UK Vintage Radio Repair and Restoration Powered By Google Custom Search Vintage Radio and TV Service Data

Go Back   UK Vintage Radio Repair and Restoration Discussion Forum > General Vintage Technology > Components and Circuits

Notices

Components and Circuits For discussions about component types, alternatives and availability, circuit configurations and modifications etc. Discussions here should be of a general nature and not about specific sets.

Closed Thread
 
Thread Tools
Old 14th Sep 2013, 6:39 am   #41
Radio Wrangler
Moderator
 
Radio Wrangler's Avatar
 
Join Date: Mar 2012
Location: Fife, Scotland, UK.
Posts: 22,800
Default Re: FET Questions

Whenever the JFET-JFET cascode was trotted-out in books there was always the advice that the upper device needed to have greater Idss than the lower device, or sometimes they said greater gm.

This advice was often repeated, though I never saw it accompanied by an explanation better than a bit of hand-waving.

Using a bipolar in the upper position handles both of these issues, but in car radios, their lower price might also be an issue. THe problem with using FETs is their large unit or batch variations in Idss, Vp and in fact most parameters. Having to stack them in order of any parameter would force a selection process on the set makers.

David
__________________
Can't afford the volcanic island yet, but the plans for my monorail and the goons' uniforms are done
Radio Wrangler is online now  
Old 14th Sep 2013, 11:45 am   #42
kalee20
Dekatron
 
Join Date: Feb 2007
Location: Lynton, N. Devon, UK.
Posts: 7,059
Default Re: FET Questions

I can't see why the upper device should need more gm, Idss etc. Obviously they'll both be passing the same current, but suitable bias on the upper device should sort that.

If the lower device has Idss of 10mA but is biased so that it's operating at 2mA, I can't see any problem in an upper device having Idss of only 5mA. It won't be operated at zero bias, so no problem!

Quote:
Originally Posted by Radio Wrangler View Post
THe problem with using FETs is their large unit or batch variations in Idss, Vp and in fact most parameters.
Yes - the variation makes me thoroughly grateful that bipolar devices exist, where the main variation is hfe (and variation in that can be catered for by design practices).
kalee20 is online now  
Old 14th Sep 2013, 1:39 pm   #43
Radio Wrangler
Moderator
 
Radio Wrangler's Avatar
 
Join Date: Mar 2012
Location: Fife, Scotland, UK.
Posts: 22,800
Default Re: FET Questions

Yeah, it mystifies me too. But it was repeated so many times it must be true

It might have been something important for one particular circuit that got believed-in out of context.

David
__________________
Can't afford the volcanic island yet, but the plans for my monorail and the goons' uniforms are done
Radio Wrangler is online now  
Old 14th Sep 2013, 11:35 pm   #44
Synchrodyne
Nonode
 
Join Date: Jan 2009
Location: Papamoa Beach, Bay of Plenty, New Zealand
Posts: 2,943
Default Re: FET Questions

Some elucidation might be provided by the following commentary from a Texas Instruments (TI) item that inter alia addressed the use of the cascode jfet configuration for VHF TV tuner RF amplifiers as a way of solving the cross-modulation problem (with bipolar devices) and matching valve (6HA5) performance. A snip of the circuit is attached.


“Using a sample of 10 transistors with Idss vvariations from 6 to 14 mA, we reached the following conclusions:

1) For best operation the Idss of Q1, (the common-source transistor) should be greater than the Idss of Q2 (the common-gate transistor), but not more than 1.5 times as great. Under these limitations the circuit gain was 15 to 18 dB and n.f. was 5.2 to 5.9 dB for all combinations.

2) Degraded but acceptable operation was found for all other combinations with but one exception. With Idss of 14 for Q1 and Idss of 6 for Q2, an AGC voltage of 8 volts drove Q2 well into the enhancement mode, resulting in excess heating of the source-gate diode, and dropping the gain to less than 11 dB from a cold start of 16 dB. With Idss of 13 for Q1 and 7 for Q2 the effect was small but observable. Eliminating the transistors with Idss less than 7 mA gave 13 dB minimum gain and 6.0 dB maximum n.f. for all combinations. The C.M. was not significantly degraded, but dynamic AGC range was for the extreme Idss ratios, degraded to about 31 dB.

3) Although limit samples were not available at the time of this evaluation, it is believed that acceptable performance of all parameters can be obtained by categorizing the SFB2484 into three Idss ranges of 5 to 8, 8 to 12, and 12 to 16 mA.”



The foregoing was empirical and probably situation-specific. One wonders though whether the first sentence in item (1) was taken out of context and generalized.

TI moved on to advocating dual-gate mosfets for TV VHF tuners fairly soon after considering jfets. I do not know whether there was much use of jfets in TV VHF tuners, but they were used in some FM tuner front ends. In respect of the previously mentioned Zenith VHF TV tuner, the use of cascoded TIS88 jfets for the RF amplifier was considered but rejected in part because of a restricted agc range (which was also noted by TI). The grounded base BF200 was chosen over the other leading candidate, the 3N140 dual-gate mosfet, apparently because of concerns about gate insulation breakdown. (It would seem that the development work was done on the cusp of the availability of protected gate devices.)


Cheers,
Attached Thumbnails
Click image for larger version

Name:	TI Cascode jfet.JPG
Views:	979
Size:	17.2 KB
ID:	84065  
Synchrodyne is offline  
Old 14th Sep 2013, 11:54 pm   #45
turretslug
Dekatron
 
turretslug's Avatar
 
Join Date: Nov 2011
Location: Surrey, UK.
Posts: 4,385
Default Re: FET Questions

Quote:
Originally Posted by Synchrodyne View Post
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.
Apologies for forgetting that! Having re-read the thread, I realised that I had read it first time but had already forgotten it by the then thread end . Too little assimilation.

