No, not really, I meant literally driving it into hard clipping, although lowish load certainly does not help. You can look up Langford-Smith's papers, I believe he covered both the overloading and varying load conditions... I got to go out now, be back tomorrow, talk to you then - it is definitely an interesting subject worth spending a bit more time on.
Last edited:
Respectfully, this is nonsense. With a pentode, you have three elements where NFB can be applied: the control grid, the cathode, and the screen grid. "Ultralinear" is simply a marketing fept. name for the application of lNFB to the screen. Regardless of how it's applied, lNFB has the same effects: improves linearity and reduces the effective plate resistance at the expense of voltage gain.
Each method has its pros and cons. Parallel ("Schade") lNFB increases the loading on the driver in the same way as Miller Effect. Cathode NFB requires an OPT with a balanced secondary (or multi-taps that accomplish the same ting) and Ultralinear doesn't play very well with pents that operate their screen grids at lower voltages (all deflection tubes, most RF finals, and a good many audio finals like the 6L6/807). For the latter types you can use OPTs with special tertiary windings, or supply the screens via a cathode follower or source follower.
The closed loop behavior is quite different than the open loop: that's the whole point of using NFB in the first place. UL doesn't create a "third type of tube" since the plate characteristics are not a third type. The more lNFB you apply, the more like a triode the plate characteristics become, as you are externally restoring the inherent lNFB of the triode that you removed when you added that extra grid to shield the input from the output. You can't tell from looking at the plate characteristics what form of lNFB is being used: parallel, cathode, or UL, or even that you're not looking at a triode's plate characteristics.
As for applying lNFB to actual triodes, lNFB can make a high-u triode into a low-u triode: a good reason to include it when using high-u RF finals like the 811 and other "zero bias" RF triodes.
The A Number One problems with UL have always been it doesn't work with types that have limited screen voltage specs, once the OPT is wound, the die is cast, and you're trusting that the manufacturer got it eight for the type you're using, and that the manufacturer knew what he was doing while designing that OPT. As for whether or not you want to use it, that depends. Some audio finals seem to need the extra help (6L6, 807, other 6L6-oids) and some don't (6V6-oids, 6BQ6s and probably most TV HD power finals).
Hmm... "ultra-linear" refers to the special condition where distortion is lower than pentode or triode mode. i.e. the distortion reduction is greater than would be predicted from the feedback factor alone. This makes it more unique than generic LNFB, so I wouldn't call it mere marketing. Of course, anything outside the ultra-linear null is simply distributed loading.
This notion of an optimum %UL for minimum distortion is somewhat nebulous I think. Look a post 296:
http://www.diyaudio.com/forums/tube...eople-dislike-ultralinear-30.html#post4462194
As the % UL increases toward triode, the "knee" of the curves drops down toward eventually 0 mA at full triode.
For UL operation, the primary Z of the OT gets increased (and B+ increased along with it), to keep the load line on the "knee" for maximum power out. This increase of primary Zload, combined with the decreasing Rp (as the Mu of the effective triode drops with higher % UL), makes for steadily decreasing distortion. (as expected from N Fdbk)
This would continue indefinitely to full triode, were it not for voltage breakdown limits of the tube, and the practicality of making the extreme high Z primary. Real triode operation just isn't ever optimised fully normally (infinite Z and infinite B+), so it suffers with higher distortion and lower power out than the intermediate (practical) UL settings can achieve.
Just as well anyway, because the screen grid (or a triode plate) has a different V to I power law than g1 (with its grid wire proximity effects). So there is a non-reduce-able distortion, no matter how much loop gain is available. The best, and most practical, N Fdbk is to go resistively back to the grid1 or the cathode, where it interacts (subtracts) with the input signal with the same power law. This also allows the pentode to use the lowest (knee optimised) primary Z with maximum power output. And one can include additional gain in the loop easily.
I see UL as a good way to get triode sound from a pentode in SE operation.
http://www.diyaudio.com/forums/tube...eople-dislike-ultralinear-30.html#post4462194
As the % UL increases toward triode, the "knee" of the curves drops down toward eventually 0 mA at full triode.
For UL operation, the primary Z of the OT gets increased (and B+ increased along with it), to keep the load line on the "knee" for maximum power out. This increase of primary Zload, combined with the decreasing Rp (as the Mu of the effective triode drops with higher % UL), makes for steadily decreasing distortion. (as expected from N Fdbk)
This would continue indefinitely to full triode, were it not for voltage breakdown limits of the tube, and the practicality of making the extreme high Z primary. Real triode operation just isn't ever optimised fully normally (infinite Z and infinite B+), so it suffers with higher distortion and lower power out than the intermediate (practical) UL settings can achieve.
