM (MF) car radio RF amplifier in which the mosfet replaced a bipolar transistor, and the other was as a mixer in an FM front end, schematic attached.

The 3N140 and 3N141 and the 4060x series appear to have been the production outcomes of the TA7151, or something like it. The 3N140 was designated as an RF amplifier and the 3N141 as a mixer, and the datasheet, as included in the RCA 1978 Linear Databook, included the following commentary:

“The 3N140, used in a common-source configuration in which gate No.2 is ac grounded, reduces oscillator feed-through to the antenna thereby minimizing oscillator radiation. The 3N141 provides excellent isolation be¬tween the oscillator and rf signals because each of the two signal frequencies being mixed has its own control element.

“The mixing function performed by the 314141 is unique in that the signal applied to gate No.2 is used to modulate the input-gate (gate No.1) transfer characteristic. This technique is superior to conventional "square law" mixing, which can only be accomplished in the non¬linear region of the device transfer characteristic.

“The use of the 3NI41 as described provides high useful conversion gains at all vhf frequencies, and the reduc¬tion in spurious responses is substantial and easily obtainable in simple circuits.”

That creates the impression that RCA saw itself as the “inventor” of gate 2 oscillator injection, and that it alone was recommending this at the time that the 3N141 was released. The test circuit in the databook (attached) shows LO injection at gate 2.

The 400600, 40601 and 40602 were designated for TV applications as RF amplifier, mixer and 1st IF amplifier respectively, and the 40603 and 40604 for FM applications as RF amplifier and mixer respectively. In both cases the mixers had gate 2 LO injection.

The same pattern was repeated with the later 40820 (TV RF), 40821 (TV mixer), 40822 (FM RF) and 40823 (FM mixer), these being gate-protected types derived from the 3N187.

So RCA’s position was quite clear; when a dual-gate mosfet was used as a mixer, the local oscillator injection was at gate 2.

But RCA also offered the 3N204/205/206 and 3N211/212/213 which were TI-origin devices. Here it seems to have followed the TI precedent in that local oscillator injection was at gate 1, along with the signal. Certainly that is the way it is shown in the databook, 3N212 case attached.

So far, so good; RCA and TI simply seems to have different views as to how dual-gate mosfets should be used as VHF TV mixers.

But when it came to FM front ends, TI changed its position, as: “Dual-gate MOSFETs in an FM tuner provide many advantages over bipolar, junction field-effect, and single-gate MOS field-effect transistors. RF amplification and mixing are aided by use of a low-feedback-capacitance transistor with low noise figure and large dynamic range. A second gate is available for either AGC or local oscillator injection. High stable RF and conversion gains are easily obtained with inexpensive, commercially available coils and without need for neutralization.”

At least that opens up the possibility that TI had reasons for doing differently in the TV and FM cases. A possible clue here is found in the RCA FM front-end design, in which it is stated that gate 2 of the mixer mosfet must be grounded at the IF, 10.7 MHz, to avoid capacitive feedback that would reduce output impedance and gain. Given that the IF is about a tenth of the signal and LO frequencies, there is no major problem in grounding gate 2 at IF whilst retaining a high impedance to ground at the LO frequencies. If it were desirable to do the same in the TV case, it would appear not to be so easy. The IF is usually just below the lowest Band I channel frequencies, and for those, the LO is only an octave or so above. In that sense, one may speculate – wildly perhaps - that having the mosfet act as a cascode with gate 2 grounded at all frequencies, and thus with LO injection on gate 1, was an easy way out of the capacitive feedback issue. Possibly that is analogous to the valve case, where pentodes were favoured as VHF TV mixers (at least after IFs moved up into the 40 MHz vicinity) even though they were contra-indicated at Band III for noise reasons, because they avoided feedback problems on the lower Band I channels, where the quieter triode would have required neutralization. Maybe too that is why cascode bipolar mixers had been found in North American VHF TV tuner practice. Just in case that speculation has any merit, a corollary question is would neutralization of the mixer, at least for the lower Band I channels, have been desirable with the RCA approach in the VHF TV case.