My take on the use of the circuit is simply that by isolating the first RF tuned circuit from different size aerials and feeder lengths, it allows the traditional aerial trimmer to be dispensed with. In the days of mechanically tuned radios with continuous and/or push-button tuning, setting the trimmer for maximum signal strength at the HF end of the MW band was a familiar ritual but by the time of electronic tuning with PLL-controlled varactors, it would have been seen as a retrograde rough edge to slick digital operation. Also, one FET, one BJT and a few passives auto-loaded on the main board might have been cheaper to an OEM than a seperate trimmer on the front panel...

The FET/BJT combo makes a high impedance "sniffer with gain"- it was got away with using a short whip aerial but I would expect it to fare poorly with high signal levels. It's often parroted that "FETs are like valves, aren't they?" but this FET/BJT strikes me as acting like a pentode, combining the input character of the FET (high input impedance, good isolation) with the output character of the BJT (high gain, high output impedance). I feel that the frequent comparison of a single transistor vs. a valve is a little unfair- if a valve is allowed to have several grids, why can't its semiconductor analogue have several junctions, i.e. a few three-terminal devices improving the overall characteristics and applicability.
turretslug is offline  
Old 15th Sep 2013, 11:34 am   #46
Radio Wrangler
Moderator
 
Radio Wrangler's Avatar
 
Join Date: Mar 2012
Location: Fife, Scotland, UK.
Posts: 22,800
Default Re: FET Questions

The dual-gate MOSFET is essentially an integrated cascode, with the benefit that the two parts undergo the same processing and so variations in area and length can set any chosen relationship between the two devices, though all the ones I've come across, I believe, have similar geometries.

The MOSFETS I've recently been playing with run at 100v peak, 45A peak, give 20dB gain and put out around 300W at 1 GHz. Not much use for front-ends unless you really want dynamic range!

David
__________________
Can't afford the volcanic island yet, but the plans for my monorail and the goons' uniforms are done
Radio Wrangler is online now  
Old 15th Sep 2013, 6:54 pm   #47
turretslug
Dekatron
 
turretslug's Avatar
 
Join Date: Nov 2011
Location: Surrey, UK.
Posts: 4,385
Default Re: FET Questions

Synchrodyne has highlighted that FETs and ICs became significant at around the same late 'sixties timeframe- my supposition is that even small-scale integration devices with just a few equivalent BJTs- say the likes of the MC1330 and MC1496- were so appealing to designers that the FET (which looked promising against the individual BJT) was a bit left on the sidelines. In not so many years, transistors had gone from strange and exotic things that cost several guineas each (so that designers had to be sparing with them and wring the most out of each one) to availability in ICs cheaply enough that elegant and effective circuits could be concocted with little worry about equivalent BJT count. Just a handful of transistors in an IC gave high input impedance, good port isolation, high gain good linearity and thermal tracking, and favourable output characteristics- much of what the FET had promised plus more. That's not to say that the FET didn't find its own, but relatively niche, place as earlier posts discuss.
turretslug is offline  
Old 15th Sep 2013, 7:00 pm   #48
turretslug
Dekatron
 
turretslug's Avatar
 
Join Date: Nov 2011
Location: Surrey, UK.
Posts: 4,385
Default Re: FET Questions

Quote:
Originally Posted by Radio Wrangler View Post
The MOSFETS I've recently been playing with run at 100v peak, 45A peak, give 20dB gain and put out around 300W at 1 GHz. Not much use for front-ends unless you really want dynamic range!
That's pretty impressive stuff, that's what I think of as Eimac 4CX type territory!
turretslug is offline  
Old 21st Sep 2013, 3:30 am   #49
Synchrodyne
Nonode
 
Join Date: Jan 2009
Location: Papamoa Beach, Bay of Plenty, New Zealand
Posts: 2,943
Default Re: FET Questions

Turretslug, I think that you’re right about the relative use of FETs and ICs, at least in consumer equipment. Essentially, FETs were used mostly for functions where they were superior to bipolar devices, discrete or integrated, but not often otherwise.

As previously suggested, hi-fi FM tuners are representative here. FETs, once available, quickly became the norm for the front end, RF and mixer stages, where they enabled performance and signal handling to match that of valves, whereas bipolar front ends had generally been a noticeable step downwards. But they were rarely seen downstream of the mixer. The IF strip quickly became IC territory, soon including the demodulator, reaching something of a plateau with the “does everything” CA3089 in 1971. Then the decoder was integrated, initially in conventional form, e.g. MC1304/5, but by 1971 in PLL form, such as the MC1310 and CA3090. A few years later, Hitachi and Toko were offering improved forms of the CA3089/3189 and MC1310.

What also helped here was that economical “block” selectivity became available at about the same time as ICs (and FETs) in the form of ceramic filters. So that minimized the exposure of bipolar stages to wideband signals, although I do not suppose it was that difficult to design bipolar first IF stages that could handle the highest signal levels that the mixer could deliver before it got itself into serious trouble.

One may look at a few FM tuner examples that I think illustrate the foregoing.