Just as well anyway, because the screen grid (or a triode plate) has a different V to I power law than g1 (with its grid wire proximity effects). So there is a non-reduce-able distortion, no matter how much loop gain is available. The best, and most practical, N Fdbk is to go resistively back to the grid1 or the cathode, where it interacts (subtracts) with the input signal with the same power law. This also allows the pentode to use the lowest (knee optimised) primary Z with maximum power output. And one can include additional gain in the loop easily.
I see UL as a good way to get triode sound from a pentode in SE operation.
Last edited:
This is the nonsense those who dismiss UL make without the full explaination. UL as defined by Acro/Dynaco has a specific definition and screen tap point designations that create the UL benefit of drastically reduced distortion and internal impedance with practically no reduction in gain that simple FB creates. It's a different phenomenon. You can simplify the language because it's a wire loop but what says it's not "feed forward" if the screen recieves the signal first inside the tube and energizes the winding ahead of plate signal.
"the UL benefit of drastically reduced distortion and internal impedance with practically no reduction in gain that simple FB creates."
I think that last part should read "practically no reduction in power", that FB has nothing to do with. The gain is certainly reduced, and the distortion is reduced, just like N FBk usually does.
The screen signal is coming from the plate. It has no direct input signal on it.
I think that last part should read "practically no reduction in power", that FB has nothing to do with. The gain is certainly reduced, and the distortion is reduced, just like N FBk usually does.
The screen signal is coming from the plate. It has no direct input signal on it.
Everyone knows the general change in the shape of the curves as you progress towards 100% triode. You can roughly see the linearity improve with intermediate k values, although you can't see the precise distortion null from the curves by eye -you have to measure it. And there is no requirement to make the load line 'follow the knee'; not sure what you're driving at there. Here's a classic diagram -you can see the UL region is around 15% impedance tapping (39% of the turns) with this particular valve type. At the very least, UL refers to the region where distortion drops to 'very nearly triode levels' but the power output is still quite high.
Last edited:
Hafler said gain... quoted from him Dec., 1953 RE You can decide how you want to relate the gain to power but maintaining the stage gain vs. common FB was his point. Are there any specs. for dB of UL?
Yes but,
It doesn't change the fact that U/L is still pants...😀 Its a compromise as with all things, if triode is considered better than U/L what is that saying about U/L?
I still say if you haven't switched between all modes then how can you comment. If U/L is so great what's the point of a triode power tube?
The question remains is triode connection superior to U/L..and why would people prefer another mode of OP connection...would it be because U/L is inferior?
ie not the power of pentode and not the sound of triode..
However I run U/L for power...but with that strange diode..
Regards
M. Gregg
It doesn't change the fact that U/L is still pants...😀 Its a compromise as with all things, if triode is considered better than U/L what is that saying about U/L?
I still say if you haven't switched between all modes then how can you comment. If U/L is so great what's the point of a triode power tube?
The question remains is triode connection superior to U/L..and why would people prefer another mode of OP connection...would it be because U/L is inferior?
ie not the power of pentode and not the sound of triode..
However I run U/L for power...but with that strange diode..
Regards
M. Gregg
Last edited:
Surly Hifi is about sound..PA is about volume and power..
If you want power there are FET amplifiers so whats the point of U/L?
If you want SS sound from a tube amp..build a FET amp.
Is the distortion of U/L less than modern SS?
Regards
M. Gregg
Last edited:
My experience is that triode mode causes a significantly lower output impedance driving the highly reactive speaker, so the typical resonance around 60 - 100HZ in most woofers will be more tightly controlled and so more eliminated. In comparison, Ultra linear causes the amp to have a higher output Z, so the woofers resonance is less controlled and therefore more prominent.
Many guitar amps now have the ultralinear/triode switch as well (including one of mine), and one way doesn't necessarily sound better or worse, just different.
Negative feedback causes harmonic distortion products to extend out further in frequency from the fundamental, which is not good for hi-fi. But the less feedback you have, the higher the output Z of the amp becomes, so it's a tradeoff with speaker damping/control.
In the EL34 Hi-Fi tube amp I designed and built, I have 0 - 15dB of neg. feedback (adjustable) which only goes back one stage to the longtail diff/ phase splitter circuit. The front end tube (6SN7) is outside of the feedback loop. It's also got an ultralinear/triode switch. From what I've read, 12dB of negative feedback seems to be the preferred amount from a listening point of view. That's what the most expensive tube amps that I looked at appear to be using. That lowers the output Z some without causing substantial extension of HD products.
