| ray_moth |
I have sometimes read that people who have tried ultralinear PP "didn't like it" and preferred triode PP operation, despite the loss of 50% of the power. I've even read cases where pentode mode sounded better than UL. However, I've never read any explanation for these findings.
If we are to believe the copious material has been published in support of UL since it became popular, it should be the "best of both worlds", in terms of distortion, power, output load flexibility and damping. Does anyone know any reasons for avoiding UL (assuming the OPT used is good enough for the job)? |
|
|
| aletheian |
| i can't say that i personally 'don't like it', but I have never gotten the results the the published data sheet circuits suggest. Maybe the topology is more sensitive to other circuit variables than other topologies.... dunno, I am just a tinkerer. i have done my best to read all the associated literature and ancient articles, but despite all the published info and internet info from other builders, a catfight seems to erupt whenever the subject is brought up. |
|
|
| Eli Duttman |
Ray,
To squeeze the absolute maximum out of ultralinear topology, a separate screen grid winding connected to a dedicated/regulated supply is needed. The rationale is the same as for pure pentode mode. Such "iron" is costly. |
|
|
| salas |
| As typically as it is implemented in amplifiers that have a triode/UL switch, the triode mode is sounding bigger, with less harshness, more depth & sonic info. Some say that UL screws the 1st Watt linearity although its total bench result seems best preferable. |
|
|
| aletheian |
| quote: | Originally posted by Eli Duttman
Ray,
To squeeze the absolute maximum out of ultralinear topology, a separate screen grid winding connected to a dedicated/regulated supply is needed. The rationale is the same as for pure pentode mode. Such "iron" is costly. |
Is the requirement lessened any with KT88 and it's variants with a regulated plate supply and strice class A? I would think so. |
|
|
| Eli Duttman |
| quote: | | Is the requirement lessened any with KT88 and it's variants with a regulated plate supply and strice class A? I would think so. |
I don't see where regulated plate B+ and Class "A" matter. When the instantaneous plate voltage drops below the screen voltage, linearity is poor. UL topology is just another form of local NFB. All NFB works best when things are reasonably linear to begin with.
Since KT88 variants were mentioned, look at the TT21. You can really max. out the plate/g2 differential with that tube. Somebody with "deep pockets" could come up with a VERY nice amp using TT21s and separate screen grid winding "iron".
TT21 data sheet here. |
|
|
| aletheian |
| quote: | Originally posted by Eli Duttman
I don't see where regulated plate B+ and Class "A" matter. When the instantaneous plate voltage drops below the screen voltage, linearity is poor. [/URL] |
Ah, i see... i was assuming that with regulation and avoiding the fluctuating current demand of class AB, the issue of screen regulation would be less important. |
|
|
| Kent Smith |
| I like UL, but it's not a perfect cure-all fix. Using the standard transformer taps, I usually have to tweak it with some series resistance to get it right.... or add global negative feedback. Pentode mode with good screen supply and some nfb can sound really good too. I dunno.... guess it depends on tube, topology, and preferences as to which path to follow. You can make a good or a bad sounding amp with any method. |
|
|
| tubelab.com |
| quote: | | As typically as it is implemented in amplifiers that have a triode/UL switch, the triode mode is sounding bigger, with less harshness, more depth & sonic info. Some say that UL screws the 1st Watt linearity although its total bench result seems best preferable. |
There may be some truth in this statement, but I think that there is more going on here. I have built an SE amplifier that has a triode - UL - pentode mode switch. It also has switchable cathode feedback in the output stage. This amp has made the rounds of several listening sessions, and amp "sound off" tests. In a few cases the same people were present, but the room, speakers, and some of the music was different. In all but one case the preferences were not the same for each listener.
I use this amp in two situations. My living room is 12 by 14 feet. My speakers are made by puting modern Silver Iris drivers into a pair of 60 year old Zenith console radios. Sensitivity is 96 db. These speakers are loud with about 5 watts, so I usually use triode mode without cathode feedback. When I want to get really loud, I use UL with CFB.
My lab has Yamaha NS-10M studio monitors the efficiency is 86db. You need maximum power to drive these,so I use UL with CFB.
We took the amp to a friends house who had some 106db speakers. UL sounded terrible on these speakers. The amp played nicely all afternoon long in triode with no feedback.
There are experiments being caried out in another thread to determine if ul could be improved upon.
See the distributed load thread. |
|
|
| ray_moth |
| I wonder if the particular speakers used could have anything to do with it? Or maybe an inferior OP transformer would give unpleasing results? |
|
|
| salas |
With sound systems in general, the good matching of drive-load-room-program, the well thought out system architectural priorities and the successful catering for them, can create synergy that can baffle our almost practically proven technical cornerstones.
But it should not baffle us at all, since these are two plains of thought that belong to a different level. |
|
|
| Cobra2 |
I see the difference between UL & triode, but what about UL vs Pentode? Where the power-difference is out of the equation...
Arne K |
|
|
| crispycircuit |
| I like pentode drive. It's more dynamic and the top end is more extented. The music has more punch. More in your face sound. That's the type of sound I like. It could be some prefer a smoother less bright amp. A laided back preference.... |
|
|
| salas |
| The pentode or to a lesser degree UL sonic mark, owes a lot to global feedback (& local screen FB for UL). There is a cap in the pf range most of times on top of the FB return resistor. Many times its ceramic. Although on a scope screen it is adjusted to eliminate ringing to the trail of a 10kHz square wave and looks technically solid, I have experienced it producing moderate midhigh pinch and blackening of the soundstage. I would suggest experimentation with the quality of that cap. |
|
|
| EC8010 |
| I don't understand where this business of "U/L screws up the first Watt" comes from, particularly when coupled with comments about the linearity of a pentode when the anode voltage falls below the screen grid voltage. The anode voltage only falls below the screen grid voltage on negative peaks at maximum power - not minimum power... |
|
|
| Bandersnatch |
| quote: | Originally posted by EC8010
I don't understand where this business of "U/L screws up the first Watt" comes from, particularly when coupled with comments about the linearity of a pentode when the anode voltage falls below the screen grid voltage. The anode voltage only falls below the screen grid voltage on negative peaks at maximum power - not minimum power... |
I am with you on this one. Most recently a friend built an open loop U-L amp. 807's with a 10k a-a load. Three taps to choose from at 20, 30 and 40%. He has fairly efficient speakers( that are comfortable with SE2A3's ), and the First Watt is often in play. Can't say I heard anything I didn't like.
I build U-L amps as well, though they're E-Linear 2-stage types that have a fair bit of FB due to the E-Linear driver/PI connection. I don't think I've found the First Watt to be troubled with these amps. The speaker sensitivity is ~100dB 1W-1m and I am often in the sub-Watt power output range.
I don't remember reading *ANYTHING* about how U-L could screw up the first Watt until fairly recently. I think it is just a bunch ov propaganda or dogma...
I think I'll use my Truck-ma to deal with *THAT* dogma...:)
cheers,
Douglas |
|
|
| SY |
Reality check- at least a million UL amps have been sold or built. They've been around for 50 years. If there were truly a fundamental flaw, you'd think it would have been noticed by now.
It really gets down to the output transformer. UL is more critical than other topologies to how tightly the primary is coupled. Use a mediocre OPT and ultralinear will sound absolutely terrible. And then you can go on the internet and tell everyone that the topology is the cause. |
|
|
| Cobra2 |
This confirms my belief; the OP trafo is 75-90 % of a tube amp...
Arne K |
|
|
| Bandersnatch |
| quote: | Originally posted by SY
Reality check- at least a million UL amps have been sold or built. They've been around for 50 years. If there were truly a fundamental flaw, you'd think it would have been noticed by now.
