GEC BC506 Non electrolytic capacitor question

[Kalee20]

I'm fairly happy that the polarity must reverse - not least because if it didn't, they wouldn't have gone to the expense of a non-polarised part

That's a good argument! (Unless they had a large stock to use up?)

With no-signal conditions, we know that there is -0.8V at the left of the cap. The other end is at -0.68V. So the positive end is on the right.

With the signal present, the voltage on the left can only be taken "down" the page, i.e. made more positive. As mentioned, I can't predict by how much, but that's the only direction of travel possible.

Yup 

Of course, VT2's emitter voltage also falls, but that can only fall to 0V worst-case.

Yup 

So if the AGC can pull the left of the cap more positive than 0V, then there's your reversal.

Yes, that will likely put the transistor into class C. But it's feeding an IFT, and that's followed by another IFT. So I suspect that will remove a lot of the harmonics, and even if not, how problematic would they be in this set with a very basic audio side?

Agree - the IFT's will indeed 'sine it back up' especially as there's a couple of them - harmonics in the IF signal reaching the detector will be negligible.

However, that's not what matters. It's the modulation envelope which is important. If a stage of the IF amplifier is operating in Class C, then when the modulation drops the amplitude instantaneously to what is insufficient to make the transistor conduct, the IF signal will drop to zero and stay there until the amplitude rises again - each cycle of the audio signal.

The detector output will be massively distorted (from an AF point of view). That's why I think that it can't be the case - however, every day is a school day... If it was my radio it would be what I'd verify or otherwise disprove!

That's why I said it needs to be tested.

I think we both agree on that 

[Mark Hennessy]

You could well be correct about that - I don't have the experience to know for sure either way - but there's two things that I wonder. The first I've already mentioned - how noticeable/objectionable would the resulting distortion be given the performance of the audio sections of the set? The second is something I've been thinking about over the course of the afternoon:

For a transistor in a conventional class A audio circuit with a resistive load, when it is conducting for something less than 360 degrees, it's easy to visualise the waveform and understand where the distortion is coming from. But doesn't this change when the collector load is inductive? In that case, the cycle is completed with the energy stored in the inductance. At least, that's what I seem to recall from learning about transmitters many moons ago.

As mentioned, I can't say either way, but it's something I'd be keen to investigate if I had the set in front of me. I don't stray too far from AF all that often, but it's good to occasionally learn something new...

I suppose if TR2 was cut off too much, the resulting loss of signal might cause the AGC voltage to drop back a bit, meaning that there might well be an upper limit to the extent to which TR2 can be biased off before a kind of equilibrium is reached? That's another interesting line of enquiry.

Some sets had an additional damping diode to cope with gross overloads, but those seemed to fall out of favour fairly early on - at least for Hacker sets

There's often rather more to "simple" circuits than meets the eye, and it can be good fun to peel back the layers!

[Mike Watterson]

It must go into class C on large signals. As the load is tuned, that should work. It works for CW and FM drivers/PA on transmitters. However it's a negative feedback loop for gain. It would be interesting to lash up such a circuit on my Rigol which has a variable power feature on the tracking generator.

[Mike Watterson]

I think all the dodgy bi-polar electrolytics I've replaced were late 1960s and also 1970s,
However the two colour plastic cased Plessey polarised ones all seem to fail too.

The oxide layer on electrolytic is the dielectric and on a regular type it will gradually dissolve / gain holes with lack of use or reverse polarity. A current limited correct DC voltage re-oxidises the surface. High ESR is usually due to the electrolyte drying out or contamination and the electrolyte limits the AC component of the current possible.

I can't see how a bipolar electrolytic can be as reliable as polarised types, and certainly any electrolytic is less reliable than ceramic or plastic dielectric.

[Mark Hennessy]

Thank you - it's good to know I wasn't dreaming! It was a long time ago, and as most of my design work is audio it's rare to use anything other than a resistive collector load - certainly not tuned. At some point I'll have a dig through my radio collection to see if I have anything that is close to the GEC circuit to investigate this further...

And I can understand why non-polar caps might seem like a bad idea - if we have to rely one oxide layer to not fail with time, then adding a second might not be the best plan

In practice, I think that the reverse-leakage of the wrongly biased oxide layer will allow enough through to help keep the correctly biased layer in good condition, as takes place in a regular polarised cap. Providing the cap sees opposite polarities from time to time, both layers should be maintained.

But modern capacitors do seem to be much better at staying healthy in scenarios where there is no bias. In audio electronics that uses split supplies, which is most these days, most capacitors have zero DC voltage across them, yet can last for decades. Similar caps acting as decouplers in the same equipment don't seem any more long-lived.

Obviously that's all rather anecdotal, but as I mentioned earlier, I'm just describing my experience

Either way, I do like non-polar caps. My opinion of them went up when Cyril Bateman published his articles about capacitor distortion in WW a while back (July 2002). Part 6 is the one that covers non-polar caps. Admittedly, that was focusing on distortion, not reliability. But as I say, in my experience at least, they seem to last just as well as their polarised equivalents. "Equivalents" is the key word there, because the non-polar types are usually 2,000 hour, 85C types. Of course we know that polarised caps are available with longer lifespans and 105C or even 125C operation, so it wouldn't be reasonable to expect the non-polar types to match those. I presume there just isn't the market for long-life, high temperature non-polar caps. Nothing stopping you make your own from two posh ones in series if needed - done that often enough...