Interesting is that the example VHF TV tuner circuit KRK-228 previously shown is an RCA design, so the RCA TV division was doing differently to RCA semiconductors.

A 1969 Zenith design for a so-called high-performance VHF TV tuner used a 3N140 as the mixer, with local oscillator injection on gate 1. The cascode form was evidently deliberately chosen. Note that the 3N140 was designated by RCA as an RF amplifier, but I suspect that there was not that much difference between the 3N140 and 3N141. The bipolar common base RF amplifier was evidently chosen because of concerns about mosfet breakdown; the 3N140 was not gate-protected.

More generally, RCA also proposed its dual-gate mosfets for use as chroma demodulators in TV receivers, but that function was quickly taken over by ICs.

And re the Leak Stereofetic front end, comparison with the previously-mentioned RCA design suggests at least the possibility that Leak modeled its design on the RCA circuit, which used a 40468 single-gate fet as RF amplifier. It would seem that RCA was moving stepwise with dual-gate mosfet applications, and at the time it had done the work for the FM mixer but not yet for RF amplfier, hence the use of athe previously developed single-gate mosfet circuit.

Cheers,

[Radio Wrangler]

I think the point that gate 2 is not cascoded with respect to the output, unlike gate 1, is interesting.

The problem becomes one of feedback via Cdg1 reducing gain. This doesn't happen if g2 is at RF ground, nor does it happen if the drain is at RF ground.

THe drain has to swing around at the IF frequency or it's dead in the water as a mixer. So, g2 needs to be low-Z at this frequency.

Similarly, the drain needs to be at low-Z at the frequency being inserted at g2, or else our mixer device starts trying to cancel its own LO drive.

And as a third condition, we need either-or-both the drain and g2 to be at low-Z at the frequency being inserted at g1 to prevent gain reduction of that path.

THese conditions can be met by having the drain able to swing in voltage only at the IF frequency. This is good because the IF doesn't have to move around. It still leaves the need pointed out above, to make g2 low-Z at the IF frequency to prevent the device's own reverse capacitance feeding the IF output back into g2 and reducing the mixing gain.

THese considerations apply to any cascode structure whether realised by a pair of discrete FETs, a pair of bipolars or an integrated cascode structure like the dual gate MOSFET.

I'm pretty certain that I saw RCA dual gate mosfets before any others.

Another route for research is Pat Hawker. Pat used to spend a lot of time in the libraries of central London, like the patent office, looking for new technologies to report in his "Technical Topics" column in Radcom/The Bulletin. He did this for 50 years. All of this is available on one CDROM. It makes addictive reading.

The RSGB did numerous surveys to monitor what interested members, and Tech Topics always came out as the first place most people turned to. Pat retired from writing it after 50 years, but they did not continue the column. They simply couldn't find anyone with the voracious appetite and the time to feed it.

Pat isn't around any longer, but he had a good innings, as they say, and he's left us a fascinating legacy.

I wonder how many other CDROMs get read from end-to-end?

David

[G0HZU_JMR]

I'm pretty sure the original TV tuner circuit you posted up used gate 1 injection simply because it made the transistion easier from 'mixer' in VHF tuning mode to 'IF amplifier' in UHF mode.

As I said before, if they had opted for gate 2 LO injection with that circuit then it gave them the design headache of removing the ac decoupling from gate 2 when the LO was needed on VHF mode. The ac decoupling at gate 2 would be needed when in UHF mode as the IF amplifier risks instability and poor transfer characteristics without it.

That circuit appears to have three modes. VHF Low mode, VHF High mode and also UHFmode.

By the way, if you look at the 3N141 'fig 2' circuit you posted up from the RCA technical document there is a glaring error in it. The circuit won't work as published because they forgot to remove the 1000pF decoupling capacitor from gate 2 where they appear to be trying to feed in an LO signal at VHF. An interesting mistake considering the topic of the conversation?