It really gets down to the output transformer. UL is more critical than other topologies to how tightly the primary is coupled. Use a mediocre OPT and ultralinear will sound absolutely terrible. And then you can go on the internet and tell everyone that the topology is the cause. |
And even with a primo output TX, taking signal from the secondary and sticking it back into the input stage( a few stages upstream from where it came out ) is still a bad idea.
cheers,
Douglas |
|
|
| SY |
| quote: | | And even with a primo output TX, taking signal from the secondary and sticking it back into the input stage( a few stages upstream from where it came out ) is still a bad idea. |
Why? |
|
|
| salas |
| quote: | Originally posted by salas
As typically as it is implemented in amplifiers that have a triode/UL switch, the triode mode is sounding bigger, with less harshness, more depth & sonic info. Some say that UL screws the 1st Watt linearity although its total bench result seems best preferable. |
''Some say'' is hardly a dogma and mostly a reference to what has been wrongly or rightly speculated. Sonic differences between very well implemented triode and UL modes are more down to amount and type of feedback IMHO. |
|
|
| bear |
Time to step in front of the whirling blades...
My feeling is that there are two separate issues here:
pentodes/tetrodes vs. triodes
and
UL vs. Pentode
In general terms, my ears tell me that pentodes and tetrodes seem to have a kind of "zing" added to the middle upper frequencies. A sort of "presence" rise, most noticeable with things like cymbals in particular.
I don't think the zing is natural nor in the original recordings.
Strapped pentodes run in triode operation seem to exhibit far less of this effect than do "real" triodes, but it is still slightly present.
What can we surmise is the cause of this? Well, clearly there must be some sort of interaction between the screen grid and the other elements. Imho, it may be entirely due to mechanical variation under load!! and not some purely electronic factor!! Regardless, this is my take on the "pentode sound".
The major reason that so many pentode amps have been made and sold is two fold:
- they make more power than triodes, and power sells
- no one made very many indirectly heated triodes, and DHTs are not as rugged, many more pentodes were made than triodes in general.
Nothing to do with quality, quality of sound or any other factors, imho.
There is a way to run two tubes together, as a "combined" output section wherein a triode and a pentode are both used in (p-p) parallel... that bears some investigation wrt the sound and vis-a-vis UL vs. Pentode...
Now, UL vs. Pentode...
I don't think you lose quite 50% power going from UL/Pentode to triode operation. In the case of a typical ST-70 you go from ~35 to >20watts... but I guess that's close to 50%. But 50% is just 3dB. So... :rolleyes:
Many of the commercial amps of today run the screens into an over current situation on peaks... not good. Adds non-linear distortions.
Of course if you "tame them" and drop the screen current to an appropriate max level, then they don't make the full "marketing snot" that they claim anymore... ;)
I tested a popular make (California mfr) amp that had "triode/pentode" switching some time ago. Removing the input circuit from the equation (ick!) and driving the grids in AB2 from a very high quality source, and doing all sorts of tests and changing of bias points, etc... (the output iron was quite surprisingly very good, btw - way better than the amp's circuit) I found that the tubes worked best and had the least non-linearity when run with a regulated screen supply that was optimized within a few volts for each particular tube pair! Either side of the optimized screen setting, not as good at all. Using the UL taps was rather less wonderful looking on the scope...
Of course, I've also worked with and on all the Dyna UL models, and most of the Acros, and many of the Heaths, and all sorts of other varied UL amps...
The best looking results I ever had from a pentode tube was with that test set up I mentioned above and the optimized screen voltage, regulated...
What I think is at play here is that the UL mode offers extra power and decreased quiescent current, plus a "cheaper build" given the fewer parts for that pesky screen supply, and has a certain harmonic structure that is inevitable (factoring out the inherent screen "zing") and different than the pure pentode.
Feedback, local or not does not cure the inherent tendencies of any given circuit's harmonic spectrum - it might reduce it or even alter the ratios, but the underlying part is still always there.
My personal experience says that wrapping feedback around an output transformer, or even an interstage transformer is a receipe for adding more problems than it usually will solve. Why? Because of two main things - phase shift that can't be undone, and the effect of most global feedback in tube amps is/was to flatten the frequency response mostly for the benefit of marketing specs over any sonic benefits.
If you want to make a tube amp flat, the way to do it is to start with really, really, really good transformers that extend out as close to (or beyond) 100kHz. as you can get, and have ample inductance to go low enough. Then if you have a clean, low distortion, well balanced driver and input stage, you've got a really good, low distortion amp to begin with.
Anyhow, I have yet to hear a UL amp that did not actually sound better with the screens tied into "triode" operation. I lay this off to the "screen zing" problem in the main.
So, the dislike of UL may stem from the above points, I would expect.
_-_-bear |
|
|
| phn |
Feedback and interstage transformers is considered a no-no. It can be done, but is difficult and ultimatelly should be avoided.
What is an output trafo but an interstage trafo? Then add "UL feedback." So we have feedback within feedback and a trafo! I think not. |
|
|
| Kent Smith |
| I guess Bear makes some good points.... UL is the cheap and easy Band-aid! Do it with Hammond transformers and you get such a soft sound, it will make any recording sound smooth and listenable! (as long as you don't miss the details!) |
|
|
| Yvesm |
| quote: | Originally posted by bear
Time to step in front of the whirling blades...
My feeling is that there are two separate issues here:
pentodes/tetrodes vs. triodes
and
UL vs. Pentode
|
It seems to me that another issue is missing : PSU regulation.
Penthode are relativly insensitive to plate voltage sag while triode are.
But penthodes ARE very sensitive to screen voltage, and - simply because they have higher Mu - most sensitive to bias shift than triodes.
Keeping screen and bias supply stable is easier than keeping plate supply stable.
Specially if PA stage is run far in AB1.
PSU is an integral part of the PA stage design.
So, my preferences go to screen and bias regulated penthodes.
UL then calls for separate screen windings and becomes less attractive !
Yves. |
|
|
| Bandersnatch |
Hey Yvesm,
What pentodes are you speaking of with higher mu? I have a few data sheets in front of me ( for excellent audio valves ), that show mu g1g2 as low as 3. It of course extends right up to 211 levels for some, but a lot are less. The value spread is pretty comparable to power triodes.
cheers,
Douglas |
|
|
| Yvesm |
Hey Douglas,
I'm not sure to understand what you mean.
I was speaking about penthodes, tied as penthodes !
Yves. |
|
|
| Bandersnatch |
| quote: | Originally posted by Yvesm
Hey Douglas,
I'm not sure to understand what you mean.
I was speaking about penthodes, tied as penthodes !
Yves. |
It would appear that I misunderstood your original post. Pentode gain is gm x Load. Mu I took as an intrinsic device property, and not one defined by, or dependant on other circuit details. For pentodes, this would be mu g1g2 and not the in-circuit gain.
Also it appears that the lower mu g1g2 valves exhibit a larger change in plate current( plate voltage and g1-k voltages held constant ) than the higher ones when g2 voltage is moved around.
cheers,
Douglas |
|
|
| bear |
Perhaps I'm missing something about the way UL works? :xeye:
The screen is set at some nominal voltage Vs, which then is modulated by the output signal, in phase with the plate signal. In effect, it is bootstrapped to the plate's AC variation. Yes?
So, on positive going peaks, the screen is driven HARD positive beyond the normal voltage limits of Pentode operation (or tries to be), no?
Is this feedback, or feedforward??
_-_-bear |
|
|
| Yvesm |
| quote: | Originally posted by bear
Perhaps I'm missing something about the way UL works? :xeye:
The screen is set at some nominal voltage Vs, which then is modulated by the output signal, in phase with the plate signal. In effect, it is bootstrapped to the plate's AC variation. Yes?
So, on positive going peaks, the screen is driven HARD positive beyond the normal voltage limits of Pentode operation (or tries to be), no?
Is this feedback, or feedforward??
_-_-bear |
Yes, screen follows the plate.
It's feedback because, plate voltage rises as the result of reduced current (G1 more negative) and this rises the screen thus trying to increase plate current.
The more I go slower, the less I go faster :D
Yves. |
|
|
| Bandersnatch |
| quote: | Originally posted by bear
Perhaps I'm missing something about the way UL works? :xeye:
The screen is set at some nominal voltage Vs, which then is modulated by the output signal, in phase with the plate signal. In effect, it is bootstrapped to the plate's AC variation. Yes?