Many guitar amps now have the ultralinear/triode switch as well (including one of mine), and one way doesn't necessarily sound better or worse, just different.
Negative feedback causes harmonic distortion products to extend out further in frequency from the fundamental, which is not good for hi-fi. But the less feedback you have, the higher the output Z of the amp becomes, so it's a tradeoff with speaker damping/control.
In the EL34 Hi-Fi tube amp I designed and built, I have 0 - 15dB of neg. feedback (adjustable) which only goes back one stage to the longtail diff/ phase splitter circuit. The front end tube (6SN7) is outside of the feedback loop. It's also got an ultralinear/triode switch. From what I've read, 12dB of negative feedback seems to be the preferred amount from a listening point of view. That's what the most expensive tube amps that I looked at appear to be using. That lowers the output Z some without causing substantial extension of HD products.
Last edited:
"And there is no requirement to make the load line 'follow the knee'; not sure what you're driving at there."
The efficiency is what limits a triode from matching the power output of the pentode or UL design. The triode (with typical load lines) can not operate down to 0V (or 50V) on the plate like a pentode. So it wastes a bunch of heat for nothing. It has a limited plate dissipation budget, so output power gets reduced.
Now, looking at the UL curves Jazbo8 posted earlier versus %UL, the plate curve "knee" moves down (lower current) and to LOWER Vp as k or UL% increases. This is signaling an opportunity to actually INCREASE efficiency ABOVE pentode no less. But it requires an increasing Zpri as k increases to intersect the "knee".
(the earlier studies did not take advantage of this beyond a very limited range, leading to INCORRECT or INCOMPLETE conclusions.)
But there is a limit to how far one can take this process, since the higher Zpri makes for a hard to make OT, and the increasing B+ required to take advantage of it (using the full Pdiss of the tube) soon exceeds the tube voltage ratings. So real triodes do not have optimised (for power) OTs. But UL can take advantage of this process up to maybe 50% UL.
The increasing Zpri with increasing UL (following the "knee") will also reduce distortion since Rp is decreasing from N Fdbk and Zpri is increasing. So increasing %UL would produce monotonically decreasing distortion and increasing power output (better efficiency) if it were practical to pursue this process. Practical OT and tube V breakdown limitations however limit this to around 40 or 50% UL.
The efficiency is what limits a triode from matching the power output of the pentode or UL design. The triode (with typical load lines) can not operate down to 0V (or 50V) on the plate like a pentode. So it wastes a bunch of heat for nothing. It has a limited plate dissipation budget, so output power gets reduced.
Now, looking at the UL curves Jazbo8 posted earlier versus %UL, the plate curve "knee" moves down (lower current) and to LOWER Vp as k or UL% increases. This is signaling an opportunity to actually INCREASE efficiency ABOVE pentode no less. But it requires an increasing Zpri as k increases to intersect the "knee".
(the earlier studies did not take advantage of this beyond a very limited range, leading to INCORRECT or INCOMPLETE conclusions.)
But there is a limit to how far one can take this process, since the higher Zpri makes for a hard to make OT, and the increasing B+ required to take advantage of it (using the full Pdiss of the tube) soon exceeds the tube voltage ratings. So real triodes do not have optimised (for power) OTs. But UL can take advantage of this process up to maybe 50% UL.
The increasing Zpri with increasing UL (following the "knee") will also reduce distortion since Rp is decreasing from N Fdbk and Zpri is increasing. So increasing %UL would produce monotonically decreasing distortion and increasing power output (better efficiency) if it were practical to pursue this process. Practical OT and tube V breakdown limitations however limit this to around 40 or 50% UL.
Attachments
Last edited:
H&K didn't invent this topology. They did, however, invent the term "Ultralinear" and admitted it was for marketing purposes.
As for the rest, this is pure audiophoolery, including special appeals to magical effects: "the distortion reduction is greater than would be predicted from the feedback factor alone". This is not supported by any sort of actual research. Even if it were true, it wouldn't hold for every type of final and every OPT that includes UL tap-offs of ~43% (which has everything to do with how the primary and secondary are sectioned). Indeed, one research paper where actual tests were conducted seemed to show a point of diminishing returns at ~25% UL tap-off. If there were some sort of point where UL works its special ju-ju, then you'd require a custom OPT for the type of final you're using in your design.
Also, i see absolutely nothing here that suggests: "the distortion reduction is greater than would be predicted from the feedback factor alone" doesn't apply for the other forms of lNFB, whether parallel or cathode.