The Rogers Ravensbourne, from early 1968, used dual-gate mosfets in the RF and mixer positions. Also, it was 4-gang, which would have been rare in the valve era, although unusually in FM practice, the bandpass circuit was in front of not following the RF amplifier, which was more reminiscent of HF/MF practice. My understanding is that a single tuned input with a bandpass interstage became the norm for 4-gang FM front ends, possibly because there is a noise/selectivity trade-off at the input. The Ravensbourne IF strip was of the LC distributed selectivity type with four simple ICs (I am not sure, but they might have been µA703), thence a discrete ratio demodulator and a discrete decoder.

The Rogers Ravensbrook in 1969 was aimed at a lower price point, and so had a 3-gang front end with a mosfet RF amplifier but a bipolar mixer. The inference is that if you could afford only one mosfet, then the RF amplifier was its best deployment. The IF strip was “more integrated” than that of the Ravensbourne, although still with LC selectivity. It used CA3012 and CA2014 ICs, the latter inclusive of the ratio demodulator diodes. The decoder was discrete.

Also in 1969 debuted the Leak Stereofetic, whose name nicely captures the theme here. It had a 3-gang front end with mosfet RF amplifier with dual-gate mosfet mixer. The IF strip was built around a pair of CA3011 ICs, with a pair of ceramic filters, and a discrete ratio demodulator. The decoder was mostly discrete, but used (both halves of) a CA3026 IC as the multiplier, essentially doing the same job as would an MC1496.

An interesting sidebar here is that IC-based FM quadrature demodulators were by then in use for TV sound IF strips, but there seemed to be a slower uptake where FM tuners were concerned.

The 1st iteration of the Quad FM3 in 1971 had a 3-gang front end with 2 dual-gate mosfets. The IF strip had a ceramic filter and 2 ICs, an L103T and a TAA661B, the latter including quadrature demodulator. The decoder used an MC1305P (non-PLL) IC. In the 2nd iteration, the MC1310P PLL decoder IC replaced the MC1305P. And in the 3rd iteration, a CA3089E replaced the two previously used IF ICs, at which stage it was on the “plateau” which might be described as a core circuit based upon dual-gate mosfets in the rf and mixer positions, ceramic IF filters, a CA3089 or derivative IF and demodulator IC, and an MC1310 or derivative decoder IC, plus individual manufacturer refinements. I guess that one may include the Toko LC block pilot tone/subcarrier post-decoder filters in this core circuitry.

I think that there are similar narratives about the division of FET and IC functions for the AM and TV cases. In the HF case, the Plessey SL600 series ICs probably stand out as an example of covering all of the functions downstream of the main selectivity. But the Japanese manufacturers of serious consumer/amateur equipment seemed to prefer discrete circuitry all the way through to the demodulator through into the 1980s. Thus from the FRG-7 (and probably before it) and through to the NRD-525 (and probably beyond it), one finds the RF and IF chains populated by FETs, interrupted by the occasional DBM.

An interesting transitional-era HF receiver example was Eddystone EC958 of 1968(?). In its original form, it predated the step-change/revolution marked by the RA1772 and in some ways followed valve receiver precedents, but using FETs instead. The 100 kHz main IF amplifier was a cascade of FETs, although it did have block rather than distributed selectivity. The EC958/12 ISB variant came a couple or so years later, and here, whilst it retained the basic core including the cascade-FET main IF amplifier, the three 100 kHz IF amplifiers in the ISB module used ICs rather than FETs. That suggests that in the early FET and IC days, the available ICs were perhaps not seen as being good enough for professional HF applications, but within a few years, that had changed. So ICs, as well as displacing discrete bipolar and valve-based circuitry, also displace some early fet applications as they developed.

In AF signal-level circuitry, FETs seem to have been confined mostly to situations where high input impedance was required, or electronically controlled variable gain, as in noise reduction systems and some forms of quadraphonic matrix decoding. And for the high impedance case, FET-input ICs seemed to become the preferred pathway. The TI TL071 and family from the later 1970s comes to mind as the obvious example, but very early in the consumer product FET and IC era came the Mullard TAA320, which I think was used in some of the late-1960s Tripletone amplifiers as a ceramic cartridge input buffer, in place of the previously used bootstrapped emitter follower.

Cheers,
Synchrodyne is offline  
Old 21st Sep 2013, 11:48 am   #50
kalee20
Dekatron
 
Join Date: Feb 2007
Location: Lynton, N. Devon, UK.
Posts: 7,059
Default Re: FET Questions

Thanks Synchrodyne for a lucid history lesson!

Personally I don't like MOSFETS, though I tolerate them if the input is sufficiently remote from the outside world (such as power MOSFETs with on-board gate drive circuitry). Junction FETs, to me are much more tolerable, though they do have a rather wide spread of characteristics. Though, I don't pretend to be an experienced RF man!
kalee20 is online now  
Old 21st Sep 2013, 7:33 pm   #51
G6Tanuki
Dekatron
 
G6Tanuki's Avatar
 
Join Date: Apr 2012
Location: Wiltshire, UK.
Posts: 13,951
Default Re: FET Questions

MOSFETs (at least the early 3Nxxx ones from the likes of RCA) were fragile and when used in VHF receiver front-ends often had a short and unhappy life if subjected to hostile levels of external RF. This was not so much of a problem here in the UK where few PMR radios ran more than around 20 Watts, but in the US a typical "Taxi/Police mobile-radio" was 100 watts.

When you're parked a few feet away from someone else on a frequency only a few tens of KHz away from your receiver, and they hit transmit, you learn a lot about receiver front-end robustness. It was also fun to try and design receivers for the old UK "P-band" 105-108MHz VHF where you were living only a few MHz down from a 50Kw FM transmitter radiating Radios 1/2/3/4.

