• WARNING: Tube/Valve amplifiers use potentially LETHAL HIGH VOLTAGES.
    Building, troubleshooting and testing of these amplifiers should only be
    performed by someone who is thoroughly familiar with
    the safety precautions around high voltages.

To vary a signal pentode to triode in stages

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Has anyone tried to make a circuit that varied a small signal pentode like EF86 or 6SJ7 to points between full triode and full pentode operation ? I would happily accept that local feedback is the better route and views on that . The application is for the input of any amplifier where loop feedback is used ( no specific amp ) . I have in the past removed the feedback loop from the output transformer and strapped G2 to the anode of the input pentode valve with good results ( triode ) . Having a distortion analyzer these days it would be nice to try options between full pentode and full triode along with less loop feedback . I suspect that this is a little more complicated than just a potential divider to G2 and there are risks to the tube . I note most circuits show G2 connected to HT and a capacitor to ground . Simply changing this resistor seems too simple ?

Some one must have asked this question before ? Any help greatly appreciated .
 
Problem is that the screen grid draws current and is not a linear resistance either (under most circumstances) versus signal or control adjustment. For a small signal pentode, operating over a limited range, the variation of resistance may be tolerable for a pot control as these schemes use. For driver or output stages you likely will need to buffer the screen with a Mosfet source follower. A series grid damper resistor and an RF (MHz, so tiny C) cap to ground or cathode may be needed to prevent parasitic oscillation.
 
Thanks to everyone who replied . The Akido is fascinating , I will have to find some time to study that in detail . I had also thought of a transformer type of UL feedback which could be used with interstage transformers . A little unlikely as it seems against the religion to do that . However my joke to a freind that EL 84 would drive a 300B correctly seems to have been done and supported . Glad to see some people think outside the box . Forgive my drawing . It is simply how Quad configured it's preamp using EF 86 . If something this simple has any merit the next question is can the anode coupling capacitor be common to both input and EL 34 ? The Blencowe's example is very useful . Also , I saw someone running a 6SJ7 at 127 V / 4mA anode and 1.1 ma g2 . That seems twice it's rating ?

AzFIu.jpg
 
If anyone returns to this thread I found a very early description of UL feedback which although probably a bit optimistic is very simple . It shows pentode and triode curves and states that they bend in opposite direction . It then shows perfectly straight curves and calls them Ultra Linear . Statements of the bleeding obvious one might say . The author was convinced that this exact point was inside every pentode one felt . I did find one example of an amplifier using EF86 with an UL intestage transformer . I found a few guitar amps with EF86 and Morph control . That is interesting as one would think a simple pentode / triode switch good enough for them , obviously not ! For the duties of a perfect UL device I would say EF184 . It is said to mimic ECC81 with greater linearity in triode . I said to my late freind Sidney Smith " inside every good triode there is a good pentode trying to get out " . He nearly crashed the car . He then spent some time with me showing triode curves . I haven't looked much at valves since his death which is obvious from my incomplete drawing . Sidney did agree and we talked about UL feedback although we both thought it a concept rather than workable ( Sid's amps had it and triode switchable ) . I have recently changed my view of UL . In SE amps it sounds rather good . Perhaps the air gap helps ? Is it only global feedback that spoils it ? Sorry about my drawing . I had rather hoped someone would show me how to do it properly . If drawn as a Leak pre-amp where feedback is taken from the output coupling cap via a resistor to g1 and a resistor to signal in ( shunt feedback ) it looks right . I feel any triode you want is available from any pentode using simple shunt feedback . I guess that perfect curve is in there somewhere ? As the capacitor is there anyway I see no extra mechanism for sound quality degradation ? I could even imagine in some conditions feed forward would happen and might be useful ? If it works who cares ? Returning to EF 184 I wondered if the Quad 11 power design using it and KT 88 might work out rather well with a bit of a redesign ? To anyone who looks at this do read all of the Akido .
 
To a first approximation a pentode will give you whatever distortion is set by signal level and bias point, due to the grid-cathode behaviour. It doesn't really notice the anode load, as g2 screens this.

To a first approximation a triode can give you very low distortion, provided that you give it a sufficiently high anode load. This is because of feedback from the anode, which can exactly counteract the non-linearity from the grid.