So, on positive going peaks, the screen is driven HARD positive beyond the normal voltage limits of Pentode operation (or tries to be), no?
Is this feedback, or feedforward??
_-_-bear |
For tapped anode winding U-L, plate V, g2 V and B+ are essentially the same. Run the plate up, and g2 goes up the tapped percentage. With a triode strapped pentode, the g2 exactly follows the anode, and g2 ratings are modified for this arrangement.
Turn the valve 'ON' all the way, and g2 is now above the anode voltage, at a point between anode V and B+ defined by the tap location.
Run U-L with a cathode FB and one can make g2 what ever one wants. Same as with a tertiary winding. Or, as in some Chicago and Plitron OPT's, one can tap g2 from the anode coil, and add the tertiary winding in the cathode to the U-L 'action'....:)
cheers,
Douglas |
|
|
| salas |
| A jumper reference from another thread really, but absolutely relative since there is brief, easy to grasp, working modes explanation plus sonic discussion, between variations of pentode, UL, triode strapped, and SET using the same basic platform. Click square #10 to start. #13,14,15 are directly relative to our discussion. |
|
|
| Dave Cigna |
| quote: | Originally posted by bear
So, on positive going peaks, the screen is driven HARD positive beyond the normal voltage limits of Pentode operation (or tries to be), no?
Is this feedback, or feedforward?? |
When the grid goes positive, which way does the plate (and screen) go? Negative. So, if it's any kind of feedback, it's the negative kind.
-- Dave |
|
|
| Wavebourn |
| I'm satisfied with pentodes when they run with low regulated g2 voltages. For example, in my 6L6 version I use 375V for plates and 315V regulated for screen grids, in my GU-50 version I use 750V plate and 300V regulated for screen grids. No need for more expensive transformers. |
|
|
| ray_moth |
It seems to me, from what has been said so far, that the separate winding cathode feedback approach may have more promise than the screen taps approach, if the results claimed for it are genuine. Williamson & Walker certainly thought so but they would say that, wouldn't they?
Cathode feedback allows the plate voltage to be much higher than the screen voltage, which might suit some tubes very well (e.g. 6L6, 807, even EL34). It also allows the screen to be regulated, which some see as critically important. A separate winding for the screens also gives screen voltage flexibility.
Transformers with separate windings, for either cathode or screen, are not rare but they're not cheap either. Perhaps that's just as well, though, assuming that the quality is reflected in the price, because the indication seems to be that low quality UL transformers give low quality results. |
|
|
| Wavebourn |
I do not know abour W&W what they would say, but back in 1960'th my dad had a tube tape recorder with SE 6P14P output that had a cathode winding. The recorder itself was so-so, but the amp was great and sounded gorgeous when I connected my DIY guitar to it. :)
I would try now such trick with GU-50 PP, but where to get output transformers? May be it is possible to order them somewhere relatively cheap?
What ratio is optimal for cathode windings?
|
|
|
| Bandersnatch |
| quote: | Originally posted by Wavebourn
I do not know abour W&W what they would say, but back in 1960'th my dad had a tube tape recorder with SE 6P14P output that had a cathode winding. The recorder itself was so-so, but the amp was great and sounded gorgeous when I connected my DIY guitar to it. :)
I would try now such trick with GU-50 PP, but where to get output transformers? May be it is possible to order them somewhere relatively cheap?
What ratio is optimal for cathode windings?
|
The GU50 looked to me like it wanted a fairly high( numeric ) load, like ~10k including the cathode winding. Say 7k5 + 15% in the cathodes. Increasing the CFB makes driving it more difficult in terms of delivering the large swing.
I am using Chicago BOH-6's in a 6V6 amp. They sound brilliant. Can't be all that difficult to acquire another pair, or a single so it can be duplicated. The Acro TO-350 is a CFB output, and IIRC Sowter can wind a good copy. I had the big Dynaco A441 and it was not a spectacular output. I wound up selling mine instead of keeping them to copy at a later date if I wanted more.
cheers,
Douglas |
|
|
| Wavebourn |
I've found too that GU-50 in P-P are happy whit 10K load with 750V on anodes, but why 15% from cathode windings?
Driving is not a problem, 6P15P in triode connection will be overkill for them. |
|
|
| Bandersnatch |
| quote: | Originally posted by Wavebourn
I've found too that GU-50 in P-P are happy whit 10K load with 750V on anodes, but why 15% from cathode windings?
Driving is not a problem, 6P15P in triode connection will be overkill for them. |
I can only answer with a question: what do you think it should have?
Second, why not get a proper triode instead of strapping a pentode. There are *MANY* that will do well.
I have built a few CFB amps with about that same size tertiary windng. While maintaining class A operating points, the B+ begins to limit the deliverable voltage swing. Less would not give enough FB, and the amp would be sensitive to reactive behaviour as is usually seen in the bass frequencies. One could always pick a speaker that was better behaved...:)
cheers,
Douglas |
|
|
| Wavebourn |
Do you mean 15% of voltage, or impedance ratio?
Speaking of pentodes... Triodes are in big demand by audiofiles, so I've found that I save *lot* of money strapping 6P15P and 12L6 tubes as triodes. Triodes of such quality are very expensive. |
|
|
| richwalters |
Strapping pentodes as triodes is one of my favourites. Despite mixed commentaries regarding UL, I've found tampering with the R/C screen snubbers makes a large difference in sound quality. Skill is required as g2 snubbers is part of the NFB network and can effect overshoot originating elsewhere in circuitry. I.M.O the 10Khz square wave test at half power into a dummy load is ** tough and highly revealing and all the triode amps I've sampled can't create decent waveform without massive slewing.
The AB2 amp I botched up from clear 6550B's is indictative how excellent the UL class is. Beat this waveform ?? |
|
|
| Bandersnatch |
| quote: | Originally posted by Wavebourn
Do you mean 15% of voltage, or impedance ratio?
|
That would be voltage ratio.
Douglas |
|
|
| Wavebourn |
| quote: | Originally posted by richwalters
Strapping pentodes as triodes is one of my favourites. Despite mixed commentaries regarding UL, I've found tampering with the R/C screen snubbers makes a large difference in sound quality. Skill is required as g2 snubbers is part of the NFB network and can effect overshoot originating elsewhere in circuitry. I.M.O the 10Khz square wave test at half power into a dummy load is ** tough and highly revealing and all the triode amps I've sampled can't create decent waveform without massive slewing.
The AB2 amp I botched up from clear 6550B's is indictative how excellent the UL class is. Beat this waveform ?? |
Unbelievable!!!
Can you draw your trick with resistors/capacitors?
Most clever that I managed to invent was putting bright red LEDs in screen grids to indicate saturation. Electro-human feedback" Hey dude, pull down faders!" :smash: |
|
|
| Dave Cigna |
An old trick to get cathode feedback is to put the secondary in the cathode circuit. You won't get a lot of feedback; connecting the cathode to the 8 ohm tap on a 5k SE transformer only puts 4% of the windings in the cathode circuit, but it's worth trying if you want to experiment.
It can be done with a PP transformer if there are 4 and 16 ohm taps. Ground the 4 ohm tap and connect the cathodes to the 16 and 0 ohm taps. Trouble is, the secondary of most PP OT's of that kind (vintage consumer grade stuff) is not going to be very well balanced either side of the 4 ohm tap, so results might be marginal.
-- Dave |
|
|
| richwalters |
Wavebourne# As you know going into AB2 with a conventional p-p UL setup carries a severe overcurrent penalty, in my case Iq = 60mA and full load (which is somewhat unreal) can hedge 200mA per tube and drastically limit tube life. (Although is tempting Don't do cont sine wave on this!) My safe hedge is to limit B+ in my amp to 350V and this provides roughly 35W from a pair of glass button gettered 6550B's. (Be honest ...I would have never expect such craving performance from such boring plain looking tubes). ??