When I was designing communications-grade VHF front-ends in the early-1980s I still put the incoming RF through a 4-stage cavity resonator before it reached a cascode pair of power-JFETs (J310) passing a few tens of Milliamps, into a hot-carrier-diode balanced-quad as the first mixer. The J310s had a *lot* of degenerative feedback - their prime purpose was to provide power (rather than voltage) gain and to isolate the LO from the atenna in order to comply with some rather serious constraints on radiation of the LO drive signal (which was provided at a really rather high level to the mixer: my standard LO output design involved a bipolar 2N3866 which needed a good heatsink).

Even to this day I reckon the two essential things for any low-crossmodulation highly-linear VHF receiver front-end are a high supply-voltage (24V or more) and plenty of current through the active, pre-filter devices.
G6Tanuki is offline  
Old 12th Oct 2013, 4:37 am   #52
Synchrodyne
Nonode
 
Join Date: Jan 2009
Location: Papamoa Beach, Bay of Plenty, New Zealand
Posts: 2,943
Default Re: FET Questions

Perhaps the fragility of the early mosfets slowed down their initial adoption in some equipment, at least until gate-protected devices became available. Once that was the case, then for domestic hi-fi FM tuners, dual-gate mosfets seem to have won the day, generally displacing earlier jfet-based designs, although there were some makers who stayed with jfets for a while, such as B&O. I think though that suitable jfets may have become available a year or so ahead of mosfets, so they had a head start.

Attached is a late 1960s FM front end circuit (Heathkit, I think) based upon jfets, with a cascode RF amplifier. And another rather curious example from the same period, make unknown, which has a single-gate mosfet + jfet cascode, with the mosfet is at the input end. Here one supposes that the combination was chosen for its performance (or performance vs cost) characteristics, and not because of jfet robustness. In both cases AC-coupled shunt cascode circuits are used, not the series-cascode suggested by TI. And also in both cases, agc is applied to the input (common source) device, whereas the TI circuit had it applied to the second (common gate) device.

On the other hand the Eddystone EC958, 1830 and 1000 series HF receivers used a jfet + mosfet series cascode (with agc applied to the mosfet gate) because of its ability to handle large signals.

Given that a dual-gate mosfet is effectively a pair of cascode-connected single-gate mosfets on the same chip, would it be possible to have a dual-gate cascode jfet as a single device?

Cheers,
Attached Thumbnails
Click image for larger version

Name:	FM jfet.jpg
Views:	1594
Size:	41.0 KB
ID:	84974   Click image for larger version

Name:	FM mosfet + jfet.jpg
Views:	3021
Size:	64.1 KB
ID:	84975   Click image for larger version

Name:	TI Cascode jfet.JPG
Views:	642
Size:	17.2 KB
ID:	84976   Click image for larger version

Name:	Eddystone 1000 Front End.jpg
Views:	1199
Size:	35.9 KB
ID:	84977  
Synchrodyne is offline  
Old 19th Oct 2013, 6:06 am   #53
Synchrodyne
Nonode
 
Join Date: Jan 2009
Location: Papamoa Beach, Bay of Plenty, New Zealand
Posts: 2,943
Default Re: FET Questions

Expanding upon my previous thought as to whether a dual-gate jfet would be possible, and perhaps reaching out over the edge of reality, then what about a triple-gate mosfet? What prompted this was re-reading Radio Wrangler’s observation that for a dual-gate mosfet used as a mixer with oscillator injection on gate 2, that gate was not cascoded with respect to the output. So adding a third gate at AC (including RF and IF) ground would appear to correct that anomaly. And I suppose it could be done with a single-gate mosfet connected in series with the dual-gate device.

Dual-gate mosfets have been variously compared with valves of the tetrode, dual-control pentode and dual-control heptode types. Thinking of the last-mentioned, it does not escape notice that there was what was in effect a triple-control valve, namely the EQ80 nonode. Very roughly, this could be thought of as something like an EH90 heptode with the addition of another control grid and another screen grid section. The objective, for its FM quadrature demodulation function, was to have two grids with suppressor grid or heptode g3 characteristics. In normal application g1 was not used for signal or reference injection, but was DC-biased to set the desired cathode current, so it was still a control grid. So the triple-control idea does have a valve-era precedent.

That said, I am not sure as to how far the comparison between valves and mosfets can be taken. In a dual-control valve, be it a pentode or heptode, the signal/bias on g3 does not materially affect cathode current, but simply apportions it between anode and screen. On the other hand, in a dual-gate mosfet, the signal/bias on g2 does exercise control on source current, so the underlying mechanisms look to be somewhat different.

Returning to FET applications, I think that the where, how and why relating to the use of small-signal fets, particularly mosfets, in FM tuners and TV RF and IF circuits is fairly clear.

In HF receivers, there was some early use of small-signal fets in professional models whose circuitry largely followed valve techniques, with for example significant RF selectivity ahead of the 1st mixer, the seeming objective to match or get closer to valve performance than could be obtained with conventional bipolar devices in similar circuits. But the advent of the Racal RA1772, with its “bomb-proof” 1st mixer, signalled a paradigm shift. It seems that in this kind of receiver, RF amplifiers, where used, were likely to be relatively high current bipolar devices, sometimes in feedback circuits.