[G0HZU_JMR]

I had a quick trawl through a patent website looking at TV tuner patents from 30 years ago and you can see this block diagram being described as 'prior art'.

i.e. reconfiguring the VHF 'mixer' as the UHF IF amplifier when in UHF mode was quite common. The VHF LO gets turned off and the mixer becomes an IF amplifier in UHF mode. So it looks like I may have initially read the circuit operation correctly for the three modes.

I think the key issue with the DG mosfet (if you tried to feed the LO at gate 2) is trying to manage the impedance at gate 2 wrt the frequency plan because it would ideally need to be very low Z at the IF frequency in UHF mode but high Z at the LO frequency in VHF mode and this requirement would be difficult to arrange cheaply/reliably/stably for both modes, especially if you also consider how much the gate 1 impedance must change with tuning voltage. This change in Z would contribute a bit to overall stability concerns because all three ports of the mixer/amp would have tuned circuits on them and one of them would have variable tuning.

So injecting the LO at gate 1 and keeping the gate 2 pin well grounded is the simplest and safest option in my opinion. Some of the patents refer to RCA tuners (and documents) so you could probably dig deeper and find out more info about this circuit from the relevant RCA patent (if it exists).

FIG. 1 is a block diagram of a conventional tuner circuit. Referring first to the conventional tuner circuit in FIG. 1, this tuner circuit comprises a VHF RF amplifier 20, a VHF mixer (or IF amplifier) 21, a UHF RF amplifier 22 and a UHF mixer 23. The VHF input signal is amplified by the VHF RF amplifier 20 and is converted into an IF output signal by VHF mixer 21. The UHF input signal is amplified by the UHF RF amplifier 22 and is converted into an IF output signal by UHF mixer 23. In receiving UHF signals, the VHF mixer 21 is used as an IF amplifier so that the IF output signal is amplified by IF amplifier 21.

[Synchrodyne]

Originally Posted by G0HZU_JMR

So injecting the LO at gate 1 and keeping the gate 2 pin well grounded is the simplest and safest option in my opinion

Even without the VHF mixer being used as an additional IF amplifier for the UHF channels, that might also have been the case, at least for the lower Band I channels.

For the System M case, channel A2 vision was 55.25 MHz and vision IF was 45.75 MHz, with LO then 101.00 MHz. The mixer FET would be looking into the 1st LC circuit in the distributed IF selectivity chain (assuming the pre-IC time period before lumped LC and then SAWFs arrived), which would be of lowish Q, centred around 44 MHz, probably including some/all of the required traps and with an overall curve that probably had non-trivial sidelobes. So the impedance at 55.25 MHz (and across channel A2) was probably not that low, allowing the possibility of Miller feedback. Maybe a trap at 55.25 MHz in the first selectivity circuit would help, but not across the whole channel. Perhaps it would be possible to include some kind of filter in the LO feed to g2 that was high impedance to ground at 100 MHz and above, but was low impedance to ground at below say 70 MHz. But reverting to g1 LO injection and grounding g2, thus making the mosfet a cascode circuit, would be a simpler solution that was evidently still satisfactory from a conversion gain viewpoint. LO feedback through the RF amplifier to the aerial input might then be more of a problem, but a cascode (or grounded base or grounded gate) RF amplifier might provide an adequate barrier.

In contrast, for the FM case, the impedance seen by the drain would be very low at signal and LO frequencies, and grounding g2 at IF was not a problem. So this allowed g2 LO injection, reducing LO feedback to the RF stage, and, as in the RCA example, allowing the use of a single-gate, grounded source RF amplifier.

At least that seems to be a plausible explanation for the use of g1 LO injection for VHF TV but g2 injection for FM.

There is a TI application note, CA-136 of October 1969, “MOSFETs in Color Television High-Frequency Section”, that might help if it could be found. I’d say there is some chance that this would discuss the trades-off involved in g1 vs. g2 LO injection.