Absolutely agree in general about film caps, but worth mentioning that certain plastic dielectrics can be less than reliable - think about how X2 caps gradually lose capacitance because of the self-healing dielectric swelling up. Often stops capacitor-dropper PSUs from working. Those might be considered historic, but we've had some quite modern bits of broadcast gear fail because of this (the capacitor dropper PSU was a start-up circuit for the main unit - long story!). Some LED lamps use capacitor droppers, of course. To be fair, they do lead a tough life in that role...

[Kalee20]

It must go into class C on large signals. As the load is tuned, that should work. It works for CW and FM drivers/PA on transmitters.

when the collector load is inductive? In that case, the cycle is completed with the energy stored in the inductance. At least, that's what I seem to recall from learning about transmitters many moons ago.

Both are correct - and the tuned LC circuit will restore a nice sinewave to a pulsed-current Class C amplifier.

But - it's immaterial really whether the IF signal is distorted - what matters is that the modulation envelope is preserved.

Drive for CW and FM is unmodulated. Even for an AM transmitter it's unmodulated - modulation happens at the anode.

Class C amplifiers are highly non-linear (ask the SSB boys and girls).

See the attached sketch for what happens in an amplifier biased beyond cutoff, with a varying signal. When the amplitude instantaneously drops below a certain level, the output goes to zero even though there is still some small input signal 'there.'

[Mike Watterson]

Oh, you can't use class C at all for SSB or Digital modes, or easily for normal AM (except of of course if the supply voltage to PA has the audio). Actually nQAM and COFDM digital needs very linear.

If the transistor is "hard off" then the IF signal is passed only by Base - Collector capacitance?

It's hard to imagine just looking at the schematic exactly how this works, but at nearly zero signal there is maximum gain and then there is a point where it gets non-linear on larger signals.

[Kalee20]

It's hard to imagine just looking at the schematic exactly how this works, but at nearly zero signal there is maximum gain and then there is a point where it gets non-linear on larger signals.

Well, the gm of a transistor is proportional to collector current - so if the transistor is biased back to a small collector current, then it'll have a low gm and hence low gain. If it is passing a high collector current then it'll have high gain.

This of course is valid if the input signal is relatively small (tens of millivolts).

But it relies on the transistor passing some collector current, which it can only do if the base is biased 'on' ie negative with respect to emitter (PNP transistors here!). It can't ever be reverse biased because that would well and truly cut off the transistor - except for super-large input signals, working in Class C, which as above is untenable for a linear amplifier.

So I'm still opining that the base won't ever go positive, and the voltage on the elcap won't therefore ever reverse.

[Mark Hennessy]

The only way to know for sure is to measure it. It would be lovely if the OP is willing to experiment for us. Just a volt meter connected across that cap is all that is needed - just to see what the range of voltages are in practice.

In the absence of data, I thought it would be useful to have an idea of the range of DC voltages one could expect from the detector diode. Most of my radios are in storage, and of the handful that are on display downstairs, they're either not very representative of the set in question, or would require too much dismantling for my liking (I've got a lot on today). So the best I could do is the Hacker Sovereign II. The AM IF PCB is under a screening cover, but it is possible to probe the audio output, which is fed almost directly from the diode - just a potential divider of 1k into 4k7.

I found this voltage varied between low 10s of mV between stations up to nearly 2V (1.95V) on the strongest stations. Radio 4 LW was nearer 1.8V. These are negative voltages (this set uses NPN transistors).

I'm not assuming these numbers are the same as you'd find in the GEC set, but if we pretend that they were for a moment, we can quickly determine that a positive bias is possible: if we ignore the base current and assume a HT of 8V, we'd have +0.7V at the base. As the BE junction is now reverse-biased, ignoring Ib was OK, probably...

This isn't a "slam dunk" - we are talking about 2 fairly different sets - but it shows that might just be possible.

I do take the points made above, of course. Perhaps this is why AGC evolved to include 2 IF stages in many cases? This GEC is a very early transistor set.

[Mike Watterson]

Also they DID use an 8uF polarised cap on the volume control. Someone got the more expensive bipolar past the cost-cutting and inventory team.
In theory I could model it in LT Spice, but likely quicker to lash it up or measure a similar set I have.

The BC 504 has has 8uF polarised to 0V (-Ve to the positive diode) and 40nF from AGC line to emitter. You'd imagine a very strong signal might be an issue. R&TVS pg 122, 1961-1962
The BC505 (which I have) is very similar, except now 100nF on AGC to emitter.

Both have 2 stage IF, as has the 1 year later BC506. I don't see any AGC voltage on 2nd IF.

The three models are very similar and change on the BC506 "saves" one capacitor. The BC505 has an extra audio pre-amp compared to the BC504 and BC506. The BC504 & BC505 use GET873 IF and GET114 for all audio. The BC506 uses 2x PXA101 for IF and PXB103 driver and PXC121 PA. All 3 use a "totem-pole" push pull output stage so there is only a driver transformer.

I'd call these 3x GEC early (1961-1962?) rather than very early (1956-1957?)?

So I can certainly hook up the BC505 and see what a strong signal does. It claims -0.8V on 1st IF base with no signal. I can monitor voltage on R8 via an RFC to see what input signal drives VT2 to cut off.

Note some Ever Ready 1950s valve sets had AGC on mixer/Osc (DK9x) and on the IF amp (DF9x), but at least one had no AGC at all!

The transistor sets with 2 IF amps are poorer than many valve portables and all mains valve sets with 1 IF due to the the 2x IFTs on the valve sets have very much higher Q, and maybe lower loss. It's not just an issue of device gain. The valve IFTs have looser coupling (two cores) and are tuned on primary & secondary. Many transistor sets (and eventually all of them) had only single tuning and a common core.