What UL provides is lowish distortion for a much lower load than would normally be needed for triode linearity. Hence UL is useful for output stages. It is less useful for small signal stages, as here it is usually easy enough to provide a high anode load and get triode linearity. In UL arrangements where g2 current contributes to the output then it is possible that partition noise will be reduced. This is less useful than it sounds because most valve audio noise is 1/f, not partition.

A useful generalisation is that if an apparently good circuit idea has not been used down the decades then it probably means that it offers no real advantage over simpler circuits. Our forerunners did know what they were doing!
 
I absolutely agree with that . I did feel the the writer of the UL explanation had hit upon a point we all of us miss . He said neither the pentode nor triode curves are ideal . Inside the valve there will always be a better curve if pentode to start with ( perhaps ) . I said to Sid that triodes are linear due to internal feedback . It is a defect which we rather like . By experimenting we can have linear or not so linear triodes if a designer of valves . Sid rather preferred EL34's strapped as triodes compared with 300B's . My friend Paul likes his TR34's as I call them and not 300 B . He likes PX25 which I think is a bit more pentode like . All very fascinating . As you sayfor an input stage probably use triodes . This curiosity of mine came about because after Sid's death I was asked to help convert a Leak TL12 ( not plus ) for a mutual friend . I had never done this in my life and had to guess some of it . I also felt it a bit wrong to alter a classic design . However it was said to be transformed . In both cases all was converted to triode . Many years ago Jean Hiraga who I long to meet one day pointed out that linear reducing harmonics is what a zero distrotion amp must have . Not absolute zero , just the reduction in harmonics on an exponential scale is more correctly a perceived zero . OK , we all know that now . Having ways to get there is useful . Ricardo Kron mistook me for someone who owed him money . I wasn't bothered and we got on rather well . Sitting together and having never met him before he says dryly " surely transistors are better " ? to which I replied " yes " . He smiled and relaxed " that's what I thought .......you know , I once made a dull emission 300B , I couldn't give them away . It was the best valve ever made " he said . Alas Ricardo is not with us now . The PX 25 I refer to is one of his . I would say EL 84 or 34 with a UL transformer might drive 300B well . I am not a fan of 300B . However I always suspected it needed more drive .
 
BTW DF96 . I/f noise in pentodes and not partition noise . That is very significant and logical . 1/f noise was first talked about by Tim de Paravacini when I heard of it . Specifically in transistors . I saw good EF 86's quoted as about 4 nV per root Hz in triode ( like a BC 108 I guess ) . If I understand correctly valve microphones use pentodes as pentodes .

To Smoking amp . I appreciate your description . In all of this the complexity gets out of hand .

I love valves because with just two we can have a proper power amplifier . Sometimes with only one capacitor in the forward signal path ( if very brave we even loose that ) . My friend is running a Fostex full range unit with just such an amp . He uses a 300 watt class D amp to drive the old Leak 12 inch Sandwich unit to give the lower octaves . It has an adapted car stereo active crossover . The Fostex only shows it's point source abilities when driven by a nice SE amplifier . The car stereo crossover has been the thing we have found most transparent , it cost pennies . It drives long cables very well . Many DAC's have failed to deliver when listened to through that crossover . Eventually the crossover could be replaced by carefully designed output transformers and passive LF bass filter . There would still be the need for a buffer amp .

My thoughts for that perfect power amp would be EF 184 and EL 36 . Less than 1% second harmonic at full power . Likelihood is KT 88 will be used . Most likely is that none will be run as absolute triodes . I suspect it will have many small feedback loops . I doubt it will have global feedback .

Thank you everyone who has replied . Hope someone keeps this going . I think it is a deeply interesting subject . In 1927 Philips gave use this chance , they lost out commersial due to pushing of the beam tetrode . The pentode is almost a box of crayons if we understand it correctly .
 
From reading so far I get the impression that you are not interested in varying over the full range from Pentode to Triode, but rather the ability to vary over a wide enough range to be able to select the optimal operating point for linear operation.

To that end, some variation of the mentioned circuits should work satisfactorily.
 
