ECC88 Cascode Questions

[Synchrodyne]

Yes, I quoted from the article on a "without prejudice" basis. I also think that “most” meant pentode based designs. But it probably excluded for example designs based upon the 12AY7 double triode. (for an example of which see “Radio News” 1949 March p.15ff, https://www.americanradiohistory.com...Page_Guide.htm).


Cheers,

[G8HQP Dave]

Thanks. This sort of information can be hard to find nowadays. Reading that article I wondered how many modern 12AY7 are actually built in the correct way for low noise and microphony, and how many are simply a dual triode which happens to have about the right mu and transconductance.

[Radio Wrangler]

I think that these days you can only expect about the right Ia at the specified grid bias and about the right gm. Mu and all the rest could be anything. Currently produced valves seem to have no goal other than audio and therefore they want them to look OK on a tester. The failure rates suggest liberties are taken with voltage/current/power ratings.

A test at full ratings would really sort them out.

Electrode capacitances, microphony and secondary emission effects don't seem to be thought of.

David

[Synchrodyne]

A couple more articles on early audio applications of the cascode:

The first was in “Electronics” for 1954 March, p234 pp.156 & 157, “Cascode Audio Amplifier has Low Noise Level”.

The heading statement includes: “Wallman amplifier circuit used as input stage of audio amplifier has pentode gain qualities with triode noise characteristics.” That was a fair summary of its qualities, and on the noise front implies that it was as quiet, but no quieter than a triode otherwise deployed.

The second was in “Electronics” for 1957 October, p.234, “Cascode Audio Preamplifier”. It runs through the basic design calculations.

Electronics magazine is available at: https://www.americanradiohistory.com...aster_Page.htm.

I have pondered as to why the cascode was not more commonly used as the input stage of domestic hi-fi control units. That had led to wondering whether the need for matching and equalizing magnetic cartridges was a factor.

A single input stage (triode, pentode or cascode) used to match a magnetic cartridge, and amplify and equalize its output required shunt feedback, with a relatively high impedance feedback network. The input arm, to meet cartridge loading specifications, would require a 47k (or in valve days, often 68k) resistor. Would this have provided a limit to the achievable signal-to noise ratio such that not much of the theoretically available increment in going from a pentode to a cascode was actually realizable?

Using series feedback would get around that problem, if indeed it was a limiting factor, but would also require a second stage, and because the first stage cathode was not at ground potential, DC heating as well. But evidently a triode pair comfortably provided enough gain for this purpose, so that there was perhaps little advantage in adopting a cascode first stage, not enough to justify using three triodes anyway. But the use of a triode pair with DC heating was not that common, I’d guess because of the difficulties of arranging for DC heating. More common I think was the use of a “flat” first-stage triode amplifier followed by a second triode with shunt feedback equalization. It was done that way in the Radford SC2 control unit for example.

It may be recalled that Wallman et al developed the cascode as the best way to deploy two triodes at the front end of radar receiver IF strips, in cases where a single triode did not provide enough gain to render negligible the noise contribution of a following pentode stage. The cascode was also found to be ideal for VHF TV tuner RF amplifiers (for which application special-purpose double triodes were developed), TV camera tube head amplifiers and some audio applications, such as microphone preamplifiers. But the empirical evidence is that it was not a universal panacea in the audio case, and that for some applications, such as domestic hi-fi amplifier magnetic cartridge input stages, other ways of using two triodes were often preferred.

That said, it occurs to me that the cascode might have been a good choice for a Baxandall tone control in situations where one wanted to extract some gain from it, although I am not aware that it was ever used for that purpose. Baxandall’s original circuit used a high-slope RF pentode and had zero gain. And although it was sometimes done around a single triode, getting some gain as well would probably be a bit of a stretch. Using a triode pair would have required series feedback, and I don’t think that Baxandall ever did such a version of his circuit. (Although the Quad QCII and 22 tone controls used triode pairs with series feedback, as best I can determine these were not simply Baxandall variants.) In the transistor era, Ellis (WW 1973 August, p.378) had proposed using a cascode pair (with bootstrapping) for the Baxandall tone control where reasonable gain was also required.


Cheers,

[Radio Wrangler]

Two other factors were also in play. The designer of an input stage for a magnetic cartridge had to find an extra 30-40 dB of gain. Cascading two stages was found to be a more economical deployment of valves than cascoding one stage, and still having to add a second stage.

Secondly, 47k in parallel with however many picofarads was the load needed to be presented to the cartridge in order to get the intended frequency response, but this has no direct relationship to the impedance the cartridge presents to the amplifier. The amplifier design needs to be planned for good noise performance considering the impedance it is presented with. The resistive component of a magnetic cartridge is much lower than the resistive component of a piezo-electric cartridge. The optimum noise performance is therefore found with higher Gm devices running somewhat more quiescent current.

Signal to noise ratio is very strictly a power ratio thing and looking at voltages can be misleading. So in general, you'd want to do careful impedance matching to get as much signal power as you can out of the transducer and into the amplifier. You then have to accept a significantly worse compromise in order to correctly load the cartridge and then it gets more complicated, in one of those feats of gross unfairness our universe is famed for, most active devices give their best noise figure with a different source impedance to that with which they give their best gain. Seen on a Smith chart you get one set of 'Gain circles' around the source Z for best gain, and another set of 'Noise circles' around the source Z for lowest noise figure. The poor design engineer has to pin the tail on the donkey and pick some source Z. It's enough to drive the poor soul into creating crop circles.

Of course this game has to be played backwards. The source Z is forced by the cartridge and the amplifier sees this in parallel with the load resistor/capacitor it is required to add. So a device has to be chosen that gives good noise behaviour with this combined source Z. This won't be gain-optimal, but the first stage gain will be enough to dilute the noise demands on the second stage. So the less than ideal gain can be made up for with an added stage.

So, the ECC88 turns out to have a significantly better input noise characteristic than the EF86, and it lacks the partition noise mechanism. But used as a single triode its anode impedance is too low to rival the mu of the EF86, but cascoding can fix this without throwing away the noise advantages. The chief downside is that the poor engineer is now under attack from a ravening horde of accountants brandishing their golf clubs in a most threatening manner. So EF86 it was, then.

David

[Herald1360]

Not quite ROTFLMAO but getting there

Prompted a quick trawl on ARH to find the difference- on a sample of one (Bentley Acoustics around 1960) the ECC88 was 8/- dearer than the EF86 (18/6 vs 10/6) so quite a mouthful for the bean counters to chew.

[G8HQP Dave]

At HF you get a significant noise advantage by using a cascode instead of a pentode, because it eliminates partition noise - the dominant noise in most HF circuits.

At audio you get less advantage, for two reasons:
1. the EF86 had been developed, and this had quite low partition noise (also 6BR7/6BS7)
2. at lower frequencies you get 1/f noise, and this is not reduced by cascode wiring

In addition, the cascode is only suitable for small signal inputs as it is less linear than the same triode used in a normal circuit. Good for small signals at moderate S/N; not so good for bigger signals at excellent S/N. It also has poor PSRR, which will be much more of a problem for audio than RF.

[Diabolical Artificer]

Some good links there Synchrodyne, thanks. I'll continue reading up on this subject, fascinating.

Thanks Dave for the sum up.

Andy.

[Syrinx1]

Some really interesting reading, thanks everyone

While researching, I also came across an interesting circuit by Dr Exely for low volume compensation... which might be useful to me as I only want a low wattage practice guitar amplifier... It's a little off topic,so I'll post the link to the Dr Exley circuit in a new thread.