Valves versus FETs

[Biggles]

It's nothing unusual to have only four valves in a battery set. Would it be possible to design a radio with the equivalent performance but using only four FETs instead? If not, why not? What are the limitations of using FETs? I am aware that your average mains set would technically employ five valves, but in a battery valve set, a self oscillating heptode mixer is used, hence only four active devices.
Alan.

Using dual gate mosfets in the RF/IF stages (yer modern tetrode/pentode) I would say no problem at all, bear in mind that valves where much more expensive than coils, so with properly designed (no expense spared) coils, yes. The output stage may require a fairly beefy cascode mosfet (big dual gate mosfets don't exist) design, we could call that one device.

It would be fairer to replicate a triode only set with FETs, all that anode to grid capacitance, just like FET drain to gate.

[G6Tanuki]

A 40673-type MOSFET can be made to work as a self-oscillating mixer at frequencies up to 200MHz or so. So no problems with the front-end. A couple more dual-gate MOSFETs as IF amps. Then are we allowed a diode-detector? (I'd go for a voltage-doubling one to get lots-of-volts to drive the output stage, remembering that FETs are intrinsically voltage-devices not current-devices).

If we're allowed to use high-ish-voltage supplies like the old battery-valve portables did then a suitable FET (VN66-type) run in class-A could deliver a decent amount (half a Watt or so) of output power with little difficulty.

Alternatively, to update the idea to the FM era: a grounded-gate RF amp followed by a superregen FM detector [even the humble 2N3819 is good to 500MHz] then 2 FET audio stages??

[kalee20]

If trying to FETise a conventional 4-valve battery superhet, the frequency changer would be the biggest challenge. A junction FET oscillator combined with a dual-gate MOSFET as (multiplicative) mixer would probably work. Whether you could squirt AVC in as well... Probably, but dunno.

The IF amplifier - a JFET.

The detector and first AF amplifier - a Shottky diode and another JFET.

The output stage - a beefy JFET (J410?) or a power MOSFET should work. I don't know of any power JFETs, which is a shame as I tend to prefer them to MOSFETs which are so inherently static sensitive!

[Radio Wrangler]

There have been some moderate power JFETs... P8000 and P8002 for example. I've tried them in grounded gate mode as buffer amps after high dynamic range mixers, but they were too hard to get.

Cree did some power JFETs in exotic semiconductor materials like silicon carbide.

Generally, FETs would give somewhat less max gain per stage than could be had from valves.

David

[turretslug]

I think it's a little unfair to expect as few as four three-terminal FETs to do the job of specialised valves with more terminals improving both their general and specific characteristics (pentodes, heptodes). As FETs are relatively cheaper in the overall cost of the set compared to valves, let's have cascode JFETS in mixer and IF stages with a separate LO, and a sort of DC-coupled discrete op-amp in the AF amp (though there was a brief trend for single-ended high voltage bipolars such as the MJE340 more-or-less in place of an output valve- not best use of battery power, though). I know that's missing the point of apparent equivalent count, but a pentode could be considered as two or more three-terminal devices combined, so relaxing the FET count restriction seems fair.

[Bazz4CQJ]

Originally Posted by G6Tanuki

[even the humble 2N3819 is good to 500MHz

That's an observation you don't often hear, but FETs are notoriously "variable"

[kalee20]

Originally Posted by Bazz4CQJ

...but FETs are notoriously "variable"

I think it was Douglas Self who wrote, "Any attempt to design a repeatable circuit with FET's is likely to make you thoroughly grateful that bipolar devices exist."

I do have a soft spot for JFETs though, ever since I learned in my mid-teens that they are very like valves from a circuit point of view. Only up to a point, though - valves don't look like variable resistors at very low anode-cathode voltages, whereas JFETs do (and bidirectionally, too).

Limitations of FETs are as already stated - and there's no equivalent to the screen-grid or the suppressor grid in a pentode. You couldn't make a FET equivalent to a transitron oscillator, for instance.

[MrBungle]

Originally Posted by kalee20

"Any attempt to design a repeatable circuit with FET's is likely to make you thoroughly grateful that bipolar devices exist."

Couldn't agree more.

About the only thing I've ever used a FET successfully and repeatably for is a follower and that's not without problems. Thank goodness they can them away inside opamps now where they are not my problem

Once I tried to build something out of a library book that used a humble MPF102 that had to be hand selected. After three consecutive muggings in Tandy which in my youth a 3 mile round trip walk as well I said "sod it" and hacked it with a boostrapped emitter follower and it worked fine with it.