I think unintentionally we have arrived at 3 states of operation . With a hypothetical minimum distortion in a fictional UL feedback arrangement for small and medium sized valves ( EF 86 EF 184 EL84 perhaps ) . If you are saying all points of operation between would be better ? Then I absolutely agree as it might find the exact sweet spot a design needs . I did get the impression reading early UL for power pentodes that we all might have lost the point . It seemed if I understood correctly that a power valve will produce it's lowest distortion if UL . This has somehow become , almost as low as triode and almost as powerfull as pentode as we now see it . If correct and not a misunderstanding by that writer ( forgive him it was early days ) there is mileage in this we never found . I never remember a UL power amp having less distrotion than when in triode . However I never had the transformer reconfigured for the same maximum output power as triode . Returning to your question . As far I can see we can have any pentode to triode curve we want if using simple shunt feedback . It does no great harm and is only doing what a triode internally does . Admittedly a triode is the fastest feedback possible . The idea we could play with the pentode curves by altering something to do with g2 seems unlikely to work very well . UL feedback is a bit like turbo chargers .It all seems very wrong yet it sort of works .
 
There may be more to the UL optimisation than just the % feedback. The % UL is really just varying the effective Mu of the resultant triode. (the pentode end is just a high Mu triode). This plays off power versus distortion for a fixed load as the effective Rp varies from the UL modded Mu.
Mu_eff = gm1/(UL% x gm2 + 1/rp)
Rp_eff = Mu_eff/gm1 (I think?)

Consider the pure triode with its heavy plate feedback. It would linearize the 3/2 power emission to Mu voltage gain if the load Z were sufficiently high. But it's varying Rp sabotages this.

Now consider the pentode, if the screen were operated at a constant % V (both AC and DC) of the plate, it would intercept a constant % of the plate current at all signal levels by symmetry (geometrical construction determines the % interception, the scaled tracking voltages determine the level of current that is being partitioned).

That would mean that the screen would be a constant % impedance of the plate Rp (actually by a higher ratio). But since the plate Rp varies with signal, we don't have a contant screen Rscrn yet. Now consider using a feedback network from the plate back to the screen that has a similar ratio of effective Rfdbk to Rscrn as the plate Rload is to the tube's effective Rp. (not just the pentode Rp any more, but the UL effective Rp which is much lower) As the Rp drops versus the Rload with signal, so does the Rscrn versus the Rfdbk with signal. So Vscrn drops disproportionately with signal as Rload drops. Could this compensate for Rload variation? Eliminate loading distortion? As the plate Rp drops, the screen V ratio drops to compensate. It's doing the right thing at least. Remember, both AC and DC must be scaled down by the same % here. A pot or R network does this conveniently.

Some math/formula solving for 3/2 power will be required to see if this is working correctly. Unfortunately this won't compensate for g1 island effect distortion, but maybe the ratios can be tweeked off some to improve that too over some range.

Note:
This scheme could be working for the linearization of small signal stage pentodes, but would lower the power output unacceptably for power stages due to the % DC drop required on the screen for tracking, unless rather high plate voltages were used, or low screen V Sweeps were used.
 
Last edited:
"The pentode end is just a high mu triode " is exactly how I see it . David Hafler put it very simply and although slightly to one side of your point reinforces it . Hafler says for UL to work current is involved and that resistor chains do not provide similar conditions . I was saying to myself what a shame we don't have the ability to do this with transistors . Then maybe we do with a cascode ? When I have time I will look at this with real life devices . No transformer's , that's for others . I think the most useful thing is to have found an interesting device ( EF184 ) which looks to me to be many different things especailly where high current is required . I suppose UL is a tracking triode connection that moves almost as bootstraps do . It has time lag and that must be where it looses out . As said a SE UL seems to work well . I suspect that the transformer air gap helps ( anyone want to say , hysteresis and constant DC bias to magnetic circuit perhaps ) ? Thank you for a journey of discovery . I have a lot of reading to do . It's been 37 years since I was taught any of this . The teacher failed to say about depletion and enhancement , however he did teach Kirchhoff . He tried to teach Fermi probability gradients without an idea why , at least he was honest . The next year never got valves at all . It's far more challenging than transistors ( forgetting JFET's that is , they are challenging ) . It's almost a religion and yet a science . I should clarify a point . I have had extreme prejudice against UL in the past .
 