On the other hand, small-signal fets retained a place in some sub-professional designs, even where these had moved away from valve-type circuits, perhaps where cost constraints indicated against using a high-level 1st mixer with high-level oscillator injection. Also, some of the Japanese makers, such as JRC, simply seemed to be averse to deploying ICs in receiver IF sections, and so used discrete circuitry based largely upon small-signal fets. There might have been an industrial factor there, as the Japanese semiconductors had developed their own FET ranges, whereas their linear ICs were often derivatives of (albeit typically improved) established Western types. But then that did not stop the hi-fi makers from using those ICs in FM tuners.

There were some all-bipolar professional HF receivers that used, I think, mostly small-signal devices, such as the Racal RA217 and Plessey PR155. Were these designed and developed before jfets and mosfets became reasonably available? Or was there a conscious decision to use bipolar rather than unipolar devices in the RF stages, etc. The PR155 I think dates from 1967, which would put it on the cusp of the FET era.

Cheers,
Synchrodyne is offline  
Old 19th Oct 2013, 9:37 am   #54
Radio Wrangler
Moderator
 
Radio Wrangler's Avatar
 
Join Date: Mar 2012
Location: Fife, Scotland, UK.
Posts: 22,800
Default Re: FET Questions

Rather a lot of things were going on at the same time, and they all influenced each other to various extents. I don't think we will ever unravel all the goings on because lots of things were people's attitudes and never recorded.

In the sixties, the highest performance HF receivers were almost all valved. However, there were strong pressures to adopt transistors. Pressures like size, heat, power consumption, price, and the level of complexity which could be stuffed into a given size of box. These pressures didn't all switch on at once. Early transistors cost more than valves and had uncertain reliability. Users were largely institutional: military, spooks, monitoring services and smaller receivers would pack more in their racks. We had an era of transistor HF receivers that were not top performers in terms of intermod etc, but were just good enough for certain defined uses. That was enough for them to be sellable. It took quite some time before the transistor-based comms receiver got ahead of the best valve ones. The RA1772 appears to have been the first generally accepted to have pulled it off.

One illustration is that the fit-more-in-a-rack-shaped RA1217 is seen in greater numbers than the boxy version, the RA217. The 217/1217 is a usable receiver, but not a star performer. Running with a -ve supply voltage WRT ground makes it inconvenient (unless you still have a positive-earth car). It was convenient for the designers when the best transistors were PNP.

Meanwhile the designers of domestic receivers wanted a piece of the action. "Transistorised" was written large and sometimes in golden letters on the outside of many boxes. It was seen as a positive marketing point. THese designers needed to get up to 100Mhz and contemporary bipolars didn't do that. Jfets could, and they were easier to make, so they got used. Added to this was a non-trivial psychological point, being a high-Z device, all the books and device manufacturers pointed out that this made tham similar to valves. Set makers were highly familiar with valves, and this made FETs seem less of a traumatic change. I'm pretty certain this had a big effect, first for the JFET, and then the dual-gate MOSFET. Some firms were a bit unsure about the JFET to MOSFET move and hung on for a while. I suspect that all the warnings about static damage and all the words printed about handling precautions created a strong perception of risk. Some firms only made the move after they'd seen others do it and survive.

The mosfets of the period worked best with highish voltages, at least 12v. Allied to this was the need of the higher the better voltages for varactors. They are non-linear devices and their intermodulation effects reduce at higher bias.

Recently I've used 5v dual gate MOSFETS which have small scale models of themselves shunting g1, so that you bias g1 with a current, and the whole beastie biasses-up as a current mirror with gain. They work fine, but need lower-Z networks... but the varactors still need every volt I can get for them.

If you wanted me to design a super-duper HF receiver today, I'd use a transformer-feedback 'Norton' amplifier as a switched RF preamp, an H-mode mixer, and Analog Devices' nice variable gain amps in the IF. There wouldn't be a dual gate MOSFET in sight. Preselection would be a gang capacitor and a stepper motor.

If it was a hifi tuner, it would be MMIC RF amps into a very high level switching FET mixer, with high voltage varactor preselection.

David
__________________
Can't afford the volcanic island yet, but the plans for my monorail and the goons' uniforms are done
Radio Wrangler is online now  
Old 24th Oct 2013, 5:36 am   #55
Synchrodyne
Nonode
 
Join Date: Jan 2009
Location: Papamoa Beach, Bay of Plenty, New Zealand
Posts: 2,943
Default Re: FET Questions

Thanks, David. That’s a good point about receiver performance being just good enough for their intended end uses, at least in respect of government and commercial applications. Not all applications would require or benefit from Racal RA1772 capability. I suppose from the receiver maker’s viewpoint, though, a single core model that does everything might offer lower production costs, and so lower user costs than an array of models with tiered performance. Still, that the later RA1792 was allegedly not as good as the RA1772 might have reflected an adjustment of performance and cost to what was for the most part being demanded.

Getting back to the early FET age, two receivers with predominantly FET-based signal paths readily come to mind, the Eddystone EC958 at the professional level and the Drake SPR-4 at the consumer level. Regarding the former, its high-stability oscillator aside, it did seem to follow the precedent of Eddystone’s valve-era practice, with for example a relatively modest (1335 kHz nominal) 1st IF tunable over a 100 kHz range, although it was not quite like any specific valved-model. At the time, Eddystone claimed that the EC958 was as good as any other receiver with a 10 kHz to 30 MHz tuning range. But that kind of tuning range, then more associated with marine main receivers, was probably not all that common back in 1968, so maybe that stipulation was something of an “out”.