As for myself, I can't use UL since I design around finals that would go into all Chernobyl if the screens were so severely over volted. Other types simply don't need the extra help. 6BQ6GAs make mainly h3, and need just enough gNFB to take the "edge" off. I also did a design based on the 6L6/807 and simply followed O. Schade's recommendation to feed back 10% of the plate voltage back to the control grids. He knew what he was doing since that, and an extra 7.0dbv cleaned up all the pentode messes and made for some damn fine sonic performance.
I think a lot of people were/are turned off by the whole "ultralinear" concept because of its history. It was basically a "Johnny come lately" idea (re: Blumlein, Walker, and McIntosh) that somehow got patented and thereafter promoted and hyped by Kereos & Hafler until the Dynaco's ST-70 and its variants/copycats became the bestselling tube audio amplifier in history. That legacy is still with us today, attests by the continuing popularity of the UL OPTs - where you can get it with any tap you want, as long as it is 43%.😀
Just for interest page 112 transformers with three windings
https://books.google.co.uk/books?id...tiary windings in output transformers&f=false
I think its interesting to note many of the old Output types and circuits used tertiary windings..and you only see standard output types which tent to limit experimentation.
PYE Mozart is one example..the transformers were or seemed more elaborate years ago. So what if anything are we missing in todays topologies.
Its interesting to note how in the article you can create F/B using a second Tx fed from the secondary of the output Tx. ie it was a work around but of course its another way to try things..
Can you still get Tx's with tertiary windings? off the shelf...we can use the secondary but its not quite the same is it..
Does anyone know of an off the shelf Tx that could be fed from the secondary of the output Tx and used as a tertiary winding?
Regards
M. Gregg
https://books.google.co.uk/books?id...tiary windings in output transformers&f=false
I think its interesting to note many of the old Output types and circuits used tertiary windings..and you only see standard output types which tent to limit experimentation.
PYE Mozart is one example..the transformers were or seemed more elaborate years ago. So what if anything are we missing in todays topologies.
Its interesting to note how in the article you can create F/B using a second Tx fed from the secondary of the output Tx. ie it was a work around but of course its another way to try things..
Can you still get Tx's with tertiary windings? off the shelf...we can use the secondary but its not quite the same is it..
Does anyone know of an off the shelf Tx that could be fed from the secondary of the output Tx and used as a tertiary winding?
Regards
M. Gregg
Last edited:
Does anyone know,
how do tertiary windings compare to U/L feedback? Or is it just cheaper to put a tap on the output Tx?
I'm just suspicious with sales and costing..😀
NB I just noticed Majestic are making a replacement quad 2 output Tx (I have no connection its just for interest)trouble is they never give an Idea of price do they..😀
Regards
M. Gregg
how do tertiary windings compare to U/L feedback? Or is it just cheaper to put a tap on the output Tx?
I'm just suspicious with sales and costing..😀
NB I just noticed Majestic are making a replacement quad 2 output Tx (I have no connection its just for interest)trouble is they never give an Idea of price do they..😀
Regards
M. Gregg
Last edited:
That's a direct result of the success of the UL connection...🙄
Yes, you can, that's 1% vs. the 99% for UL. Plitron, Sowter, One-Electron, Tango (old stock if you can find them), and a few others do offer OPT's with tertiary windings.
Not sure exactly what you meant, perhaps you can post an example?
Tertiary windings usually are superior to the UL tap, but they also cost more to make, and that's why the manufacturers in general love the UL design, the OPT's were/are cheaper to produce, so more profit for them, why they are also "ULTRA-linear", surely that's better than just linear!😀
Last edited:
Additional windings make for higher leakage L, so are not so hot. Using an extra OT for the tertiary winding or FB winding will really mess up the phase response.
The uTracer data from a real tube, fixed to include the k x Ig2 factor, provides a super tool for investigating UL. If the data can be manipulated mathematically, one could plot load lines (from the "knee" say, tangent to the max Pdiss hyperbola) and then plot the gain (delta Vout) versus Vg1 steps. All these questions of linearity and power output versus %UL can be answered definitively. (without having to wind one turn on an OT)
One might want to also do the analysis for a beam tube as well.
A super (new?) feature for the uTracer would be to take data along a specified load line directly, and plot the gain variation.
The uTracer data from a real tube, fixed to include the k x Ig2 factor, provides a super tool for investigating UL. If the data can be manipulated mathematically, one could plot load lines (from the "knee" say, tangent to the max Pdiss hyperbola) and then plot the gain (delta Vout) versus Vg1 steps. All these questions of linearity and power output versus %UL can be answered definitively. (without having to wind one turn on an OT)
One might want to also do the analysis for a beam tube as well.
A super (new?) feature for the uTracer would be to take data along a specified load line directly, and plot the gain variation.
Last edited:
- Status
- Not open for further replies.