The MF and HF examples I can find use g2 LO injection. Taking the single-conversion Eddystone 1000 series as an example, one assumes that the impedance seen by the drain is high at the 455 kHz IF and the associated passband but declines rapidly above and below, so is low enough at signal and LO frequencies that feedback is not a significant problem. As an aside, but still on topic, the FM signal path for the Eddystone 1002 FM-AM variant looks rather odd; a bipolar front end (possibly a Mullard module) feeding into an IF chain that has two LC-coupled dual-gate mosfet stages ahead of a ceramic filter and a CA3089 IC. IF gain ahead of the CA3089 was not unusual, and one can understand the need for care in respect of linearity when the gain stages were ahead the main selectivity block. But one suspects that the front end would have been the limiting feature when it came to signal handling, such that bipolar devices would have been satisfactory in the early IF stages. RCA itself shows bipolar pre-stage in its CA3089 (and CA3189) data.

Cheers,

[G0HZU_JMR]

Perhaps it would be possible to include some kind of filter in the LO feed to g2 that was high impedance to ground at 100 MHz and above, but was low impedance to ground at below say 70 MHz.

Yes, it's certainly possible but I'd be very nervous about having tuned circuits on three of the ports of that device because of the risk of instability

Especially when the Gate 1 preselection is tuned up and down wrt frequency.

But then I don't really have much experience doing serious design work with dual gate mosfets. The last ones I used were for wideband high reverse isolation buffers and that was many years ago.

One thing that isn't so good about gate 1 injection is that the combined peak voltage of the signals plus the LO drive level must never be allowed to cause pinchoff or saturation but I guess this probably won't happen with a typical TV tuner anyway and even if it did the person would be told to fit an inline attenuator in the aerial feeder to combat any overloading symptoms.

[Synchrodyne]

If the Zenith VHF TV Tuner example is anything to go by, the RF stage was designed to overload before the mixer, thus protecting the latter from receiving too much signal.

Zenith’s RF amplifier desiderata included:

• Large signal handling capability; Over l00mV of Vund (fund: ±2 channels away from fDES) for 1% cross-modulation. At least 400mV of desired signal voltage without overloading.
• AGC Range; over 50dB of maximum gain reduc¬tion is desirable to protect the mixer stage from overloading.

And for the mixer:

• Sufficiently high conversion gain; over 12 dB.
• Fairly constant conversion gain and output band¬width over the range of oscillator injection voltage variation and device variation.
• Large signal handling capability.
• Low cross-modulation; i.e. low or no cross-modulation contribution by the mixer stage for the overall cross-modulation performance of the tuner.
• Low intermodulation and spurious responses.

I have since found the previously mentioned TI paper on mosfet FM Tuner design. Contrary to what I expected from the previously quoted abstract, in fact it considers both the gate 1 and gate 2 local oscillator injection cases for the mixer. To quote from the concluding remarks: “The dual-gate MOSFET is well-suited to FM tuner service as an RF amplifier and as a mixer. As an amplifier it provides high gain, low noise, and a large agc range without overload. As a mixer the dual-gate MOSFET has large conversion gain and high spurious rejection. Gate 2 L.O. injection offers 10 dB better spurious rejection but requires three times the signal. Gate 2 local-oscillator injection does have an average of 6 dB less gain than Gate 1, but the gain loss necessary to eliminate all signs of RF-L.O. interaction could easily make up the difference.”

So it would seem that the choice of gate 1 or gate 2 LO injection is to some extent situational and would rest with where the equipment designer wanted to make the trades-off. For the FM case, that the LO differs from the RF by only around 10%, but that the IF is only around 10% of the RF, might favour gate 2 injection. The TI example had a 10.7 MHz filter at gate 1 to trap out any IF that found its way back there.

In view of what has found in the TI FM paper, conceivably the TI Application Report CA-136 “MOSFETs in Color Television High-Frequency Section” of October 1969 would include some discussion about gate 1 vs. gate 2 LO injection. But so far I have not been able to find this document.

Cheers,

[G0HZU_JMR]

I've always used gate 2 injection but this has only ever been for hobby use rather than at work. The reason I would normally shy away from gate 1 LO injection is because of the Vpk limitations mentioned in my previous post.