The Drake SPR-4 was more-or-less a solid state version of the R-4C valved receiver, and allegedly had similar performance.

As a QED to one of your comments, at both ends of the scale, the makers in these two cases seemed to be staying in their valve-era comfort areas with high impedance semiconductors. Although it may be noted that by then, Eddystone had built up some experience with bipolar HF receiver circuitry.

Cheers,
Synchrodyne is offline  
Old 17th Nov 2013, 2:33 am   #56
Synchrodyne
Nonode
 
Join Date: Jan 2009
Location: Papamoa Beach, Bay of Plenty, New Zealand
Posts: 2,943
Default Re: FET Questions

Recently I came across some interesting comments in one of the Plessey manuals, written by James M. Bryant, and I think from the early 1970s, which adds some background to this topic.

In discussing mixers, in the context of receiver design that included minimum RF gain and maximum mixer resistance to cross-modulation as desiderata, he runs through a hierarchy of mixer types, from bipolar (worst) to diode ring (best), with FETs in the middle ranking.

JFETs were said to be far better than bipolar transistors, but did not make ideal mixers when used alone. But a pair in a single-balanced circuit could have useful gain, low noise, and high cross-modulation resistance, but required quite high local oscillator power.

The dual-gate mosfet was said to be another useful mixer, with the fact that it was more often used at VHF than HF resulting more from custom than any real technical reason. It was said to provide useful gain at low noise with modest local oscillator requirements.

I think that puts into perspective the use by JRC and others of push-pull JFET mixers in receivers that might be described as having mid-range performance – and price points - as compared with the top-end commercial examples.

Cheers,
Synchrodyne is offline  
Old 17th Nov 2013, 9:44 am   #57
Radio Wrangler
Moderator
 
Radio Wrangler's Avatar
 
Join Date: Mar 2012
Location: Fife, Scotland, UK.
Posts: 22,800
Default Re: FET Questions

Sitting by my right elbow is an Icom IC765, a good example of the push-pull mixer species. It has quite decent intermod performance and is actually limited more by the switching diodes for the sub-octave RX input filters than it is by the 3SK125 FETs in the first mixer.

The Japanese manufacturers went down their own path as far as a standardised architecture for HF receivers went. I think it was driven strongly by the need to hit certain marketing bullet points. They wanted shiftable IF bandwidths and also variable IF bandwidth. No doubt they agonised over thoughts of people sitting at home comparing brochures and noticing that the Ikensucom TRX7444 lacks true variable bandwidth, so it's going to have to be the Suken 732 that goes on the wish list.

The variable bandwidth business means having multiple IFs and having wider actual filters than the effective bandwidth in play. This allows large signals close to a wanted small signal to sometimes penetrate further into the receiver. This locks out the usual process of having progressively increasing signal levels down the receiver for small signals. So multiple conversions, filters, all operate close to the input level. This is generally a disaster for noise performance which is not very critical on HF, but it is also a big problem for dynamic range at the large signal end. Only after the defining selectivity has been finally achieved can gain be thrown at the business of radio reception.

Once past the defining selectivity, the designers went wild. THey threw lots of gain in, and did it cheaply. The resulting IF strips have pretty poor linearity, but that doesn't matter, they thought, there is only one signal so there is nothing to intermodulate with.

This was a bad assumption, those receivers with good linearity right through sound a lot clearer. Receivers without the variable IF trick have better close-in intermod performance.
So there is a lot to be said for a bank of switched filters.

Now that the big 3 in Japan have gone for IF ADCs and DSP, a lot of the nastiness in the late IFs has gone. I still think they're a bit poor in the audio department, but they now have the issue of ADC dynamic range. Sure there's the business of oversampling which after decimation creates some more bits of resolution, but it doesn't work quite as the classic theory if your signal isn't surrounded by random noise, but is rather suffering from noisy neighbours. So the designers have taken to adding mode-specific roofing filters at the high IF needed to put the general coverage bullet point in the brochure.

Of course, through all of this the synthesiser phase noise is also a significant limitation. Over the past few decades, synthesisers have got smaller, cheaper and more agile for jumping around, but they haven't got significantly cleaner.

I have a Racal RA1792 in the attic. It is a poor example of how to do a nice receiver.

It has no selectivity ahead of the mixer. The mixer is good, though, but not good enough to do that to it.

It has an early delta-sigma fractional N synthesiser which puts very broad noise sidebands on the LO. Even the tuning knob has a wear-prone nylon bush as a bearing. And by now most examples are blighted with LCD rot. Interesting to show people, but junk bond status. You can also hit spot frequencies where the fractional-N scrambling gets it wrong and the pseudo-random noise collapses into strong LO sidebands. (I did a bit of work on these sorts of synths some time ago… I think the pat number was US 6509800).

I saw a lot of 1792s being sold at high prices by several traders in the UK, and the buyers were attracted by the thought of a late model Racal receiver. They didn't know it was a cheapo model, cut to the hilt, and the prices outstripped the 1772.

So the late end of the all-analogue HF architecture had balanced JFET mixers and the latest generation £8000 boxes use MOSFETS in the H-mode configuration. We're back to AM, SSB and CW switched filters, but now at 60MHz first IFs, and because of the limited skirt selectivity ahead of a mix down to a very low IF, they're using image reject mixers.

The key to these receivers are the 60MHz ultra-narrow filters. I just happen to think they can be homebrewed. One of those things to be played with sometime.

