Comparing OPS.

[seppoa]

Also NFB is of NO use if the output elements are biased so that there is a portion with no current thru them. This portion without current is equal to zero amplificatin, and as there has to be amplification for NFB to function, the distortion from the crossover is much more pronounced with NFB, compared to no NFB. This is the reason to much of the bad sounding distortion caused by crossover distortion, together with the nature of the distortion. It is very interesting to feed a amplifier with a pure sinus signal around 1 kHz, and varying the bias, The change in sound around a certain point is very hearable, in fact the change is chocking when you hear the difference the first time 🙂

 

[seppoa]

UL is in sort a mix between penthode and triode connection, BUT more important is the feedback to the screen, that makes UL giving more power, nearly as much as pentode coupling but ALSO at the same time the low distortion and low output impedance of triode coupling, so imagine the best of two worlds 🙂

 

[seppoa]

Nearly ALL OTs have a UL tap. The most preferred tap is at 43%, but this is not a figure carved in stone. Lundahl transformers have the tap at 33%, and the measured figures in the link given shows impressive results, in my opinion.
In fact I am nearly finishing an amp made to this design, albeit with other iron...

 

[seppoa]

And last, here an exerpt from wikipedia about UL:
By judicious choice of the screen-grid percentage-tap, the benefits of both triode and pentode vacuum-tubes can be realised. Over a very narrow range of percentage-tapping, distortion is found to fall to an unusually low value—sometimes less than for either triode or pentode operation[2]—while power efficiency is only slightly reduced compared with full pentode operation. The optimum percentage-tap to achieve ultra-linear operation depends mainly on the type of valve used; a commonly seen percentage is 43% (of the number of transformer primary turns on the plate-circuit) which applies to the KT88, although many other valve types have optimum values close to this. A value of 20% was recommended for 6V6GTs. Mullard circuits such as the 5-20 also used 20% distributed loading (but did not achieve ultra-linear operation), while LEAK amplifiers used 50%).

 

[Johan Potgieter]

Am I correct? So I should buy OT with UL tap? If I get UL tap, should I get 70% or over?

Your points are all correct, but one must not look at this in a linear way. '70%' towards triode does not mean that 70% triode characterisitcs and 30% pentode characteristics exist. The advantage lies in that proximity to favourable triode-operation is reached at some screen tap long before favourable pentode advantages diminishes. (See the KT88 figure referred to below.)

I did not find a comparison between different screen taps in the Byrith-article. A better read is found in ultra-linear It is long, and the previously referred to KT88 characteristic graphs are found about 75% down (no figure number; just before fig. 3). Another good reference given there is the articles by Langford-Smith round about 1955.

It will be seen that the behaviour of beam tubes is rather different to that of classic pentodes, as is the optimum screen tap for different tubes (this is pointed out in the Langford-Smith papers). Note: One must keep in mind that mostly the tap ratio is given as a voltage one (i.e. percentage of transformer primary turns), whereas Langford-Smith and others refer to an impedance ratio. This is sometimes not clear, leading to mis-interpretation of ratios.

Also, why you loose power if you connect the pentode in triode mode? Voltage is the same, PT is the same, why the power is so much lower?

Because the available plate voltage swing for triodes is lower than for pentodes. Perhaps you got used to the fact that transistors can mostly 'swing' through >90% of available rail voltage. This is not nearly so for triodes.

P.S: OT, but while it has been touched on: I cannot agree that EL34s are a good replacement in the Quad II. Their characteristics are quite different from KT66, apart from a higher heater current known to have overloaded the already skimpy power transformer. The 6L6GC or 5881 is within <10% of the KT66 characteristics and would be a proper replacement.

 

[Johan Potgieter]

/wiki/Mullard"]Mullard[/URL] circuits such as the 5-20 also used 20% distributed loading (but did not achieve ultra-linear operation).

In the references I have, 43% is given for the 5-20 design, perhaps a different article to yours. But I do not understand why you say 20% does not achieve UL operation? The 'optimal' %-screen tap is not a sharply defined ratio (you in fact said as much in your post #23) but can lie over a region, depending on what compromise between triode and pentode characteristics one desires.

Marginal note: Not to deviate too much from the topic, but the term 'ultra-linear' is somewhat of a misnomer: I prefer 'distributed-load'. Nothing much is particularly 'more linear'. Simply, the best of both pentode and triode operation is achieved. (See the article 'Amplifiers and Superlatives' by Williamson and Walker.)