The usual UL output mode allows the plate voltage to drop below the screen voltage, causing noticeable screen current distortion. The OT xfmr tap helps ameliorate that problem by making some use of the screen current at least. I suspect the "optimum" UL tapping is a result of playing off the screen current distortion against the 3/2 power distortion of the pentode mode. They are in opposition. Also the effective Rp is being lowered in UL mode, which lowers loading distortion. Power versus distortion tradeoffs as well. Unfortunately, the pentode 3/2 power distortion (variable Rp and emission) is current related and the screen current distortion is voltage related. A resistive load puts them in phase to cancel. A reactive load causes them to go out of phase and not cancel.

By scaling BOTH the DC and AC %UL, as I mentioned in the above scheme, the screen current distortion factor is eliminated. (constant fraction of plate current diverted to the screen, plate V always above the screen V by a fixed ratio) By using the same R loading distortion factor in the screen feedback, as the output already suffers from, it should generate a compensating correction of the plate loading distortion, if adjusted correctly. (hopefully anyway, that's the idea at least)

Of course, this only works for a resistive load also. But if the load is a fairly constant reactance, then a similar reactive screen feedback network could be used. (which brings up the interesting idea of phase compensating standard UL too. Either scheme would want a speaker reactance model in the screen feedback path.)
 
Last edited:
That is very interesting . I must put that on a scope and look at it . I will try a real speaker ( something inexpensive ) and the 8R 2uF . I will be looking at the UL tap ( hopefully it stays stable ) . Hafler stated clipping distortion is superior in UL . I never found that . One thing the teacher said which is interesting " you audio people talk about things in the signal path , you obviously don't know Kirchhoff current laws , everything is in the signal path " . How true . As the UL tap has a series resistor I will look in the current mode also . As anyone will guess it is not always possible to measure without influencing results . I am cautions of current mode . in a trans-conductance transistor amplifier the voltage waveform will seem to show gross distortion . However at the collector of the transistor we see nothing . Inserting a small resistor in the transconductance path allows us to see it was a ghost in the machine . However it never seems completely satisfactory as evidence . For what it's worth I don't like transconducatce amplifiers and still operate a 1: 5 ratio of impedance ( or the Williamson 1 to 3 if forced ) . If you don't mind a deviation that is an interesting point . A British amplifier of the 1970's had an output resistance of 2K4 feeding the next stage . That stage the voltage amplifier stage ( VAS ) was at 8 mA and Beta of 100 . The input impedance of the VAS 312 ohms .The emitter sat on a diode at 0.6V . The designer must have thought it is like a 75 ohms resistor which would offer 7K8 input impedance . Replacing the diode with that resistor transformed that amplifier ( measured IM distortion reduced ) . Yet DC conditions remained the same . I suspect the ghosts we find are not ghost and why not eliminate them . That amplifier had the usual bootstrapped constant current source to the collector which works a treat . I notice the ratio of triode output inpedance to transformer is about the same ratio as an optimum . Probably a coincidence ?
 
smoking-amp said:
the pentode end is just a high Mu triode)
No, not really. The pentode is more like a triode in which the functions of majority current collection (anode) and voltage feedback (g2) have been separated. You don't get low distortion by giving a pentode an infinite anode load, as the distortion is set by the grid-cathode behaviour. Remember, pentode curves go the opposite way - increasing mu will not make a triode curve like a pentode.
 
That's a very good way to put it and logical . I was debating something . Supposing we could easily make 100 % anode to g1 feedback on a pentode some will say we don't get exactly what we get from joining g2 to the anode . Could I be incredibly simplistic and say that's because g2 is closer to the anode ? I can see an attractive argument for using a high mu triode in parallel with a pentode . As it's a reasonable idea I am sure it's been done . I am glad you said about high mu triodes . Without looking at any curves I assumed them to look more like pentodes . My friend likes his triode connected EL 34 and PX 25 triodes . I always assumed because ra was similar . Apparently sending signal to g2 of EL 34 makes a low mu triode more like 300 B . Taking my last point an EL34 triode strapped and a EL 34 pentode might be getting somewhere in paired parallel . Might we have just described UL feedback ? The difference is no time lag if done this way ?
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.