David
__________________
Can't afford the volcanic island yet, but the plans for my monorail and the goons' uniforms are done
Radio Wrangler is online now  
Old 1st Dec 2013, 5:10 am   #58
Synchrodyne
Nonode
 
Join Date: Jan 2009
Location: Papamoa Beach, Bay of Plenty, New Zealand
Posts: 2,943
Default Re: FET Questions

Quote:
Originally Posted by Radio Wrangler View Post
The Japanese manufacturers went down their own path as far as a standardised architecture for HF receivers went. I think it was driven strongly by the need to hit certain marketing bullet points. They wanted shiftable IF bandwidths and also variable IF bandwidth. No doubt they agonised over thoughts of people sitting at home comparing brochures and noticing that the Ikensucom TRX7444 lacks true variable bandwidth, so it's going to have to be the Suken 732 that goes on the wish list.

The variable bandwidth business means having multiple IFs and having wider actual filters than the effective bandwidth in play. This allows large signals close to a wanted small signal to sometimes penetrate further into the receiver. This locks out the usual process of having progressively increasing signal levels down the receiver for small signals. So multiple conversions, filters, all operate close to the input level. This is generally a disaster for noise performance which is not very critical on HF, but it is also a big problem for dynamic range at the large signal end. Only after the defining selectivity has been finally achieved can gain be thrown at the business of radio reception.

Once past the defining selectivity, the designers went wild. THey threw lots of gain in, and did it cheaply. The resulting IF strips have pretty poor linearity, but that doesn't matter, they thought, there is only one signal so there is nothing to intermodulate with.

This was a bad assumption, those receivers with good linearity right through sound a lot clearer. Receivers without the variable IF trick have better close-in intermod performance.
So there is a lot to be said for a bank of switched filters.
Yes, it does seem that the Japanese makers were chasing marketing features, particularly in the 1980s and a bit beyond. And those who reviewed and wrote magazine articles on HF receivers for the consumer market often seemed to judge by feature count, and also “DX”, rather than listening performance. Some of those features, such as passband shift, were not often found on higher level professional receivers, either. I suppose one might rationalize this on the basis that units going into the consumer market might be used for very difficult DX-chasing under conditions in which professional communications would be abandoned. But as a counter, very linear receivers without passband shift might offer clear and clean enough reception not to need that feature.

Quote:
Originally Posted by Radio Wrangler View Post

Of course, through all of this the synthesiser phase noise is also a significant limitation. Over the past few decades, synthesisers have got smaller, cheaper and more agile for jumping around, but they haven't got significantly cleaner.
I remember hearing about the synthesizer phase noise problem back in the late 1980s from the proprietor of Phase Track Ltd, who made the Liniplex short wave listening receivers, mostly aimed at folk who listened to the BBC World Service, and not at all at the DX market. These had a tracking PLL form of fully synchronous demodulation, also with a selectable sideband feature obtained by phase shifting and matrixing. The first model, the F1, had six crystal controlled frequencies only, because at the time, synthesizers with low enough phase noise were not available at the appropriate cost level. The phase noise was particularly apparent when operating in selectable sideband mode. The following F2 had the same, and also provision for use with an external synthesizer, on the basis that such with low enough phase noise would soon be available. Thus Phase Track soon offered as the OSC-1. Initially the latter ran only to 22 MHz (i.e. up to the 13-metre band), as at the time extending it to 26 MHz (to cover the 11-metre band) would have been too costly. A few years later, the range was extended to 26 MHz, as improved technology allowed this without cost penalty. Fast frequency resetting was not a requirement here, as setting was done by thumbwheel switches. If I remember rightly, the OSC-1 synthesizer was priced not much below the F2 receiver.

Returning to FET-based circuits, I have attached a marked-up signal level chart for the JRC NRD-525, a well-respected consumer/amateur semi-professional synthesized HF receiver from the 2nd half of the 1980s. It used FETs throughout the signal circuit from the RF amplifier right through to the final IF amplifier. The RF amplifier and the two mixers each used push-pull pairs of 2SK125 JFETs, whilst all of the four IF amplifiers used 3SK77 dual-gate MOSFETs, with agc via gate 2.

The 2SK125 was described as a UHF device, and the 3SK77 as a VHF device. This would seem to reinforce the notion that FET devices were primarily developed for use at VHF or UHF, but also used at HF. In fact the precedent of using nominally VHF devices at HF had been established in the valve era. For example, the ECC189 double triode, designed for VHF TV tuner RF stages, was employed as a cascode 1st RF amplifier in HF receivers from the late 1950s. And the EF80 VHF pentode was also used at HF where a high slope valve was required.

The gain at the second IF, 455 kHz, from 3 MOSFET stages, totals 63 dB, which I think might also have been attainable with a standard IC such as the MC1349. I am not sure how the FET and IC options would compare in terms of agc range, though.

The IF output, at 100 mV, seems low, but then the NRD-525 uses an IC-based quasi-synchronous demodulator, a species that is sometimes referred to as a low-level demodulator and is often reckoned to allow elimination of the final IF stage as compared with an otherwise similar circuit with a diode demodulator.

During its currency, the NRD-525 was generally reckoned by the reviewers to be a bit better than the best HF receivers from ICOM, Kenwood and Yaesu, but on the other hand it was not, and was not intended to be in the same league as the fully professional receivers such as the Racal RA1772 and like models. Possibly for its time it represented about the best that could be obtained from a signal-path circuit using standard small-signal FETs, and without resorting to more exotic devices, or mixer configurations that required very high drive levels. Presumably it did better than would have a similar signal path based upon standard small-signal BJTs. So back then, the rationale for using FETs in this class of HF receiver might not have been too much different to the rationale for their use in other consumer equipment, such as FM tuners.