And thus, FETs have always annoyed me.

Bipolar transistor, minimal base voltage variation (although still voltage driven) and at a low impedance, miller effect quite low. FET, voltage driven, high impedance, lots of miller effect. A tetrode or pentode, very small miller capacitance, easy to use hence my avocation of cascode (or dual gate) FETs. This applies to the output stage too.

[Biggles]

I hadn't really thought about the fact that the equivalent to a JFET is going to be a triode, so you would really need more devices to replicate the performance of "modern" valves. I must admit I prefer bipolars if I am designing something. Nice and predictable and I am familiar with a lot of the common types. Enough to cover most of the tasks I require them to do anyway. I would only use FETs if I was working to a tried and tested circuit. Otherwise I have only really used dual gate mosfets in receiver front ends and mixers. The 40673 was a wonderful device, as was the 3N201.

[Radio Wrangler]

Not only is a FET a triode, but a FET cascode or a dual-gate MOSFET act. like a tetrode/pentode in haveng an effectively grounded screen to prevent Miller effect due to capacitive current going straight fro output to input. As said above, the transitron effect is missing.

Fets have wide variation in threshold or pinch-off voltage, a very wide variation in saturation current and basically have nothing that's very well tied down.

Bipolars have well defined base-emitter voltage/current characteristics, have wide variation in current gain, and no limit short of destruction in max current. At least they have one reliable parameter.

Biasing FETs to get reliable control of quiescent current is a bit of a problem. Bipolars are a lot easier in this respect.

NXP have recently put out a huge document listing swathes and swathes of small transistors which are now to be no longer available. It looks like all the dual gate MOSFETs have bought the farm. For maximum fun and errors, the list starts with lots of devices in alphanumerical order, but then there isa pile of stuff in scattered order. I've just found a critical part in the scrambled area having thought for a while that they'd missed it when I didn't find it in the orderly section.

David

[Craig Sawyers]

Originally Posted by Radio Wrangler

NXP have recently put out a huge document listing swathes and swathes of small transistors which are now to be no longer available. It looks like all the dual gate MOSFETs have bought the farm.
David

Semiconductor manufacturers are only interested in semiconductors which have huge markets. So anything that is viewed as being lower volume or niche, or tricky to fabricate (like p-channel jFETS, or PNP small signal transistors with low and specified rbb' supported by noise contours) is likely to get discontinued.

They are driven by the mobile phone, automotive and home entertainment markets.

[Argus25]

Originally Posted by Biggles

It's nothing unusual to have only four valves in a battery set. Would it be possible to design a radio with the equivalent performance but using only four FETs instead? If not, why not?

A single fet and most valves (pentodes) for radio work cannot be directly compared, because the fet is a three electrode device with Miller (feedback) capacitance operating between the drain and the gate, much like the triode's anode and grid. The pentode is essentially a 4 element device, the screen grid acts as an input to output shield between its control grid & plate.

In radios at least, any amplifier operating at radio frequencies with a tuned circuit in its input and output, of the same frequency, is unstable due to Miller feedback when it iss constructed with one device like the fet, or a triode. This is why in RF stages and IF stages pentodes are used. Oscillators and audio stages only require a three element device.

For a fet stage to act as a 4 element device with "input to output isolation" like a pentode (circumventing the Miller effect) then TWO fets need to be wired in cascode, to electrically simulate a pentode function.

So consider a radio with valves 1) RF pentode, 2) mixer osc, 3) IF amp pentode, 4) detector diode, 5) 1st audio amp 6) audio output,

All could be done with one fet each, except items 1 & 3 which would ideally require two fets each in cascode. It might be possible to get away with just one fet in each of these stages if the fet used had a low enough feedback capacitance, but instability is a risk, neutralisation can help.

As an analogy, early transistors like OC45's had a high feedback capacitance and RF and IF stages that used these always required neutralisation, but later transistors like the AF117/127, had such a low feedback capacitance that neutralisation was avoided.

If you look up the Fetron TS6AK5, you will see that these were two fets wired in a variation of the cascode configuration so as to simulate the function of a 6AK5 pentode, and if you want to see an all fetron radio there is one here:

http://worldphaco.com/uploads/WORLDFETRON.pdf

[Synchrodyne]

I was of the impression that fets were not much used for battery-powered portable circuits because they had higher current consumption for about the same performance as bipolar circuits. Still, that said, Hayward & Damm reinvented the jfet/bipolar cascode in part to provide a circuit that was less sensitive to battery voltage drop than the jfet cascode, as noted here: https://www.vintage-radio.net/forum/...=98398&page=6; see post #104.

