Hello!
I have buit a double push-pull (4x KT88) Williamson based.
Ecc82 Pre-amplifier and concertina fase splitter.
Ecc82 Text book differential amplifier driver.
The output impedance of the driver is about 10KOms.
When I only use 2 output valves, sounds better: More presence, better definition of sounding stage.
The bass are less grease and there is a subjective sensation leading to a bigger emotion felt when we listen to the records.
This effect is unproportional towards the expected because of the increasing Miller capacitance.
Does anyone know why this happens?
I have buit a double push-pull (4x KT88) Williamson based.
Ecc82 Pre-amplifier and concertina fase splitter.
Ecc82 Text book differential amplifier driver.
The output impedance of the driver is about 10KOms.
When I only use 2 output valves, sounds better: More presence, better definition of sounding stage.
The bass are less grease and there is a subjective sensation leading to a bigger emotion felt when we listen to the records.
This effect is unproportional towards the expected because of the increasing Miller capacitance.
Does anyone know why this happens?
jrarauio,
i have observed the same, on all signal levels. And so have my buddies. I do not parallel tubes or tube sections, rather i use a juicier tube.
Be it phono input, line stage, power amp.
I do not know the reason and my tube guru does not know it either. Theoretically, a triode consists of elementary triode systems wired in parallel (due to grid spacing variations and to varaitions of the electric field at the ends of the triode system). So it should not make a difference. But it does. I do not have to understand it, i simply care not to parallel tubes.
i have observed the same, on all signal levels. And so have my buddies. I do not parallel tubes or tube sections, rather i use a juicier tube.
Be it phono input, line stage, power amp.
I do not know the reason and my tube guru does not know it either. Theoretically, a triode consists of elementary triode systems wired in parallel (due to grid spacing variations and to varaitions of the electric field at the ends of the triode system). So it should not make a difference. But it does. I do not have to understand it, i simply care not to parallel tubes.
I am certainly in the belief that a single pair of devices is best.
Did you compensate for the different load requirement for a pr or a quad?
dave
Did you compensate for the different load requirement for a pr or a quad?
dave
Thank you for the tips!
No, I did not compensate the load.
In this unconpensed conditions the slew rate increase from 6,2us to 10us.
Obviuosly the bandwith and the dumping factor should decrease in this conditions.
So, here we have all the factors for a naff sound, but the final results are totaly unexpected. It is the oposite!
I remain without knowing why this happens, and this is bad for my mind!
😀
Regards
jraraujo
No, I did not compensate the load.
In this unconpensed conditions the slew rate increase from 6,2us to 10us.
Obviuosly the bandwith and the dumping factor should decrease in this conditions.
So, here we have all the factors for a naff sound, but the final results are totaly unexpected. It is the oposite!
I remain without knowing why this happens, and this is bad for my mind!
😀
Regards
jraraujo
Hi,
One other possibiltiy could be the added Miller effect the driver stage can not cope with.Higher powered triodes need to be compensated for this effect if not HF roll will occur.
Paralleled tubes :
.
The reason you probably won't like it is that the distortion characteristic has predominance for uneven order harmonics.
Try this without ( or partially bypassed ) cathode bypass caps and things look a lot better already.
In other words introduce some degenerative feedback.
On a general note : if you parallel tubes you should know your stuff and if you have a plotter ( or a software sim at least ) trace those curves with both sections paralleled.
You'll notice the difference in linearity.
In a nuttshell: // tubes aren't necessarily a bad thing.
Rgds,
Frank
One other possibiltiy could be the added Miller effect the driver stage can not cope with.Higher powered triodes need to be compensated for this effect if not HF roll will occur.
Paralleled tubes :
.
The reason you probably won't like it is that the distortion characteristic has predominance for uneven order harmonics.
Try this without ( or partially bypassed ) cathode bypass caps and things look a lot better already.
In other words introduce some degenerative feedback.
On a general note : if you parallel tubes you should know your stuff and if you have a plotter ( or a software sim at least ) trace those curves with both sections paralleled.