 

[Alan0354]

Nearly ALL OTs have a UL tap. The most preferred tap is at 43%, but this is not a figure carved in stone. Lundahl transformers have the tap at 33%, and the measured figures in the link given shows impressive results, in my opinion.
In fact I am nearly finishing an amp made to this design, albeit with other iron...

Thanks for your detail replies. I have been reading UL, that's the reason I have not come back here. I am interested in using EL34. So a 43% tap is what I need?

From reading some articles, since the screen grid can control the current of the tube just like the control grid. In UL connection, the voltage driving the screen grid is % of the plate voltage and is in phase with the plate voltage. Which means the screen voltage is OPPOSITE phase with the input signal at the control grid. This is almost like a negative feedback and it helps linearity. Is that true?

What link are you referring to about the 33%. I went back to page 3 and look at the link you provide of Lundahl, I did not see it.

Is Lundahl the best on the market?What OT are you using?

Thanks

 

[DF96]

If impedance is not the reason, why is valve has less crossover distortion?

See post 6. Output valves need volts of signal to make any big change; a BJT needs just a few mV. Hence SS crossover distortion tends to be spiky and hard for the NFB to correct; valve crossover is smoother and easier to correct - and not too bad even when uncorrected.

 

[Alan0354]

See post 6. Output valves need volts of signal to make any big change; a BJT needs just a few mV. Hence SS crossover distortion tends to be spiky and hard for the NFB to correct; valve crossover is smoother and easier to correct - and not too bad even when uncorrected.

Thanks

Do you have any link to explain this a little bit more? I did googled about tube crossover distortion and can't find anything.

 

[DF96]

Just compare the transconductance of a valve with a BJT, both at typical output stage currents. 5-20mA/V at 20-50mA vs. Ic/26mV (say, 1A/V at 26mA). Then remember that gm varies slowly with current for a valve, and linearly with current for a BJT.

 

[Alan0354]

If that is the case, MOSFET output stage have advantage over BJT as you can say the same thing about MOSFET? Cordell even have graphs on the output voltage at the crossover region of both BJT ande MOSFET. MOSFET definitely have a smoother transition than a kink of the BJT.

For listening, what distortion is more noticeable and objectionable? I know odd harmonics are not good for audiophile amps. Cordell kind of implies crossover distortion is not pleasant to hear and is hard to get rid of using NFB.

 

[Johan Potgieter]

If I might chime in again:

Firstly in comparing valves with BJTs one must look at pentodes (valves), not triodes (not that circuits fot the latter cannot have cross-over distortion).

Then I would shift the accent slightly for BJTs, from voltage input to current input behaviour, which is actually what 'feeds' a BJT (but not disagreeing with DF96's explanation). As intimated in the remark regarding the gm, BJTs simply react 'sharper' to changes than valves. And in the cross-over region design cannot be sloppy as higher order harmonic products pop up very quickly with non-linearity and as said before, minute quantities and combinations of those are very persistent 'garden weeds'.*

NFB cannot cure cross-over distortion; in fact it makes any change in gm more rapid, assisting in the generation of higher order (odd) harmonics, although visually (e.g. on a scope) the impression is given of lower distortion. (Basics: First design a proper amplifier before the application of NFB or whatever other tricks!) With present knowledge, however, cross-over distortion should no longer be a factor in amplifier design - such is proved by a number of proper designs, both for valves and BJTs. Low output spectrum analyses for both topologies confirm that. (The works of Douglas Self and Bob Cordell among many more have covered that.)

*For those not in the know, high-order odd harmonics of 7th - 11th order and their combinations, can 'sound' so strident that amplitudes of those even at the threshold-of-hearing and below can be perceptable. Our brains do not only react to phenomena which are consciously audible!

 

[Alan0354]

Hi Johan, Please post your thoughts. I am just very slow in digesting the materials. I still yet to go over your post. I need to read before I post back.

I don't want to agree or disagree with the assertion that tube go through a more graduate transition compare to BJT. The reason is because the transition is respect to the input ( primary) side. Sure you need a few volts to go through the transition.....BUT remember the output of the tube has to go through the OT that STEP DOWN. So the transition with respect to the speaker is a lot smaller and might not be any better than the BJT. I still don't understand, I still need to read more.

Could it be other reason why tube is better in regarding to crossover. Bottom line, I want to compare between SS and tube. If the tube really is superior in the crossover department, there might be a scientific reason tube is better......rather than just say people holding onto the old believes.