The trick... since I use ECL82 pents (as triodes) Lowish'' Z phasesplitter drivers and used the surplus ECL82 triodes signal section as diodes which behave soft. The caveat is to run the o/p tubes with an over impedanced primary 4.5KA_A. (43% taps). For the 6550 stats this would appear to break the rule books regarding power throughput and forcing a 5Khz sinewave through the block gave an overload waveform too ugly i.e it was creating foldback rectification voltage limiting by upsetting the real DC component of g1. SO I did the old AB2 guitar amp trick of strapping the ""tube diode"" from g1 to fixed bias neg via resistor. (cathode to signal) This forces the o/p tubes to respond and care must chosen with series resistor value to actual power throughput.
The initial problem was high THD at 10Khz. The output stage impedance reflection match was wrong ( I knew it was) so I tried AB2 forcing and it worked and reduced thd by more than half. A masssive improvement. It can only work if the power supply can hold up such enormous signal transients and not voltage droop. Bear-in-mind AB2 condition is really intermittant i.e music and speech, and doing sine wave tests can also Jarr up the output tranny ! Fit a fuse !!
To the crux.The g2- anode snubber R/C value is critical for optimal waveform squarewave ripple BUT the Global NFB (my case 20dB)also throws the towel in as well as the phase compensation in the first stage in a 4 stage amp. This is the surprise jewel....A good amp design of nom 20dB global NFB should be able to take an extra 15dB NFB before instability sets in. I found improved transient reproduction with correct g2 to anode RC components, in my case with a good quality o/p tranny with low leakage L , I arrived at the values of 1K5 and 330pF (silvered mica). This was a vast improvement to the recommended 1nF/1K which created considerable overshoot. The low freq end, using a 100W p-p o/p tranny on 35W should enable a single frequency figure power throughput. In my case 7Hz. This is too low so I reduced phasesplitter coupling cap values to create 1% thd at 10Hz. This created enough loop gain (15dB) at 25Hz to effectively dampen the LS. Got the Jist ?? With single shot LF square wave YOu might find the amplifer transient response also coincides with optimum Q. Yes.
More anon
richj. |
|
|
| gingertube |
Rich,
Yes those zobel networks from screen to anode are "required". They make a huge difference to the Ultralinear Sound. I routinely put them into every Ultralinear Amp I build and so it was one of those things I forgot to mention because I do it without thinking about it. Another local constructor reported a huge improvement as well when he followed my suggestion to him to put them in.
On one amp using 4 x KT88 in Parallel Push Pull Ultralinear into a Hammond 1650T tranny I actually borrowed some gear from work and did full magnitude and phase plots (Bode Plots) before and after fitting the zobels and it was quite clear that it was effectively suppressing transformer resonances. I used voltseconds methods to optimise the zobels. Here is the link
http://www.siteswithstyle.com/VoltS...g_in_xfmrs.html
Interestingly I ended up with different zobels on the push and pull sides and that is something that I've never seen before. I used teh Rd_opt_Zi values.
Method (in brief) was to build the output tubes stage (i.e no phase splitter/driver) and drive one side at a time from the signal generator while looking at the anode signal on the CRO, with the otherside at DC idle. Then identified the side with the lowest resonance frequency and tackled that first.
A big clue as to why zobels might be required was given in my experience with the Menno Vandervenne VDV70/100 amps I built some years ago. (Parallel Push Pull EL34 into a Toroidal Output Tranny with badwidth to >200kHz). The amps had 150R series screen resistors. It worked fine in triode mode or pentode mode but suffered "squegging" (bursts of high frequency oscillation) when in ultralinear mode. In that case I did'nt get around to fitting zobels but I did confirm that increasing the screen series resistor from 150R to 1K cured the problem. Menno since posted this "fix" on his web pages.
Cheers,
Ian |
|
|
| richwalters |
In a well balanced/ made o/p p-p tranny the Zobel should be identical each side. One little trick I often do is to run a 5Khz square wave at low power through the amp into a dummy and substitute the global nfb resistor with a pot and see how much extra nfb the amp can accept before instability sets in. This is a pretty good test and at the same time one can tweak front end zobel and other values. Oscill'Probe earthing is also important.
Bear in mind (big admission) with big amps it can be frustrating switching on and off and waiting for warm up adjusting components but I always fit a hot switch which can turn the B+ off. I've been in the tube business for some time and acquired the (bad) habit fiddling with circuits while B+ is up......and poking with an unearthed soldering iron..(who hasn't done this ????).
richj |
|
|
| salas |
| quote: | Originally posted by Wavebourn
What ratio is optimal for cathode windings?
|
LUNDAHL uses 25% in its LL1620CFB model. |
|
|
| Wavebourn |
| quote: | Originally posted by salas
LUNDAHL uses 25% in its LL1620CFB model. |
25% of impedance or voltage? |
|
|
| salas |
They state: ''One primary winding on each coil has been split in half to support push - pull cathode feedback applications with 25% feedback''
And here is how they connect it: |
|
|
| Kent Smith |
| In reading others experiences and methods, it strikes me how much effort is required to get UL working right! Maybe straight pentode or triode mode is actually easier and less sensitive to circuit variations!?? |
|
|
| SY |
| Having designed and built multiple examples of all three (and a few more) I don't think so. Each one has its own set of issues that has to be dealt with. Triodes have reactive inputs and need regulated plate supplies (or can only be run class A), pentodes need tightly regulated screens and tend to be very load sensitive, UL needs tight coupling in the transformer and better plate regulation than pentode. Circlotron needs lots of drive and two floating power supplies per channel. Unity coupled needs a transformer with tight multifilar winding. And so on and so on. |
|
|
| ray_moth |
SY, it still seems to me that UL has its unique features. Only in UL does the screen current go through some part of the primary winding; it doesn't with either pentode or triode. It follows that the OPT should therefore be more of a factor in the quality and stability of a UL amp than either of the other two.
Of course, a good quality OPT is still important with any of these topologies, but I can't help thinking that triode offers the simplest solution. You need to add little or no NFB with triodes (depending on the needs of your speaker), so instability at the extremes of the frequency spectrum is unlikely to pose a problem. |
|
|
| SY |
| quote: | | SY, it still seems to me that UL has its unique features. |
Yes it does. That was exactly my point. So do triode and pentode, just different ones in each case. Triodes have input capacitance issues and (to operate properly) need tightly regulated plate supplies. Those are not trivial considerations. |
|
|
| Bandersnatch |
| quote: | Originally posted by SY
Yes it does. That was exactly my point. So do triode and pentode, just different ones in each case. Triodes have input capacitance issues and (to operate properly) need tightly regulated plate supplies. Those are not trivial considerations. |
Hey Sy,
You keep saying that triode amps need tightly regulated plate supplies. How tightly regulated? Perhaps it is just our definitions of 'trivial considerations' differ. An LC filter for a Class A PP amp is just a minimum starting point. If it needs to be really stiff, try Hg vapour diodes...:)
It seems that the triode is the least concerned with its exact operating point. Shift it around a little bit and nothing big hapens. OTOH, move a pentode amp's plate supply a bit and all on a sudden you're driving the load line out through the knee and odd HD starts happening in big way.
Anyway, I just rigged the output stage of my latest amp 40% U-L. I have the choice of 20 and 30%( though also the trouble of dealing with the unused taps ). 30% taps feed the E-Linear driver/FB stage. The OPT is a decent one that I've used previously. It's winding geometry is from the Peerless S-265. 10 layers on each side of the CT leaves convenient end-of-layer tap points.
cheers,
Douglas |
|
|
| SY |
One of the big reasons that mediocre triode amps can sound mushy and indistinct is the rails moving around with signal, especially if you run the amp sensibly in AB (which will challenge the Hg/LC passive approach). It will still work after a fashion, but will be suboptimal- analogous to UL with a less-than-stellar transformer.
The single biggest improvement to the triode amps I've built was putting in low impedance active regulation on the plates. That did more to unmush and tighten the sound than any other tweak I've tried, but is complex if it is designed to actually survive the rigors of real-world use. |
|
|
| gingertube |
Mucked about with the zobels on the 6V6 Ultralinear over the last couple of days.
Transformer is 8K : 4 Ohms
Np/Ns = 44.72
For Anode to Screen one side
44.72 / 2 * 0.57 = 12.75
So for my nominally 6 Ohm speaker on the 4 Ohm secondary tap we reflect 974.6 Ohms between anode and screen taps.