You'll notice the difference in linearity.
In a nuttshell: // tubes aren't necessarily a bad thing.
Rgds,
Frank
jraraujo said:
Are your output KT88s triode connected as is in original Williamson?
If so, I know the answer.
Denis.
The KT88 are conected without degenerative kathode
Thank you for your repply
No they aren't.
The KT88 have an invdividual kathode c.c.s., 87mA bias, bypassed with a 100uF capacitor.
The ccs is tunabled type for a fine DC cancelation on OPT.
The rest is more or less like williamson or better like the Leak 50.
But I've some friends that have found this same phenomenon on other topologies.
Please inform me if you know the reason why sounds better with only two valves.
Regards
🙂 🙂
Thank you for your repply
No they aren't.
The KT88 have an invdividual kathode c.c.s., 87mA bias, bypassed with a 100uF capacitor.
The ccs is tunabled type for a fine DC cancelation on OPT.
The rest is more or less like williamson or better like the Leak 50.
But I've some friends that have found this same phenomenon on other topologies.
Please inform me if you know the reason why sounds better with only two valves.
Regards
🙂 🙂
Re: The KT88 are conected without degenerative kathode
But I suppose the screen grids of KT88s are connected to their plates as in original Williamson so we see four parallel push-pull triodes.
Any push-pull triode stage with a center-tapped primary of the OPT (not like a circlotron) is prone to a specific kind of odd order distortion, which is caused by a finite mutual leakage inductance between the halves of a primary. Naturally, this mutual leakage inductance cannot be zero in any real output transformer, although it can be reduced by a proper design.
The physics behind this distortion mechanism is quite complex, so I cannot describe it just by few words, but the result is following:
The EMF induced at this mentioned mutual leakage inductance (its waveform looks like sharp spikes) simultaneously modulates the instant plate voltages of both triodes of a PP pair in such a way that the amount of odd harmonics generated by the power stage is increased. As is known, the PP circuit cancels even harmonics, while the odd ones are expected to be not influenced.
Indeed, the mentioned effect results in more odd harmonics production in comparison to parallel SE operation of the same two tubes and same plate current swing. Nothing to say this smears fine structure of the audio signal and results in an appreciable loss of resolution. What is most intriguing, this effect increases with lowering of the internal resistance of the triodes, i.e., when you put two PP pairs instead of one to the same OPT, this will result in greater resolution loss!
Historically this effect was first observed in large class AB2 PP triode modulators of AM broadcast transmitters. In those stages this effect was so pronounced that an easely observable sharp crossover step appeared in the output waveform when the modulator was fed by reasonably high frequency. Of course, nothing like is expected in the case of a class A triode PP stage, by the increased amount of odd order nonlinearity is still there. This may be a reason why SE triode amplifiers achieved so high praise.
A really good PP OPT for a triode amplifier is very difficult to design and make. Indeed, the abovedescribed distortions are unlikelyu to appear if the -3dB point of full power frequency response is somewhere at 200kHz and both the amplitude and phase responses are essentially the same from the both halves of a primary to a secondary. Such PP topologies as the Circlotron requiring no central tap are free of such distortion mechanism, but also there can be other sources of distortion in the circuits with multiple power tubes.
The better or maybe even ultimate way to introduce a local NFB to the pentode output stage is a cathode feedback (CFB) circuit as that first employed in QUAD II amplifier. This circuit is free of such distortions and possess many other great advantages.
The UL topology is somewhat less prone to these distortions than a triode connection, but has its own drawback.
The above distortion can appear in PP output stages with any bias circuit.
Regards,
Denis
jraraujo said:
But I suppose the screen grids of KT88s are connected to their plates as in original Williamson so we see four parallel push-pull triodes.
Any push-pull triode stage with a center-tapped primary of the OPT (not like a circlotron) is prone to a specific kind of odd order distortion, which is caused by a finite mutual leakage inductance between the halves of a primary. Naturally, this mutual leakage inductance cannot be zero in any real output transformer, although it can be reduced by a proper design.