Cheers,
Attached Thumbnails
Click image for larger version

Name:	JRC NRD-525 Level Diagram.jpg
Views:	805
Size:	48.5 KB
ID:	86569  
Synchrodyne is offline  
Old 10th Dec 2013, 12:13 am   #59
Synchrodyne
Nonode
 
Join Date: Jan 2009
Location: Papamoa Beach, Bay of Plenty, New Zealand
Posts: 2,943
Default Re: FET Questions

In comparing professional HF receivers (such as the Racal RA1772) with the more consumer-oriented and feature-laden models such as the JRC NRD-525 and its ilk, I suppose that the latter were more likely to be used by knowledgeable enthusiasts who wanted or appreciated their multiple features and were prepared to “fiddle around” with the like of passband shift controls (and fit and use accessories) to get the best results. On the other hand, professional receivers were not always used by those with a specific interest in radio receiver technology, but by folk who were not experts in that field, and whose objective was simple and reliable operation of communications circuits. Also, professional receivers were usually connected to large aerial systems, such as rhombics, that delivered huge aggregate signals, maybe as high as 100 mV, whereas consumer HF receivers were mostly operated with much more modest aerials delivering much lower signal levels.

Another Japanese example with a mostly FET-based signal path was the Yaesu FRG-7, item attached, from the mid-1970s. FETs, variously 3SK40 dual-gate mosfets and 2SK19 jfets, were used from the RF amplifier through to the 3rd mixer, with the 455 kHz IF amplifier using bipolar devices. The 3SK40 and 2SK19 appear to have been standard VHF devices. Interesting was the use of single 2SK19 devices for the 2nd and 3rd mixers.

A look at the corresponding Kenwood and ICOM models shows that this cluster of consumer-oriented Japanese HF receivers released in the decade or so from 1975 had signal paths that were based largely upon standard small-signal “VHF/UHF” FETs. ICOM stood aside slightly in its use of four-diode DBM for the earlier mixers, and bipolar IC DBMs as 4th mixers in its R70 and R71 models. But whereas the R70 had had a packaged diode DBM 1st mixer, the succeeding R71 had the push-pull jfet type, using a pair of 2SK125. Both the R70 and R71 used a 3SK74 dual-gate mosfet 3rd mixer, though. Kenwood made some interesting changes with its R-5000 of 1987 as compared with its earlier R-1000 and R-2000 models. These had used a dual-gate mosfet RF amplifier and a push-pull dual-gate mosfet 1st and 2nd mixers. The R5000 had a shunt cascode jfet RF amplifier and push-pull jfet 1st and 2nd mixers, all using the 2SK125 device, which seems to have been quite popular amongst these makers.

Looking at earlier and simpler Kenwood HF receivers, such as the R599, one finds dual-gate mosfets used in the RF and 1st and 2nd mixer stages, with bipolar devices thereafter. The QR-666 and R-300 were interesting in that there were dual-conversion only on their respective highest frequency bands. Thus the 2nd mosfet in the chain acted as either an agc’d IF amplifier or a 2nd mixer according to the band selected. Gate 2 was used for both agc bias and oscillator injection, with fixed bias in the latter case. This compares with the previously discussed Zenith VHF TV tuner case where the VHF mixer mosfet was also used as a (non-agc’d) IF amplifier on UHF, in this case with VHF oscillator injection on gate 1.

Anyway, it would appear that these Japanese manufacturers saw standard small-signal VHF/UHF FET devices as capable of delivering the performance they required of their consumer-oriented HF receivers. Early in the FET era, it may have been that these devices were more costly than bipolar transistors, hence their more sparing use ahead of the main IF selectivity, with bipolar devices following. For whatever reasons, they were not persuaded of the virtues of bipolar ICs for use following the main selectivity. Possibly that was because Japanese semiconductor makers were not offering items comparable to say the Plessey SL600 series.

Cheers,
Attached Thumbnails
Click image for larger version

Name:	Yaesu FRG7.jpg
Views:	795
Size:	78.8 KB
ID:	86760  
Synchrodyne is offline  
Old 10th Dec 2013, 12:45 am   #60
Radio Wrangler
Moderator
 
Radio Wrangler's Avatar
 
Join Date: Mar 2012
Location: Fife, Scotland, UK.
Posts: 22,800
Default Re: FET Questions

Racal resisted the Plessey SL600 IC family as well. Most of their signal handling stages were discrete. The Radar people loved their loggers, but the ICs went into midrange HF telephones and things like that. The two guys at Plessey that stirred this lot up were designer Barrie Gilbert (of Gilbert cell fame) And applications guy James Bryant G4CLF who published the 9MHz transceiver core design using the parts. A lot of amateurs built G4CLF boards and built full transceivers around them. Not seen James for a while, he used to be ADI's European apps manager and visited South Queensferry occasionally.

David
__________________
Can't afford the volcanic island yet, but the plans for my monorail and the goons' uniforms are done
Radio Wrangler is online now  
Closed Thread

Thread Tools



All times are GMT +1. The time now is 6:01 pm.


All information and advice on this forum is subject to the WARNING AND DISCLAIMER located at https://www.vintage-radio.net/rules.html.
Failure to heed this warning may result in death or serious injury to yourself and/or others.


Powered by vBulletin®
Copyright ©2000 - 2024, vBulletin Solutions, Inc.
Copyright ©2002 - 2023, Paul Stenning.