On somewhat unrelated subject. I am working on a clean booster for guitar. I am comparing one with only opamps vs one with a simple JFET common source first stage. The later stage is exactly the same. I can tell you there is a certain "organic" sound ( 2H maybe ) about the JFET frontend that opamp doesn't have. I still question whether no distortion is everything in audiophile.

Thanks

 

[smoking-amp]

Unfortunately there don't seem to be any "wingspread" type gm plots around for P-P tube circuits (as Cordell and Self have for BJTs and Mosfets). This is a major major deficiency in the tube amplifier literature. It is even difficult to find simple gm plots for power tubes. Fortunately some of the signal type tubes have gm plots available. If you look at the 6JC6 datasheet, page 5, you can see a single ended gm plot.

For P-P you would take two of these curves (one of them flipped around) in various amounts of overlap (bias voltage controlled) and then sum them to get the gm "wingspread" for crossover. You want a constant gm sum for the full signal range of course. The tube (SE) gm plots look similar to Mosfet gm plots except for the "tail" near cutoff. This "tail" helps some for class AB, but is clearly not a cure-all for crossover distortion.

A real "Blameless" HiFi tube amp would include some gm compensation control (aka, differential current feedback) or Error Correction (EC) to fix this. None such exist in the classic tube amp designs. (even though EC was actually invented initially many years ago for tube amps. Then re-invented for SS amps later.)

Most class AB tube amps just push the output idle currents up enough so that they are mostly in "pseudo class A" for moderate sound levels. (Making for a flat bottomed "V" gm wingspread plot.)

http://tubedata.milbert.com/sheets/135/6/6JC6A.pdf

 

[smoking-amp]

The presence of the gm "tail" for tubes does at least avoid sudden sharp transitions in the gm wing curve (which do appear for BJTs and Mosfets). So global feedback is considerably more effective in cleaning that up. (ie, lower order crossover distortions occur in the tube case)

Local feedback to a class A differential driver stage can also be used to fix the crossover dist., so OT limitations are not so limiting. Probably why the respected Citation II Amp. used them.

 

[Alan0354]

The presence of the gm "tail" for tubes does at least avoid sudden sharp transitions in the gm wing curve (which do appear for BJTs and Mosfets). So global feedback is considerably more effective in cleaning that up. (ie, lower order crossover distortions occur in the tube case)

Local feedback to a class A differential driver stage can also be used to fix the crossover dist., so OT limitations are not so limiting. Probably why the respected Citation II Amp. used them.

I understand the crossover is not as sharp with tubes at the input of the power tubes. But the OT is step down, if you refer the crossover to the speaker, the crossover might become sharper because the signal at the input of the power tubes are likely to be higher amplitude than at the speaker. So the crossover region is being compressed due to the step down of voltage.

But I think you are right that the crossover is still going to be more gentle than the BJT. How does that compare to MOSFET?

Also, it crossover distortion more noticeable than harmonic distortion to the ear?

 

[Alan0354]

Y

I did not find a comparison between different screen taps in the Byrith-article. A better read is found in ultra-linear It is long, and the previously referred to KT88 characteristic graphs are found about 75% down (no figure number; just before fig. 3). Another good reference given there is the articles by Langford-Smith round about 1955.

Thanks
I finished reading the article, it's a good article, a little long winded at the beginning. But it's interesting to learn a little of the history.

 

[scopeboy]

I can tell you there is a certain "organic" sound ( 2H maybe ) about the JFET frontend that opamp doesn't have. I still question whether no distortion is everything in audiophile.

This question is valid and there is no logical correct answer to it, it is a matter of belief and personal taste. There are some studies showing that some listeners preferred a small amount of low order harmonic distortion to no distortion.

But do bear in mind that adding harmonic distortion to a single instrument in the studio is quite different to distorting the whole mix in your home. Hendrix's guitar sounded great through a Fuzz Face, octave pedal and cranked Marshall stack, but I wouldn't want to listen to the whole album through that system. 🙂

 

[DF96]

If that is the case, MOSFET output stage have advantage over BJT as you can say the same thing about MOSFET?

MOSFETs have a smoother characteristic than BJT. The big snag with MOSFETs is that you can't buy any two (same or complementary) which have the same characteristics, whereas (to a first approximation) all BJT are exactly the same.

 

[SY]

MOSFETs have a smoother characteristic than BJT. The big snag with MOSFETs is that you can't buy any two (same or complementary) which have the same characteristics, whereas (to a first approximation) all BJT are exactly the same.

The other big snag is the lower gm compared to bipolars. Output stages are very likely to be emitter/source followers.