The physics behind this distortion mechanism is quite complex, so I cannot describe it just by few words, but the result is following:
The EMF induced at this mentioned mutual leakage inductance (its waveform looks like sharp spikes) simultaneously modulates the instant plate voltages of both triodes of a PP pair in such a way that the amount of odd harmonics generated by the power stage is increased. As is known, the PP circuit cancels even harmonics, while the odd ones are expected to be not influenced.
Indeed, the mentioned effect results in more odd harmonics production in comparison to parallel SE operation of the same two tubes and same plate current swing. Nothing to say this smears fine structure of the audio signal and results in an appreciable loss of resolution. What is most intriguing, this effect increases with lowering of the internal resistance of the triodes, i.e., when you put two PP pairs instead of one to the same OPT, this will result in greater resolution loss!
Historically this effect was first observed in large class AB2 PP triode modulators of AM broadcast transmitters. In those stages this effect was so pronounced that an easely observable sharp crossover step appeared in the output waveform when the modulator was fed by reasonably high frequency. Of course, nothing like is expected in the case of a class A triode PP stage, by the increased amount of odd order nonlinearity is still there. This may be a reason why SE triode amplifiers achieved so high praise.
A really good PP OPT for a triode amplifier is very difficult to design and make. Indeed, the abovedescribed distortions are unlikelyu to appear if the -3dB point of full power frequency response is somewhere at 200kHz and both the amplitude and phase responses are essentially the same from the both halves of a primary to a secondary. Such PP topologies as the Circlotron requiring no central tap are free of such distortion mechanism, but also there can be other sources of distortion in the circuits with multiple power tubes.
The better or maybe even ultimate way to introduce a local NFB to the pentode output stage is a cathode feedback (CFB) circuit as that first employed in QUAD II amplifier. This circuit is free of such distortions and possess many other great advantages.
The UL topology is somewhat less prone to these distortions than a triode connection, but has its own drawback.
The above distortion can appear in PP output stages with any bias circuit.
Regards,
Denis
Denis,
Very interesting! It seems like toroidal OPT generally have high bandwidth due to the topology; how (in general) is the mutual leakage inductance of toroidal vs. standard OPT? Would toroidal OPT be desirable in a paralleled design?
Very interesting! It seems like toroidal OPT generally have high bandwidth due to the topology; how (in general) is the mutual leakage inductance of toroidal vs. standard OPT? Would toroidal OPT be desirable in a paralleled design?
tiroth said:
Certainly not! The toroidal cores are the worst ones for OPTs. They give lower leakage inductance in respect to conventional ones only if we speak of simple winding without interleaving. This is feasible for mains transformers reducing their stray field. Indeed, the properly interleaved IE- or, the better, double-C cored transformer is much better. I use only double-C cores of high-quality grain-oriented electric steel. The flat and uniform surface is the must for precise turn layout. This allows one to get virtualy any degree of suppression of both leakage inductance and shunt capacitance. It is a pity these design rules are seldom followed by transformer makers.
Another great drawback of the toroids is their DC succeptibility. An airgap is a must for a PP OPT too. I've found the natural airgap (25 um or so) present between the halves of a double-C core is just what Doctor has prescribed.
Keep It Simple!
Denis.
Hmm...I was under the impression that DC saturation was a nonissue with PP provided the design allows sufficient bias adjustment to balance the two halves of the primary. Not true?
Thanks for your input...I have seldom seen any rationale to back up OPT design.
Thanks for your input...I have seldom seen any rationale to back up OPT design.
tiroth said:
Certainly not true.
A subsonic component of plate current also appears in PP stages due to slightly asymmetrical clipping. This can be called dynamic DC offset.
The operating points of tubes of a PP pair also can drift more or less.
The good PP OPT should tolerate a DC imbalance of at least +-20% of one tube idle current.
Moreover, a small airgap linearises primary inductance.
Denis.