To get the mixer to work well here requires the LO signal to be big enough to really exploit the square law characteristic of the device but this eats into the peak voltages that can be coped with before it hits saturation or pinchoff (and the square law response is lost)

So there's a tradeoff in performance here and I would usually prefer to use gate 2 injection. But I wouldn't want to use gate 2 if I also had to use the stage as an amplifier with the LO removed, especially with that frequency plan. Also, gate 2 LO drive normally requires a different bias arrangement to get full benefi from the ac swing and this again isn't ideal for operating as an amplifier in UHF mode.

At a guess, the biasing for 'amplifier' and 'gate1 LO' mode can be very similar because the mixing in this configuration comes from the inherent square law response of the device and this allows 'regular' biasing to support both modes. So to me it's clear why gate 1 biasing was chosen here (but I am beginning to feel a bit like a stuck record in giving my reasons for this and I can't say for sure if my reasons are accurate )

I would also say that there isn't going to be a defining answer anywhere as to why designers opt for either option. Even if one designer produced a technical note to say why they chose either option for their tuner design it doesn't mean their choice was a panacea or even a good choice.

[Synchrodyne]

Yes, I think that the evidence unearthed to date indicates that choice of gate 1 or gate 2 for local oscillator injection is situational and subject to individual designer preference.

Thinking about it some more, in US TV tuner case, it was the norm to route the output of the UHF tuner through the whole VHF tuner, both RF amplifier and mixer, this going back to the valve era. US UHF tuners were typically quite simply, with a bandpass input feeding a diode mixer fed by a valve or bipolar transistor local oscillator. It was desirable to follow this with a stage of low noise amplification, which in valve days was the cascode RF amplifier. Looking again at the data, the Zenith VHF TV tuner was intended to be used in this way for UHF reception, the “Channel 1” position on the turret accommodating this. So Zenith’s choice of gate 1 LO injection may have been informed in part by this need, and the same likely would have applied to other setmakers. Possibly there is some discussion of this aspect in the TI Application Note.

The RCA case represented a change, in that the KRK 226 UHF tuner included both an RF amplifier (fet) and an IF preamplifier (bipolar), so that only the mixer stage of the KRK 228 VHF tuner was used as an IF amplifier on UHF.

From bits and pieces of information found on the web, it would appear that the RCA CTC 74 and CTC 81 TV receivers in which the KRK 226 and KRK 228 tuners were used date from 1975. Also, the KRK 226 was said to use a mosfet rf amplifier (i.e. not a GaAsfet). So it would appear that mosfets suitable (both technically and economically) for consumer UHF applications were available from at least 1975. Interestingly, none of the mosfets listed in the RCA 1978 manual is claimed for use above 500 MHz, so perhaps the RCA TV receiver division was using a third party mosfet.

The RCA CTC 74 and CTC 81 TV receivers also had three-stage dual-gate mosfet IF strips. So on UHF there were both bipolar and mosfet pre-IF stages, the former in the UHF tuner and the latter being the VHF mixer. It does seem rather odd though, to have a bipolar stage ahead of four mosfet stages in a distributed selectivity IF system where selectivity was probably quite minimal until just ahead of the 1st stage of the IF strip proper. On the other hand, Zenith chose to use a bipolar RF amplifier ahead of a mosfet mixer in its 1969 VHF tuner. And in the SAW filter age, TV IF pre-amplifiers/SAW drivers were usually bipolar, whether discrete or of the IC type.

Along a somewhat different vector, the Eddystone EB35 Mk III shortwave receiver of the mid-1970s had a dual-gate mosfet mixer with gate 2 LO injection and with agc (and signal) applied to gate 1, which I suspect was not a common arrangement. And the agc was of the wideband type, obtained from a jfet amplifier followed by a rectifier fed from the mixer output. Unlike its predecessors, the EB35 Mk III lacked an RF amplifier and had a two-gang not a three-gang front end, so presumably the mixer was vulnerable to overload from strong out-of-band signals. The mosfet in this case was the 40841, which in its 1978 manual RCA billed as an “economy”, DC-to-500 MHz device. Also, it appears to be the only dual-gate device in the manual for which AM receivers were included in the list of typical applications.