So choose 1K for the Zobel resistor.
With zero feedback do a frequency response sweep - noted that the first transformer resonance creates a peak at 74kHz.
for R = 1K and F = 74kHz then C = 2.15nF
BUT we are adding 2 of these things - one each side and they will be effectively in AC parallel so actually want 2 off 148kHz networks so that in parallel they give 74kHz.
So I fitted 1K + 1nF zobels between anode and screen each side. A quick check of the no feedback, full power frequency response shows that the resonance peak is gone and -3dB point which was 36kHz is now about 34kHz.
Put the feedback back in and sit down for a listen.
Much more detailed and refined, smooth, easy top end. A big improvement.
Cheers,
Ian |
|
|
| wrenchone |
| Interesting - I'm building an SE ultralinear amp using 1625 outputs. Has anyone tried plate to screen zobel networks in an SE amp? |
|
|
| richwalters |
| Ultimate check for zobel and general compensation is to slam a square wave between 5-10Khz though the amp about 2/3 power out into a dummy load and look at the quality of it. |
|
|
| Bandersnatch |
Hey Ginger!,
Does the class of operation have anything to do with your calculations? and is your output tap percentage exactly 43%? Seems like for an amp running Class A, you'd want 43% of 4k.
cheers,
Douglas |
|
|
| gingertube |
I'm running Class AB.
I did check the 10kHz squarewave at 6.5 Watts output which is about 75% of maximum power. Damping looked close to critical - that is no undershoot, very slight single half cycle of overshoot, edge speeds show evidence of slewrate limiting somewhere in the amp - most likely in the OT itself. I based this opinion on the fact that I could see the edge speed change part way up the slope as the amp transitions from Class A to Class B and the effective drive impedance changes.
I did originally try 1K + 2n2 for the zobels and there was evidence of over damping - serious undershoot which lead me to believe the zobels are in AC parallel. This also "closed down" the sound of the amp both in terms of apparent speed (attack) and in terms of the stereo image. This also stacks up with another theory of mine which is that a lot of the stereo imaging information is in the phase response which is why lots of global NFB kills image stone dead.
I've got to admit that I'm trying to fit theory to what I discovered empirically here.
Cheers,
Ian |
|
|
| Wavebourn |
| You may try Zobel between anodes to test your theory... |
|
|
| Miles Prower |
The only reason I don't like ultralinear is that once the OPT is wound, the die is cast. You take what they gave you. OTOH, parallel local feedback does the exact same thing, but it's tweakable so that it's NBD to determine what amount of local feedback works best. Don't like the initial results? Simply change one resistor and wash, rinse, repeat.
Parallel local feedback also works best with VTs where G2 requirements are significantly lower than the Vpk requirement (807s, 6L6s, most TV horizontal deflection PAs).
Other than that, it's just two different ways to accomplish the same thing. |
|
|
| gingertube |
Miles,
What do you mean by parallel local feedback.
In the 6V6 Ultralinear I've just finished I use Ultralinear connection plus balanced shunt feedback from the output tube anodes back to the diffamp front end (Baby Huey scheme with source follower MOSFETs in between diffamp outputs and output tube inputs).
Very similar to this one I posted yonks ago (post #58) except higher shunt feedback (27K in lieu of 13K shunt feedback set resistor) and some additional supplies so I could use my favourite little E-Line Mosfet - a 600mW Zetex ZVN0545A
http://www.diyaudio.com/forums/show...t=&pagenumber=6
Cheers,
Ian |
|
|
| Bandersnatch |
| quote: | Originally posted by Miles Prower
The only reason I don't like ultralinear is that once the OPT is wound, the die is cast. You take what they gave you. |
Not sure what you mean by 'the die is cast'. It is like that with*ANY* output. It is up to you to make it as useful as possible when designing it. Like using at least two sets of taps so the U-L ratio can be varied. Getting a secondary with a CT so a little bit ov CFB can be intorduced.
I'd say that the least useful would not even have U-L taps, and secondary taps. One primary, and one secondary...though a fine amp could be built with it and a few pentodes.
cheers,
Douglas |
|
|
| Yvesm |
| quote: | Originally posted by Bandersnatch
Not sure what you mean by 'the die is cast'. It is like that with*ANY* output. It is up to you to make it as useful as possible when designing it. Like using at least two sets of taps so the U-L ratio can be varied. Getting a secondary with a CT so a little bit ov CFB can be intorduced.
|
Mmmmh ?
I'm firmly convinced than taps in an OPT are diabolic :hot:
You can't have the same leakage inductance for all taps so, whatever you do, you are playing with more than one parameter.
This has to do with the interleaving scheme which is usually unknow, but LL is measurable any way.
Also consider than, the higher the screen tap are, the higher the screens swing and thus produce higher "Miller" effect. :bawling:
And so, we become more dependant of the driver internal impedance.
I just mean that what is observed in a specific rig may not be reproduced in another one even with same output tubes.
Yves. |
|
|
| richwalters |
| quote: | Originally posted by Yvesm
Mmmmh ?
I'm firmly convinced than taps in an OPT are diabolic :hot:
You can't have the same leakage inductance for all taps so, whatever you do, you are playing with more than one parameter.
Yves. |
Yes okay,....but that's the art in mastering it. It wasn't designed without a bonus. If you want power and negative feedback within the o/p stage then somewhere has to give ground for gain and performance. Part of the problem is that no two manufacturers come up with identical o/p trannies. All have their little quirks. One is for certain, the bigger the beast je more problems one comes up with.
richj |
|
|
| richwalters |
| Below is a common waveform problem of many typ UL o/p stages, poor square wave slewing caused by high interwinding capacitances. In this case, a high power 150W wide b/w has large core (10kg) so gets a large area copper exposure. Leakage inductance isn't a problem and maker claims self resonance around 60Khz. The problem gets worse with high drive at high frequencies and leads to waveform ripple. The cure is a Zobel across whole primary, but other parts with compensation i.e front end phase and cap fitted in global nfb line can also effect. |
|
|
| richwalters |
the 2nd photo is after correction.
Not every UL amp will require correction but in this case as the transformer has a 14 section design only a light Zobel compenation was required 500pF and 3K resistor. At full sine or square wave drive (rare in music) considerable power is dissipated and components rated accordingly. I used polypropl cap and metal film resistors.
G2 to anode snubber deficiency wil cause ringing on each waveform half.
richj |
|
|
| Miles Prower |
| quote: | Originally posted by gingertube
Miles,
What do you mean by parallel local feedback. |
See attached.
| quote: | | Not sure what you mean by 'the die is cast'. |
You can't change it once the OPT is wound short of ripping out the primary and rewinding it.
| quote: | | It is like that with*ANY* output. It is up to you to make it as useful as possible when designing it. Like using at least two sets of taps so the U-L ratio can be varied. Getting a secondary with a CT so a little bit ov CFB can be intorduced. |
Nice in theory, and wonderful when everything works just like you anticipated it would. Doesn't always happen, and electronic design is as much art as science. Do it long enough, and sure enough, some unanticipated gremlins will rear their ugly heads. In those cases, it pays to be able to improvise. |
|
|
| Wavebourn |
| A parallel feedback as shown is nice in theory (it is applied exactly to the same point where input is), and it is indeed very good for the output stage, but what about requirement for much lower resistance to drive it? |
|
|
| Miles Prower |
| quote: | Originally posted by Wavebourn
A parallel feedback as shown is nice in theory (it is applied exactly to the same point where input is), and it is indeed very good for the output stage, but what about requirement for much lower resistance to drive it? |
It's not a problem since I take care of it by applying the same "Rule of Five" that I use with solid state designs. |
|
|
| richwalters |
I wonder how many other p-p UL stages look this bad ?? Debugging a well-known kit with a squarewave reveals what a shoddy design it was.
richj |
|
|
| Miles Prower |
| quote: | Originally posted by richwalters
I wonder how many other p-p UL stages look this bad ?? Debugging a well-known kit with a squarewave reveals what a shoddy design it was. |
It also looks like there's some nonlinearity at the zero crossings. Were they driving it into deep Class AB2? Or is that just a cheap OPT? |
|
|
| richwalters |
Miles # That particular distorted UL waveform was within 2dB from rated power headroom in an output stage behaving goodenough class A i.e Iq 95mA quies was 105mA at 20W o/p(mere 10mA increase). Thd at 20W 1Khz was 0.03% very good.. but at 10Khz = 1%(!!) and there was the problem. Increasing primary current to 110mA per tube reduced thd to 0.5% is a cheap solution but wasteful.