Hello
You can interleaving windings on toroidal opts too. With care so do not have a large winding built.
You must Interleaving so that you maximize the contact cooper area between primary and secondary.
With this you garentee a reduction of leakage indutance.
Bifilar windings reduce shunt capacitance.
On toroidal opts all magnetic path is used by the windinds, and that is the reason why you use all bebefits of grain oriented.
On conventional and C cores, the magnetic particles of the non wind magnetic path do not change poles at the same time.
The particles magnetic polarity of the non wind magnetic path has a delay in what concerns the wind magnetic path.
This is the reason why toroidal opts saturate more suddenly than the others.
Regards
You can interleaving windings on toroidal opts too. With care so do not have a large winding built.
You must Interleaving so that you maximize the contact cooper area between primary and secondary.
With this you garentee a reduction of leakage indutance.
Bifilar windings reduce shunt capacitance.
On toroidal opts all magnetic path is used by the windinds, and that is the reason why you use all bebefits of grain oriented.
On conventional and C cores, the magnetic particles of the non wind magnetic path do not change poles at the same time.
The particles magnetic polarity of the non wind magnetic path has a delay in what concerns the wind magnetic path.
This is the reason why toroidal opts saturate more suddenly than the others.
Regards
jraraujo said:
In any case it is impossible to attain the same degree of precision in turns layout with toroidals.
The C cores allow one to utilise the benefits of grain-oriented steel as toroids, but the provide flat and uniform surface for winding. In additional, they possess a natural airgap. It is this latter that makes them saturate rather gradually.
I cannot see why bifilar winding decrase shunt capacitance. It can rather increase it to surely untolerable value. In contrary, the proper interleaving technique with flat layers and rather thick insulation allows one to attain almost any ratio between leakage inductance and shunt capacitance.
And finally, the good OPT must behave equally well over quite wide range of resistive and reactive loads. This requires optimising of the above ratio.
Therefore I'm sure double-C cores are the best.
Denis.
[That is true, but you do not need Interleaving 10 layers on toroidal.
4 o 5 primary layers is enoug to reduce leakage indutance to a very low value.
Yes you have a minimum natural air gap and a uniform surface. And it is also true they saturate gradually.
This happens because the windignds do not cover all magnetic path. Like I have exposed.
The portoin of magnetic path not wound, has an intrinsec delay of particles magnetization.
That is why you van not use all benefits of G.O. silicon steel.
With Toroidal you have more conssistent lower full power band. And you have not a increase of indutive reactance at upper band.
On C cores you assist to a progressive degradation of sound at full power lower band.
Easely you can make toroidal ops with a high Quality factor with low Lsp (3 to 10 mH) for a 300 to 500 Henry Prim. Indutance
And values of 500 pF of Efecctive primary capacitante. That makes a QF = 300.000.
I have optimised tranformers in this conditions.
Please refer to "Modern High-End Valve Amplifiers based on toroidal output transformers" from Ir Memnno van der Veen
If you do not see why bifilar windings reduce shunt capacitance, Its very simple, try it like myself.
But if you need teorical expanations, please refer to "Transformer and Inductor Design Handbook" from Colonel Wm. T. McLyman. this is the "bible" of Aeronautical and Spaceships transformers tecnology.
Tehe ratio you refer is the Quality factor of a OPT, do you make it 300.000 with C-Cores?
Double C-cores, simple C-Cores and E I cores are certainly easy to do.
Regards
jraraujo
4 o 5 primary layers is enoug to reduce leakage indutance to a very low value.
Yes you have a minimum natural air gap and a uniform surface. And it is also true they saturate gradually.
This happens because the windignds do not cover all magnetic path. Like I have exposed.
The portoin of magnetic path not wound, has an intrinsec delay of particles magnetization.
That is why you van not use all benefits of G.O. silicon steel.
With Toroidal you have more conssistent lower full power band. And you have not a increase of indutive reactance at upper band.
On C cores you assist to a progressive degradation of sound at full power lower band.