Cheers,

[Synchrodyne]

I have attached the schematic (unknown make) for another mosfet-based VHF TV tuner, in this case of the turret type and with the UHF IF routed through the whole tuner. Again, g1 LO injection is used for the mixer stage.

And re the RCA 40841 as being the designated device for MF/HF applications, in the 1978 manual it showed a 40841 MF RF amplifier used ahead of the CA3088 AM radio IC. On the other hand, Tandberg used a 40822 as MF RF amplifier in its initial TR2075 tuner-amplifier. Maybe in this case it had also used a 40822 as an FM RF amplifier, and wanted to rationalize components.

There was a 196907 Wireless World article on an amateur HF SSB single-conversion receiver in which dual-gate mosfets were used for the RF and mixer stages. It was noted that several devices were tested and that at frequencies up to 30 MHz, the differences between them were small. It was also observed that these were essentially VHF devices, which at HF had the advantage of having constant output impedance over the range 1 to 30 MHz.

Thus it would appear that even in the early days of dual-gate mosfets, they were seen as being essentially VHF devices. One may wonder then whether the use of mosfets at MF/HF was akin to the valve era use of “VHF” valves, such as the EF85, at MF/HF - as discussed in the concurrent thread: https://www.vintage-radio.net/forum/...ad.php?t=99046 - and if so, whether similar care was necessary to obtain adequate stability.

Also, were there bipolar solutions that could match mosfet performance in MF/HF applications without the disadvantages of using essentially VHF devices at lower frequencies? The early RCA work positioned dual-gate mosfets as being better than bipolar transistors or jfets in terms of cross-modulation, etc. But would small-scale bipolar ICs, such as the CA3028 type, have done the job equally as well.

Looking more at valve analogies, a dual-gate mosfet RF or IF amplifier with agc on g2 looks somewhat like a dual-control pentode with agc applied to g3. But with the pentode, g3 does not have any material effect on cathode current; rather it apportions the current between g2 and the anode. On the other hand in the mosfet case, both g1 and g2 control source current. So the underlying mechanisms are different. One could also make the comparison with a cascode double triode. But here I am not sure that agc could be applied to the grid of the second triode without upsetting operation of the input triode, and even if it could, adding a standing positive DC voltage to the AGC voltage might not be without difficulties. (Maybe an anode bend AGC rectifier with the no-signal anode voltage at the desired standing bias.)

A dual-gate mosfet mixer with g2 LO injection looks like a dual-control pentode with g3 LO injection, although as with the RF/IF amplifier, the underlying mechanisms are be different. When the signal and LO are both on g1, the mosfet looks analogous to a pentode mixer with signal and LO on g1, or more closely to a cascode double triode with signal and LO on gt1 – although I don’t know that such was actually used in practice.

Cheers,

[turretslug]

I have found this to be a most absorbing and informative discussion, it's been fascinating to see the depth of understanding and hands-on experience here. Thanks definitely go to Synchrodyne for initiating the thread and continuing to post deeply researched and evidently time-consuming contributions and to Radio Wrangler, G0HZU JMR, G8HQP Dave and others for their lengthy and informed offerings.

I suppose that we all draw our own inferences from what we read and, for me, it has offered further confirmation as to why valved HF communication receivers continue to have credibility (and not just as objets d'art) some 80 years after they appeared and half a century after they ceased to be made in significant quantity. Yes, the RA1772 showed the way to go with FETs, but it's still a large, heavy radio that fetches a premium- and to many hobbyists, it seems to be fearsomely complex compared with its predecessors. If you understand and can fix a domestic valve table radio, that'll set you in good stead for approaching an HRO or an AR88, the same doesn't apply between a six-transistor portable and an RA1772! Interesting to see that "Circuits using EF85" is trending towards the subject, also...