The difficulties associated with actually using fets was covered in several publications from early in the fet era. Perhaps this, taken from Pat Hawker’s “Technical Topics” for 1969 January, in which he relayed a commentary from Eddystone on its then-new marine MF/HF spot frequency receiver (evidently the EC964), provides a concise summary:

“On FETs versus bipolar transistors, the choice was determined by the fact that although the FET is capable of better large-signal performance, the spread in parameters makes it difficult to take full advantage of this, unless complex circuitry is used to restrict the effect of the variations (this factor is less important for the home-constructor than for quantity production). For this reason, FETs were used in all signal-handling circuits up to the main selectivity (i.e. where strong unwanted signals may be present), and bipolars in the later stages.”

(That was on hand because it was intended to be part of a larger posting for the “FET Questions” thread, but one that I did not complete before the thread closed; I think it had anyway run its course by then.)

Originally Posted by kalee20

If trying to FETise a conventional 4-valve battery superhet, the frequency changer would be the biggest challenge. A junction FET oscillator combined with a dual-gate MOSFET as (multiplicative) mixer would probably work. Whether you could squirt AVC in as well... Probably, but dunno.

Well, it has been done, agc on a dual-gate mosfet mixer that is. An example was the Eddystone EB35 Mk III, AM front end schematic here:



AGC bias was applied to gate 1, along with the signal, with the oscillator applied to gate 2. Whether it could also be made self-oscillating whilst retaining AGC, I don’t know.

A range of fet-based AM demodulator circuits was described in Wireless World 1978 September, page 45ff.


Cheers,

[Radio Wrangler]

The Mullard "Transistor Audio and Radio Circuits" little blue book of 1969 has a section propounding the benefits of bipolar transistors, run at very low currents, over FETs for battery radios.

David

[G6Tanuki]

Originally Posted by Synchrodyne

A range of fet-based AM demodulator circuits was described in Wireless World 1978 September, page 45ff.

I recall building a shortwave regenerative receiver using a FET as an infinite-impedance detector.

The infinite-impedance detector is basically a "cathode-follower" with a lot of negative bias so it operates on the non-linear part of the transconductance curve. Giving near-100% audio negative-feedback it gives lower distortion than diodes etc. and the [near] infinite input-impedance means you can easily hook it directly across a parallel-tuned circuit.

However it provides less-than-unity gain. So in my receiver I used what amounted to a Q-multiplier [using a a BF115 transistor] around the aerial tuned-circuit - this could be wound-up to the point of oscillation if needed.

There's much to be said for separating the "regeneration" and "detection" functions in simple receivers - it means you can adjust them independently for best results rather than having to compromise.

I also FET-ized the detector in a cheap 'marine band' transistor radio I used to listen to 80 and 160 metre band amateurs - again using an infinite-impedance JFET design. Of course since this replaced the original diode there was no source for an AGC voltage - until I realised that the infinite-impedance detector would have an amplified version of the input-signal at its drain! So, a 4.7K resistor in the drain-feed, a 100pF capacitor and an OA91 and I had AGC/S-meter action again.

[kalee20]

Originally Posted by G6Tanuki

I also FET-ized the detector in a cheap 'marine band' transistor radio I used to listen to 80 and 160 metre band amateurs - again using an infinite-impedance JFET design. Of course since this replaced the original diode there was no source for an AGC voltage - until I realised that the infinite-impedance detector would have an amplified version of the input-signal at its drain! So, a 4.7K resistor in the drain-feed, a 100pF capacitor and an OA91 and I had AGC/S-meter action again.

Though, in such a circuit, there'll be RF voltage at the drain, and with a resistor as drain load, it will have shunt capacitance.

Enter the non-zero Cdg, and with the partly capacitive drain load, the input will look resistive by Miller effect! So, unfortunately it's no longer an infinite-impedance detector.

It's why anode-bend detectors, at least with a triode, aren't quite as damping-free on the input tuned ciircuit as an infinite-impedance detector.

[G6Tanuki]

True. I agree it's not perfect [a tetrode/pentode "FET" would be better] but it did work.

I wonder how easy it would be to use two JFETs in a "cascode infinite-impedance detector" to neutralise Miller effect?

Or decouple the drain to RF but use a small-enough capacitor so there'd still be demodulated audio there to rectify for the AGC? Audio-derived AGC can work well if you get the time constants right.

A pentode FET isn't needed, there are no secondary emission electrons in a FET!