The problem was an o/p tranny with high leakage capacitance, i.e somewhat oversized for the amplifier design. The result is an extended LF response bonus to single figures, I measured thd at 20Hz at around 0.3% confirming high prim inductance combined with an overkill of iron -> to reduce copper area which still cripples the HF response. So there is an equilibrium where oversize can ruin other parameters.
what I'm implying is there's no point using an expensive 18 sectioned o/p tranny where low power is used. Case of going over the books to see the tradeoff between core area, leakage inductance and capacitive losses. No easy task and law of diminishing returns prevails. So there are unskilled manufacturers still around.
I personally like to use o/p tranny size in a circuit where LF cutoff -3dB is around 20Hz giving 1% thd which is around the old industry standard and most designers work to this.
> The result of Zobel trimming is the Penalty for stability by reducing the upper b/w to -3dB at 30Khz not really acceptable for digital reprod but the circuit now accepts the electrostatic loudspeaker whereas before it didn't.
richj |
|
|
| smoking-amp |
"Optimised?"
Using a 43% lower DC screen voltage will just get you less power output like the triode case. (or a lot of distortion if power is not reduced)
My working hypothesis (working on data measurements to verify this yet) is that the "magic" 43% tap is just the point where screen grid current distortion (using B+ DC on the screens) = the 3/2 power law plate current distortion component, due to a low load impedance. Since they are out of phase they cancel largely. The 43% figure is specific to tube construction (mainly screen grid spacing or equiv. triode configured Mu), so will vary with tube. Nothing magic about 43%.
The effect of the screen current distortion is to syphon off plate current when plate voltage is lower than the screen V. (it is put back into the xfmr primary at a less effective tap) This effectively rounds off the usuall pentode knees in the curves on the left side. A load line then has near equal spacing of plate curves crossing it, effectively eliminating the usual 3/2 power inflation of the curves at high current.
Using a tertiary winding to reduce the DC screen voltage will eliminate the large screen current distortion component leaving one with an un-cancelled large 3/2 power plate distortion at the same power. So power output has to be reduced (back to triode level) to eliminate this high plate distortion.
Screen current distortion is a voltage output dependant factor (and a DC screen V factor). Plate current 3/2 power law distortion is a current output dependant factor (and load Z factor). When a real speaker is connected, inductive reactance will cause voltage and current to go out of phase, and the two distortion components will no longer cancel. So UL tests great on the test bench with a dummy load, but falls flat with most real speakers.
UL was just a marketing trick. There is however some hope that it can be fixed.
Don |
|
|
| smoking-amp |
Active Ultralinear design attached:
Half of the CCS current, times RL, determines the V DC drop on the screens from the plates for the main tubes V1 and V2. Can play with the connection of RL to the xfmr UL tap, instead of to the plate tap, to get correct bootstrap voltage across it. It will then act as an effective CCS load for V3 or V4 and give very high gain to the correction loop from V3 and V4 LTP. Ratio of R2 to R1 sets expected gain or Mu for the output tubes. When the output tubes are operating correctly at this Mu gain, the grids of V3 and V4 are stationary. Any deviation from this Mu rule upsets the V3 and V4 grids and gets corrected by LTP V3 & V4 varying the main tube screen voltages thru the Mosfet buffers.
Don |
|
|
| Johan Potgieter |
Hi Richwalters,
Is that waveform on your post #73 really that bad?
Examination (as well as I am able to from the drawing) shows a mainly 260 Khz damped ringing. The "serrated" cross-overs indicate the same ringing still, plus an over-compensation by some phase correction cap, probably the lead one over a feedback resistor. (It is not cross-over distortion although it looks similar.) Decreasing this or whatever other move is necessary, will probably show up the full 260 KHz damped ringing with little else wrong. If such ringing is bothersome it can probably be better cured in a somewhat different way.
The "bother" factor is also dependant on a stimulus being there to excite it; in this case the fast rise time of the signal generator. Would such stimuli occur in real life? If so, where is the supersonic filter that should keep it out? The presence of signals there can lead to other problems - no amplifier can be bother-free into the MHz region, any proper design is going to go reactive somewhere. And why should it be bother-free there - cut the frequencies that will stimulate that and forget about it. The same thing, even more so, is relevant to semiconductor circuits. Looking after what happens outside the amplifier pass-band is quite as important in good amplifier design as what is audible.
As you intimate, the rest of the design might well be questionable; I am commenting only with regard to output stages and ringing generated there (which it usually is).
Lastly, I have found in output transformer design a tendency to over-accentuate leakage inductance at the expense of interwinding capacitance. Many products happily go to 100s of KHz because of low leakage, but starts cutting not far above 20 Khz due to interwinding capacitance, especially with pentode circuits. It heavily depends on the design, naturally, but from practical product experience over a wide range of designs, I start running into primarily capacitive attenuation with more than 4 secondaries or so, especially on larger transformers (and I use C-cores which necessitate fewer windings).
Regards. |
|
|
| Johan Potgieter |
| quote: | Originally posted by smoking-amp
Using a 43% lower DC screen voltage will just get you less power output like the triode case. (or a lot of distortion if power is not reduced)
UL was just a marketing trick. There is however some hope that it can be fixed. |
Hi Don,
How to reply to this (again) ......
I am on record stating that I respect your analyses and heaps of work done on this subject, and I am looking forward to results of your particular topology. It will be useful.
But with respect, we seem to be going in circles about certain matters. I have been away from this thread for some time, and there still seems to be something on your mind that (dc) screen voltage must be lower than anode voltage to begin with, and that UL as is, is somewhat of a deception.
I am not also going to repeat, and will just suggest that such a stance does really simply not correspond with practice. There has been so many proper measurements over decades (let us forget about the few poor ones), indicating that UL operation gives most of the good characteristics of pentode and triode stages. I can again only refer to the first graph somewhat down in the reference given in post #76 (for KT88 operation). I have used and measured (on proper instrumentation - I hope you accept an HP spectrum analyser as good enough) similar results in countless cases. Without intending to offend, I must respectfully object to your sweeping simplification that UL "was just a marketing trick". Sorry, friend!
Abundant data and measurements indicate otherwise! Not my fancy; I am just the messenger.
Sure there are less successful designs, etc. etc. - they are not a criterion. And that it can be improved; I would be the first one to laud and adopt any such circuit that anyone can come up with. But for now I get the low distortion (almost triode but at almost pentode outputs and efficiency) shown in manufacturer's data sheets, and I honestly cannot see why I would need much else.
Generally, for the most, screen taps around 40% appear to give lowest distortion/watt output. Lower for 6V6s, but I have not used them much. In my output circuit based on the Quad II topology I need to stay at about 20 - 25% equivalent taps otherwise at the extra amplitude required by the drivers, their distortion begins to become dominent. (I use 6L6s.)
Just some comment on the often maligned taps: As someone said there is leakage reactance - but the effect of that is small compared to primary/secondary leakage because of the different turns ratio. That is part of transformer design anyway, also as said (Richwalters?) - I have never found that a problem. Not to present my results as laudible (I only know them best); is it not satisfactory that a 100W output stage (4 x 6L6GC) with quite a substantial output transformer can give a -3 dB point of 90 KHz after a slight peak and then roll off smoothly (output stage only) at 110W? Etc. etc. for reactive loads, not to bore members.