One avenue that hasn't been mentioned is FETs in car radios- to me, there would seem to have been a window of oppurtunity starting from the late 'sixties, perhaps, for such devices for L/M/HF use where good sensitivity would be a boon and strong signal handling less critical- i.e. when using a metre or so of whip aerial. Car radios have long struck me as relatively unassuming exponents of technology that was "useful", as opposed to "showey" and can incorporate interesting techniques. In my workshop, aural wallpaper comes courtesy of a mono Clarion radio (RN-9089) whose AM RF stage is a cascode JFET/bipolar arrangement, giving hi-Z input for the short aerial and varactor-tuned output into the standard consumer AM processing IC. VHF/FM input is to dual-gate MOSFET- no surprises here and seen in other car radios of the period. Was AM car radio use a common place for FETs, or just in the odd application like this? I suppose bipolar AM radio circuitry was well understood and developed in other fields and consumer manufacturers are rarely motivated beyond what is "sufficient".

[emeritus]

I was using the RCA 3N140 and 3N141 devices in the early 1970's, the 3N141 as a mixer, groups of four 3N140s arranged in a bridge configuration as change-over switches for selective electronic switching of different bandwidth crystal filters in the IF strip (23MHz) of a military radio were were developing. I probably still have a couple lurking in my odds and ends box.

As I recall the data sheets gave dire warnings about the potential for damage by static elelctricity, and the devices shipped with coiled springs shorting the leads together, but we usually didn't bother using the springs when reworking and reusing them, and never had any failures. Perhaps static wasn't so much of a problem in the damp British climate!

My recollection is that, at that time, RCA were the only source of discrete dual gate FETS that were useable at that sort of frequency.

[Synchrodyne]

Originally Posted by turretslug

One avenue that hasn't been mentioned is FETs in car radios- to me, there would seem to have been a window of oppurtunity starting from the late 'sixties, perhaps, for such devices for L/M/HF use where good sensitivity would be a boon and strong signal handling less critical- i.e. when using a metre or so of whip aerial. Car radios have long struck me as relatively unassuming exponents of technology that was "useful", as opposed to "showey" and can incorporate interesting techniques. In my workshop, aural wallpaper comes courtesy of a mono Clarion radio (RN-9089) whose AM RF stage is a cascode JFET/bipolar arrangement, giving hi-Z input for the short aerial and varactor-tuned output into the standard consumer AM processing IC. VHF/FM input is to dual-gate MOSFET- no surprises here and seen in other car radios of the period. Was AM car radio use a common place for FETs, or just in the odd application like this? I suppose bipolar AM radio circuitry was well understood and developed in other fields and consumer manufacturers are rarely motivated beyond what is "sufficient".

There is some evidence that fets were used in the AM circuits of car radios. The early RCA MF RF amplifier using a dual-gate mosfet, mentioned in post #28, was intended to replace a bipolar transistor in a car radio. Also, the CA3088 plus dual-gate mosfet example was probably oriented towards car radio applications. I have attached a snip from a schematic for a mid-1980s Delo AM Stereo/FM Stereo car radio that shows a jfet/bipolar cascode used as a broadband AM RF amplifier. This idea seems to have either come from or have been promoted by the AM radio IC makers, as it shows up in the data for both the National LM1863 and the Sanyo LA1235. These ICs supplied wideband agc for the external RF amplifier.

The AM front end for the Carver TX-11a hi-fi tuner used a derivative of the Sanyo circuit. Having used one of these when I was in the USA, I found that Its behaviour was a bit odd, in that with a largish external aerial and so one assumes a large delivered signal, the RF agc appeared to depress gain to the point where the signal-to-noise ratio suffered. I suspect that the untuned mixed cascode RF stage was really intended to work with the lower signal levels found in car radio applications. Looking at comparable wideband hi-fi AM stereo tuners – which I have not used - the later Denon TU-680NAB had a tuned RF input feeding directly into a LA1247 IC, with no external amplification, whilst the earlier Sansui TU-D99AMX had a tuned input feeding a jfet amplifier, thence into a Sanyo LA1247.

Cheers.