Easely you can make toroidal ops with a high Quality factor with low Lsp (3 to 10 mH) for a 300 to 500 Henry Prim. Indutance
And values of 500 pF of Efecctive primary capacitante. That makes a QF = 300.000.
I have optimised tranformers in this conditions.
Please refer to "Modern High-End Valve Amplifiers based on toroidal output transformers" from Ir Memnno van der Veen
If you do not see why bifilar windings reduce shunt capacitance, Its very simple, try it like myself.
But if you need teorical expanations, please refer to "Transformer and Inductor Design Handbook" from Colonel Wm. T. McLyman. this is the "bible" of Aeronautical and Spaceships transformers tecnology.
Tehe ratio you refer is the Quality factor of a OPT, do you make it 300.000 with C-Cores?
Double C-cores, simple C-Cores and E I cores are certainly easy to do.
Regards
jraraujo
This is absolutely wrong statement from the viewpoint of physics.
Nonhing to say more.
OK. If you say so!
If have nothing to say, I do not desire arge with you.
But for respect to other people visitor of this forum, yes I have something to say:
Physics is not axiomatic or dogmatic.
Physics is phenomenologic, and due this we have theorical physics.
Physics phenomenon are based on time and space.
There are no instantaneos phenomenon of propagation of magnetic fields or other fields.
From poit "A" to point "B" there is always a "delta t" in what concerns of propagation.
This afirmation is not axiomatic or dogmatic but phenomenologic.
Due this anyone can see why magnetic path of an OPT is not magnetised at the some time.
In other words, core elementary magnets do not realign instantaneosly.
Grain Oriented cores have elementary magnets oriented on direction of magnetic path.
Let your conscious imagination see what's happen on a core when we aply a magnetic force.
Progressively al elementary magnets rotate and align in a specific "delta t".
Now imagine a the movement of magnetic particles due A.C. on a windind.
Only with all magnetic path wond you extract al beneficts of Grain Oriented cores.
In other words you have less hysteresis losses.
The theme of this forum is completely put out of the way of inicial issue: "Double push-pull versus single push-pull"
The originally description of the phenomenon, happens with all types of transformers. Toroidal or Conventional.
And until now, nobody post a satisfactory explanation for the fact.
Regards
Nonhing to say more.
OK. If you say so!
If have nothing to say, I do not desire arge with you.
But for respect to other people visitor of this forum, yes I have something to say:
Physics is not axiomatic or dogmatic.
Physics is phenomenologic, and due this we have theorical physics.
Physics phenomenon are based on time and space.
There are no instantaneos phenomenon of propagation of magnetic fields or other fields.
From poit "A" to point "B" there is always a "delta t" in what concerns of propagation.
This afirmation is not axiomatic or dogmatic but phenomenologic.
Due this anyone can see why magnetic path of an OPT is not magnetised at the some time.
In other words, core elementary magnets do not realign instantaneosly.
Grain Oriented cores have elementary magnets oriented on direction of magnetic path.
Let your conscious imagination see what's happen on a core when we aply a magnetic force.
Progressively al elementary magnets rotate and align in a specific "delta t".
Now imagine a the movement of magnetic particles due A.C. on a windind.
Only with all magnetic path wond you extract al beneficts of Grain Oriented cores.
In other words you have less hysteresis losses.
The theme of this forum is completely put out of the way of inicial issue: "Double push-pull versus single push-pull"
The originally description of the phenomenon, happens with all types of transformers. Toroidal or Conventional.
And until now, nobody post a satisfactory explanation for the fact.
Regards
Hi there,
I would like to put my limited knowledge 2 cents in here, if it would help:
All other things being equal, could it merely be the difference in how two tubes see the primary impedance versus four? I mean, if two tubes need to optimally have a primary impedance of 6K, wouldn't four need a primary impedance of 3K to compensate for the paralleled impedance of the tubes?