Lastly, the equivalency of local resistive NFB round a simple pentode stage; it could give similar results, yes. But such designs mostly include the driver stage, and it has been indicated before in an analysis by Prof Matti Otala, that there can be transient distortion problems in an output stage with inherently high internal impedance, reduced to acceptable values by external NFB. Some will argue that UL is also NFB; sure, but it is an inherent output circuit thing, compared to external voltage NFB including a previous stage.
Regards. |
|
|
| Tubes4e4 |
Hi smoking-amp,
| quote: | Originally posted by smoking-amp
"Optimised?"
My working hypothesis (working on data measurements to verify this yet) is that the "magic" 43% tap is just the point where screen grid current distortion (using B+ DC on the screens) = the 3/2 power law plate current distortion component, due to a low load impedance.
Since they are out of phase they cancel largely. The 43% figure is specific to tube construction (mainly screen grid spacing or equiv. triode configured Mu), so will vary with tube. Nothing magic about 43%.
|
My hypothesis is that all those usual values like 20%, 25%, 40% and even 43% just fit very nicely to equal turn number interleaving schemes (f.e, 43% is at 3rd connection of a 7 parts (equal turns!) primary winding scheme; of course 14 parts for PP).
Fits very nicely to effective production methods, eh. Honi soit qui mal y pense. Really, nothing magic about 43%.
Regards,
Tom |
|
|
| ray_moth |
| quote: | | Using a 43% lower DC screen voltage will just get you less power output like the triode case. (or a lot of distortion if power is not reduced) |
Johan, I think Smoking-amp's response was to the link provided by Cobra-2, namely, the 'Web Ace' site of Dennis Grimwood on the subject of what he calls 'Optimised Ultralinear'. Have you seen it?
Dennis has strong beliefs about how the voltages and taps applied to screens should be optimised. He says that the ratio of screen voltage and screen primary turns to the voltage and turns at the plates should be in proportion to the inter-electrode distances in the construction of the tube itself, so as to conform to the voltage gradient across the tube. Of course, that necessitates a separate winding for the screens, in order that a lesser voltage can be applied to them.
For example, according to Dennis, if the distance from the cathode to the screen is 50% of the distance from the cathode to the plate, then both the screen turns and the screen voltage should be 50% of those at the plate.
The way Dennis argues it makes it seem reasonable but, of course, that doesn't necessarily mean it's right! If you have read what he says, do you agree with him? |
|
|
| Miles Prower |
| quote: | Originally posted by ray_moth
The way Dennis argues it makes it seem reasonable but, of course, that doesn't necessarily mean it's right! If you have read what he says, do you agree with him? |
Yes, it does seem reasonable, however, I strongly suspect that this is a bit of self promotion. Here's what you'll find at the very bottom of the page:
| quote: |
IMPORTANT NOTICE
THE AUTHOR MAKES NO CLAIM WHATSOEVER AS TO THE VALIDITY OR ACCURACY OF ANY STATEMENT, INFORMATION OR OPINION CONTAINED IN THESE PAGES AND NO LIABILITY WILL BE ACCEPTED FOR ANY ERROR OR OMISSION OF ANY KIND WHATSOEVER.
|
That says it all, dontachathink?
Now, we've had some 100 years of experience with VTs. I highly doubt that there is something radically new and undiscovered just waiting out there. As for maintaining a plate voltage to screen voltage ratio that's porportional to the ratios of cathode to element spacing, the suggested Q-Point for the least THD operation of the 807/6L6 defies this:
Vpk= 365Vdc
Vsgk= 270Vdc
Rl= 6K6 (P-2-P)
THD= 1.8% (which is spot-on, as that's very close to what I actually measured with a pair of PP 807s)
270/365= ~74%, and he claims that the spacing ratio is closer to 50%. You also have a whole boat load of audio power finals where the screen and plate are suggested to operate at the same voltage: 6V6, 6AQ5, 50C5, 50L6, 6CU5, etc. The 6V6 & Co. are some of the best sounding VTs going. Of course, sometimes you can do better if you get that screen voltage down. Busting the specs a bit, and running at Vpk= 150Vdc -- 180Vdc, and getting Vsgk down to 90Vdc will improve the linearity of a 50C5 considerably. Given that the 50C5 is most frequently operated off a transformerless power supply connected directly to the AC mains, it's not surprising that the Q-Point for max audio power under those conditions isn't the same as that for least THD.
Here he claims:
| quote: |
Another and previously unpublished option to creating an operating environment where the Screen Grids will be at a DC potential sufficiently high enough to attract and accelerate electrons towards the Plates but, to maximise power output, not to collect and divert them to earth through the B+ supply, is the humble silicon diode semi-conductor rectifier.
By inserting a standard half-wave silicon rectifier diode in series with the Grid Stopper resistor, an electronic control circuit is created whereby the Screen Grid will be able to be energised at DC potential attracting and accelerating electrons towards the Plate - still electrostatically controlling current flow in the normal way - but blocking the flow of AC current from the Screen Grid back to the DC source - ie "one way traffic"
|
It's been tried and what it mainly does is poof VTs.
It's all part of his schtick to make you think he has something grand and glorious so's you'll buy his stuff. Self promotion, pure and simple. |
|
|
| Johan Potgieter |
| quote: | Originally posted by ray_moth
The way Dennis argues it makes it seem reasonable but, of course, that doesn't necessarily mean it's right! If you have read what he says, do you agree with him? |
Thanks Ray - and Miles!
To start with a guilty feeling ... I have read the Dennis document some time ago (before this correspondence), and should really have done so again! In that sense Smoking Amp may have been influenced and also been a messenger like I claimed before, and perhaps I owe him an apology for being harsh - not intended.
To keep this brief and yet say something meaningful..... The best I can do is to refer to Dennis' remarks by page - I have printed it all out over 26 pages; hope that helps.
No, I carefully do not agree with some of his major assumptions; I simply don't see the grounds for some, after I sat and pondered the document this evening. (Us losing the cricket against the no. 9 seed did not encourage further tv watching!) Perhaps the "bottom line" is that from tube basics he makes assumptions, and then go on from there.
His departing point, that geometric position of the G2 must correspond with the amplitude of signal entered there, does not necessarily follow; I can't see that. His argument is sound (and I must compliment him on an excellent description of what goes on inside a tube, for novices) as far as it goes, but the effect of an electrostatic field as generated by G2 is more complex than its geometric position. We are familiar with rp, the so-called plate resistance. Similarly there will also be an r.g2, measureable in the same way- that will play a role. G2 is not just another control grid at very high input impedance because it is negative; it has different characteristics. There is the grid wire pitch, wire thickness, etc. (It is not new to feed an analogue signal into G2, e.g. for control purposes, and one has to feed it right.) Since inside a tube everything is rather interrelated, I will not be surprised if part of his thoughts are true, but not in that simple way. [One can generate an electrostatic field without drawing a single mA (very high impedance electrode), or drawing quite a current. Consider e.g. the difference between the effect of G2 in pentodes and beam tubes, not to mention electrometer tubes. It's a subject on its own.]
Then there is the blank disagreement with published tube data. He veers towards transmitter tubes where Va of 600 - 1KV is normal: subsequently Vg2 must be lower. But in his eventual list there is not one audio tube! Have everybody been dense all these years?? Sorry. He stated that lower max. Vg2 than Va, so happening to be 43%, is common in tube specs. That is plainly not so for any audio tube, they are mostly equal! Not that max. specs have anything to do with typical operation - there he again goes awry. He alluded that as a result ratings are exceeded - again NO! That clearly depends on the designer; there is no danger to having Vg2 = Va provided the max. dissipation is not exceeded, which is basic design.
Regarding the "necessity" to never let Va go lower than Vg2 (signal-wise), also nonsense. It does so all the time in straight pentode designs, and proper designs never go unstable or berserk as a result of that alone. Again this practice veers toward transmitter-type voltages. You keep to that philosophy with any audio power tube as we know and use them, and you will be hard-put to get over 20W even from a pair of 6550s!
Then (where do I stop) there is the optimum of 43%. This has been debunked before by Smoking Amp - just to confirm, the GEC graph on p. 3 for KT88 shows different: The respective max. outputs as well as D remains constant from 20% - 41% taps. One tends towards the 40% side simply because the rp is lower there, other things being equal; nothing to do with distortion. The graph on p.17, also by GEC, is again contradictory to the above - strange; probably different operating conditions, not given.
