Yes, Qualityten, there is a circuit description somewhere - which caused some heated discussion in the past - and as a result of which I have vowed not to engage in in future . . . oh well ;D
The circuit appears to have a standard pentode input stage, followed by a somewhat strange Schmitt (long-tailed pair) phase inverter - consisting in this case not of the usual twin-triode format but combining a pentode-triode configuration. The puzzle starts when one notices that neither of the pentode stages are connected as classic pentodes, but something which might smack of a misunderstanding of the classic pentode connection.
To refresh: A pentode is connected with cathode and screen grid both bypassed to signal common or 'earth'. In the Radford however , neither stage follows convention. One would expect the EF86 screen bypass capacitor (C3) to go to the EF86 signal common, but it does not . An equivalent circuit reveals that there is in fact quite a large signal between G2 and signal common - some 10x the input signal. This is because the NFB signal goes to EF86 cathode, placing all of the NFB signal between G2 and cathode.
Going to the phase inverter, similarly the screen bypass capacitor (C5) does not go to stage signal common (the cathodes of both valves) but directly to amplifier common, again placing external signal between cathode and G1 , again compromising pentode operation of the 6U8 phase splitter.
The resultant operation has never been subjected to measurements as far as I know, though the well-known international commentator Stuart Yanniger (SY) has taken part in a previous discussion years ago.
Apology to the OP Mudda for an unexpected message length! Also it was perhaps somewhat O.T. I hope it will have become clear why this is not a simple circuit, as all would perhaps have thought at first glance. There is also suspicion that the designer had erroneously connected the screen bypass capacitors as per classic way, overlooking the particular operation of this circuit*. As it stands, it reminds of a tapped screen operation (as per UL) but such cannot be valid with an RC circuit.
* Note: This deviation from true pentode mode occurred more often, notably a.o. in some Leak designs. As said it was not clear at the previous instance what exactly the consequences were distortion-wise. It might not been serious at low signal amplitude - suspicion remains as to the effect of the 'error signal', which in these cases is an order of magnitude higher than the stage input signal.
So there Qualityten - sometimes surprises creep in! I would like to find an analyses of the circuit by someone with both knowledge and measuring instruments to provide the answer.
(Perhaps one can persuade 'Handsome' or 'Ufudu' to contribute ? )
(Might perhaps add: Use of a pentode input in the phase inverter was perhaps for extended open-loop gain; the capacitive load on the anode of the input EF86 will have been lessened. Manley followed a similar logic by inserting a cathode follower after the high impedance anode resistor of the input stage.)
[But we are way off topic now. Perhaps a new thread for those interested in the specific mentioned design deviation - ]
"To cut a long story short the problem turned out to be the ECF82. Arthur Bailey's phase-splitter design, using this valve, is unlike any other. All other two-valve splitters use two identical valves - usually triodes and often both in the same envelope. An advantage of this is that it's relatively easy to ensure that these valves, which are produced on the same machine at the same time, are well matched. The ECF82, by contrast, contains one triode ? and one pentode (F). They were designed for use in TVs and radios with the triode and the pentode doing different jobs (the triode was typically an oscillator and the pentode a mixer). "
"Strange that given the apparently well documented technical merits of the BPS circuit in the radio and hI-fi press of the time, that it was never adopted to my knowledge by other contemporary leading amplifier manufacturers. Was there a patent in force or did the majority of amp designers realise that the original 5-20 spec really didn't need 'improving' after all ?."
I'm not really sure why no-one else tried this. The ECF82 was widely available and pretty cheap (only 30-40% more than the ECC83 in 1962, when the circuit was made public). And there's no mention of a patent in Bailey's Sept '62 Wireless World article or in Radford's ads for the MA15 amp. He does, however, make the point that distortion in the phase-splitter only becomes important if you're using the very best output transformers. Radford prided themselves on being extremely good at transformer design and construction and maybe he was having a dig at the other amp manufacturers ? Equally maybe they didn't want to be seen to have been outsmarted by Radford and so rejected the design through a mixture of pride and the 'not invented here' mindset."