Here is why I say this: I have played with just a single pair of pentodes. I connected them straight pentode and UL. I have used 8K primary and 6k6 primary. I find that the 6k6 primary in UL mode sounded more musical, broader soundstage, and sweeter. UL sounded this way (more musical, etc) versus pentode. But UL sounded best with the 6k6 primary impedance. From what I understood about UL, the plate resistance drops. So I chose 6k6 right off, but had the higher impedance ones just to try.
Now, granted, you are actually increasing total plate resistance when you pull out two of the tubes. Which is why the sound changes, IMHO. The single pair may be better matched to the transformers primary impedance.
There should be a difference going from parallel to non parallel, due to this difference in impedance. It is just as when (I have read reviews where they do this) one uses the other taps than the required one for their speakers. " The sound seemed more restricted when I used the 16 ohm taps than the 8 ohm taps..." and the like. This is because, as I am sure you know, the reflected impedance to the output tubes changes, changing the characteristics of the tubes. I would suspect the reverse is true with the tubes and impedance through the transformer.
So if you are using a certain value primary impedance, may I suggest using half the value when paralleling? The impedance curve might make all the difference in sound quality, because then the tubes will be performing within their best plate curve parameters.
Just a thought.
Gabe
I would like to put my limited knowledge 2 cents in here, if it would help:
All other things being equal, could it merely be the difference in how two tubes see the primary impedance versus four? I mean, if two tubes need to optimally have a primary impedance of 6K, wouldn't four need a primary impedance of 3K to compensate for the paralleled impedance of the tubes?
Here is why I say this: I have played with just a single pair of pentodes. I connected them straight pentode and UL. I have used 8K primary and 6k6 primary. I find that the 6k6 primary in UL mode sounded more musical, broader soundstage, and sweeter. UL sounded this way (more musical, etc) versus pentode. But UL sounded best with the 6k6 primary impedance. From what I understood about UL, the plate resistance drops. So I chose 6k6 right off, but had the higher impedance ones just to try.
Now, granted, you are actually increasing total plate resistance when you pull out two of the tubes. Which is why the sound changes, IMHO. The single pair may be better matched to the transformers primary impedance.
There should be a difference going from parallel to non parallel, due to this difference in impedance. It is just as when (I have read reviews where they do this) one uses the other taps than the required one for their speakers. " The sound seemed more restricted when I used the 16 ohm taps than the 8 ohm taps..." and the like. This is because, as I am sure you know, the reflected impedance to the output tubes changes, changing the characteristics of the tubes. I would suspect the reverse is true with the tubes and impedance through the transformer.
So if you are using a certain value primary impedance, may I suggest using half the value when paralleling? The impedance curve might make all the difference in sound quality, because then the tubes will be performing within their best plate curve parameters.
Just a thought.
Gabe
Hi
Thanks for your post.
I'm using a 2K to 5 Ohms U.Linear transformer.
The four KT88 see 2K Anode to Anode if load = 5 Ohms; or 3200 Ohms if load = 8 Ohms; or 1600 Ohms if load = 4 Ohms.
If my theory is correct, 8 Ohms speakers have 4-5 Ohms at low frequencies.
There also a large amount of feedback (12dB).
I think that values are all right for four KT88 and not correct for only a pair of the same tubes. Don't you agree?
Regards
Thanks for your post.
I'm using a 2K to 5 Ohms U.Linear transformer.
The four KT88 see 2K Anode to Anode if load = 5 Ohms; or 3200 Ohms if load = 8 Ohms; or 1600 Ohms if load = 4 Ohms.
If my theory is correct, 8 Ohms speakers have 4-5 Ohms at low frequencies.
There also a large amount of feedback (12dB).
I think that values are all right for four KT88 and not correct for only a pair of the same tubes. Don't you agree?
Regards
For a single pair, 4K is recommended. So... you have it close.
Perhaps going for a smaller primary? Like 1500?
At this point I would be guessing.
Gabe
Perhaps going for a smaller primary? Like 1500?
At this point I would be guessing.
Gabe
- Status
- Not open for further replies.