By the way, that is another major problem I have with this epistle: Not a single result is given that I could find; it simply sounded excellent to the writer. Sorry, that does not constitute proof. At least the operations that he "disapproved" of, is well-documented in tube data from several manufacturers. One is inclined to ask: Debunk those and give your own results, or else.
Finally perhaps, I don't want to sound wise, but I cannot see the great mystery of how UL works - it is quite intuitive! Taking G2 from pentode to triode operation, must surely follow about the way it happens? The exact physics/maths might interest some - fair enough. But there are times that an engineering approach by measuring and drawing graphs like the KT88-example (they also exist for EL34), give a better picture than symbols. ("A picture is worth ... ") It not only shows tendencies, but also sharp changes/parameter (to stay clear of), etc.- I think you know what I mean.
There are other irritating points; one must stop some time. To sum up, regrettably Dennis' arguments start to unravel from about p. 16. As I said before, Smoking Amp seems to investigate all this, his methodology looks sound, and I for one would appreciate seeing his findings. I would just urge him to keep an open mind, not just tied to "Vg2-must-be-lower-than Va".
There - see what you have done, Ray_Moth. My longest contribution. Time for others to take it further.
Regards |
|
|
| ray_moth |
Thanks, Johan! This is well on-topic, since it examines someone's ideas about how to do UL right (in his view). I've been wondering, ever since I first saw Dennis's articles a few years ago (he's written several), what other practitioners would think of his ideas. As you say, he doesn't provide any proof that he's correct.
Another Dennis Grimwood article is on 'Optimised Electron Stream', where he ropes the suppressor into the UL argument on the grounds that it, too, sits part way along the path from cathode to plate. Any thoughts on that? |
|
|
| Miles Prower |
| quote: | Originally posted by ray_moth
Another Dennis Grimwood article is on 'Optimised Electron Stream', where he ropes the suppressor into the UL argument on the grounds that it, too, sits part way along the path from cathode to plate. Any thoughts on that? |
"It's ridiculous", comes to mind here. :D |
|
|
| tubelab.com |
I read through this web site a while back. I tried the diode in the screen grid trick, and it does provoke s particularly nasty sounding distortion near full power. I have not tested the optimized electron stream technology ( I am not sure he has either), but I would expect distortion due to secondary emision.
But I found these humorous quotes" It will be seen also from tube plate characteristic curves that the values of grid bias within the above range are such that most tube types cannot operate linearly in Class A - ie the positive alternation of signal will not produce the same plate current change as an equal negative alternation of signal.
Furthermore, any transient peak signal having an amplitude several times the RMS value will not be reproduced in the output stage load unless it is in the positive alternation of the AC signal - because any signal greater than plate current cutoff cannot be reproduced - ie the negative signal alternation will be truncated.
This is why a single-ended amplifier is fundamentally a waste of time and money for hi-fi.
This is also why a Class A push-pull amplifier is undesirable - unless it has sufficient headroom power to cater for all power output conditions required for high fidelity reproduction at the required listening level.
See paper by Williamson and Walker on the above topics.
I guess that I have been wasting my time and money for the last few years. So have many of us! |
|
|
| Tweeker |
| Well, it is a good reason to strive for fast overload recovery in any amp. But, theres a price to pay for more watts, and not just economic. Ei- Cmiller growing as you parallel devices and then having to get them biased right etc. In AB you have gm doubling or crossover distortion to contend with. |
|
|
| Dave Cigna |
| quote: | Originally posted by Johan Potgieter
By the way, that is another major problem I have with this epistle: Not a single result is given that I could find; it simply sounded excellent to the writer. |
Exactly! It seems that he has developed in his own mind an image of what the electric field looks like inside a pentode. Unfortunately, as evidenced by his conclusions, (UL tap ratio, Vg2 must be less than Va, etc...) his mental image is wrong and too simplistic. There are many textbooks that give good descriptions with diagrams of the voltage gradient, in particular how it is changed by grid wires. Apparently he has not seen them or did not understand them, or simply chooses to believe that he knows better!
The RCA Engineering Bulletin on Beam Power Tubes by O.H. Shade has the diagrams. A scan can be downloaded from tubezone.net. (Scroll down to the bottom of the page.) Just don't fall for the advertising hype. The publication was meant to sell RCA's new beam tubes, so it paints them in especially pretty colors.
-- Dave |
|
|
| salas |
| After all this excellent discussion I tend to conclude that what we like or don't like about UL sound is the sound of local feedback and OPT quality which becomes even more central? |
|
|
| Johan Potgieter |
MA!!! Tubelab has stolen my thunder! :bawling:
Seriously, thanks for saving me much trouble.
Ray,
I did glance through that G3-Grimbeek paper, until I came to the diode-in-G2 arguments and stopped. (Frankly, did not remember that those were also from this gentleman.)
Since you asked me, the G3 argument is much the same as the G2 one; again unraveling somewhere down the line. Since I got a salary for being inquisitive and going down roads less travelled (as a research engineer I was not paid to re-invent the wheel), I was rather like Dennis. Only one cannot simply declare the work of others also having been of that inclination void - I am referring to manufacturer's data. I cannot simply call such data "a barrier to further development". They want to sell their product after all! They must give reliable data regarding normal (safe) use. The only barrier to further developmind is in your mind, not their data.
After that philosophy, my remarks would be the same as before. Again I find funny statements, like "electrons attracted to G3 because it is at lower potential than the plate" (or similar). In normal operation, electrons are not attracted to G3 - how does a negative electrode attract negatively charged electrons?? (We then seem to need a whole lot more re-research than about where one simple grid should go. Like Tubelab said.) Electrons which bounce off the anode because of arriving there at speed, are prevented from flying back to the positively charged screen lurking just round the corner, by G3 exerting a repelling field to send them back home (the anode). In the process a few do get stuck on G3, theoretically causing a current, but that is totally negligible.
And that is about it! Again, not desiring to sound like an oracle, I also used G3 for signal purposes long ago. (The Mu is about 1!).
And extra grids could be used for a purpose - we all know about frequency converters. In exactly the same way the normal G3 was also designed for purpose: to prevent secondary or "back" emission from the anode. Fiddle around by all means, but do not expect to suddenly find a greater use for G3 than just what it was designed for. All engineers in the hey-day of tubes (never mind me) were not fools.
Again enough other statements popped up to keep a more or less fixed frown on this forehead. Sorry, no go.
Salas,
No; well not so critical as to be frightening. UL has been used for so long with mostly laudible results. Sure, there are the poor designs/output transformers; they are not the norm. Also, yet again, the "liking/not liking" of the sound of UL, ditto the above, as well as: You are also listening to a complete circuit, etc. etc. One cannot single out one element without examining the whole.
Mainly, there is no reason to be scared of it - least of all the urban legends that aboud. (I used to call them old wives' tales, but I suppose these days one must be politically correct ....)
Regards.
Edited: Typos galore - sorry |
|
|
| ray_moth |
| quote: | | The only barrier to further developmind is in your mind, not their data. |
Why pick on me? :D
Again, thanks, Johan. I find your response to those articles by Dennis Grimwood (not Grimbeek!), together with Tubelab's response, to make good sense. I suppose it all boils down to Dennis's assertion that engineers over the decades have got it wrong, while at the same time failing to provide any acceptable justification of why one should believe that he's got it right. (There's someone like that at AA, with whom I have had several altercations, but let's not go there!)
Dennis's 'Optimised UL' is, perhaps, somewhat plausible; however, his 'Optimised Electron Stream' is definitely not. Connecting the suppressor to a positive potential between plate and screen turns a pentode into a tetrode, and back comes the dreaded kink! Of course, he can apply his OES approach only to pure pentodes that have a separate pin for g3, so his beloved 6146s and 807s and other beam power tubes can't be considered.
Another proponent of UL with somewhat unusual ideas is Stan White, who designed the POWRTRON amplifier over 55 years ago. In his | | | |