"The problem with all the other phase-splitters (at least according to Bailey) is the low HF limit for the NFB loop, set by problems of phase shift at high frequencies. His design, using a pentode, eliminates a significant contribution to this phase shift i.e. the Miller capacitance of the phase-splitter's input triode. It works a treat. He argued that a double pentode would have worked even better and perhaps it would also have been better-matched than the triode-pentodes seem to be (then again maybe my sample is just too small to judge from). Unfortunately there was no other application for a double pentode and Bailey's phase-splitter wasn't popular enough to justify any of the valve manufacturers putting one into production just for him."
The Radford designs were high-performance designs driven mainly by Radford's high-quality transformers. The odd use of two (very) dissimilar tubes for the phase-splitter, as noted above, minimised the effect of the input pentode's high output impedance and the phase-splitter triode's miller capacitance. The result was both bandwidth and phase deviations being moved much higher up in frequency. This allowed more feedback to be utilised and thus more advantage to be taken of the higher-performing transformer. No doubt the higher feedback then ironed out the the increased 2nd harmonic distortion of the "mismatched" phase-splitter. This was quite prescient thinking for it's time: essentially the "make an amplifier linear BEFORE the addition of feedback" route. Stu Hegeman's Harmon Kardon Citation took this thinking to the extreme using all pentodes in the input/phase-splitter and, was then able to implement nested feedback loops and get extremely wide-bandwidth. If I am not mistaken I believe our own Alan Hobkirk designed and marketed an amplifier that modified the classic Mullard/Leak input stage to the same end. Hobkirk placed a buffer between input tube and phase-splitter - that way retaining a matched phase-splitter but side-stepping the Miller effect.
[member=144]Ampdog[/member] interestingly the original STA12 Radfords did use a "proper" pentode connection with screen bypassed to the cathode...I haven't read all the articles on their design but perhaps the reasoning for that move is mentioned somewhere?
"[member=144]Ampdog[/member] interestingly the original STA12 Radfords did use a "proper" pentode connection with screen bypassed to the cathode...I haven't read all the articles on their design but perhaps the reasoning for that move is mentioned somewhere?""
This is not correct, there is sometimes some misconception regarding the Radford's model numbers.
The Radford STA12 is a derivative of the Mullard 5/20 amplifier. It uses: EF86, ECC83 and EL34 valves.
The STA15 MkII was the first Radford that utilised the Triode/Pentode configuration. The circuit diagram is basically the same as the MKIII. Its output transformers differs in size, it is physically smaller and used different plugs and sockets compared to the MKIII. It has its own pre-amplifier.
On the forum is a posting regarding the Radford STA12 restoration.
Thanks for 'joining' on this topic, which yielded some contradictions in the past, but never an acceptable explanation! As mentioned before, this surprised me as other forums had many E.E.s as members.
Continuing with some thread-jacking: The Schmitt phase inverter was often used without an input buffer (if I may call it thus). Many successful circuits exist (e.g. all the Leaks and the classic Mullard circuits) without such a buffer. In my own former life I built several like circuits without requiring cathode followers after the 1st stage - but no quibble. Again the un-pentode pentode circuits are the puzzle.
I will put up the subject again in other major forums, hoping to get better replies than that it 'does not seem' to matter while Spice analysis clearly shows that it does! Perhaps academic - still . . . .
I stand corrected: Radford never used the ?traditional? pentode configuration in any of their designs they always connected screen bypass to 0V.
The advantage of the ?buffered Schmitt? and ?pentode-triode Schmitt? was simply wider bandwidth which I presume allowed those designers to use more feedback. But as [member=144]Ampdog[/member] notes it wasn?t really a problem that anyone else found. Interestingly the Radford design still utilised a step network on the input pentode despite its bandwidth enhancing splitter.
In those days of course LOTS of feedback was normal as was ridiculously high sensitivity, making for rather noisy amplifiers from Leak, Mullard et al (unlike the Quad II which refreshingly, was designed far more appropriately ~1.4V sensitivity) All of these amplifiers were power amplifiers after all and had matching preamps, so didnt need all that sensitivity . substituting the 12AX7 in the splitter with a 12AU7 would have mitigated the bandwidth issue and lowered the gain assisting with both the noise issue and providing a far more useful input sensitivity.